repo
stringlengths
2
99
file
stringlengths
13
225
code
stringlengths
0
18.3M
file_length
int64
0
18.3M
avg_line_length
float64
0
1.36M
max_line_length
int64
0
4.26M
extension_type
stringclasses
1 value
mmvae-public
mmvae-public/src/report/analyse_ms.py
"""Calculate cross and joint coherence of trained model on MNIST-SVHN dataset. Train and evaluate a linear model for latent space digit classification.""" import argparse import os import sys import torch import torch.nn as nn import torch.optim as optim # relative import hacks (sorry) import inspect currentdir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(currentdir) sys.path.insert(0, parentdir) # for bash user os.chdir(parentdir) # for pycharm user import models from helper import Latent_Classifier, SVHN_Classifier, MNIST_Classifier from utils import Logger, Timer torch.backends.cudnn.benchmark = True parser = argparse.ArgumentParser(description='Analysing MM-DGM results') parser.add_argument('--save-dir', type=str, default="", metavar='N', help='save directory of results') parser.add_argument('--no-cuda', action='store_true', default=False, help='disables CUDA use') cmds = parser.parse_args() runPath = cmds.save_dir sys.stdout = Logger('{}/ms_acc.log'.format(runPath)) args = torch.load(runPath + '/args.rar') # cuda stuff needs_conversion = cmds.no_cuda and args.cuda conversion_kwargs = {'map_location': lambda st, loc: st} if needs_conversion else {} args.cuda = not cmds.no_cuda and torch.cuda.is_available() device = torch.device("cuda" if args.cuda else "cpu") torch.manual_seed(args.seed) modelC = getattr(models, 'VAE_{}'.format(args.model)) model = modelC(args) if args.cuda: model.cuda() model.load_state_dict(torch.load(runPath + '/model.rar', **conversion_kwargs), strict=False) B = 256 # rough batch size heuristic train_loader, test_loader = model.getDataLoaders(B, device=device) N = len(test_loader.dataset) def classify_latents(epochs, option): model.eval() vae = unpack_model(option) if '_' not in args.model: epochs *= 10 # account for the fact the mnist-svhn has more examples (roughly x10) classifier = Latent_Classifier(args.latent_dim, 10).to(device) criterion = nn.CrossEntropyLoss() optimizer = optim.Adam(classifier.parameters(), lr=0.001) for epoch in range(epochs): # loop over the dataset multiple times running_loss = 0.0 total_iters = len(train_loader) print('\n====> Epoch: {:03d} '.format(epoch)) for i, data in enumerate(train_loader): # get the inputs x, targets = unpack_data_mlp(data, option) x, targets = x.to(device), targets.to(device) with torch.no_grad(): qz_x_params = vae.enc(x) zs = vae.qz_x(*qz_x_params).rsample() optimizer.zero_grad() outputs = classifier(zs) loss = criterion(outputs, targets) loss.backward() optimizer.step() # print statistics running_loss += loss.item() if (i + 1) % 1000 == 0: print('iteration {:04d}/{:d}: loss: {:6.3f}'.format(i + 1, total_iters, running_loss / 1000)) running_loss = 0.0 print('Finished Training, calculating test loss...') classifier.eval() total = 0 correct = 0 with torch.no_grad(): for i, data in enumerate(test_loader): x, targets = unpack_data_mlp(data, option) x, targets = x.to(device), targets.to(device) qz_x_params = vae.enc(x) zs = vae.qz_x(*qz_x_params).rsample() outputs = classifier(zs) _, predicted = torch.max(outputs.data, 1) total += targets.size(0) correct += (predicted == targets).sum().item() print('The classifier correctly classified {} out of {} examples. Accuracy: ' '{:.2f}%'.format(correct, total, correct / total * 100)) def _maybe_train_or_load_digit_classifier_img(path, epochs): options = [o for o in ['mnist', 'svhn'] if not os.path.exists(path.format(o))] for option in options: print("Cannot find trained {} digit classifier in {}, training...". format(option, path.format(option))) classifier = globals()['{}_Classifier'.format(option.upper())]().to(device) criterion = nn.CrossEntropyLoss() optimizer = optim.Adam(classifier.parameters(), lr=0.001) for epoch in range(epochs): # loop over the dataset multiple times running_loss = 0.0 total_iters = len(train_loader) print('\n====> Epoch: {:03d} '.format(epoch)) for i, data in enumerate(train_loader): # get the inputs x, targets = unpack_data_mlp(data, option) x, targets = x.to(device), targets.to(device) optimizer.zero_grad() outputs = classifier(x) loss = criterion(outputs, targets) loss.backward() optimizer.step() # print statistics running_loss += loss.item() if (i + 1) % 1000 == 0: print('iteration {:04d}/{:d}: loss: {:6.3f}'.format(i + 1, total_iters, running_loss / 1000)) running_loss = 0.0 print('Finished Training, calculating test loss...') classifier.eval() total = 0 correct = 0 with torch.no_grad(): for i, data in enumerate(test_loader): x, targets = unpack_data_mlp(data, option) x, targets = x.to(device), targets.to(device) outputs = classifier(x) _, predicted = torch.max(outputs.data, 1) total += targets.size(0) correct += (predicted == targets).sum().item() print('The classifier correctly classified {} out of {} examples. Accuracy: ' '{:.2f}%'.format(correct, total, correct / total * 100)) torch.save(classifier.state_dict(), path.format(option)) mnist_net, svhn_net = MNIST_Classifier().to(device), SVHN_Classifier().to(device) mnist_net.load_state_dict(torch.load(path.format('mnist'))) svhn_net.load_state_dict(torch.load(path.format('svhn'))) return mnist_net, svhn_net def cross_coherence(epochs): model.eval() mnist_net, svhn_net = _maybe_train_or_load_digit_classifier_img("../data/{}_model.pt", epochs=epochs) mnist_net.eval() svhn_net.eval() total = 0 corr_m = 0 corr_s = 0 with torch.no_grad(): for i, data in enumerate(test_loader): mnist, svhn, targets = unpack_data_mlp(data, option='both') mnist, svhn, targets = mnist.to(device), svhn.to(device), targets.to(device) _, px_zs, _ = model([mnist, svhn], 1) mnist_mnist = mnist_net(px_zs[1][0].mean.squeeze(0)) svhn_svhn = svhn_net(px_zs[0][1].mean.squeeze(0)) _, pred_m = torch.max(mnist_mnist.data, 1) _, pred_s = torch.max(svhn_svhn.data, 1) total += targets.size(0) corr_m += (pred_m == targets).sum().item() corr_s += (pred_s == targets).sum().item() print('Cross coherence: \n SVHN -> MNIST {:.2f}% \n MNIST -> SVHN {:.2f}%'.format( corr_m / total * 100, corr_s / total * 100)) def joint_coherence(): model.eval() mnist_net, svhn_net = MNIST_Classifier().to(device), SVHN_Classifier().to(device) mnist_net.load_state_dict(torch.load('../data/mnist_model.pt')) svhn_net.load_state_dict(torch.load('../data/svhn_model.pt')) mnist_net.eval() svhn_net.eval() total = 0 corr = 0 with torch.no_grad(): pzs = model.pz(*model.pz_params).sample([10000]) mnist = model.vaes[0].dec(pzs) svhn = model.vaes[1].dec(pzs) mnist_mnist = mnist_net(mnist[0].squeeze(1)) svhn_svhn = svhn_net(svhn[0].squeeze(1)) _, pred_m = torch.max(mnist_mnist.data, 1) _, pred_s = torch.max(svhn_svhn.data, 1) total += pred_m.size(0) corr += (pred_m == pred_s).sum().item() print('Joint coherence: {:.2f}%'.format(corr / total * 100)) def unpack_data_mlp(dataB, option='both'): if len(dataB[0]) == 2: if option == 'both': return dataB[0][0], dataB[1][0], dataB[1][1] elif option == 'svhn': return dataB[1][0], dataB[1][1] elif option == 'mnist': return dataB[0][0], dataB[0][1] else: return dataB def unpack_model(option='svhn'): if 'mnist_svhn' in args.model: return model.vaes[1] if option == 'svhn' else model.vaes[0] else: return model if __name__ == '__main__': with Timer('MM-VAE analysis') as t: print('-' * 25 + 'latent classification accuracy' + '-' * 25) print("Calculating latent classification accuracy for single MNIST VAE...") classify_latents(epochs=30, option='mnist') # # print("\n Calculating latent classification accuracy for single SVHN VAE...") classify_latents(epochs=30, option='svhn') # print('\n' + '-' * 45 + 'cross coherence' + '-' * 45) cross_coherence(epochs=30) # print('\n' + '-' * 45 + 'joint coherence' + '-' * 45) joint_coherence()
9,192
36.831276
113
py
mmvae-public
mmvae-public/src/report/helper.py
import json import os import pickle from collections import Counter, OrderedDict import numpy as np import torch import torch.nn as nn import torch.nn.functional as F from gensim.models import FastText from nltk.tokenize import sent_tokenize, word_tokenize from scipy.linalg import eig from skimage.filters import threshold_yen as threshold class OrderedCounter(Counter, OrderedDict): """Counter that remembers the order elements are first encountered.""" def __repr__(self): return '%s(%r)' % (self.__class__.__name__, OrderedDict(self)) def __reduce__(self): return self.__class__, (OrderedDict(self),) def cca(views, k=None, eps=1e-12): """Compute (multi-view) CCA Args: views (list): list of views where each view `v_i` is of size `N x o_i` k (int): joint projection dimension | if None, find using Otsu eps (float): regulariser [default: 1e-12] Returns: correlations: correlations along each of the k dimensions projections: projection matrices for each view """ V = len(views) # number of views N = views[0].size(0) # number of observations (same across views) os = [v.size(1) for v in views] kmax = np.min(os) ocum = np.cumsum([0] + os) os_sum = sum(os) A, B = np.zeros([os_sum, os_sum]), np.zeros([os_sum, os_sum]) for i in range(V): v_i = views[i] v_i_bar = v_i - v_i.mean(0).expand_as(v_i) # centered, N x o_i C_ij = (1.0 / (N - 1)) * torch.mm(v_i_bar.t(), v_i_bar) # A[ocum[i]:ocum[i + 1], ocum[i]:ocum[i + 1]] = C_ij B[ocum[i]:ocum[i + 1], ocum[i]:ocum[i + 1]] = C_ij for j in range(i + 1, V): v_j = views[j] # N x o_j v_j_bar = v_j - v_j.mean(0).expand_as(v_j) # centered C_ij = (1.0 / (N - 1)) * torch.mm(v_i_bar.t(), v_j_bar) A[ocum[i]:ocum[i + 1], ocum[j]:ocum[j + 1]] = C_ij A[ocum[j]:ocum[j + 1], ocum[i]:ocum[i + 1]] = C_ij.t() A[np.diag_indices_from(A)] += eps B[np.diag_indices_from(B)] += eps eigenvalues, eigenvectors = eig(A, B) # TODO: sanity check to see that all eigenvalues are e+0i idx = eigenvalues.argsort()[::-1] # sort descending eigenvalues = eigenvalues[idx] # arrange in descending order if k is None: t = threshold(eigenvalues.real[:kmax]) k = np.abs(np.asarray(eigenvalues.real[0::10]) - t).argmin() * 10 # closest k % 10 == 0 idx print('k unspecified, (auto-)choosing:', k) eigenvalues = eigenvalues[idx[:k]] eigenvectors = eigenvectors[:, idx[:k]] correlations = torch.from_numpy(eigenvalues.real).type_as(views[0]) proj_matrices = torch.split(torch.from_numpy(eigenvectors.real).type_as(views[0]), os) return correlations, proj_matrices def fetch_emb(lenWindow, minOccur, emb_path, vocab_path, RESET): if not os.path.exists(emb_path) or RESET: with open('../data/cub/text_trainvalclasses.txt', 'r') as file: text = file.read() sentences = sent_tokenize(text) texts = [] for i, line in enumerate(sentences): words = word_tokenize(line) texts.append(words) model = FastText(size=300, window=lenWindow, min_count=minOccur) model.build_vocab(sentences=texts) model.train(sentences=texts, total_examples=len(texts), epochs=10) with open(vocab_path, 'rb') as file: vocab = json.load(file) i2w = vocab['i2w'] base = np.ones((300,), dtype=np.float32) emb = [base * (i - 1) for i in range(3)] for word in list(i2w.values())[3:]: emb.append(model[word]) emb = np.array(emb) with open(emb_path, 'wb') as file: pickle.dump(emb, file) else: with open(emb_path, 'rb') as file: emb = pickle.load(file) return emb def fetch_weights(weights_path, vocab_path, RESET, a=1e-3): if not os.path.exists(weights_path) or RESET: with open('../data/cub/text_trainvalclasses.txt', 'r') as file: text = file.read() sentences = sent_tokenize(text) occ_register = OrderedCounter() for i, line in enumerate(sentences): words = word_tokenize(line) occ_register.update(words) with open(vocab_path, 'r') as file: vocab = json.load(file) w2i = vocab['w2i'] weights = np.zeros(len(w2i)) total_occ = sum(list(occ_register.values())) exc_occ = 0 for w, occ in occ_register.items(): if w in w2i.keys(): weights[w2i[w]] = a / (a + occ / total_occ) else: exc_occ += occ weights[0] = a / (a + exc_occ / total_occ) with open(weights_path, 'wb') as file: pickle.dump(weights, file) else: with open(weights_path, 'rb') as file: weights = pickle.load(file) return weights def fetch_pc(emb, weights, train_loader, pc_path, RESET): sentences = torch.cat([d[1][0] for d in train_loader]).int() emb_dataset = apply_weights(emb, weights, sentences) if not os.path.exists(pc_path) or RESET: _, _, V = torch.svd(emb_dataset - emb_dataset.mean(dim=0), some=True) v = V[:, 0].unsqueeze(-1) u = v.mm(v.t()) with open(pc_path, 'wb') as file: pickle.dump(u, file) else: with open(pc_path, 'rb') as file: u = pickle.load(file) return u def apply_weights(emb, weights, data): fn_trun = lambda s: s[:np.where(s == 2)[0][0] + 1] if 2 in s else s batch_emb = [] for sent_i in data: emb_stacked = torch.stack([emb[idx] for idx in fn_trun(sent_i)]) weights_stacked = torch.stack([weights[idx] for idx in fn_trun(sent_i)]) batch_emb.append(torch.sum(emb_stacked * weights_stacked.unsqueeze(-1), dim=0) / emb_stacked.shape[0]) return torch.stack(batch_emb, dim=0) def apply_pc(weighted_emb, u): return torch.cat([e - torch.matmul(u, e.unsqueeze(-1)).squeeze() for e in weighted_emb.split(2048, 0)]) class Latent_Classifier(nn.Module): """ Generate latent parameters for SVHN image data. """ def __init__(self, in_n, out_n): super(Latent_Classifier, self).__init__() self.mlp = nn.Linear(in_n, out_n) def forward(self, x): return self.mlp(x) class SVHN_Classifier(nn.Module): def __init__(self): super(SVHN_Classifier, self).__init__() self.conv1 = nn.Conv2d(3, 10, kernel_size=5) self.conv2 = nn.Conv2d(10, 20, kernel_size=5) self.conv2_drop = nn.Dropout2d() self.fc1 = nn.Linear(500, 50) self.fc2 = nn.Linear(50, 10) def forward(self, x): x = F.relu(F.max_pool2d(self.conv1(x), 2)) x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2)) x = x.view(-1, 500) x = F.relu(self.fc1(x)) x = F.dropout(x, training=self.training) x = self.fc2(x) return F.log_softmax(x, dim=-1) class MNIST_Classifier(nn.Module): def __init__(self): super(MNIST_Classifier, self).__init__() self.conv1 = nn.Conv2d(1, 10, kernel_size=5) self.conv2 = nn.Conv2d(10, 20, kernel_size=5) self.conv2_drop = nn.Dropout2d() self.fc1 = nn.Linear(320, 50) self.fc2 = nn.Linear(50, 10) def forward(self, x): x = F.relu(F.max_pool2d(self.conv1(x), 2)) x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2)) x = x.view(-1, 320) x = F.relu(self.fc1(x)) x = F.dropout(x, training=self.training) x = self.fc2(x) return F.log_softmax(x, dim=-1)
7,712
32.977974
110
py
mmvae-public
mmvae-public/src/report/calculate_likelihoods.py
"""Calculate data marginal likelihood p(x) evaluated on the trained generative model.""" import os import sys import argparse import numpy as np import torch from torchvision.utils import save_image # relative import hacks (sorry) import inspect currentdir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe()))) parentdir = os.path.dirname(currentdir) sys.path.insert(0, parentdir) # for bash user os.chdir(parentdir) # for pycharm user import models from utils import Logger, Timer, unpack_data, log_mean_exp torch.backends.cudnn.benchmark = True parser = argparse.ArgumentParser(description='Analysing MM-DGM results') parser.add_argument('--save-dir', type=str, default="", metavar='N', help='save directory of results') parser.add_argument('--iwae-samples', type=int, default=1000, metavar='I', help='number of samples to estimate marginal log likelihood (default: 1000)') parser.add_argument('--no-cuda', action='store_true', default=False, help='disables CUDA use') cmds = parser.parse_args() runPath = cmds.save_dir sys.stdout = Logger('{}/llik.log'.format(runPath)) args = torch.load(runPath + '/args.rar') # cuda stuff needs_conversion = cmds.no_cuda and args.cuda conversion_kwargs = {'map_location': lambda st, loc: st} if needs_conversion else {} args.cuda = not cmds.no_cuda and torch.cuda.is_available() device = torch.device("cuda" if args.cuda else "cpu") torch.manual_seed(args.seed) modelC = getattr(models, 'VAE_{}'.format(args.model)) model = modelC(args) if args.cuda: model.cuda() model.load_state_dict(torch.load(runPath + '/model.rar', **conversion_kwargs), strict=False) B = 12000 // cmds.iwae_samples # rough batch size heuristic train_loader, test_loader = model.getDataLoaders(B, device=device) N = len(test_loader.dataset) def m_iwae(qz_xs, px_zs, zss, x): """IWAE estimate for log p_\theta(x) for multi-modal vae -- fully vectorised""" lws = [] for r, qz_x in enumerate(qz_xs): lpz = model.pz(*model.pz_params).log_prob(zss[r]).sum(-1) lqz_x = log_mean_exp(torch.stack([qz_x.log_prob(zss[r]).sum(-1) for qz_x in qz_xs])) lpx_z = [px_z.log_prob(x[d]).view(*px_z.batch_shape[:2], -1) .mul(model.vaes[d].llik_scaling).sum(-1) for d, px_z in enumerate(px_zs[r])] lpx_z = torch.stack(lpx_z).sum(0) lw = lpz + lpx_z - lqz_x lws.append(lw) return log_mean_exp(torch.cat(lws)).sum() def iwae(qz_x, px_z, zs, x): """IWAE estimate for log p_\theta(x) -- fully vectorised.""" lpz = model.pz(*model.pz_params).log_prob(zs).sum(-1) lpx_z = px_z.log_prob(x).view(*px_z.batch_shape[:2], -1) * model.llik_scaling lqz_x = qz_x.log_prob(zs).sum(-1) return log_mean_exp(lpz + lpx_z.sum(-1) - lqz_x).sum() @torch.no_grad() def joint_elbo(K): model.eval() llik = 0 obj = locals()[('m_' if hasattr(model, 'vaes') else '') + 'iwae']() for dataT in test_loader: data = unpack_data(dataT, device=device) llik += obj(model, data, K).item() print('Marginal Log Likelihood of joint {} (IWAE, K = {}): {:.4f}' .format(model.modelName, K, llik / N)) def cross_iwaes(qz_xs, px_zs, zss, x): lws = [] for e, _px_zs in enumerate(px_zs): # rows are encoders lpz = model.pz(*model.pz_params).log_prob(zss[e]).sum(-1) lqz_x = qz_xs[e].log_prob(zss[e]).sum(-1) _lpx_zs = [px_z.log_prob(x[d]).view(*px_z.batch_shape[:2], -1).sum(-1) for d, px_z in enumerate(_px_zs)] lws.append([log_mean_exp(_lpx_z + lpz - lqz_x).sum() for _lpx_z in _lpx_zs]) return lws def individual_iwaes(qz_xs, px_zs, zss, x): lws = [] for d, _px_zs in enumerate(np.array(px_zs).T): # rows are decoders now lw = [px_z.log_prob(x[d]).view(*px_z.batch_shape[:2], -1).sum(-1) + model.pz(*model.pz_params).log_prob(zss[e]).sum(-1) - log_mean_exp(torch.stack([qz_x.log_prob(zss[e]).sum(-1) for qz_x in qz_xs])) for e, px_z in enumerate(_px_zs)] lw = torch.cat(lw) lws.append(log_mean_exp(lw).sum()) return lws @torch.no_grad() def m_llik_eval(K): model.eval() llik_joint = 0 llik_synergy = np.array([0 for _ in model.vaes]) lliks_cross = np.array([[0 for _ in model.vaes] for _ in model.vaes]) for dataT in test_loader: data = unpack_data(dataT, device=device) qz_xs, px_zs, zss = model(data, K) objs = individual_iwaes(qz_xs, px_zs, zss, data) objs_cross = cross_iwaes(qz_xs, px_zs, zss, data) llik_joint += m_iwae(qz_xs, px_zs, zss, data) llik_synergy = llik_synergy + np.array(objs) lliks_cross = lliks_cross + np.array(objs_cross) print('Marginal Log Likelihood of joint {} (IWAE, K = {}): {:.4f}' .format(model.modelName, K, llik_joint / N)) print('-' * 89) for i, llik in enumerate(llik_synergy): print('Marginal Log Likelihood of {} from {} (IWAE, K = {}): {:.4f}' .format(model.vaes[i].modelName, model.modelName, K, (llik / N).item())) print('-' * 89) for e, _lliks_cross in enumerate(lliks_cross): for d, llik_cross in enumerate(_lliks_cross): print('Marginal Log Likelihood of {} from {} (IWAE, K = {}): {:.4f}' .format(model.vaes[d].modelName, model.vaes[e].modelName, K, (llik_cross / N).item())) print('-' * 89) @torch.no_grad() def llik_eval(K): model.eval() llik_joint = 0 for dataT in test_loader: data = unpack_data(dataT, device=device) qz_xs, px_zs, zss = model(data, K) llik_joint += iwae(qz_xs, px_zs, zss, data) print('Marginal Log Likelihood of joint {} (IWAE, K = {}): {:.4f}' .format(model.modelName, K, llik_joint / N)) @torch.no_grad() def generate_sparse(D, steps, J): """generate `steps` perturbations for all `D` latent dimensions on `J` datapoints. """ model.eval() for i, dataT in enumerate(test_loader): data = unpack_data(dataT, require_length=(args.projection == 'Sft'), device=device) qz_xs, _, zss = model(data, args.K) for i, (qz_x, zs) in enumerate(zip(qz_xs, zss)): embs = [] # for delta in torch.linspace(0.01, 0.99, steps=steps): for delta in torch.linspace(-5, 5, steps=steps): for d in range(D): mod_emb = qz_x.mean + torch.zeros_like(qz_x.mean) mod_emb[:, d] += model.vaes[i].pz(*model.vaes[i].pz_params).stddev[:, d] * delta embs.append(mod_emb) embs = torch.stack(embs).transpose(0, 1).contiguous() for r in range(2): samples = model.vaes[r].px_z(*model.vaes[r].dec(embs.view(-1, D)[:((J) * steps * D)])).mean save_image(samples.cpu(), os.path.join(runPath, 'latent-traversals-{}x{}.png'.format(i, r)), nrow=D) break if __name__ == '__main__': with Timer('MM-VAE analysis') as t: # likelihood evaluation print('-' * 89) eval = locals()[('m_' if hasattr(model, 'vaes') else '') + 'llik_eval'] eval(cmds.iwae_samples) print('-' * 89)
7,240
38.785714
116
py
mmvae-public
mmvae-public/src/models/vae_svhn.py
# SVHN model specification import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F from numpy import sqrt from torch.utils.data import DataLoader from torchvision import transforms, datasets from torchvision.utils import save_image, make_grid from utils import Constants from vis import plot_embeddings, plot_kls_df from .vae import VAE # Constants dataSize = torch.Size([3, 32, 32]) imgChans = dataSize[0] fBase = 32 # base size of filter channels # Classes class Enc(nn.Module): """ Generate latent parameters for SVHN image data. """ def __init__(self, latent_dim): super(Enc, self).__init__() self.enc = nn.Sequential( # input size: 3 x 32 x 32 nn.Conv2d(imgChans, fBase, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase) x 16 x 16 nn.Conv2d(fBase, fBase * 2, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 2) x 8 x 8 nn.Conv2d(fBase * 2, fBase * 4, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 4) x 4 x 4 ) self.c1 = nn.Conv2d(fBase * 4, latent_dim, 4, 1, 0, bias=True) self.c2 = nn.Conv2d(fBase * 4, latent_dim, 4, 1, 0, bias=True) # c1, c2 size: latent_dim x 1 x 1 def forward(self, x): e = self.enc(x) lv = self.c2(e).squeeze() return self.c1(e).squeeze(), F.softmax(lv, dim=-1) * lv.size(-1) + Constants.eta class Dec(nn.Module): """ Generate a SVHN image given a sample from the latent space. """ def __init__(self, latent_dim): super(Dec, self).__init__() self.dec = nn.Sequential( nn.ConvTranspose2d(latent_dim, fBase * 4, 4, 1, 0, bias=True), nn.ReLU(True), # size: (fBase * 4) x 4 x 4 nn.ConvTranspose2d(fBase * 4, fBase * 2, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 2) x 8 x 8 nn.ConvTranspose2d(fBase * 2, fBase, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase) x 16 x 16 nn.ConvTranspose2d(fBase, imgChans, 4, 2, 1, bias=True), nn.Sigmoid() # Output size: 3 x 32 x 32 ) def forward(self, z): z = z.unsqueeze(-1).unsqueeze(-1) # fit deconv layers out = self.dec(z.view(-1, *z.size()[-3:])) out = out.view(*z.size()[:-3], *out.size()[1:]) # consider also predicting the length scale return out, torch.tensor(0.75).to(z.device) # mean, length scale class SVHN(VAE): """ Derive a specific sub-class of a VAE for SVHN """ def __init__(self, params): super(SVHN, self).__init__( dist.Laplace, # prior dist.Laplace, # likelihood dist.Laplace, # posterior Enc(params.latent_dim), Dec(params.latent_dim), params ) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'svhn' self.dataSize = dataSize self.llik_scaling = 1. @property def pz_params(self): return self._pz_params[0], F.softmax(self._pz_params[1], dim=1) * self._pz_params[1].size(-1) @staticmethod def getDataLoaders(batch_size, shuffle=True, device='cuda'): kwargs = {'num_workers': 1, 'pin_memory': True} if device == 'cuda' else {} tx = transforms.ToTensor() train = DataLoader(datasets.SVHN('../data', split='train', download=True, transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) test = DataLoader(datasets.SVHN('../data', split='test', download=True, transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) return train, test def generate(self, runPath, epoch): N, K = 64, 9 samples = super(SVHN, self).generate(N, K).cpu() # wrangle things so they come out tiled samples = samples.view(K, N, *samples.size()[1:]).transpose(0, 1) s = [make_grid(t, nrow=int(sqrt(K)), padding=0) for t in samples] save_image(torch.stack(s), '{}/gen_samples_{:03d}.png'.format(runPath, epoch), nrow=int(sqrt(N))) def reconstruct(self, data, runPath, epoch): recon = super(SVHN, self).reconstruct(data[:8]) comp = torch.cat([data[:8], recon]).data.cpu() save_image(comp, '{}/recon_{:03d}.png'.format(runPath, epoch)) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(SVHN, self).analyse(data, K=10) labels = ['Prior', self.modelName.lower()] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch))
5,053
37
101
py
mmvae-public
mmvae-public/src/models/mmvae_cub_images_sentences.py
# cub multi-modal model specification import matplotlib.pyplot as plt import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from numpy import sqrt, prod from torch.utils.data import DataLoader from torchnet.dataset import TensorDataset, ResampleDataset from torchvision.utils import save_image, make_grid from utils import Constants from vis import plot_embeddings, plot_kls_df from .mmvae import MMVAE from .vae_cub_image import CUB_Image from .vae_cub_sent import CUB_Sentence # Constants maxSentLen = 32 minOccur = 3 # This is required because there are 10 captions per image. # Allows easier reuse of the same image for the corresponding set of captions. def resampler(dataset, idx): return idx // 10 class CUB_Image_Sentence(MMVAE): def __init__(self, params): super(CUB_Image_Sentence, self).__init__(dist.Laplace, params, CUB_Image, CUB_Sentence) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.vaes[0].llik_scaling = self.vaes[1].maxSentLen / prod(self.vaes[0].dataSize) \ if params.llik_scaling == 0 else params.llik_scaling for vae in self.vaes: vae._pz_params = self._pz_params self.modelName = 'cubIS' self.i2w = self.vaes[1].load_vocab() @property def pz_params(self): return self._pz_params[0], \ F.softmax(self._pz_params[1], dim=1) * self._pz_params[1].size(1) + Constants.eta def getDataLoaders(self, batch_size, shuffle=True, device='cuda'): # load base datasets t1, s1 = self.vaes[0].getDataLoaders(batch_size, shuffle, device) t2, s2 = self.vaes[1].getDataLoaders(batch_size, shuffle, device) kwargs = {'num_workers': 2, 'pin_memory': True} if device == 'cuda' else {} train_loader = DataLoader(TensorDataset([ ResampleDataset(t1.dataset, resampler, size=len(t1.dataset) * 10), t2.dataset]), batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = DataLoader(TensorDataset([ ResampleDataset(s1.dataset, resampler, size=len(s1.dataset) * 10), s2.dataset]), batch_size=batch_size, shuffle=shuffle, **kwargs) return train_loader, test_loader def generate(self, runPath, epoch): N = 8 samples = super(CUB_Image_Sentence, self).generate(N) images, captions = [sample.data.cpu() for sample in samples] captions = self._sent_preprocess(captions) fig = plt.figure(figsize=(8, 6)) for i, (image, caption) in enumerate(zip(images, captions)): fig = self._imshow(image, caption, i, fig, N) plt.savefig('{}/gen_samples_{:03d}.png'.format(runPath, epoch)) plt.close() def reconstruct(self, raw_data, runPath, epoch): N = 8 recons_mat = super(CUB_Image_Sentence, self).reconstruct([d[:N] for d in raw_data]) fns = [lambda images: images.data.cpu(), lambda sentences: self._sent_preprocess(sentences)] for r, recons_list in enumerate(recons_mat): for o, recon in enumerate(recons_list): data = fns[r](raw_data[r][:N]) recon = fns[o](recon.squeeze()) if r != o: fig = plt.figure(figsize=(8, 6)) for i, (_data, _recon) in enumerate(zip(data, recon)): image, caption = (_data, _recon) if r == 0 else (_recon, _data) fig = self._imshow(image, caption, i, fig, N) plt.savefig('{}/recon_{}x{}_{:03d}.png'.format(runPath, r, o, epoch)) plt.close() else: if r == 0: comp = torch.cat([data, recon]) save_image(comp, '{}/recon_{}x{}_{:03d}.png'.format(runPath, r, o, epoch)) else: with open('{}/recon_{}x{}_{:03d}.txt'.format(runPath, r, o, epoch), "w+") as txt_file: for r_sent, d_sent in zip(recon, data): txt_file.write('[DATA] ==> {}\n'.format(' '.join(self.i2w[str(i)] for i in d_sent))) txt_file.write('[RECON] ==> {}\n\n'.format(' '.join(self.i2w[str(i)] for i in r_sent))) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(CUB_Image_Sentence, self).analyse(data, K=10) labels = ['Prior', *[vae.modelName.lower() for vae in self.vaes]] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch)) def _sent_preprocess(self, sentences): """make sure raw data is always passed as dim=2 to avoid argmax. last dimension must always be word embedding.""" if len(sentences.shape) > 2: sentences = sentences.argmax(-1).squeeze() return [self.vaes[1].fn_trun(s) for s in self.vaes[1].fn_2i(sentences)] def _imshow(self, image, caption, i, fig, N): """Imshow for Tensor.""" ax = fig.add_subplot(N // 2, 4, i * 2 + 1) ax.axis('off') image = image.numpy().transpose((1, 2, 0)) # plt.imshow(image) ax = fig.add_subplot(N // 2, 4, i * 2 + 2) pos = ax.get_position() ax.axis('off') plt.text( x=0.5 * (pos.x0 + pos.x1), y=0.5 * (pos.y0 + pos.y1), ha='left', s='{}'.format( ' '.join(self.i2w[str(i)] + '\n' if (n + 1) % 5 == 0 else self.i2w[str(i)] for n, i in enumerate(caption))), fontsize=6, verticalalignment='center', horizontalalignment='center' ) return fig
6,015
42.912409
119
py
mmvae-public
mmvae-public/src/models/vae.py
# Base VAE class definition import torch import torch.nn as nn from utils import get_mean, kl_divergence from vis import embed_umap, tensors_to_df class VAE(nn.Module): def __init__(self, prior_dist, likelihood_dist, post_dist, enc, dec, params): super(VAE, self).__init__() self.pz = prior_dist self.px_z = likelihood_dist self.qz_x = post_dist self.enc = enc self.dec = dec self.modelName = None self.params = params self._pz_params = None # defined in subclass self._qz_x_params = None # populated in `forward` self.llik_scaling = 1.0 @property def pz_params(self): return self._pz_params @property def qz_x_params(self): if self._qz_x_params is None: raise NameError("qz_x params not initalised yet!") return self._qz_x_params @staticmethod def getDataLoaders(batch_size, shuffle=True, device="cuda"): # handle merging individual datasets appropriately in sub-class raise NotImplementedError def forward(self, x, K=1): self._qz_x_params = self.enc(x) qz_x = self.qz_x(*self._qz_x_params) zs = qz_x.rsample(torch.Size([K])) px_z = self.px_z(*self.dec(zs)) return qz_x, px_z, zs def generate(self, N, K): self.eval() with torch.no_grad(): pz = self.pz(*self.pz_params) latents = pz.rsample(torch.Size([N])) px_z = self.px_z(*self.dec(latents)) data = px_z.sample(torch.Size([K])) return data.view(-1, *data.size()[3:]) def reconstruct(self, data): self.eval() with torch.no_grad(): qz_x = self.qz_x(*self.enc(data)) latents = qz_x.rsample() # no dim expansion px_z = self.px_z(*self.dec(latents)) recon = get_mean(px_z) return recon def analyse(self, data, K): self.eval() with torch.no_grad(): qz_x, _, zs = self.forward(data, K=K) pz = self.pz(*self.pz_params) zss = [pz.sample(torch.Size([K, data.size(0)])).view(-1, pz.batch_shape[-1]), zs.view(-1, zs.size(-1))] zsl = [torch.zeros(zs.size(0)).fill_(i) for i, zs in enumerate(zss)] kls_df = tensors_to_df( [kl_divergence(qz_x, pz).cpu().numpy()], head='KL', keys=[r'KL$(q(z|x)\,||\,p(z))$'], ax_names=['Dimensions', r'KL$(q\,||\,p)$'] ) return embed_umap(torch.cat(zss, 0).cpu().numpy()), \ torch.cat(zsl, 0).cpu().numpy(), \ kls_df
2,674
32.024691
89
py
mmvae-public
mmvae-public/src/models/vae_cub_image.py
# CUB Image model specification import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from numpy import sqrt from torchvision import datasets, transforms from torchvision.utils import make_grid, save_image from utils import Constants from vis import plot_embeddings, plot_kls_df from .vae import VAE # Constants imgChans = 3 fBase = 64 # Classes class Enc(nn.Module): """ Generate latent parameters for CUB image data. """ def __init__(self, latentDim): super(Enc, self).__init__() modules = [ # input size: 3 x 128 x 128 nn.Conv2d(imgChans, fBase, 4, 2, 1, bias=True), nn.ReLU(True), # input size: 1 x 64 x 64 nn.Conv2d(fBase, fBase * 2, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 2) x 32 x 32 nn.Conv2d(fBase * 2, fBase * 4, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 4) x 16 x 16 nn.Conv2d(fBase * 4, fBase * 8, 4, 2, 1, bias=True), nn.ReLU(True)] # size: (fBase * 8) x 4 x 4 self.enc = nn.Sequential(*modules) self.c1 = nn.Conv2d(fBase * 8, latentDim, 4, 1, 0, bias=True) self.c2 = nn.Conv2d(fBase * 8, latentDim, 4, 1, 0, bias=True) # c1, c2 size: latentDim x 1 x 1 def forward(self, x): e = self.enc(x) return self.c1(e).squeeze(), F.softplus(self.c2(e)).squeeze() + Constants.eta class Dec(nn.Module): """ Generate an image given a sample from the latent space. """ def __init__(self, latentDim): super(Dec, self).__init__() modules = [nn.ConvTranspose2d(latentDim, fBase * 8, 4, 1, 0, bias=True), nn.ReLU(True), ] modules.extend([ nn.ConvTranspose2d(fBase * 8, fBase * 4, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 4) x 16 x 16 nn.ConvTranspose2d(fBase * 4, fBase * 2, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase * 2) x 32 x 32 nn.ConvTranspose2d(fBase * 2, fBase, 4, 2, 1, bias=True), nn.ReLU(True), # size: (fBase) x 64 x 64 nn.ConvTranspose2d(fBase, imgChans, 4, 2, 1, bias=True), nn.Sigmoid() # Output size: 3 x 128 x 128 ]) self.dec = nn.Sequential(*modules) def forward(self, z): z = z.unsqueeze(-1).unsqueeze(-1) # fit deconv layers out = self.dec(z.view(-1, *z.size()[-3:])) out = out.view(*z.size()[:-3], *out.size()[1:]) return out, torch.tensor(0.01).to(z.device) class CUB_Image(VAE): """ Derive a specific sub-class of a VAE for a CNN sentence model. """ def __init__(self, params): super(CUB_Image, self).__init__( dist.Laplace, # prior dist.Laplace, # likelihood dist.Laplace, # posterior Enc(params.latent_dim), Dec(params.latent_dim), params ) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'cubI' self.dataSize = torch.Size([3, 64, 64]) self.llik_scaling = 1. @property def pz_params(self): return self._pz_params[0], F.softplus(self._pz_params[1]) + Constants.eta # remember that when combining with captions, this should be x10 def getDataLoaders(self, batch_size, shuffle=True, device="cuda"): kwargs = {'num_workers': 1, 'pin_memory': True} if device == "cuda" else {} tx = transforms.Compose([transforms.Resize([64, 64]), transforms.ToTensor()]) train_loader = torch.utils.data.DataLoader( datasets.ImageFolder('../data/cub/train', transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = torch.utils.data.DataLoader( datasets.ImageFolder('../data/cub/test', transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) return train_loader, test_loader def generate(self, runPath, epoch): N, K = 64, 9 samples = super(CUB_Image, self).generate(N, K).data.cpu() # wrangle things so they come out tiled samples = samples.view(K, N, *samples.size()[1:]).transpose(0, 1) s = [make_grid(t, nrow=int(sqrt(K)), padding=0) for t in samples.data.cpu()] save_image(torch.stack(s), '{}/gen_samples_{:03d}.png'.format(runPath, epoch), nrow=int(sqrt(N))) def reconstruct(self, data, runPath, epoch): recon = super(CUB_Image, self).reconstruct(data[:8]) comp = torch.cat([data[:8], recon]) save_image(comp.data.cpu(), '{}/recon_{:03d}.png'.format(runPath, epoch)) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(CUB_Image, self).analyse(data, K=10) labels = ['Prior', self.modelName.lower()] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch))
5,350
37.221429
91
py
mmvae-public
mmvae-public/src/models/vae_mnist.py
# MNIST model specification import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F from numpy import prod, sqrt from torch.utils.data import DataLoader from torchvision import datasets, transforms from torchvision.utils import save_image, make_grid from utils import Constants from vis import plot_embeddings, plot_kls_df from .vae import VAE # Constants dataSize = torch.Size([1, 28, 28]) data_dim = int(prod(dataSize)) hidden_dim = 400 def extra_hidden_layer(): return nn.Sequential(nn.Linear(hidden_dim, hidden_dim), nn.ReLU(True)) # Classes class Enc(nn.Module): """ Generate latent parameters for MNIST image data. """ def __init__(self, latent_dim, num_hidden_layers=1): super(Enc, self).__init__() modules = [] modules.append(nn.Sequential(nn.Linear(data_dim, hidden_dim), nn.ReLU(True))) modules.extend([extra_hidden_layer() for _ in range(num_hidden_layers - 1)]) self.enc = nn.Sequential(*modules) self.fc21 = nn.Linear(hidden_dim, latent_dim) self.fc22 = nn.Linear(hidden_dim, latent_dim) def forward(self, x): e = self.enc(x.view(*x.size()[:-3], -1)) # flatten data lv = self.fc22(e) return self.fc21(e), F.softmax(lv, dim=-1) * lv.size(-1) + Constants.eta class Dec(nn.Module): """ Generate an MNIST image given a sample from the latent space. """ def __init__(self, latent_dim, num_hidden_layers=1): super(Dec, self).__init__() modules = [] modules.append(nn.Sequential(nn.Linear(latent_dim, hidden_dim), nn.ReLU(True))) modules.extend([extra_hidden_layer() for _ in range(num_hidden_layers - 1)]) self.dec = nn.Sequential(*modules) self.fc3 = nn.Linear(hidden_dim, data_dim) def forward(self, z): p = self.fc3(self.dec(z)) d = torch.sigmoid(p.view(*z.size()[:-1], *dataSize)) # reshape data d = d.clamp(Constants.eta, 1 - Constants.eta) return d, torch.tensor(0.75).to(z.device) # mean, length scale class MNIST(VAE): """ Derive a specific sub-class of a VAE for MNIST. """ def __init__(self, params): super(MNIST, self).__init__( dist.Laplace, # prior dist.Laplace, # likelihood dist.Laplace, # posterior Enc(params.latent_dim, params.num_hidden_layers), Dec(params.latent_dim, params.num_hidden_layers), params ) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'mnist' self.dataSize = dataSize self.llik_scaling = 1. @property def pz_params(self): return self._pz_params[0], F.softmax(self._pz_params[1], dim=1) * self._pz_params[1].size(-1) @staticmethod def getDataLoaders(batch_size, shuffle=True, device="cuda"): kwargs = {'num_workers': 1, 'pin_memory': True} if device == "cuda" else {} tx = transforms.ToTensor() train = DataLoader(datasets.MNIST('../data', train=True, download=True, transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) test = DataLoader(datasets.MNIST('../data', train=False, download=True, transform=tx), batch_size=batch_size, shuffle=shuffle, **kwargs) return train, test def generate(self, runPath, epoch): N, K = 64, 9 samples = super(MNIST, self).generate(N, K).cpu() # wrangle things so they come out tiled samples = samples.view(K, N, *samples.size()[1:]).transpose(0, 1) # N x K x 1 x 28 x 28 s = [make_grid(t, nrow=int(sqrt(K)), padding=0) for t in samples] save_image(torch.stack(s), '{}/gen_samples_{:03d}.png'.format(runPath, epoch), nrow=int(sqrt(N))) def reconstruct(self, data, runPath, epoch): recon = super(MNIST, self).reconstruct(data[:8]) comp = torch.cat([data[:8], recon]).data.cpu() save_image(comp, '{}/recon_{:03d}.png'.format(runPath, epoch)) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(MNIST, self).analyse(data, K=10) labels = ['Prior', self.modelName.lower()] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch))
4,623
37.857143
101
py
mmvae-public
mmvae-public/src/models/vae_cub_sent_ft.py
# Sentence model specification - CUB image feature version import json import os import numpy as np import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from torch.utils.data import DataLoader from datasets import CUBSentences from utils import Constants, FakeCategorical from .vae import VAE maxSentLen = 32 # max length of any description for birds dataset minOccur = 3 embeddingDim = 128 lenWindow = 3 fBase = 32 vocabSize = 1590 vocab_path = '../data/cub/oc:{}_sl:{}_s:{}_w:{}/cub.vocab'.format(minOccur, maxSentLen, 300, lenWindow) # Classes class Enc(nn.Module): """ Generate latent parameters for sentence data. """ def __init__(self, latentDim): super(Enc, self).__init__() self.embedding = nn.Embedding(vocabSize, embeddingDim, padding_idx=0) self.enc = nn.Sequential( # input size: 1 x 32 x 128 nn.Conv2d(1, fBase, 4, 2, 1, bias=True), nn.BatchNorm2d(fBase), nn.ReLU(True), # size: (fBase) x 16 x 64 nn.Conv2d(fBase, fBase * 2, 4, 2, 1, bias=True), nn.BatchNorm2d(fBase * 2), nn.ReLU(True), # size: (fBase * 2) x 8 x 32 nn.Conv2d(fBase * 2, fBase * 4, 4, 2, 1, bias=True), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # # size: (fBase * 4) x 4 x 16 nn.Conv2d(fBase * 4, fBase * 8, (1, 4), (1, 2), (0, 1), bias=True), nn.BatchNorm2d(fBase * 8), nn.ReLU(True), # size: (fBase * 8) x 4 x 8 nn.Conv2d(fBase * 8, fBase * 16, (1, 4), (1, 2), (0, 1), bias=True), nn.BatchNorm2d(fBase * 16), nn.ReLU(True), # size: (fBase * 8) x 4 x 4 ) self.c1 = nn.Conv2d(fBase * 16, latentDim, 4, 1, 0, bias=True) self.c2 = nn.Conv2d(fBase * 16, latentDim, 4, 1, 0, bias=True) # c1, c2 size: latentDim x 1 x 1 def forward(self, x): e = self.enc(self.embedding(x.long()).unsqueeze(1)) mu, logvar = self.c1(e).squeeze(), self.c2(e).squeeze() return mu, F.softplus(logvar) + Constants.eta class Dec(nn.Module): """ Generate a sentence given a sample from the latent space. """ def __init__(self, latentDim): super(Dec, self).__init__() self.dec = nn.Sequential( nn.ConvTranspose2d(latentDim, fBase * 16, 4, 1, 0, bias=True), nn.BatchNorm2d(fBase * 16), nn.ReLU(True), # size: (fBase * 8) x 4 x 4 nn.ConvTranspose2d(fBase * 16, fBase * 8, (1, 4), (1, 2), (0, 1), bias=True), nn.BatchNorm2d(fBase * 8), nn.ReLU(True), # size: (fBase * 8) x 4 x 8 nn.ConvTranspose2d(fBase * 8, fBase * 4, (1, 4), (1, 2), (0, 1), bias=True), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 4) x 8 x 32 nn.ConvTranspose2d(fBase * 4, fBase * 2, 4, 2, 1, bias=True), nn.BatchNorm2d(fBase * 2), nn.ReLU(True), # size: (fBase * 2) x 16 x 64 nn.ConvTranspose2d(fBase * 2, fBase, 4, 2, 1, bias=True), nn.BatchNorm2d(fBase), nn.ReLU(True), # size: (fBase) x 32 x 128 nn.ConvTranspose2d(fBase, 1, 4, 2, 1, bias=True), nn.ReLU(True) # Output size: 1 x 64 x 256 ) # inverts the 'embedding' module upto one-hotness self.toVocabSize = nn.Linear(embeddingDim, vocabSize) def forward(self, z): z = z.unsqueeze(-1).unsqueeze(-1) # fit deconv layers out = self.dec(z.view(-1, *z.size()[-3:])).view(-1, embeddingDim) return self.toVocabSize(out).view(*z.size()[:-3], maxSentLen, vocabSize), class CUB_Sentence_ft(VAE): """ Derive a specific sub-class of a VAE for a sentence model. """ def __init__(self, params): super(CUB_Sentence_ft, self).__init__( prior_dist=dist.Normal, likelihood_dist=FakeCategorical, post_dist=dist.Normal, enc=Enc(params.latent_dim), dec=Dec(params.latent_dim), params=params) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'cubSft' self.llik_scaling = 1. self.tie_modules() self.fn_2i = lambda t: t.cpu().numpy().astype(int) self.fn_trun = lambda s: s[:np.where(s == 2)[0][0] + 1] if 2 in s else s self.vocab_file = vocab_path self.maxSentLen = maxSentLen self.vocabSize = vocabSize def tie_modules(self): # This looks dumb, but is actually dumber than you might realise. # A linear(a, b) module has a [b x a] weight matrix, but an embedding(a, b) # module has a [a x b] weight matrix. So when we want the transpose at # decoding time, we just use the weight matrix as is. self.dec.toVocabSize.weight = self.enc.embedding.weight @property def pz_params(self): return self._pz_params[0], F.softplus(self._pz_params[1]) + Constants.eta @staticmethod def getDataLoaders(batch_size, shuffle=True, device="cuda"): kwargs = {'num_workers': 1, 'pin_memory': True} if device == "cuda" else {} tx = lambda data: torch.Tensor(data) t_data = CUBSentences('../data', split='train', transform=tx, max_sequence_length=maxSentLen) s_data = CUBSentences('../data', split='test', transform=tx, max_sequence_length=maxSentLen) train_loader = DataLoader(t_data, batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = DataLoader(s_data, batch_size=batch_size, shuffle=shuffle, **kwargs) return train_loader, test_loader def reconstruct(self, data, runPath, epoch): recon = super(CUB_Sentence_ft, self).reconstruct(data[:8]).argmax(dim=-1).squeeze() recon, data = self.fn_2i(recon), self.fn_2i(data[:8]) recon, data = [self.fn_trun(r) for r in recon], [self.fn_trun(d) for d in data] i2w = self.load_vocab() print("\n Reconstruction examples (excluding <PAD>):") for r_sent, d_sent in zip(recon[:3], data[:3]): print('[DATA] ==> {}'.format(' '.join(i2w[str(i)] for i in d_sent))) print('[RECON] ==> {}\n'.format(' '.join(i2w[str(i)] for i in r_sent))) with open('{}/recon_{:03d}.txt'.format(runPath, epoch), "w+") as txt_file: for r_sent, d_sent in zip(recon, data): txt_file.write('[DATA] ==> {}\n'.format(' '.join(i2w[str(i)] for i in d_sent))) txt_file.write('[RECON] ==> {}\n\n'.format(' '.join(i2w[str(i)] for i in r_sent))) def generate(self, runPath, epoch): N, K = 5, 4 i2w = self.load_vocab() samples = super(CUB_Sentence_ft, self).generate(N, K).argmax(dim=-1).squeeze() samples = samples.view(K, N, samples.size(-1)).transpose(0, 1) # N x K x 64 samples = [[self.fn_trun(s) for s in ss] for ss in self.fn_2i(samples)] # samples = [self.fn_trun(s) for s in samples] print("\n Generated examples (excluding <PAD>):") for s_sent in samples[0][:3]: print('[GEN] ==> {}'.format(' '.join(i2w[str(i)] for i in s_sent if i != 0))) with open('{}/gen_samples_{:03d}.txt'.format(runPath, epoch), "w+") as txt_file: for s_sents in samples: for s_sent in s_sents: txt_file.write('{}\n'.format(' '.join(i2w[str(i)] for i in s_sent))) txt_file.write('\n') def analyse(self, data, runPath, epoch): pass def load_vocab(self): # call dataloader function to create vocab file if not os.path.exists(self.vocab_file): _, _ = self.getDataLoaders(256) with open(self.vocab_file, 'r') as vocab_file: vocab = json.load(vocab_file) return vocab['i2w']
8,185
40.135678
103
py
mmvae-public
mmvae-public/src/models/mmvae.py
# Base MMVAE class definition from itertools import combinations import torch import torch.nn as nn from utils import get_mean, kl_divergence from vis import embed_umap, tensors_to_df class MMVAE(nn.Module): def __init__(self, prior_dist, params, *vaes): super(MMVAE, self).__init__() self.pz = prior_dist self.vaes = nn.ModuleList([vae(params) for vae in vaes]) self.modelName = None # filled-in per sub-class self.params = params self._pz_params = None # defined in subclass @property def pz_params(self): return self._pz_params @staticmethod def getDataLoaders(batch_size, shuffle=True, device="cuda"): # handle merging individual datasets appropriately in sub-class raise NotImplementedError def forward(self, x, K=1): qz_xs, zss = [], [] # initialise cross-modal matrix px_zs = [[None for _ in range(len(self.vaes))] for _ in range(len(self.vaes))] for m, vae in enumerate(self.vaes): qz_x, px_z, zs = vae(x[m], K=K) qz_xs.append(qz_x) zss.append(zs) px_zs[m][m] = px_z # fill-in diagonal for e, zs in enumerate(zss): for d, vae in enumerate(self.vaes): if e != d: # fill-in off-diagonal px_zs[e][d] = vae.px_z(*vae.dec(zs)) return qz_xs, px_zs, zss def generate(self, N): self.eval() with torch.no_grad(): data = [] pz = self.pz(*self.pz_params) latents = pz.rsample(torch.Size([N])) for d, vae in enumerate(self.vaes): px_z = vae.px_z(*vae.dec(latents)) data.append(px_z.mean.view(-1, *px_z.mean.size()[2:])) return data # list of generations---one for each modality def reconstruct(self, data): self.eval() with torch.no_grad(): _, px_zs, _ = self.forward(data) # cross-modal matrix of reconstructions recons = [[get_mean(px_z) for px_z in r] for r in px_zs] return recons def analyse(self, data, K): self.eval() with torch.no_grad(): qz_xs, _, zss = self.forward(data, K=K) pz = self.pz(*self.pz_params) zss = [pz.sample(torch.Size([K, data[0].size(0)])).view(-1, pz.batch_shape[-1]), *[zs.view(-1, zs.size(-1)) for zs in zss]] zsl = [torch.zeros(zs.size(0)).fill_(i) for i, zs in enumerate(zss)] kls_df = tensors_to_df( [*[kl_divergence(qz_x, pz).cpu().numpy() for qz_x in qz_xs], *[0.5 * (kl_divergence(p, q) + kl_divergence(q, p)).cpu().numpy() for p, q in combinations(qz_xs, 2)]], head='KL', keys=[*[r'KL$(q(z|x_{})\,||\,p(z))$'.format(i) for i in range(len(qz_xs))], *[r'J$(q(z|x_{})\,||\,q(z|x_{}))$'.format(i, j) for i, j in combinations(range(len(qz_xs)), 2)]], ax_names=['Dimensions', r'KL$(q\,||\,p)$'] ) return embed_umap(torch.cat(zss, 0).cpu().numpy()), \ torch.cat(zsl, 0).cpu().numpy(), \ kls_df
3,238
37.105882
92
py
mmvae-public
mmvae-public/src/models/__init__.py
from .mmvae_cub_images_sentences import CUB_Image_Sentence as VAE_cubIS from .mmvae_cub_images_sentences_ft import CUB_Image_Sentence_ft as VAE_cubISft from .mmvae_mnist_svhn import MNIST_SVHN as VAE_mnist_svhn from .vae_cub_image import CUB_Image as VAE_cubI from .vae_cub_image_ft import CUB_Image_ft as VAE_cubIft from .vae_cub_sent import CUB_Sentence as VAE_cubS from .vae_mnist import MNIST as VAE_mnist from .vae_svhn import SVHN as VAE_svhn __all__ = [VAE_mnist_svhn, VAE_mnist, VAE_svhn, VAE_cubIS, VAE_cubS, VAE_cubI, VAE_cubISft, VAE_cubIft]
565
46.166667
79
py
mmvae-public
mmvae-public/src/models/vae_cub_sent.py
# Sentence model specification - real CUB image version import os import json import numpy as np import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from torch.utils.data import DataLoader from datasets import CUBSentences from utils import Constants, FakeCategorical from .vae import VAE # Constants maxSentLen = 32 # max length of any description for birds dataset minOccur = 3 embeddingDim = 128 lenWindow = 3 fBase = 32 vocabSize = 1590 vocab_path = '../data/cub/oc:{}_sl:{}_s:{}_w:{}/cub.vocab'.format(minOccur, maxSentLen, 300, lenWindow) # Classes class Enc(nn.Module): """ Generate latent parameters for sentence data. """ def __init__(self, latentDim): super(Enc, self).__init__() self.embedding = nn.Embedding(vocabSize, embeddingDim, padding_idx=0) self.enc = nn.Sequential( # input size: 1 x 32 x 128 nn.Conv2d(1, fBase, 4, 2, 1, bias=False), nn.BatchNorm2d(fBase), nn.ReLU(True), # size: (fBase) x 16 x 64 nn.Conv2d(fBase, fBase * 2, 4, 2, 1, bias=False), nn.BatchNorm2d(fBase * 2), nn.ReLU(True), # size: (fBase * 2) x 8 x 32 nn.Conv2d(fBase * 2, fBase * 4, 4, 2, 1, bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # # size: (fBase * 4) x 4 x 16 nn.Conv2d(fBase * 4, fBase * 4, (1, 4), (1, 2), (0, 1), bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 8) x 4 x 8 nn.Conv2d(fBase * 4, fBase * 4, (1, 4), (1, 2), (0, 1), bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 8) x 4 x 4 ) self.c1 = nn.Conv2d(fBase * 4, latentDim, 4, 1, 0, bias=False) self.c2 = nn.Conv2d(fBase * 4, latentDim, 4, 1, 0, bias=False) # c1, c2 size: latentDim x 1 x 1 def forward(self, x): e = self.enc(self.embedding(x.long()).unsqueeze(1)) mu, logvar = self.c1(e).squeeze(), self.c2(e).squeeze() return mu, F.softplus(logvar) + Constants.eta class Dec(nn.Module): """ Generate a sentence given a sample from the latent space. """ def __init__(self, latentDim): super(Dec, self).__init__() self.dec = nn.Sequential( nn.ConvTranspose2d(latentDim, fBase * 4, 4, 1, 0, bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 8) x 4 x 4 nn.ConvTranspose2d(fBase * 4, fBase * 4, (1, 4), (1, 2), (0, 1), bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 8) x 4 x 8 nn.ConvTranspose2d(fBase * 4, fBase * 4, (1, 4), (1, 2), (0, 1), bias=False), nn.BatchNorm2d(fBase * 4), nn.ReLU(True), # size: (fBase * 4) x 8 x 32 nn.ConvTranspose2d(fBase * 4, fBase * 2, 4, 2, 1, bias=False), nn.BatchNorm2d(fBase * 2), nn.ReLU(True), # size: (fBase * 2) x 16 x 64 nn.ConvTranspose2d(fBase * 2, fBase, 4, 2, 1, bias=False), nn.BatchNorm2d(fBase), nn.ReLU(True), # size: (fBase) x 32 x 128 nn.ConvTranspose2d(fBase, 1, 4, 2, 1, bias=False), nn.ReLU(True) # Output size: 1 x 64 x 256 ) # inverts the 'embedding' module upto one-hotness self.toVocabSize = nn.Linear(embeddingDim, vocabSize) def forward(self, z): z = z.unsqueeze(-1).unsqueeze(-1) # fit deconv layers out = self.dec(z.view(-1, *z.size()[-3:])).view(-1, embeddingDim) return self.toVocabSize(out).view(*z.size()[:-3], maxSentLen, vocabSize), class CUB_Sentence(VAE): """ Derive a specific sub-class of a VAE for a sentence model. """ def __init__(self, params): super(CUB_Sentence, self).__init__( prior_dist=dist.Normal, likelihood_dist=FakeCategorical, post_dist=dist.Normal, enc=Enc(params.latent_dim), dec=Dec(params.latent_dim), params=params) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'cubS' self.llik_scaling = 1. self.tie_modules() self.fn_2i = lambda t: t.cpu().numpy().astype(int) self.fn_trun = lambda s: s[:np.where(s == 2)[0][0] + 1] if 2 in s else s self.vocab_file = vocab_path self.maxSentLen = maxSentLen self.vocabSize = vocabSize def tie_modules(self): # This looks dumb, but is actually dumber than you might realise. # A linear(a, b) module has a [b x a] weight matrix, but an embedding(a, b) # module has a [a x b] weight matrix. So when we want the transpose at # decoding time, we just use the weight matrix as is. self.dec.toVocabSize.weight = self.enc.embedding.weight @property def pz_params(self): return self._pz_params[0], F.softplus(self._pz_params[1]) + Constants.eta @staticmethod def getDataLoaders(batch_size, shuffle=True, device="cuda"): kwargs = {'num_workers': 1, 'pin_memory': True} if device == "cuda" else {} tx = lambda data: torch.Tensor(data) t_data = CUBSentences('../data', split='train', transform=tx, max_sequence_length=maxSentLen) s_data = CUBSentences('../data', split='test', transform=tx, max_sequence_length=maxSentLen) train_loader = DataLoader(t_data, batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = DataLoader(s_data, batch_size=batch_size, shuffle=shuffle, **kwargs) return train_loader, test_loader def reconstruct(self, data, runPath, epoch): recon = super(CUB_Sentence, self).reconstruct(data[:8]).argmax(dim=-1).squeeze() recon, data = self.fn_2i(recon), self.fn_2i(data[:8]) recon, data = [self.fn_trun(r) for r in recon], [self.fn_trun(d) for d in data] i2w = self.load_vocab() print("\n Reconstruction examples (excluding <PAD>):") for r_sent, d_sent in zip(recon[:3], data[:3]): print('[DATA] ==> {}'.format(' '.join(i2w[str(i)] for i in d_sent))) print('[RECON] ==> {}\n'.format(' '.join(i2w[str(i)] for i in r_sent))) with open('{}/recon_{:03d}.txt'.format(runPath, epoch), "w+") as txt_file: for r_sent, d_sent in zip(recon, data): txt_file.write('[DATA] ==> {}\n'.format(' '.join(i2w[str(i)] for i in d_sent))) txt_file.write('[RECON] ==> {}\n\n'.format(' '.join(i2w[str(i)] for i in r_sent))) def generate(self, runPath, epoch): N, K = 5, 4 i2w = self.load_vocab() samples = super(CUB_Sentence, self).generate(N, K).argmax(dim=-1).squeeze() samples = samples.view(K, N, samples.size(-1)).transpose(0, 1) # N x K x 64 samples = [[self.fn_trun(s) for s in ss] for ss in self.fn_2i(samples)] # samples = [self.fn_trun(s) for s in samples] print("\n Generated examples (excluding <PAD>):") for s_sent in samples[0][:3]: print('[GEN] ==> {}'.format(' '.join(i2w[str(i)] for i in s_sent if i != 0))) with open('{}/gen_samples_{:03d}.txt'.format(runPath, epoch), "w+") as txt_file: for s_sents in samples: for s_sent in s_sents: txt_file.write('{}\n'.format(' '.join(i2w[str(i)] for i in s_sent))) txt_file.write('\n') def analyse(self, data, runPath, epoch): pass def load_vocab(self): # call dataloader function to create vocab file if not os.path.exists(self.vocab_file): _, _ = self.getDataLoaders(256) with open(self.vocab_file, 'r') as vocab_file: vocab = json.load(vocab_file) return vocab['i2w']
8,186
39.935
103
py
mmvae-public
mmvae-public/src/models/mmvae_cub_images_sentences_ft.py
# cub multi-modal model specification import matplotlib.pyplot as plt import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from numpy import sqrt, prod from torch.utils.data import DataLoader from torchnet.dataset import TensorDataset, ResampleDataset from torchvision.utils import save_image, make_grid from utils import Constants from vis import plot_embeddings, plot_kls_df from .mmvae import MMVAE from .vae_cub_image_ft import CUB_Image_ft from .vae_cub_sent_ft import CUB_Sentence_ft # Constants maxSentLen = 32 minOccur = 3 # This is required because there are 10 captions per image. # Allows easier reuse of the same image for the corresponding set of captions. def resampler(dataset, idx): return idx // 10 class CUB_Image_Sentence_ft(MMVAE): def __init__(self, params): super(CUB_Image_Sentence_ft, self).__init__(dist.Normal, params, CUB_Image_ft, CUB_Sentence_ft) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.vaes[0].llik_scaling = self.vaes[1].maxSentLen / prod(self.vaes[0].dataSize) \ if params.llik_scaling == 0 else params.llik_scaling for vae in self.vaes: vae._pz_params = self._pz_params self.modelName = 'cubISft' self.i2w = self.vaes[1].load_vocab() @property def pz_params(self): return self._pz_params[0], \ F.softplus(self._pz_params[1]) + Constants.eta def getDataLoaders(self, batch_size, shuffle=True, device='cuda'): # load base datasets t1, s1 = self.vaes[0].getDataLoaders(batch_size, shuffle, device) t2, s2 = self.vaes[1].getDataLoaders(batch_size, shuffle, device) kwargs = {'num_workers': 2, 'pin_memory': True} if device == 'cuda' else {} train_loader = DataLoader(TensorDataset([ ResampleDataset(t1.dataset, resampler, size=len(t1.dataset) * 10), t2.dataset]), batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = DataLoader(TensorDataset([ ResampleDataset(s1.dataset, resampler, size=len(s1.dataset) * 10), s2.dataset]), batch_size=batch_size, shuffle=shuffle, **kwargs) return train_loader, test_loader def generate(self, runPath, epoch): N = 8 samples = super(CUB_Image_Sentence_ft, self).generate(N) samples[0] = self.vaes[0].unproject(samples[0], search_split='train') images, captions = [sample.data.cpu() for sample in samples] captions = self._sent_preprocess(captions) fig = plt.figure(figsize=(8, 6)) for i, (image, caption) in enumerate(zip(images, captions)): fig = self._imshow(image, caption, i, fig, N) plt.savefig('{}/gen_samples_{:03d}.png'.format(runPath, epoch)) plt.close() def reconstruct(self, raw_data, runPath, epoch): N = 8 recons_mat = super(CUB_Image_Sentence_ft, self).reconstruct([d[:N] for d in raw_data]) fns = [lambda images: images.data.cpu(), lambda sentences: self._sent_preprocess(sentences)] for r, recons_list in enumerate(recons_mat): for o, recon in enumerate(recons_list): data = fns[r](raw_data[r][:N]) recon = fns[o](recon.squeeze()) if r != o: fig = plt.figure(figsize=(8, 6)) for i, (_data, _recon) in enumerate(zip(data, recon)): image, caption = (_data, _recon) if r == 0 else (_recon, _data) search_split = 'test' if r == 0 else 'train' image = self.vaes[0].unproject(image.unsqueeze(0), search_split=search_split) fig = self._imshow(image, caption, i, fig, N) plt.savefig('{}/recon_{}x{}_{:03d}.png'.format(runPath, r, o, epoch)) plt.close() else: if r == 0: data_ = self.vaes[0].unproject(data, search_split='test') recon_ = self.vaes[0].unproject(recon, search_split='train') comp = torch.cat([data_, recon_]) save_image(comp, '{}/recon_{}x{}_{:03d}.png'.format(runPath, r, o, epoch)) else: with open('{}/recon_{}x{}_{:03d}.txt'.format(runPath, r, o, epoch), "w+") as txt_file: for r_sent, d_sent in zip(recon, data): txt_file.write('[DATA] ==> {}\n'.format(' '.join(self.i2w[str(i)] for i in d_sent))) txt_file.write('[RECON] ==> {}\n\n'.format(' '.join(self.i2w[str(i)] for i in r_sent))) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(CUB_Image_Sentence_ft, self).analyse(data, K=10) labels = ['Prior', *[vae.modelName.lower() for vae in self.vaes]] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch)) def _sent_preprocess(self, sentences): """make sure raw data is always passed as dim=2 to avoid argmax. last dimension must always be word embedding.""" if len(sentences.shape) > 2: sentences = sentences.argmax(-1).squeeze() return [self.vaes[1].fn_trun(s) for s in self.vaes[1].fn_2i(sentences)] def _imshow(self, image, caption, i, fig, N): """Imshow for Tensor.""" ax = fig.add_subplot(N // 2, 4, i * 2 + 1) ax.axis('off') image = image.numpy().transpose((1, 2, 0)) # plt.imshow(image) ax = fig.add_subplot(N // 2, 4, i * 2 + 2) pos = ax.get_position() ax.axis('off') plt.text( x=0.5 * (pos.x0 + pos.x1), y=0.5 * (pos.y0 + pos.y1), ha='left', s='{}'.format( ' '.join(self.i2w[str(i)] + '\n' if (n + 1) % 5 == 0 else self.i2w[str(i)] for n, i in enumerate(caption))), fontsize=6, verticalalignment='center', horizontalalignment='center' ) return fig
6,432
44.302817
119
py
mmvae-public
mmvae-public/src/models/vae_cub_image_ft.py
# CUB Image feature model specification import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F import torch.utils.data from numpy import sqrt from torchvision.utils import make_grid, save_image from datasets import CUBImageFt from utils import Constants, NN_lookup from vis import plot_embeddings, plot_kls_df from .vae import VAE # Constants imgChans = 3 fBase = 64 class Enc(nn.Module): """ Generate latent parameters for CUB image feature. """ def __init__(self, latent_dim, n_c): super(Enc, self).__init__() dim_hidden = 256 self.enc = nn.Sequential() for i in range(int(torch.tensor(n_c / dim_hidden).log2())): self.enc.add_module("layer" + str(i), nn.Sequential( nn.Linear(n_c // (2 ** i), n_c // (2 ** (i + 1))), nn.ELU(inplace=True), )) # relies on above terminating at dim_hidden self.fc21 = nn.Linear(dim_hidden, latent_dim) self.fc22 = nn.Linear(dim_hidden, latent_dim) def forward(self, x): e = self.enc(x) return self.fc21(e), F.softplus(self.fc22(e)) + Constants.eta class Dec(nn.Module): """ Generate a CUB image feature given a sample from the latent space. """ def __init__(self, latent_dim, n_c): super(Dec, self).__init__() self.n_c = n_c dim_hidden = 256 self.dec = nn.Sequential() for i in range(int(torch.tensor(n_c / dim_hidden).log2())): indim = latent_dim if i == 0 else dim_hidden * i outdim = dim_hidden if i == 0 else dim_hidden * (2 * i) self.dec.add_module("out_t" if i == 0 else "layer" + str(i) + "_t", nn.Sequential( nn.Linear(indim, outdim), nn.ELU(inplace=True), )) # relies on above terminating at n_c // 2 self.fc31 = nn.Linear(n_c // 2, n_c) def forward(self, z): p = self.dec(z.view(-1, z.size(-1))) mean = self.fc31(p).view(*z.size()[:-1], -1) return mean, torch.tensor([0.01]).to(mean.device) class CUB_Image_ft(VAE): """ Derive a specific sub-class of a VAE for a CNN sentence model. """ def __init__(self, params): super(CUB_Image_ft, self).__init__( dist.Normal, # prior dist.Laplace, # likelihood dist.Normal, # posterior Enc(params.latent_dim, 2048), Dec(params.latent_dim, 2048), params ) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.modelName = 'cubIft' self.dataSize = torch.Size([2048]) self.llik_scaling = 1. @property def pz_params(self): return self._pz_params[0], \ F.softplus(self._pz_params[1]) + Constants.eta # remember that when combining with captions, this should be x10 def getDataLoaders(self, batch_size, shuffle=True, device="cuda"): kwargs = {'num_workers': 1, 'pin_memory': True} if device == "cuda" else {} train_dataset = CUBImageFt('../data', 'train', device) test_dataset = CUBImageFt('../data', 'test', device) train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=shuffle, **kwargs) test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=batch_size, shuffle=shuffle, **kwargs) train_dataset._load_data() test_dataset._load_data() self.unproject = lambda emb_h, search_split='train', \ te=train_dataset.ft_mat, td=train_dataset.data_mat, \ se=test_dataset.ft_mat, sd=test_dataset.data_mat: \ NN_lookup(emb_h, te, td) if search_split == 'train' else NN_lookup(emb_h, se, sd) return train_loader, test_loader def generate(self, runPath, epoch): N, K = 64, 9 samples = super(CUB_Image_ft, self).generate(N, K).data.cpu() samples = self.unproject(samples, search_split='train') samples = samples.view(K, N, *samples.size()[1:]).transpose(0, 1) s = [make_grid(t, nrow=int(sqrt(K)), padding=0) for t in samples.data.cpu()] save_image(torch.stack(s), '{}/gen_samples_{:03d}.png'.format(runPath, epoch), nrow=int(sqrt(N))) def reconstruct(self, data, runPath, epoch): recon = super(CUB_Image_ft, self).reconstruct(data[:8]) data_ = self.unproject(data[:8], search_split='test') recon_ = self.unproject(recon, search_split='train') comp = torch.cat([data_, recon_]) save_image(comp.data.cpu(), '{}/recon_{:03d}.png'.format(runPath, epoch)) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(CUB_Image_ft, self).analyse(data, K=10) labels = ['Prior', self.modelName.lower()] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch))
5,311
38.348148
100
py
mmvae-public
mmvae-public/src/models/mmvae_mnist_svhn.py
# MNIST-SVHN multi-modal model specification import os import torch import torch.distributions as dist import torch.nn as nn import torch.nn.functional as F from numpy import sqrt, prod from torch.utils.data import DataLoader from torchnet.dataset import TensorDataset, ResampleDataset from torchvision.utils import save_image, make_grid from vis import plot_embeddings, plot_kls_df from .mmvae import MMVAE from .vae_mnist import MNIST from .vae_svhn import SVHN class MNIST_SVHN(MMVAE): def __init__(self, params): super(MNIST_SVHN, self).__init__(dist.Laplace, params, MNIST, SVHN) grad = {'requires_grad': params.learn_prior} self._pz_params = nn.ParameterList([ nn.Parameter(torch.zeros(1, params.latent_dim), requires_grad=False), # mu nn.Parameter(torch.zeros(1, params.latent_dim), **grad) # logvar ]) self.vaes[0].llik_scaling = prod(self.vaes[1].dataSize) / prod(self.vaes[0].dataSize) \ if params.llik_scaling == 0 else params.llik_scaling self.modelName = 'mnist-svhn' @property def pz_params(self): return self._pz_params[0], F.softmax(self._pz_params[1], dim=1) * self._pz_params[1].size(-1) def getDataLoaders(self, batch_size, shuffle=True, device='cuda'): if not (os.path.exists('../data/train-ms-mnist-idx.pt') and os.path.exists('../data/train-ms-svhn-idx.pt') and os.path.exists('../data/test-ms-mnist-idx.pt') and os.path.exists('../data/test-ms-svhn-idx.pt')): raise RuntimeError('Generate transformed indices with the script in bin') # get transformed indices t_mnist = torch.load('../data/train-ms-mnist-idx.pt') t_svhn = torch.load('../data/train-ms-svhn-idx.pt') s_mnist = torch.load('../data/test-ms-mnist-idx.pt') s_svhn = torch.load('../data/test-ms-svhn-idx.pt') # load base datasets t1, s1 = self.vaes[0].getDataLoaders(batch_size, shuffle, device) t2, s2 = self.vaes[1].getDataLoaders(batch_size, shuffle, device) train_mnist_svhn = TensorDataset([ ResampleDataset(t1.dataset, lambda d, i: t_mnist[i], size=len(t_mnist)), ResampleDataset(t2.dataset, lambda d, i: t_svhn[i], size=len(t_svhn)) ]) test_mnist_svhn = TensorDataset([ ResampleDataset(s1.dataset, lambda d, i: s_mnist[i], size=len(s_mnist)), ResampleDataset(s2.dataset, lambda d, i: s_svhn[i], size=len(s_svhn)) ]) kwargs = {'num_workers': 2, 'pin_memory': True} if device == 'cuda' else {} train = DataLoader(train_mnist_svhn, batch_size=batch_size, shuffle=shuffle, **kwargs) test = DataLoader(test_mnist_svhn, batch_size=batch_size, shuffle=shuffle, **kwargs) return train, test def generate(self, runPath, epoch): N = 64 samples_list = super(MNIST_SVHN, self).generate(N) for i, samples in enumerate(samples_list): samples = samples.data.cpu() # wrangle things so they come out tiled samples = samples.view(N, *samples.size()[1:]) save_image(samples, '{}/gen_samples_{}_{:03d}.png'.format(runPath, i, epoch), nrow=int(sqrt(N))) def reconstruct(self, data, runPath, epoch): recons_mat = super(MNIST_SVHN, self).reconstruct([d[:8] for d in data]) for r, recons_list in enumerate(recons_mat): for o, recon in enumerate(recons_list): _data = data[r][:8].cpu() recon = recon.squeeze(0).cpu() # resize mnist to 32 and colour. 0 => mnist, 1 => svhn _data = _data if r == 1 else resize_img(_data, self.vaes[1].dataSize) recon = recon if o == 1 else resize_img(recon, self.vaes[1].dataSize) comp = torch.cat([_data, recon]) save_image(comp, '{}/recon_{}x{}_{:03d}.png'.format(runPath, r, o, epoch)) def analyse(self, data, runPath, epoch): zemb, zsl, kls_df = super(MNIST_SVHN, self).analyse(data, K=10) labels = ['Prior', *[vae.modelName.lower() for vae in self.vaes]] plot_embeddings(zemb, zsl, labels, '{}/emb_umap_{:03d}.png'.format(runPath, epoch)) plot_kls_df(kls_df, '{}/kl_distance_{:03d}.png'.format(runPath, epoch)) def resize_img(img, refsize): return F.pad(img, (2, 2, 2, 2)).expand(img.size(0), *refsize)
4,479
45.185567
101
py
mmvae-public
mmvae-public/bin/make-mnist-svhn-idx.py
import torch from torchvision import datasets, transforms def rand_match_on_idx(l1, idx1, l2, idx2, max_d=10000, dm=10): """ l*: sorted labels idx*: indices of sorted labels in original list """ _idx1, _idx2 = [], [] for l in l1.unique(): # assuming both have same idxs l_idx1, l_idx2 = idx1[l1 == l], idx2[l2 == l] n = min(l_idx1.size(0), l_idx2.size(0), max_d) l_idx1, l_idx2 = l_idx1[:n], l_idx2[:n] for _ in range(dm): _idx1.append(l_idx1[torch.randperm(n)]) _idx2.append(l_idx2[torch.randperm(n)]) return torch.cat(_idx1), torch.cat(_idx2) if __name__ == '__main__': max_d = 10000 # maximum number of datapoints per class dm = 30 # data multiplier: random permutations to match # get the individual datasets tx = transforms.ToTensor() train_mnist = datasets.MNIST('../data', train=True, download=True, transform=tx) test_mnist = datasets.MNIST('../data', train=False, download=True, transform=tx) train_svhn = datasets.SVHN('../data', split='train', download=True, transform=tx) test_svhn = datasets.SVHN('../data', split='test', download=True, transform=tx) # svhn labels need extra work train_svhn.labels = torch.LongTensor(train_svhn.labels.squeeze().astype(int)) % 10 test_svhn.labels = torch.LongTensor(test_svhn.labels.squeeze().astype(int)) % 10 mnist_l, mnist_li = train_mnist.targets.sort() svhn_l, svhn_li = train_svhn.labels.sort() idx1, idx2 = rand_match_on_idx(mnist_l, mnist_li, svhn_l, svhn_li, max_d=max_d, dm=dm) print('len train idx:', len(idx1), len(idx2)) torch.save(idx1, '../data/train-ms-mnist-idx.pt') torch.save(idx2, '../data/train-ms-svhn-idx.pt') mnist_l, mnist_li = test_mnist.targets.sort() svhn_l, svhn_li = test_svhn.labels.sort() idx1, idx2 = rand_match_on_idx(mnist_l, mnist_li, svhn_l, svhn_li, max_d=max_d, dm=dm) print('len test idx:', len(idx1), len(idx2)) torch.save(idx1, '../data/test-ms-mnist-idx.pt') torch.save(idx2, '../data/test-ms-svhn-idx.pt')
2,087
44.391304
90
py
MinkLoc3D
MinkLoc3D-master/training/train.py
# Author: Jacek Komorowski # Warsaw University of Technology import argparse import torch from training.trainer import do_train from misc.utils import MinkLocParams from datasets.dataset_utils import make_dataloaders if __name__ == '__main__': parser = argparse.ArgumentParser(description='Train Minkowski Net embeddings using BatchHard negative mining') parser.add_argument('--config', type=str, required=True, help='Path to configuration file') parser.add_argument('--model_config', type=str, required=True, help='Path to the model-specific configuration file') parser.add_argument('--debug', dest='debug', action='store_true') parser.set_defaults(debug=False) parser.add_argument('--visualize', dest='visualize', action='store_true') parser.set_defaults(visualize=False) args = parser.parse_args() print('Training config path: {}'.format(args.config)) print('Model config path: {}'.format(args.model_config)) print('Debug mode: {}'.format(args.debug)) print('Visualize: {}'.format(args.visualize)) params = MinkLocParams(args.config, args.model_config) params.print() if args.debug: torch.autograd.set_detect_anomaly(True) dataloaders = make_dataloaders(params, debug=args.debug) do_train(dataloaders, params, debug=args.debug, visualize=args.visualize)
1,341
37.342857
120
py
MinkLoc3D
MinkLoc3D-master/training/trainer.py
# Author: Jacek Komorowski # Warsaw University of Technology # Train on Oxford dataset (from PointNetVLAD paper) using BatchHard hard negative mining. import os from datetime import datetime import numpy as np import torch import pickle import tqdm import pathlib from torch.utils.tensorboard import SummaryWriter from eval.evaluate import evaluate, print_eval_stats from misc.utils import MinkLocParams, get_datetime from models.loss import make_loss from models.model_factory import model_factory def print_stats(stats, phase): if 'num_pairs' in stats: # For batch hard contrastive loss s = '{} - Mean loss: {:.6f} Avg. embedding norm: {:.4f} Pairs per batch (all/non-zero pos/non-zero neg): {:.1f}/{:.1f}/{:.1f}' print(s.format(phase, stats['loss'], stats['avg_embedding_norm'], stats['num_pairs'], stats['pos_pairs_above_threshold'], stats['neg_pairs_above_threshold'])) elif 'num_triplets' in stats: # For triplet loss s = '{} - Mean loss: {:.6f} Avg. embedding norm: {:.4f} Triplets per batch (all/non-zero): {:.1f}/{:.1f}' print(s.format(phase, stats['loss'], stats['avg_embedding_norm'], stats['num_triplets'], stats['num_non_zero_triplets'])) elif 'num_pos' in stats: s = '{} - Mean loss: {:.6f} Avg. embedding norm: {:.4f} #positives/negatives: {:.1f}/{:.1f}' print(s.format(phase, stats['loss'], stats['avg_embedding_norm'], stats['num_pos'], stats['num_neg'])) s = '' l = [] if 'mean_pos_pair_dist' in stats: s += 'Pos dist (min/mean/max): {:.4f}/{:.4f}/{:.4f} Neg dist (min/mean/max): {:.4f}/{:.4f}/{:.4f}' l += [stats['min_pos_pair_dist'], stats['mean_pos_pair_dist'], stats['max_pos_pair_dist'], stats['min_neg_pair_dist'], stats['mean_neg_pair_dist'], stats['max_neg_pair_dist']] if 'pos_loss' in stats: if len(s) > 0: s += ' ' s += 'Pos loss: {:.4f} Neg loss: {:.4f}' l += [stats['pos_loss'], stats['neg_loss']] if len(l) > 0: print(s.format(*l)) def tensors_to_numbers(stats): stats = {e: stats[e].item() if torch.is_tensor(stats[e]) else stats[e] for e in stats} return stats def do_train(dataloaders, params: MinkLocParams, debug=False, visualize=False): # Create model class s = get_datetime() model = model_factory(params) model_name = 'model_' + params.model_params.model + '_' + s print('Model name: {}'.format(model_name)) weights_path = create_weights_folder() model_pathname = os.path.join(weights_path, model_name) if hasattr(model, 'print_info'): model.print_info() else: n_params = sum([param.nelement() for param in model.parameters()]) print('Number of model parameters: {}'.format(n_params)) # Move the model to the proper device before configuring the optimizer if torch.cuda.is_available(): device = "cuda" model.to(device) else: device = "cpu" print('Model device: {}'.format(device)) loss_fn = make_loss(params) # Training elements if params.weight_decay is None or params.weight_decay == 0: optimizer = torch.optim.Adam(model.parameters(), lr=params.lr) else: optimizer = torch.optim.Adam(model.parameters(), lr=params.lr, weight_decay=params.weight_decay) if params.scheduler is None: scheduler = None else: if params.scheduler == 'CosineAnnealingLR': scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=params.epochs+1, eta_min=params.min_lr) elif params.scheduler == 'MultiStepLR': scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, params.scheduler_milestones, gamma=0.1) else: raise NotImplementedError('Unsupported LR scheduler: {}'.format(params.scheduler)) ########################################################################### # Initialize TensorBoard writer ########################################################################### now = datetime.now() logdir = os.path.join("../tf_logs", now.strftime("%Y%m%d-%H%M%S")) writer = SummaryWriter(logdir) ########################################################################### # ########################################################################### is_validation_set = 'val' in dataloaders if is_validation_set: phases = ['train', 'val'] else: phases = ['train'] # Training statistics stats = {'train': [], 'val': [], 'eval': []} for epoch in tqdm.tqdm(range(1, params.epochs + 1)): for phase in phases: if phase == 'train': model.train() else: model.eval() running_stats = [] # running stats for the current epoch count_batches = 0 for batch, positives_mask, negatives_mask in dataloaders[phase]: # batch is (batch_size, n_points, 3) tensor # labels is list with indexes of elements forming a batch count_batches += 1 batch_stats = {} if debug and count_batches > 2: break batch = {e: batch[e].to(device) for e in batch} n_positives = torch.sum(positives_mask).item() n_negatives = torch.sum(negatives_mask).item() if n_positives == 0 or n_negatives == 0: # Skip a batch without positives or negatives print('WARNING: Skipping batch without positive or negative examples') continue optimizer.zero_grad() if visualize: #visualize_batch(batch) pass with torch.set_grad_enabled(phase == 'train'): # Compute embeddings of all elements embeddings = model(batch) loss, temp_stats, _ = loss_fn(embeddings, positives_mask, negatives_mask) temp_stats = tensors_to_numbers(temp_stats) batch_stats.update(temp_stats) batch_stats['loss'] = loss.item() if phase == 'train': loss.backward() optimizer.step() running_stats.append(batch_stats) torch.cuda.empty_cache() # Prevent excessive GPU memory consumption by SparseTensors # ******* PHASE END ******* # Compute mean stats for the epoch epoch_stats = {} for key in running_stats[0].keys(): temp = [e[key] for e in running_stats] epoch_stats[key] = np.mean(temp) stats[phase].append(epoch_stats) print_stats(epoch_stats, phase) # ******* EPOCH END ******* if scheduler is not None: scheduler.step() loss_metrics = {'train': stats['train'][-1]['loss']} if 'val' in phases: loss_metrics['val'] = stats['val'][-1]['loss'] writer.add_scalars('Loss', loss_metrics, epoch) if 'num_triplets' in stats['train'][-1]: nz_metrics = {'train': stats['train'][-1]['num_non_zero_triplets']} if 'val' in phases: nz_metrics['val'] = stats['val'][-1]['num_non_zero_triplets'] writer.add_scalars('Non-zero triplets', nz_metrics, epoch) elif 'num_pairs' in stats['train'][-1]: nz_metrics = {'train_pos': stats['train'][-1]['pos_pairs_above_threshold'], 'train_neg': stats['train'][-1]['neg_pairs_above_threshold']} if 'val' in phases: nz_metrics['val_pos'] = stats['val'][-1]['pos_pairs_above_threshold'] nz_metrics['val_neg'] = stats['val'][-1]['neg_pairs_above_threshold'] writer.add_scalars('Non-zero pairs', nz_metrics, epoch) if params.batch_expansion_th is not None: # Dynamic batch expansion epoch_train_stats = stats['train'][-1] if 'num_non_zero_triplets' not in epoch_train_stats: print('WARNING: Batch size expansion is enabled, but the loss function is not supported') else: # Ratio of non-zero triplets rnz = epoch_train_stats['num_non_zero_triplets'] / epoch_train_stats['num_triplets'] if rnz < params.batch_expansion_th: dataloaders['train'].batch_sampler.expand_batch() print('') # Save final model weights final_model_path = model_pathname + '_final.pth' torch.save(model.state_dict(), final_model_path) stats = {'train_stats': stats, 'params': params} # Evaluate the final model model.eval() final_eval_stats = evaluate(model, device, params) print('Final model:') print_eval_stats(final_eval_stats) stats['eval'] = {'final': final_eval_stats} print('') # Pickle training stats and parameters pickle_path = model_pathname + '_stats.pickle' pickle.dump(stats, open(pickle_path, "wb")) # Append key experimental metrics to experiment summary file model_params_name = os.path.split(params.model_params.model_params_path)[1] config_name = os.path.split(params.params_path)[1] _, model_name = os.path.split(model_pathname) prefix = "{}, {}, {}".format(model_params_name, config_name, model_name) export_eval_stats("experiment_results.txt", prefix, final_eval_stats) def export_eval_stats(file_name, prefix, eval_stats): s = prefix ave_1p_recall_l = [] ave_recall_l = [] # Print results on the final model with open(file_name, "a") as f: for ds in ['oxford', 'university', 'residential', 'business']: ave_1p_recall = eval_stats[ds]['ave_one_percent_recall'] ave_1p_recall_l.append(ave_1p_recall) ave_recall = eval_stats[ds]['ave_recall'][0] ave_recall_l.append(ave_recall) s += ", {:0.2f}, {:0.2f}".format(ave_1p_recall, ave_recall) mean_1p_recall = np.mean(ave_1p_recall_l) mean_recall = np.mean(ave_recall_l) s += ", {:0.2f}, {:0.2f}\n".format(mean_1p_recall, mean_recall) f.write(s) def create_weights_folder(): # Create a folder to save weights of trained models this_file_path = pathlib.Path(__file__).parent.absolute() temp, _ = os.path.split(this_file_path) weights_path = os.path.join(temp, 'weights') if not os.path.exists(weights_path): os.mkdir(weights_path) assert os.path.exists(weights_path), 'Cannot create weights folder: {}'.format(weights_path) return weights_path
10,845
39.17037
139
py
MinkLoc3D
MinkLoc3D-master/eval/evaluate.py
# Author: Jacek Komorowski # Warsaw University of Technology # Evaluation code adapted from PointNetVlad code: https://github.com/mikacuy/pointnetvlad from sklearn.neighbors import KDTree import numpy as np import pickle import os import argparse import torch import tqdm import MinkowskiEngine as ME import random from misc.utils import MinkLocParams from models.model_factory import model_factory def evaluate(model, device, params, silent=True): # Run evaluation on all eval datasets assert len(params.eval_database_files) == len(params.eval_query_files) stats = {} for database_file, query_file in zip(params.eval_database_files, params.eval_query_files): # Extract location name from query and database files location_name = database_file.split('_')[0] temp = query_file.split('_')[0] assert location_name == temp, 'Database location: {} does not match query location: {}'.format(database_file, query_file) p = os.path.join(params.dataset_folder, database_file) with open(p, 'rb') as f: database_sets = pickle.load(f) p = os.path.join(params.dataset_folder, query_file) with open(p, 'rb') as f: query_sets = pickle.load(f) temp = evaluate_dataset(model, device, params, database_sets, query_sets, silent=silent) stats[location_name] = temp return stats def evaluate_dataset(model, device, params, database_sets, query_sets, silent=True): # Run evaluation on a single dataset recall = np.zeros(25) count = 0 similarity = [] one_percent_recall = [] database_embeddings = [] query_embeddings = [] model.eval() for set in tqdm.tqdm(database_sets, disable=silent): database_embeddings.append(get_latent_vectors(model, set, device, params)) for set in tqdm.tqdm(query_sets, disable=silent): query_embeddings.append(get_latent_vectors(model, set, device, params)) for i in range(len(query_sets)): for j in range(len(query_sets)): if i == j: continue pair_recall, pair_similarity, pair_opr = get_recall(i, j, database_embeddings, query_embeddings, query_sets, database_sets) recall += np.array(pair_recall) count += 1 one_percent_recall.append(pair_opr) for x in pair_similarity: similarity.append(x) ave_recall = recall / count average_similarity = np.mean(similarity) ave_one_percent_recall = np.mean(one_percent_recall) stats = {'ave_one_percent_recall': ave_one_percent_recall, 'ave_recall': ave_recall, 'average_similarity': average_similarity} return stats def load_pc(file_name, params): # returns Nx3 matrix file_path = os.path.join(params.dataset_folder, file_name) pc = np.fromfile(file_path, dtype=np.float64) # coords are within -1..1 range in each dimension assert pc.shape[0] == params.num_points * 3, "Error in point cloud shape: {}".format(file_path) pc = np.reshape(pc, (pc.shape[0] // 3, 3)) pc = torch.tensor(pc, dtype=torch.float) return pc def get_latent_vectors(model, set, device, params): # Adapted from original PointNetVLAD code model.eval() embeddings_l = [] for elem_ndx in set: x = load_pc(set[elem_ndx]["query"], params) with torch.no_grad(): # coords are (n_clouds, num_points, channels) tensor coords = ME.utils.sparse_quantize(coordinates=x, quantization_size=params.model_params.mink_quantization_size) bcoords = ME.utils.batched_coordinates([coords]) # Assign a dummy feature equal to 1 to each point # Coords must be on CPU, features can be on GPU - see MinkowskiEngine documentation feats = torch.ones((bcoords.shape[0], 1), dtype=torch.float32) batch = {'coords': bcoords.to(device), 'features': feats.to(device)} embedding = model(batch) # embedding is (1, 1024) tensor if params.normalize_embeddings: embedding = torch.nn.functional.normalize(embedding, p=2, dim=1) # Normalize embeddings embedding = embedding.detach().cpu().numpy() embeddings_l.append(embedding) embeddings = np.vstack(embeddings_l) return embeddings def get_recall(m, n, database_vectors, query_vectors, query_sets, database_sets): # Original PointNetVLAD code database_output = database_vectors[m] queries_output = query_vectors[n] # When embeddings are normalized, using Euclidean distance gives the same # nearest neighbour search results as using cosine distance database_nbrs = KDTree(database_output) num_neighbors = 25 recall = [0] * num_neighbors top1_similarity_score = [] one_percent_retrieved = 0 threshold = max(int(round(len(database_output)/100.0)), 1) num_evaluated = 0 for i in range(len(queries_output)): # i is query element ndx query_details = query_sets[n][i] # {'query': path, 'northing': , 'easting': } true_neighbors = query_details[m] if len(true_neighbors) == 0: continue num_evaluated += 1 distances, indices = database_nbrs.query(np.array([queries_output[i]]), k=num_neighbors) for j in range(len(indices[0])): if indices[0][j] in true_neighbors: if j == 0: similarity = np.dot(queries_output[i], database_output[indices[0][j]]) top1_similarity_score.append(similarity) recall[j] += 1 break if len(list(set(indices[0][0:threshold]).intersection(set(true_neighbors)))) > 0: one_percent_retrieved += 1 one_percent_recall = (one_percent_retrieved/float(num_evaluated))*100 recall = (np.cumsum(recall)/float(num_evaluated))*100 return recall, top1_similarity_score, one_percent_recall def print_eval_stats(stats): for database_name in stats: print('Dataset: {}'.format(database_name)) t = 'Avg. top 1% recall: {:.2f} Avg. similarity: {:.4f} Avg. recall @N:' print(t.format(stats[database_name]['ave_one_percent_recall'], stats[database_name]['average_similarity'])) print(stats[database_name]['ave_recall']) if __name__ == "__main__": parser = argparse.ArgumentParser(description='Evaluate model on PointNetVLAD (Oxford) dataset') parser.add_argument('--config', type=str, required=True, help='Path to configuration file') parser.add_argument('--model_config', type=str, required=True, help='Path to the model-specific configuration file') parser.add_argument('--weights', type=str, required=False, help='Trained model weights') args = parser.parse_args() print('Config path: {}'.format(args.config)) print('Model config path: {}'.format(args.model_config)) if args.weights is None: w = 'RANDOM WEIGHTS' else: w = args.weights print('Weights: {}'.format(w)) print('') params = MinkLocParams(args.config, args.model_config) params.print() if torch.cuda.is_available(): device = "cuda" else: device = "cpu" print('Device: {}'.format(device)) model = model_factory(params) if args.weights is not None: assert os.path.exists(args.weights), 'Cannot open network weights: {}'.format(args.weights) print('Loading weights: {}'.format(args.weights)) model.load_state_dict(torch.load(args.weights, map_location=device)) model.to(device) stats = evaluate(model, device, params, silent=False) print_eval_stats(stats)
7,883
36.542857
120
py
MinkLoc3D
MinkLoc3D-master/models/minkloc.py
# Author: Jacek Komorowski # Warsaw University of Technology import torch import MinkowskiEngine as ME from models.minkfpn import MinkFPN from models.netvlad import MinkNetVladWrapper import layers.pooling as pooling class MinkLoc(torch.nn.Module): def __init__(self, model, in_channels, feature_size, output_dim, planes, layers, num_top_down, conv0_kernel_size): super().__init__() self.model = model self.in_channels = in_channels self.feature_size = feature_size # Size of local features produced by local feature extraction block self.output_dim = output_dim # Dimensionality of the global descriptor self.backbone = MinkFPN(in_channels=in_channels, out_channels=self.feature_size, num_top_down=num_top_down, conv0_kernel_size=conv0_kernel_size, layers=layers, planes=planes) self.n_backbone_features = output_dim if model == 'MinkFPN_Max': assert self.feature_size == self.output_dim, 'output_dim must be the same as feature_size' self.pooling = pooling.MAC() elif model == 'MinkFPN_GeM': assert self.feature_size == self.output_dim, 'output_dim must be the same as feature_size' self.pooling = pooling.GeM() elif model == 'MinkFPN_NetVlad': self.pooling = MinkNetVladWrapper(feature_size=self.feature_size, output_dim=self.output_dim, cluster_size=64, gating=False) elif model == 'MinkFPN_NetVlad_CG': self.pooling = MinkNetVladWrapper(feature_size=self.feature_size, output_dim=self.output_dim, cluster_size=64, gating=True) else: raise NotImplementedError('Model not implemented: {}'.format(model)) def forward(self, batch): # Coords must be on CPU, features can be on GPU - see MinkowskiEngine documentation x = ME.SparseTensor(batch['features'], coordinates=batch['coords']) x = self.backbone(x) # x is (num_points, n_features) tensor assert x.shape[1] == self.feature_size, 'Backbone output tensor has: {} channels. Expected: {}'.format(x.shape[1], self.feature_size) x = self.pooling(x) assert x.dim() == 2, 'Expected 2-dimensional tensor (batch_size,output_dim). Got {} dimensions.'.format(x.dim()) assert x.shape[1] == self.output_dim, 'Output tensor has: {} channels. Expected: {}'.format(x.shape[1], self.output_dim) # x is (batch_size, output_dim) tensor return x def print_info(self): print('Model class: MinkLoc') n_params = sum([param.nelement() for param in self.parameters()]) print('Total parameters: {}'.format(n_params)) n_params = sum([param.nelement() for param in self.backbone.parameters()]) print('Backbone parameters: {}'.format(n_params)) n_params = sum([param.nelement() for param in self.pooling.parameters()]) print('Aggregation parameters: {}'.format(n_params)) if hasattr(self.backbone, 'print_info'): self.backbone.print_info() if hasattr(self.pooling, 'print_info'): self.pooling.print_info()
3,242
50.47619
141
py
MinkLoc3D
MinkLoc3D-master/models/model_factory.py
# Author: Jacek Komorowski # Warsaw University of Technology import models.minkloc as minkloc def model_factory(params): in_channels = 1 if 'MinkFPN' in params.model_params.model: model = minkloc.MinkLoc(params.model_params.model, in_channels=in_channels, feature_size=params.model_params.feature_size, output_dim=params.model_params.output_dim, planes=params.model_params.planes, layers=params.model_params.layers, num_top_down=params.model_params.num_top_down, conv0_kernel_size=params.model_params.conv0_kernel_size) else: raise NotImplementedError('Model not implemented: {}'.format(params.model_params.model)) return model
793
38.7
113
py
MinkLoc3D
MinkLoc3D-master/models/resnet.py
# Copyright (c) Chris Choy (chrischoy@ai.stanford.edu). # # Permission is hereby granted, free of charge, to any person obtaining a copy of # this software and associated documentation files (the "Software"), to deal in # the Software without restriction, including without limitation the rights to # use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies # of the Software, and to permit persons to whom the Software is furnished to do # so, subject to the following conditions: # # The above copyright notice and this permission notice shall be included in all # copies or substantial portions of the Software. # # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR # IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, # FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE # AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER # LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE # SOFTWARE. # # Please cite "4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural # Networks", CVPR'19 (https://arxiv.org/abs/1904.08755) if you use any part # of the code. import torch.nn as nn import MinkowskiEngine as ME from MinkowskiEngine.modules.resnet_block import BasicBlock, Bottleneck class ResNetBase(nn.Module): block = None layers = () init_dim = 64 planes = (64, 128, 256, 512) def __init__(self, in_channels, out_channels, D=3): nn.Module.__init__(self) self.D = D assert self.block is not None self.network_initialization(in_channels, out_channels, D) self.weight_initialization() def network_initialization(self, in_channels, out_channels, D): self.inplanes = self.init_dim self.conv1 = ME.MinkowskiConvolution( in_channels, self.inplanes, kernel_size=5, stride=2, dimension=D) self.bn1 = ME.MinkowskiBatchNorm(self.inplanes) self.relu = ME.MinkowskiReLU(inplace=True) self.pool = ME.MinkowskiAvgPooling(kernel_size=2, stride=2, dimension=D) self.layer1 = self._make_layer( self.block, self.planes[0], self.layers[0], stride=2) self.layer2 = self._make_layer( self.block, self.planes[1], self.layers[1], stride=2) self.layer3 = self._make_layer( self.block, self.planes[2], self.layers[2], stride=2) self.layer4 = self._make_layer( self.block, self.planes[3], self.layers[3], stride=2) self.conv5 = ME.MinkowskiConvolution( self.inplanes, self.inplanes, kernel_size=3, stride=3, dimension=D) self.bn5 = ME.MinkowskiBatchNorm(self.inplanes) self.glob_avg = ME.MinkowskiGlobalMaxPooling() self.final = ME.MinkowskiLinear(self.inplanes, out_channels, bias=True) def weight_initialization(self): for m in self.modules(): if isinstance(m, ME.MinkowskiConvolution): ME.utils.kaiming_normal_(m.kernel, mode='fan_out', nonlinearity='relu') if isinstance(m, ME.MinkowskiBatchNorm): nn.init.constant_(m.bn.weight, 1) nn.init.constant_(m.bn.bias, 0) def _make_layer(self, block, planes, blocks, stride=1, dilation=1, bn_momentum=0.1): downsample = None if stride != 1 or self.inplanes != planes * block.expansion: downsample = nn.Sequential( ME.MinkowskiConvolution( self.inplanes, planes * block.expansion, kernel_size=1, stride=stride, dimension=self.D), ME.MinkowskiBatchNorm(planes * block.expansion)) layers = [] layers.append( block( self.inplanes, planes, stride=stride, dilation=dilation, downsample=downsample, dimension=self.D)) self.inplanes = planes * block.expansion for i in range(1, blocks): layers.append( block( self.inplanes, planes, stride=1, dilation=dilation, dimension=self.D)) return nn.Sequential(*layers) def forward(self, x): x = self.conv1(x) x = self.bn1(x) x = self.relu(x) x = self.pool(x) x = self.layer1(x) x = self.layer2(x) x = self.layer3(x) x = self.layer4(x) x = self.conv5(x) x = self.bn5(x) x = self.relu(x) x = self.glob_avg(x) return self.final(x) class ResNet14(ResNetBase): BLOCK = BasicBlock LAYERS = (1, 1, 1, 1) class ResNet18(ResNetBase): BLOCK = BasicBlock LAYERS = (2, 2, 2, 2) class ResNet34(ResNetBase): BLOCK = BasicBlock LAYERS = (3, 4, 6, 3) class ResNet50(ResNetBase): BLOCK = Bottleneck LAYERS = (3, 4, 6, 3) class ResNet101(ResNetBase): BLOCK = Bottleneck LAYERS = (3, 4, 23, 3)
5,315
31.814815
87
py
MinkLoc3D
MinkLoc3D-master/models/minkfpn.py
# Author: Jacek Komorowski # Warsaw University of Technology import torch.nn as nn import MinkowskiEngine as ME from MinkowskiEngine.modules.resnet_block import BasicBlock from models.resnet import ResNetBase class MinkFPN(ResNetBase): # Feature Pyramid Network (FPN) architecture implementation using Minkowski ResNet building blocks def __init__(self, in_channels, out_channels, num_top_down=1, conv0_kernel_size=5, block=BasicBlock, layers=(1, 1, 1), planes=(32, 64, 64)): assert len(layers) == len(planes) assert 1 <= len(layers) assert 0 <= num_top_down <= len(layers) self.num_bottom_up = len(layers) self.num_top_down = num_top_down self.conv0_kernel_size = conv0_kernel_size self.block = block self.layers = layers self.planes = planes self.lateral_dim = out_channels self.init_dim = planes[0] ResNetBase.__init__(self, in_channels, out_channels, D=3) def network_initialization(self, in_channels, out_channels, D): assert len(self.layers) == len(self.planes) assert len(self.planes) == self.num_bottom_up self.convs = nn.ModuleList() # Bottom-up convolutional blocks with stride=2 self.bn = nn.ModuleList() # Bottom-up BatchNorms self.blocks = nn.ModuleList() # Bottom-up blocks self.tconvs = nn.ModuleList() # Top-down tranposed convolutions self.conv1x1 = nn.ModuleList() # 1x1 convolutions in lateral connections # The first convolution is special case, with kernel size = 5 self.inplanes = self.planes[0] self.conv0 = ME.MinkowskiConvolution(in_channels, self.inplanes, kernel_size=self.conv0_kernel_size, dimension=D) self.bn0 = ME.MinkowskiBatchNorm(self.inplanes) for plane, layer in zip(self.planes, self.layers): self.convs.append(ME.MinkowskiConvolution(self.inplanes, self.inplanes, kernel_size=2, stride=2, dimension=D)) self.bn.append(ME.MinkowskiBatchNorm(self.inplanes)) self.blocks.append(self._make_layer(self.block, plane, layer)) # Lateral connections for i in range(self.num_top_down): self.conv1x1.append(ME.MinkowskiConvolution(self.planes[-1 - i], self.lateral_dim, kernel_size=1, stride=1, dimension=D)) self.tconvs.append(ME.MinkowskiConvolutionTranspose(self.lateral_dim, self.lateral_dim, kernel_size=2, stride=2, dimension=D)) # There's one more lateral connection than top-down TConv blocks if self.num_top_down < self.num_bottom_up: # Lateral connection from Conv block 1 or above self.conv1x1.append(ME.MinkowskiConvolution(self.planes[-1 - self.num_top_down], self.lateral_dim, kernel_size=1, stride=1, dimension=D)) else: # Lateral connection from Con0 block self.conv1x1.append(ME.MinkowskiConvolution(self.planes[0], self.lateral_dim, kernel_size=1, stride=1, dimension=D)) self.relu = ME.MinkowskiReLU(inplace=True) def forward(self, x): # *** BOTTOM-UP PASS *** # First bottom-up convolution is special (with bigger stride) feature_maps = [] x = self.conv0(x) x = self.bn0(x) x = self.relu(x) if self.num_top_down == self.num_bottom_up: feature_maps.append(x) # BOTTOM-UP PASS for ndx, (conv, bn, block) in enumerate(zip(self.convs, self.bn, self.blocks)): x = conv(x) # Decreases spatial resolution (conv stride=2) x = bn(x) x = self.relu(x) x = block(x) if self.num_bottom_up - 1 - self.num_top_down <= ndx < len(self.convs) - 1: feature_maps.append(x) assert len(feature_maps) == self.num_top_down x = self.conv1x1[0](x) # TOP-DOWN PASS for ndx, tconv in enumerate(self.tconvs): x = tconv(x) # Upsample using transposed convolution x = x + self.conv1x1[ndx+1](feature_maps[-ndx - 1]) return x
4,364
44.947368
125
py
MinkLoc3D
MinkLoc3D-master/models/loss.py
# Author: Jacek Komorowski # Warsaw University of Technology import numpy as np import torch from pytorch_metric_learning import losses, miners, reducers from pytorch_metric_learning.distances import LpDistance def make_loss(params): if params.loss == 'BatchHardTripletMarginLoss': # BatchHard mining with triplet margin loss # Expects input: embeddings, positives_mask, negatives_mask loss_fn = BatchHardTripletLossWithMasks(params.margin, params.normalize_embeddings) elif params.loss == 'BatchHardContrastiveLoss': loss_fn = BatchHardContrastiveLossWithMasks(params.pos_margin, params.neg_margin, params.normalize_embeddings) else: print('Unknown loss: {}'.format(params.loss)) raise NotImplementedError return loss_fn class HardTripletMinerWithMasks: # Hard triplet miner def __init__(self, distance): self.distance = distance # Stats self.max_pos_pair_dist = None self.max_neg_pair_dist = None self.mean_pos_pair_dist = None self.mean_neg_pair_dist = None self.min_pos_pair_dist = None self.min_neg_pair_dist = None def __call__(self, embeddings, positives_mask, negatives_mask): assert embeddings.dim() == 2 d_embeddings = embeddings.detach() with torch.no_grad(): hard_triplets = self.mine(d_embeddings, positives_mask, negatives_mask) return hard_triplets def mine(self, embeddings, positives_mask, negatives_mask): # Based on pytorch-metric-learning implementation dist_mat = self.distance(embeddings) (hardest_positive_dist, hardest_positive_indices), a1p_keep = get_max_per_row(dist_mat, positives_mask) (hardest_negative_dist, hardest_negative_indices), a2n_keep = get_min_per_row(dist_mat, negatives_mask) a_keep_idx = torch.where(a1p_keep & a2n_keep) a = torch.arange(dist_mat.size(0)).to(hardest_positive_indices.device)[a_keep_idx] p = hardest_positive_indices[a_keep_idx] n = hardest_negative_indices[a_keep_idx] self.max_pos_pair_dist = torch.max(hardest_positive_dist).item() self.max_neg_pair_dist = torch.max(hardest_negative_dist).item() self.mean_pos_pair_dist = torch.mean(hardest_positive_dist).item() self.mean_neg_pair_dist = torch.mean(hardest_negative_dist).item() self.min_pos_pair_dist = torch.min(hardest_positive_dist).item() self.min_neg_pair_dist = torch.min(hardest_negative_dist).item() return a, p, n def get_max_per_row(mat, mask): non_zero_rows = torch.any(mask, dim=1) mat_masked = mat.clone() mat_masked[~mask] = 0 return torch.max(mat_masked, dim=1), non_zero_rows def get_min_per_row(mat, mask): non_inf_rows = torch.any(mask, dim=1) mat_masked = mat.clone() mat_masked[~mask] = float('inf') return torch.min(mat_masked, dim=1), non_inf_rows class BatchHardTripletLossWithMasks: def __init__(self, margin, normalize_embeddings): self.margin = margin self.normalize_embeddings = normalize_embeddings self.distance = LpDistance(normalize_embeddings=normalize_embeddings, collect_stats=True) # We use triplet loss with Euclidean distance self.miner_fn = HardTripletMinerWithMasks(distance=self.distance) reducer_fn = reducers.AvgNonZeroReducer(collect_stats=True) self.loss_fn = losses.TripletMarginLoss(margin=self.margin, swap=True, distance=self.distance, reducer=reducer_fn, collect_stats=True) def __call__(self, embeddings, positives_mask, negatives_mask): hard_triplets = self.miner_fn(embeddings, positives_mask, negatives_mask) dummy_labels = torch.arange(embeddings.shape[0]).to(embeddings.device) loss = self.loss_fn(embeddings, dummy_labels, hard_triplets) stats = {'loss': loss.item(), 'avg_embedding_norm': self.loss_fn.distance.final_avg_query_norm, 'num_non_zero_triplets': self.loss_fn.reducer.triplets_past_filter, 'num_triplets': len(hard_triplets[0]), 'mean_pos_pair_dist': self.miner_fn.mean_pos_pair_dist, 'mean_neg_pair_dist': self.miner_fn.mean_neg_pair_dist, 'max_pos_pair_dist': self.miner_fn.max_pos_pair_dist, 'max_neg_pair_dist': self.miner_fn.max_neg_pair_dist, 'min_pos_pair_dist': self.miner_fn.min_pos_pair_dist, 'min_neg_pair_dist': self.miner_fn.min_neg_pair_dist } return loss, stats, hard_triplets class BatchHardContrastiveLossWithMasks: def __init__(self, pos_margin, neg_margin, normalize_embeddings): self.pos_margin = pos_margin self.neg_margin = neg_margin self.distance = LpDistance(normalize_embeddings=normalize_embeddings, collect_stats=True) self.miner_fn = HardTripletMinerWithMasks(distance=self.distance) # We use contrastive loss with squared Euclidean distance reducer_fn = reducers.AvgNonZeroReducer(collect_stats=True) self.loss_fn = losses.ContrastiveLoss(pos_margin=self.pos_margin, neg_margin=self.neg_margin, distance=self.distance, reducer=reducer_fn, collect_stats=True) def __call__(self, embeddings, positives_mask, negatives_mask): hard_triplets = self.miner_fn(embeddings, positives_mask, negatives_mask) dummy_labels = torch.arange(embeddings.shape[0]).to(embeddings.device) loss = self.loss_fn(embeddings, dummy_labels, hard_triplets) stats = {'loss': loss.item(), 'avg_embedding_norm': self.loss_fn.distance.final_avg_query_norm, 'pos_pairs_above_threshold': self.loss_fn.reducer.reducers['pos_loss'].pos_pairs_above_threshold, 'neg_pairs_above_threshold': self.loss_fn.reducer.reducers['neg_loss'].neg_pairs_above_threshold, 'pos_loss': self.loss_fn.reducer.reducers['pos_loss'].pos_loss.item(), 'neg_loss': self.loss_fn.reducer.reducers['neg_loss'].neg_loss.item(), 'num_pairs': 2*len(hard_triplets[0]), 'mean_pos_pair_dist': self.miner_fn.mean_pos_pair_dist, 'mean_neg_pair_dist': self.miner_fn.mean_neg_pair_dist, 'max_pos_pair_dist': self.miner_fn.max_pos_pair_dist, 'max_neg_pair_dist': self.miner_fn.max_neg_pair_dist, 'min_pos_pair_dist': self.miner_fn.min_pos_pair_dist, 'min_neg_pair_dist': self.miner_fn.min_neg_pair_dist } return loss, stats, hard_triplets
6,696
49.353383
118
py
MinkLoc3D
MinkLoc3D-master/models/netvlad.py
# Code taken from PointNetVLAD Pytorch implementation: https://github.com/cattaneod/PointNetVlad-Pytorch # Adapted by Jacek Komorowski import torch import torch.nn as nn import torch.utils.data import torch.nn.functional as F import math # NOTE: The toolbox can only pool lists of features of the same length. It was specifically optimized to efficiently # o so. One way to handle multiple lists of features of variable length is to create, via a data augmentation # technique, a tensor of shape: 'batch_size'x'max_samples'x'feature_size'. Where 'max_samples' would be the maximum # number of feature per list. Then for each list, you would fill the tensor with 0 values. class NetVLADLoupe(nn.Module): def __init__(self, feature_size, cluster_size, output_dim, gating=True, add_batch_norm=True): super().__init__() self.feature_size = feature_size self.output_dim = output_dim self.gating = gating self.add_batch_norm = add_batch_norm self.cluster_size = cluster_size self.softmax = nn.Softmax(dim=-1) self.cluster_weights = nn.Parameter(torch.randn(feature_size, cluster_size) * 1 / math.sqrt(feature_size)) self.cluster_weights2 = nn.Parameter(torch.randn(1, feature_size, cluster_size) * 1 / math.sqrt(feature_size)) self.hidden1_weights = nn.Parameter( torch.randn(cluster_size * feature_size, output_dim) * 1 / math.sqrt(feature_size)) if add_batch_norm: self.cluster_biases = None self.bn1 = nn.BatchNorm1d(cluster_size) else: self.cluster_biases = nn.Parameter(torch.randn(cluster_size) * 1 / math.sqrt(feature_size)) self.bn1 = None self.bn2 = nn.BatchNorm1d(output_dim) if gating: self.context_gating = GatingContext(output_dim, add_batch_norm=add_batch_norm) def forward(self, x): # Expects (batch_size, num_points, channels) tensor assert x.dim() == 3 num_points = x.shape[1] activation = torch.matmul(x, self.cluster_weights) if self.add_batch_norm: # activation = activation.transpose(1,2).contiguous() activation = activation.view(-1, self.cluster_size) activation = self.bn1(activation) activation = activation.view(-1, num_points, self.cluster_size) # activation = activation.transpose(1,2).contiguous() else: activation = activation + self.cluster_biases activation = self.softmax(activation) activation = activation.view((-1, num_points, self.cluster_size)) a_sum = activation.sum(-2, keepdim=True) a = a_sum * self.cluster_weights2 activation = torch.transpose(activation, 2, 1) x = x.view((-1, num_points, self.feature_size)) vlad = torch.matmul(activation, x) vlad = torch.transpose(vlad, 2, 1) vlad = vlad - a vlad = F.normalize(vlad, dim=1, p=2) vlad = vlad.reshape((-1, self.cluster_size * self.feature_size)) vlad = F.normalize(vlad, dim=1, p=2) vlad = torch.matmul(vlad, self.hidden1_weights) vlad = self.bn2(vlad) if self.gating: vlad = self.context_gating(vlad) return vlad class GatingContext(nn.Module): def __init__(self, dim, add_batch_norm=True): super(GatingContext, self).__init__() self.dim = dim self.add_batch_norm = add_batch_norm self.gating_weights = nn.Parameter( torch.randn(dim, dim) * 1 / math.sqrt(dim)) self.sigmoid = nn.Sigmoid() if add_batch_norm: self.gating_biases = None self.bn1 = nn.BatchNorm1d(dim) else: self.gating_biases = nn.Parameter( torch.randn(dim) * 1 / math.sqrt(dim)) self.bn1 = None def forward(self, x): gates = torch.matmul(x, self.gating_weights) if self.add_batch_norm: gates = self.bn1(gates) else: gates = gates + self.gating_biases gates = self.sigmoid(gates) activation = x * gates return activation class MinkNetVladWrapper(torch.nn.Module): # Wrapper around NetVlad class to process sparse tensors from Minkowski networks def __init__(self, feature_size, output_dim, cluster_size=64, gating=True): super().__init__() self.feature_size = feature_size self.output_dim = output_dim self.net_vlad = NetVLADLoupe(feature_size=feature_size, cluster_size=cluster_size, output_dim=output_dim, gating=gating, add_batch_norm=True) def forward(self, x): # x is SparseTensor assert x.F.shape[1] == self.feature_size features = x.decomposed_features # features is a list of (n_points, feature_size) tensors with variable number of points batch_size = len(features) features = torch.nn.utils.rnn.pad_sequence(features, batch_first=True) # features is (batch_size, n_points, feature_size) tensor padded with zeros x = self.net_vlad(features) assert x.shape[0] == batch_size assert x.shape[1] == self.output_dim return x # Return (batch_size, output_dim) tensor
5,287
38.17037
118
py
MinkLoc3D
MinkLoc3D-master/datasets/dataset_utils.py
# Author: Jacek Komorowski # Warsaw University of Technology import numpy as np import torch from torch.utils.data import DataLoader import MinkowskiEngine as ME from datasets.oxford import OxfordDataset, TrainTransform, TrainSetTransform from datasets.samplers import BatchSampler from misc.utils import MinkLocParams def make_datasets(params: MinkLocParams, debug=False): # Create training and validation datasets datasets = {} train_transform = TrainTransform(params.aug_mode) train_set_transform = TrainSetTransform(params.aug_mode) datasets['train'] = OxfordDataset(params.dataset_folder, params.train_file, train_transform, set_transform=train_set_transform) val_transform = None if params.val_file is not None: datasets['val'] = OxfordDataset(params.dataset_folder, params.val_file, val_transform) return datasets def make_collate_fn(dataset: OxfordDataset, mink_quantization_size=None): # set_transform: the transform to be applied to all batch elements def collate_fn(data_list): # Constructs a batch object clouds = [e[0] for e in data_list] labels = [e[1] for e in data_list] batch = torch.stack(clouds, dim=0) # Produces (batch_size, n_points, 3) tensor if dataset.set_transform is not None: # Apply the same transformation on all dataset elements batch = dataset.set_transform(batch) if mink_quantization_size is None: # Not a MinkowskiEngine based model batch = {'cloud': batch} else: coords = [ME.utils.sparse_quantize(coordinates=e, quantization_size=mink_quantization_size) for e in batch] coords = ME.utils.batched_coordinates(coords) # Assign a dummy feature equal to 1 to each point # Coords must be on CPU, features can be on GPU - see MinkowskiEngine documentation feats = torch.ones((coords.shape[0], 1), dtype=torch.float32) batch = {'coords': coords, 'features': feats} # Compute positives and negatives mask # Compute positives and negatives mask positives_mask = [[in_sorted_array(e, dataset.queries[label].positives) for e in labels] for label in labels] negatives_mask = [[not in_sorted_array(e, dataset.queries[label].non_negatives) for e in labels] for label in labels] positives_mask = torch.tensor(positives_mask) negatives_mask = torch.tensor(negatives_mask) # Returns (batch_size, n_points, 3) tensor and positives_mask and # negatives_mask which are batch_size x batch_size boolean tensors return batch, positives_mask, negatives_mask return collate_fn def make_dataloaders(params: MinkLocParams, debug=False): """ Create training and validation dataloaders that return groups of k=2 similar elements :param train_params: :param model_params: :return: """ datasets = make_datasets(params, debug=debug) dataloders = {} train_sampler = BatchSampler(datasets['train'], batch_size=params.batch_size, batch_size_limit=params.batch_size_limit, batch_expansion_rate=params.batch_expansion_rate) # Collate function collates items into a batch and applies a 'set transform' on the entire batch train_collate_fn = make_collate_fn(datasets['train'], params.model_params.mink_quantization_size) dataloders['train'] = DataLoader(datasets['train'], batch_sampler=train_sampler, collate_fn=train_collate_fn, num_workers=params.num_workers, pin_memory=True) if 'val' in datasets: val_sampler = BatchSampler(datasets['val'], batch_size=params.batch_size) # Collate function collates items into a batch and applies a 'set transform' on the entire batch # Currently validation dataset has empty set_transform function, but it may change in the future val_collate_fn = make_collate_fn(datasets['val'], params.model_params.mink_quantization_size) dataloders['val'] = DataLoader(datasets['val'], batch_sampler=val_sampler, collate_fn=val_collate_fn, num_workers=params.num_workers, pin_memory=True) return dataloders def in_sorted_array(e: int, array: np.ndarray) -> bool: pos = np.searchsorted(array, e) if pos == len(array) or pos == -1: return False else: return array[pos] == e
4,547
44.48
125
py
MinkLoc3D
MinkLoc3D-master/datasets/oxford.py
# Author: Jacek Komorowski # Warsaw University of Technology # Dataset wrapper for Oxford laser scans dataset from PointNetVLAD project # For information on dataset see: https://github.com/mikacuy/pointnetvlad import os import pickle import numpy as np import math from scipy.linalg import expm, norm import random import torch from torch.utils.data import Dataset import torchvision.transforms as transforms import tqdm class OxfordDataset(Dataset): """ Dataset wrapper for Oxford laser scans dataset from PointNetVLAD project. """ def __init__(self, dataset_path: str, query_filename: str, image_path: str = None, lidar2image_ndx=None, transform=None, set_transform=None, image_transform=None, use_cloud=True): assert os.path.exists(dataset_path), 'Cannot access dataset path: {}'.format(dataset_path) self.dataset_path = dataset_path self.query_filepath = os.path.join(dataset_path, query_filename) assert os.path.exists(self.query_filepath), 'Cannot access query file: {}'.format(self.query_filepath) self.transform = transform self.set_transform = set_transform self.queries: Dict[int, TrainingTuple] = pickle.load(open(self.query_filepath, 'rb')) self.image_path = image_path self.lidar2image_ndx = lidar2image_ndx self.image_transform = image_transform self.n_points = 4096 # pointclouds in the dataset are downsampled to 4096 points self.image_ext = '.png' self.use_cloud = use_cloud print('{} queries in the dataset'.format(len(self))) def __len__(self): return len(self.queries) def __getitem__(self, ndx): # Load point cloud and apply transform file_pathname = os.path.join(self.dataset_path, self.queries[ndx].rel_scan_filepath) query_pc = self.load_pc(file_pathname) if self.transform is not None: query_pc = self.transform(query_pc) return query_pc, ndx def get_positives(self, ndx): return self.queries[ndx].positives def get_non_negatives(self, ndx): return self.queries[ndx].non_negatives def load_pc(self, filename): # Load point cloud, does not apply any transform # Returns Nx3 matrix file_path = os.path.join(self.dataset_path, filename) pc = np.fromfile(file_path, dtype=np.float64) # coords are within -1..1 range in each dimension assert pc.shape[0] == self.n_points * 3, "Error in point cloud shape: {}".format(file_path) pc = np.reshape(pc, (pc.shape[0] // 3, 3)) pc = torch.tensor(pc, dtype=torch.float) return pc class TrainingTuple: # Tuple describing an element for training/validation def __init__(self, id: int, timestamp: int, rel_scan_filepath: str, positives: np.ndarray, non_negatives: np.ndarray, position: np.ndarray): # id: element id (ids start from 0 and are consecutive numbers) # ts: timestamp # rel_scan_filepath: relative path to the scan # positives: sorted ndarray of positive elements id # negatives: sorted ndarray of elements id # position: x, y position in meters (northing, easting) assert position.shape == (2,) self.id = id self.timestamp = timestamp self.rel_scan_filepath = rel_scan_filepath self.positives = positives self.non_negatives = non_negatives self.position = position class TrainTransform: def __init__(self, aug_mode): # 1 is default mode, no transform self.aug_mode = aug_mode if self.aug_mode == 1: t = [JitterPoints(sigma=0.001, clip=0.002), RemoveRandomPoints(r=(0.0, 0.1)), RandomTranslation(max_delta=0.01), RemoveRandomBlock(p=0.4)] else: raise NotImplementedError('Unknown aug_mode: {}'.format(self.aug_mode)) self.transform = transforms.Compose(t) def __call__(self, e): if self.transform is not None: e = self.transform(e) return e class TrainSetTransform: def __init__(self, aug_mode): # 1 is default mode, no transform self.aug_mode = aug_mode self.transform = None t = [RandomRotation(max_theta=5, max_theta2=0, axis=np.array([0, 0, 1])), RandomFlip([0.25, 0.25, 0.])] self.transform = transforms.Compose(t) def __call__(self, e): if self.transform is not None: e = self.transform(e) return e class RandomFlip: def __init__(self, p): # p = [p_x, p_y, p_z] probability of flipping each axis assert len(p) == 3 assert 0 < sum(p) <= 1, 'sum(p) must be in (0, 1] range, is: {}'.format(sum(p)) self.p = p self.p_cum_sum = np.cumsum(p) def __call__(self, coords): r = random.random() if r <= self.p_cum_sum[0]: # Flip the first axis coords[..., 0] = -coords[..., 0] elif r <= self.p_cum_sum[1]: # Flip the second axis coords[..., 1] = -coords[..., 1] elif r <= self.p_cum_sum[2]: # Flip the third axis coords[..., 2] = -coords[..., 2] return coords class RandomRotation: def __init__(self, axis=None, max_theta=180, max_theta2=15): self.axis = axis self.max_theta = max_theta # Rotation around axis self.max_theta2 = max_theta2 # Smaller rotation in random direction def _M(self, axis, theta): return expm(np.cross(np.eye(3), axis / norm(axis) * theta)).astype(np.float32) def __call__(self, coords): if self.axis is not None: axis = self.axis else: axis = np.random.rand(3) - 0.5 R = self._M(axis, (np.pi * self.max_theta / 180) * 2 * (np.random.rand(1) - 0.5)) if self.max_theta2 is None: coords = coords @ R else: R_n = self._M(np.random.rand(3) - 0.5, (np.pi * self.max_theta2 / 180) * 2 * (np.random.rand(1) - 0.5)) coords = coords @ R @ R_n return coords class RandomTranslation: def __init__(self, max_delta=0.05): self.max_delta = max_delta def __call__(self, coords): trans = self.max_delta * np.random.randn(1, 3) return coords + trans.astype(np.float32) class RandomScale: def __init__(self, min, max): self.scale = max - min self.bias = min def __call__(self, coords): s = self.scale * np.random.rand(1) + self.bias return coords * s.astype(np.float32) class RandomShear: def __init__(self, delta=0.1): self.delta = delta def __call__(self, coords): T = np.eye(3) + self.delta * np.random.randn(3, 3) return coords @ T.astype(np.float32) class JitterPoints: def __init__(self, sigma=0.01, clip=None, p=1.): assert 0 < p <= 1. assert sigma > 0. self.sigma = sigma self.clip = clip self.p = p def __call__(self, e): """ Randomly jitter points. jittering is per point. Input: BxNx3 array, original batch of point clouds Return: BxNx3 array, jittered batch of point clouds """ sample_shape = (e.shape[0],) if self.p < 1.: # Create a mask for points to jitter m = torch.distributions.categorical.Categorical(probs=torch.tensor([1 - self.p, self.p])) mask = m.sample(sample_shape=sample_shape) else: mask = torch.ones(sample_shape, dtype=torch.int64 ) mask = mask == 1 jitter = self.sigma * torch.randn_like(e[mask]) if self.clip is not None: jitter = torch.clamp(jitter, min=-self.clip, max=self.clip) e[mask] = e[mask] + jitter return e class RemoveRandomPoints: def __init__(self, r): if type(r) is list or type(r) is tuple: assert len(r) == 2 assert 0 <= r[0] <= 1 assert 0 <= r[1] <= 1 self.r_min = float(r[0]) self.r_max = float(r[1]) else: assert 0 <= r <= 1 self.r_min = None self.r_max = float(r) def __call__(self, e): n = len(e) if self.r_min is None: r = self.r_max else: # Randomly select removal ratio r = random.uniform(self.r_min, self.r_max) mask = np.random.choice(range(n), size=int(n*r), replace=False) # select elements to remove e[mask] = torch.zeros_like(e[mask]) return e class RemoveRandomBlock: """ Randomly remove part of the point cloud. Similar to PyTorch RandomErasing but operating on 3D point clouds. Erases fronto-parallel cuboid. Instead of erasing we set coords of removed points to (0, 0, 0) to retain the same number of points """ def __init__(self, p=0.5, scale=(0.02, 0.33), ratio=(0.3, 3.3)): self.p = p self.scale = scale self.ratio = ratio def get_params(self, coords): # Find point cloud 3D bounding box flattened_coords = coords.view(-1, 3) min_coords, _ = torch.min(flattened_coords, dim=0) max_coords, _ = torch.max(flattened_coords, dim=0) span = max_coords - min_coords area = span[0] * span[1] erase_area = random.uniform(self.scale[0], self.scale[1]) * area aspect_ratio = random.uniform(self.ratio[0], self.ratio[1]) h = math.sqrt(erase_area * aspect_ratio) w = math.sqrt(erase_area / aspect_ratio) x = min_coords[0] + random.uniform(0, 1) * (span[0] - w) y = min_coords[1] + random.uniform(0, 1) * (span[1] - h) return x, y, w, h def __call__(self, coords): if random.random() < self.p: x, y, w, h = self.get_params(coords) # Fronto-parallel cuboid to remove mask = (x < coords[..., 0]) & (coords[..., 0] < x+w) & (y < coords[..., 1]) & (coords[..., 1] < y+h) coords[mask] = torch.zeros_like(coords[mask]) return coords
10,148
33.40339
115
py
MinkLoc3D
MinkLoc3D-master/datasets/samplers.py
# Author: Jacek Komorowski # Warsaw University of Technology import random import copy from torch.utils.data import Sampler from datasets.oxford import OxfordDataset class ListDict(object): def __init__(self, items=None): if items is not None: self.items = copy.deepcopy(items) self.item_to_position = {item: ndx for ndx, item in enumerate(items)} else: self.items = [] self.item_to_position = {} def add(self, item): if item in self.item_to_position: return self.items.append(item) self.item_to_position[item] = len(self.items)-1 def remove(self, item): position = self.item_to_position.pop(item) last_item = self.items.pop() if position != len(self.items): self.items[position] = last_item self.item_to_position[last_item] = position def choose_random(self): return random.choice(self.items) def __contains__(self, item): return item in self.item_to_position def __iter__(self): return iter(self.items) def __len__(self): return len(self.items) class BatchSampler(Sampler): # Sampler returning list of indices to form a mini-batch # Samples elements in groups consisting of k=2 similar elements (positives) # Batch has the following structure: item1_1, ..., item1_k, item2_1, ... item2_k, itemn_1, ..., itemn_k def __init__(self, dataset: OxfordDataset, batch_size: int, batch_size_limit: int = None, batch_expansion_rate: float = None, max_batches: int = None): if batch_expansion_rate is not None: assert batch_expansion_rate > 1., 'batch_expansion_rate must be greater than 1' assert batch_size <= batch_size_limit, 'batch_size_limit must be greater or equal to batch_size' self.batch_size = batch_size self.batch_size_limit = batch_size_limit self.batch_expansion_rate = batch_expansion_rate self.max_batches = max_batches self.dataset = dataset self.k = 2 # Number of positive examples per group must be 2 if self.batch_size < 2 * self.k: self.batch_size = 2 * self.k print('WARNING: Batch too small. Batch size increased to {}.'.format(self.batch_size)) self.batch_idx = [] # Index of elements in each batch (re-generated every epoch) self.elems_ndx = list(self.dataset.queries) # List of point cloud indexes def __iter__(self): # Re-generate batches every epoch self.generate_batches() for batch in self.batch_idx: yield batch def __len(self): return len(self.batch_idx) def expand_batch(self): if self.batch_expansion_rate is None: print('WARNING: batch_expansion_rate is None') return if self.batch_size >= self.batch_size_limit: return old_batch_size = self.batch_size self.batch_size = int(self.batch_size * self.batch_expansion_rate) self.batch_size = min(self.batch_size, self.batch_size_limit) print('=> Batch size increased from: {} to {}'.format(old_batch_size, self.batch_size)) def generate_batches(self): # Generate training/evaluation batches. # batch_idx holds indexes of elements in each batch as a list of lists self.batch_idx = [] unused_elements_ndx = ListDict(self.elems_ndx) current_batch = [] assert self.k == 2, 'sampler can sample only k=2 elements from the same class' while True: if len(current_batch) >= self.batch_size or len(unused_elements_ndx) == 0: # Flush out batch, when it has a desired size, or a smaller batch, when there's no more # elements to process if len(current_batch) >= 2*self.k: # Ensure there're at least two groups of similar elements, otherwise, it would not be possible # to find negative examples in the batch assert len(current_batch) % self.k == 0, 'Incorrect bach size: {}'.format(len(current_batch)) self.batch_idx.append(current_batch) current_batch = [] if (self.max_batches is not None) and (len(self.batch_idx) >= self.max_batches): break if len(unused_elements_ndx) == 0: break # Add k=2 similar elements to the batch selected_element = unused_elements_ndx.choose_random() unused_elements_ndx.remove(selected_element) positives = self.dataset.get_positives(selected_element) if len(positives) == 0: # Broken dataset element without any positives continue unused_positives = [e for e in positives if e in unused_elements_ndx] # If there're unused elements similar to selected_element, sample from them # otherwise sample from all similar elements if len(unused_positives) > 0: second_positive = random.choice(unused_positives) unused_elements_ndx.remove(second_positive) else: second_positive = random.choice(list(positives)) current_batch += [selected_element, second_positive] for batch in self.batch_idx: assert len(batch) % self.k == 0, 'Incorrect bach size: {}'.format(len(batch))
5,532
38.805755
114
py
MinkLoc3D
MinkLoc3D-master/layers/pooling.py
# Code taken from: https://github.com/filipradenovic/cnnimageretrieval-pytorch # and ported to MinkowskiEngine by Jacek Komorowski import torch import torch.nn as nn import MinkowskiEngine as ME class MAC(nn.Module): def __init__(self): super().__init__() self.f = ME.MinkowskiGlobalMaxPooling() def forward(self, x: ME.SparseTensor): x = self.f(x) return x.F # Return (batch_size, n_features) tensor class SPoC(nn.Module): def __init__(self): super().__init__() self.f = ME.MinkowskiGlobalAvgPooling() def forward(self, x: ME.SparseTensor): x = self.f(x) return x.F # Return (batch_size, n_features) tensor class GeM(nn.Module): def __init__(self, p=3, eps=1e-6): super(GeM, self).__init__() self.p = nn.Parameter(torch.ones(1) * p) self.eps = eps self.f = ME.MinkowskiGlobalAvgPooling() def forward(self, x: ME.SparseTensor): # This implicitly applies ReLU on x (clamps negative values) temp = ME.SparseTensor(x.F.clamp(min=self.eps).pow(self.p), coordinates=x.C) temp = self.f(temp) # Apply ME.MinkowskiGlobalAvgPooling return temp.F.pow(1./self.p) # Return (batch_size, n_features) tensor
1,280
30.243902
84
py
MinkLoc3D
MinkLoc3D-master/misc/utils.py
# Author: Jacek Komorowski # Warsaw University of Technology import os import configparser import time import numpy as np class ModelParams: def __init__(self, model_params_path): config = configparser.ConfigParser() config.read(model_params_path) params = config['MODEL'] self.model_params_path = model_params_path self.model = params.get('model') self.output_dim = params.getint('output_dim', 256) # Size of the final descriptor # Add gating as the last step if 'vlad' in self.model.lower(): self.cluster_size = params.getint('cluster_size', 64) # Size of NetVLAD cluster self.gating = params.getboolean('gating', True) # Use gating after the NetVlad ####################################################################### # Model dependent ####################################################################### if 'MinkFPN' in self.model: # Models using MinkowskiEngine self.mink_quantization_size = params.getfloat('mink_quantization_size') # Size of the local features from backbone network (only for MinkNet based models) # For PointNet-based models we always use 1024 intermediary features self.feature_size = params.getint('feature_size', 256) if 'planes' in params: self.planes = [int(e) for e in params['planes'].split(',')] else: self.planes = [32, 64, 64] if 'layers' in params: self.layers = [int(e) for e in params['layers'].split(',')] else: self.layers = [1, 1, 1] self.num_top_down = params.getint('num_top_down', 1) self.conv0_kernel_size = params.getint('conv0_kernel_size', 5) def print(self): print('Model parameters:') param_dict = vars(self) for e in param_dict: print('{}: {}'.format(e, param_dict[e])) print('') def get_datetime(): return time.strftime("%Y%m%d_%H%M") def xyz_from_depth(depth_image, depth_intrinsic, depth_scale=1000.): # Return X, Y, Z coordinates from a depth map. # This mimics OpenCV cv2.rgbd.depthTo3d() function fx = depth_intrinsic[0, 0] fy = depth_intrinsic[1, 1] cx = depth_intrinsic[0, 2] cy = depth_intrinsic[1, 2] # Construct (y, x) array with pixel coordinates y, x = np.meshgrid(range(depth_image.shape[0]), range(depth_image.shape[1]), sparse=False, indexing='ij') X = (x - cx) * depth_image / (fx * depth_scale) Y = (y - cy) * depth_image / (fy * depth_scale) xyz = np.stack([X, Y, depth_image / depth_scale], axis=2) xyz[depth_image == 0] = np.nan return xyz class MinkLocParams: """ Params for training MinkLoc models on Oxford dataset """ def __init__(self, params_path, model_params_path): """ Configuration files :param path: General configuration file :param model_params: Model-specific configuration """ assert os.path.exists(params_path), 'Cannot find configuration file: {}'.format(params_path) assert os.path.exists(model_params_path), 'Cannot find model-specific configuration file: {}'.format(model_params_path) self.params_path = params_path self.model_params_path = model_params_path self.model_params_path = model_params_path config = configparser.ConfigParser() config.read(self.params_path) params = config['DEFAULT'] self.num_points = params.getint('num_points', 4096) self.dataset_folder = params.get('dataset_folder') params = config['TRAIN'] self.num_workers = params.getint('num_workers', 0) self.batch_size = params.getint('batch_size', 128) # Set batch_expansion_th to turn on dynamic batch sizing # When number of non-zero triplets falls below batch_expansion_th, expand batch size self.batch_expansion_th = params.getfloat('batch_expansion_th', None) if self.batch_expansion_th is not None: assert 0. < self.batch_expansion_th < 1., 'batch_expansion_th must be between 0 and 1' self.batch_size_limit = params.getint('batch_size_limit', 256) # Batch size expansion rate self.batch_expansion_rate = params.getfloat('batch_expansion_rate', 1.5) assert self.batch_expansion_rate > 1., 'batch_expansion_rate must be greater than 1' else: self.batch_size_limit = self.batch_size self.batch_expansion_rate = None self.lr = params.getfloat('lr', 1e-3) self.scheduler = params.get('scheduler', 'MultiStepLR') if self.scheduler is not None: if self.scheduler == 'CosineAnnealingLR': self.min_lr = params.getfloat('min_lr') elif self.scheduler == 'MultiStepLR': scheduler_milestones = params.get('scheduler_milestones') self.scheduler_milestones = [int(e) for e in scheduler_milestones.split(',')] else: raise NotImplementedError('Unsupported LR scheduler: {}'.format(self.scheduler)) self.epochs = params.getint('epochs', 20) self.weight_decay = params.getfloat('weight_decay', None) self.normalize_embeddings = params.getboolean('normalize_embeddings', True) # Normalize embeddings during training and evaluation self.loss = params.get('loss') if 'Contrastive' in self.loss: self.pos_margin = params.getfloat('pos_margin', 0.2) self.neg_margin = params.getfloat('neg_margin', 0.65) elif 'Triplet' in self.loss: self.margin = params.getfloat('margin', 0.4) # Margin used in loss function else: raise 'Unsupported loss function: {}'.format(self.loss) self.aug_mode = params.getint('aug_mode', 1) # Augmentation mode (1 is default) self.train_file = params.get('train_file') self.val_file = params.get('val_file', None) self.eval_database_files = ['oxford_evaluation_database.pickle', 'business_evaluation_database.pickle', 'residential_evaluation_database.pickle', 'university_evaluation_database.pickle'] self.eval_query_files = ['oxford_evaluation_query.pickle', 'business_evaluation_query.pickle', 'residential_evaluation_query.pickle', 'university_evaluation_query.pickle'] assert len(self.eval_database_files) == len(self.eval_query_files) # Read model parameters self.model_params = ModelParams(self.model_params_path) self._check_params() def _check_params(self): assert os.path.exists(self.dataset_folder), 'Cannot access dataset: {}'.format(self.dataset_folder) def print(self): print('Parameters:') param_dict = vars(self) for e in param_dict: if e != 'model_params': print('{}: {}'.format(e, param_dict[e])) self.model_params.print() print('')
7,153
39.88
140
py
MinkLoc3D
MinkLoc3D-master/generating_queries/generate_test_sets.py
# PointNetVLAD datasets: based on Oxford RobotCar and Inhouse # Code adapted from PointNetVLAD repo: https://github.com/mikacuy/pointnetvlad import numpy as np import os import pandas as pd from sklearn.neighbors import KDTree import pickle import argparse # For training and test data splits X_WIDTH = 150 Y_WIDTH = 150 # For Oxford P1 = [5735712.768124, 620084.402381] P2 = [5735611.299219, 620540.270327] P3 = [5735237.358209, 620543.094379] P4 = [5734749.303802, 619932.693364] # For University Sector P5 = [363621.292362, 142864.19756] P6 = [364788.795462, 143125.746609] P7 = [363597.507711, 144011.414174] # For Residential Area P8 = [360895.486453, 144999.915143] P9 = [362357.024536, 144894.825301] P10 = [361368.907155, 145209.663042] P_DICT = {"oxford": [P1, P2, P3, P4], "university": [P5, P6, P7], "residential": [P8, P9, P10], "business": []} def check_in_test_set(northing, easting, points): in_test_set = False for point in points: if point[0] - X_WIDTH < northing < point[0] + X_WIDTH and point[1] - Y_WIDTH < easting < point[1] + Y_WIDTH: in_test_set = True break return in_test_set def output_to_file(output, base_path, filename): file_path = os.path.join(base_path, filename) with open(file_path, 'wb') as handle: pickle.dump(output, handle, protocol=pickle.HIGHEST_PROTOCOL) print("Done ", filename) def construct_query_and_database_sets(base_path, runs_folder, folders, pointcloud_fols, filename, p, output_name): database_trees = [] test_trees = [] for folder in folders: print(folder) df_database = pd.DataFrame(columns=['file', 'northing', 'easting']) df_test = pd.DataFrame(columns=['file', 'northing', 'easting']) df_locations = pd.read_csv(os.path.join(base_path, runs_folder, folder, filename), sep=',') # df_locations['timestamp']=runs_folder+folder+pointcloud_fols+df_locations['timestamp'].astype(str)+'.bin' # df_locations=df_locations.rename(columns={'timestamp':'file'}) for index, row in df_locations.iterrows(): # entire business district is in the test set if output_name == "business": df_test = df_test.append(row, ignore_index=True) elif check_in_test_set(row['northing'], row['easting'], p): df_test = df_test.append(row, ignore_index=True) df_database = df_database.append(row, ignore_index=True) database_tree = KDTree(df_database[['northing', 'easting']]) test_tree = KDTree(df_test[['northing', 'easting']]) database_trees.append(database_tree) test_trees.append(test_tree) test_sets = [] database_sets = [] for folder in folders: database = {} test = {} df_locations = pd.read_csv(os.path.join(base_path, runs_folder, folder, filename), sep=',') df_locations['timestamp'] = runs_folder + folder + pointcloud_fols + \ df_locations['timestamp'].astype(str) + '.bin' df_locations = df_locations.rename(columns={'timestamp': 'file'}) for index, row in df_locations.iterrows(): # entire business district is in the test set if output_name == "business": test[len(test.keys())] = {'query': row['file'], 'northing': row['northing'], 'easting': row['easting']} elif check_in_test_set(row['northing'], row['easting'], p): test[len(test.keys())] = {'query': row['file'], 'northing': row['northing'], 'easting': row['easting']} database[len(database.keys())] = {'query': row['file'], 'northing': row['northing'], 'easting': row['easting']} database_sets.append(database) test_sets.append(test) for i in range(len(database_sets)): tree = database_trees[i] for j in range(len(test_sets)): if i == j: continue for key in range(len(test_sets[j].keys())): coor = np.array([[test_sets[j][key]["northing"], test_sets[j][key]["easting"]]]) index = tree.query_radius(coor, r=25) # indices of the positive matches in database i of each query (key) in test set j test_sets[j][key][i] = index[0].tolist() output_to_file(database_sets, base_path, output_name + '_evaluation_database.pickle') output_to_file(test_sets, base_path, output_name + '_evaluation_query.pickle') if __name__ == '__main__': parser = argparse.ArgumentParser(description='Generate evaluation datasets') parser.add_argument('--dataset_root', type=str, required=True, help='Dataset root folder') args = parser.parse_args() print('Dataset root: {}'.format(args.dataset_root)) assert os.path.exists(args.dataset_root), f"Cannot access dataset root folder: {args.dataset_root}" base_path = args.dataset_root # For Oxford folders = [] runs_folder = "oxford/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) index_list = [5, 6, 7, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 22, 24, 31, 32, 33, 38, 39, 43, 44] print(len(index_list)) for index in index_list: folders.append(all_folders[index]) print(folders) construct_query_and_database_sets(base_path, runs_folder, folders, "/pointcloud_20m/", "pointcloud_locations_20m.csv", P_DICT["oxford"], "oxford") # For University Sector folders = [] runs_folder = "inhouse_datasets/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) uni_index = range(10, 15) for index in uni_index: folders.append(all_folders[index]) print(folders) construct_query_and_database_sets(base_path, runs_folder, folders, "/pointcloud_25m_25/", "pointcloud_centroids_25.csv", P_DICT["university"], "university") # For Residential Area folders = [] runs_folder = "inhouse_datasets/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) res_index = range(5, 10) for index in res_index: folders.append(all_folders[index]) print(folders) construct_query_and_database_sets(base_path, runs_folder, folders, "/pointcloud_25m_25/", "pointcloud_centroids_25.csv", P_DICT["residential"], "residential") # For Business District folders = [] runs_folder = "inhouse_datasets/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) bus_index = range(5) for index in bus_index: folders.append(all_folders[index]) print(folders) construct_query_and_database_sets(base_path, runs_folder, folders, "/pointcloud_25m_25/", "pointcloud_centroids_25.csv", P_DICT["business"], "business")
6,937
40.54491
119
py
MinkLoc3D
MinkLoc3D-master/generating_queries/generate_training_tuples_baseline.py
# PointNetVLAD datasets: based on Oxford RobotCar and Inhouse # Code adapted from PointNetVLAD repo: https://github.com/mikacuy/pointnetvlad import numpy as np import os import pandas as pd from sklearn.neighbors import KDTree import pickle import argparse import tqdm from datasets.oxford import TrainingTuple # Import test set boundaries from generating_queries.generate_test_sets import P1, P2, P3, P4, check_in_test_set # Test set boundaries P = [P1, P2, P3, P4] RUNS_FOLDER = "oxford/" FILENAME = "pointcloud_locations_20m_10overlap.csv" POINTCLOUD_FOLS = "/pointcloud_20m_10overlap/" def construct_query_dict(df_centroids, base_path, filename, ind_nn_r, ind_r_r=50): # ind_nn_r: threshold for positive examples # ind_r_r: threshold for negative examples # Baseline dataset parameters in the original PointNetVLAD code: ind_nn_r=10, ind_r=50 # Refined dataset parameters in the original PointNetVLAD code: ind_nn_r=12.5, ind_r=50 tree = KDTree(df_centroids[['northing', 'easting']]) ind_nn = tree.query_radius(df_centroids[['northing', 'easting']], r=ind_nn_r) ind_r = tree.query_radius(df_centroids[['northing', 'easting']], r=ind_r_r) queries = {} for anchor_ndx in range(len(ind_nn)): anchor_pos = np.array(df_centroids.iloc[anchor_ndx][['northing', 'easting']]) query = df_centroids.iloc[anchor_ndx]["file"] # Extract timestamp from the filename scan_filename = os.path.split(query)[1] assert os.path.splitext(scan_filename)[1] == '.bin', f"Expected .bin file: {scan_filename}" timestamp = int(os.path.splitext(scan_filename)[0]) positives = ind_nn[anchor_ndx] non_negatives = ind_r[anchor_ndx] positives = positives[positives != anchor_ndx] # Sort ascending order positives = np.sort(positives) non_negatives = np.sort(non_negatives) # Tuple(id: int, timestamp: int, rel_scan_filepath: str, positives: List[int], non_negatives: List[int]) queries[anchor_ndx] = TrainingTuple(id=anchor_ndx, timestamp=timestamp, rel_scan_filepath=query, positives=positives, non_negatives=non_negatives, position=anchor_pos) file_path = os.path.join(base_path, filename) with open(file_path, 'wb') as handle: pickle.dump(queries, handle, protocol=pickle.HIGHEST_PROTOCOL) print("Done ", filename) if __name__ == '__main__': parser = argparse.ArgumentParser(description='Generate Baseline training dataset') parser.add_argument('--dataset_root', type=str, required=True, help='Dataset root folder') args = parser.parse_args() print('Dataset root: {}'.format(args.dataset_root)) assert os.path.exists(args.dataset_root), f"Cannot access dataset root folder: {args.dataset_root}" base_path = args.dataset_root all_folders = sorted(os.listdir(os.path.join(base_path, RUNS_FOLDER))) folders = [] # All runs are used for training (both full and partial) index_list = range(len(all_folders) - 1) print("Number of runs: " + str(len(index_list))) for index in index_list: folders.append(all_folders[index]) print(folders) df_train = pd.DataFrame(columns=['file', 'northing', 'easting']) df_test = pd.DataFrame(columns=['file', 'northing', 'easting']) for folder in tqdm.tqdm(folders): df_locations = pd.read_csv(os.path.join(base_path, RUNS_FOLDER, folder, FILENAME), sep=',') df_locations['timestamp'] = RUNS_FOLDER + folder + POINTCLOUD_FOLS + df_locations['timestamp'].astype(str) + '.bin' df_locations = df_locations.rename(columns={'timestamp': 'file'}) for index, row in df_locations.iterrows(): if check_in_test_set(row['northing'], row['easting'], P): df_test = df_test.append(row, ignore_index=True) else: df_train = df_train.append(row, ignore_index=True) print("Number of training submaps: " + str(len(df_train['file']))) print("Number of non-disjoint test submaps: " + str(len(df_test['file']))) # ind_nn_r is a threshold for positive elements - 10 is in original PointNetVLAD code for refined dataset construct_query_dict(df_train, base_path, "training_queries_baseline.pickle", ind_nn_r=10) construct_query_dict(df_test, base_path, "test_queries_baseline.pickle", ind_nn_r=10)
4,386
43.765306
123
py
MinkLoc3D
MinkLoc3D-master/generating_queries/generate_training_tuples_refine.py
# PointNetVLAD datasets: based on Oxford RobotCar and Inhouse # Code adapted from PointNetVLAD repo: https://github.com/mikacuy/pointnetvlad import os import pandas as pd import argparse import tqdm # Import test set boundaries from generating_queries.generate_test_sets import P1, P2, P3, P4, P5, P6, P7, P8, P9, P10, check_in_test_set from generating_queries.generate_training_tuples_baseline import construct_query_dict # Test set boundaries P = [P1, P2, P3, P4, P5, P6, P7, P8, P9, P10] if __name__ == '__main__': parser = argparse.ArgumentParser(description='Generate Refined training dataset') parser.add_argument('--dataset_root', type=str, required=True, help='Dataset root folder') args = parser.parse_args() print('Dataset root: {}'.format(args.dataset_root)) assert os.path.exists(args.dataset_root), f"Cannot access dataset root folder: {args.dataset_root}" base_path = args.dataset_root runs_folder = "inhouse_datasets/" filename = "pointcloud_centroids_10.csv" pointcloud_fols = "/pointcloud_25m_10/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) folders = [] index_list = range(5, 15) for index in index_list: folders.append(all_folders[index]) print(folders) ####Initialize pandas DataFrame df_train = pd.DataFrame(columns=['file', 'northing', 'easting']) for folder in tqdm.tqdm(folders): df_locations = pd.read_csv(os.path.join(base_path, runs_folder, folder, filename), sep=',') df_locations['timestamp'] = runs_folder + folder + pointcloud_fols + df_locations['timestamp'].astype(str) + '.bin' df_locations = df_locations.rename(columns={'timestamp': 'file'}) for index, row in df_locations.iterrows(): if check_in_test_set(row['northing'], row['easting'], P): continue else: df_train = df_train.append(row, ignore_index=True) print(len(df_train['file'])) ##Combine with Oxford data runs_folder = "oxford/" filename = "pointcloud_locations_20m_10overlap.csv" pointcloud_fols = "/pointcloud_20m_10overlap/" all_folders = sorted(os.listdir(os.path.join(base_path, runs_folder))) folders = [] index_list = range(len(all_folders) - 1) for index in index_list: folders.append(all_folders[index]) print(folders) for folder in folders: df_locations = pd.read_csv(os.path.join(base_path, runs_folder, folder, filename), sep=',') df_locations['timestamp'] = runs_folder + folder + pointcloud_fols + df_locations['timestamp'].astype(str) + '.bin' df_locations = df_locations.rename(columns={'timestamp': 'file'}) for index, row in df_locations.iterrows(): if check_in_test_set(row['northing'], row['easting'], P): continue else: df_train = df_train.append(row, ignore_index=True) print("Number of training submaps: " + str(len(df_train['file']))) # ind_nn_r is a threshold for positive elements - 12.5 is in original PointNetVLAD code for refined dataset construct_query_dict(df_train, base_path, "training_queries_refine.pickle", ind_nn_r=12.5)
3,216
38.716049
123
py
RadarCommDataset
RadarCommDataset-main/load_dataset.py
import h5py dkeys = [] # labels of each entry in the h5 W = [] # dataset container with h5py.File('RadComDynamic.hdf5', 'r') as f: for key in f.keys(): dkeys.append(key) W.append(f[key][:]) f.close() # W can be now split into training, validation,and testing sets to run your ML algorithm on it
324
24
94
py
RadarCommDataset
RadarCommDataset-main/visualize.py
import h5py import sys import argparse import matplotlib.pyplot as plt def parse_args(): "Parse the command line arguments" parser = argparse.ArgumentParser() parser.add_argument("-num", type=int,default="0", help="Which sample to pick. 0 to 699") parser.add_argument("-snr", type=int,default="10", help="SNR: -20 to 18 in 2 step increments") parser.add_argument("-mod",default="pulsed", help="Modulation options: pulsed,fmcw,bpsk,amdsb,amssb,ask") parser.add_argument("-sig",default="Airborne-detection", help="Signal type options: Airborne-detection,Airborne-range,Air-Ground-MTI,Ground mapping,Radar-Altimeter,Satcom,AM radio,short-range") return parser.parse_args() def main(): args = parse_args() with h5py.File('RadComDynamic.hdf5', 'r') as f: key = args.mod,args.sig,args.snr,args.num waveform = f[str(key)][:] real = waveform[0:128] imag = waveform[128:] f.close() # Plot and visualize the selected sample plt.figure(figsize=[8, 6]) plt.plot(real,'-go') plt.plot(imag,'-bo') plt.title(str(key), fontsize=16) plt.show() if __name__ == "__main__": sys.exit(not main())
1,281
32.736842
160
py
pydeps
pydeps-master/setup.py
#!/usr/bin/env python # -*- coding: utf-8 -*- """pydeps - Python module dependency visualization """ # pragma: nocover import io import sys import setuptools from setuptools.command.test import test as TestCommand version='1.12.12' class PyTest(TestCommand): user_options = [('pytest-args=', 'a', "Arguments to pass to py.test")] def initialize_options(self): TestCommand.initialize_options(self) self.pytest_args = [] def finalize_options(self): TestCommand.finalize_options(self) self.test_args = [] self.test_suite = True def run_tests(self): # import here, cause outside the eggs aren't loaded import pytest errno = pytest.main(self.pytest_args) sys.exit(errno) setuptools.setup( name='pydeps', version=version, packages=setuptools.find_packages(exclude=['tests*']), install_requires=[ 'enum34; python_version < "3.4"', 'stdlib_list', ], long_description=io.open('README.rst', encoding='utf8').read(), entry_points={ 'console_scripts': [ 'pydeps = pydeps.pydeps:pydeps', ] }, url='https://github.com/thebjorn/pydeps', cmdclass={'test': PyTest}, license='BSD', author='bjorn', author_email='bp@datakortet.no', description='Display module dependencies', keywords='Python Module Dependency graphs', classifiers=[ 'Development Status :: 5 - Production/Stable', 'Intended Audience :: Developers', 'Natural Language :: English', 'License :: OSI Approved :: BSD License', 'Operating System :: OS Independent', 'Programming Language :: Python', 'Programming Language :: Python :: 2', 'Programming Language :: Python :: 2.7', 'Programming Language :: Python :: 3', 'Topic :: Software Development :: Libraries :: Python Modules', ] )
1,918
26.811594
74
py
pydeps
pydeps-master/tasks.py
# pragma: nocover from invoke import Collection, task from dktasklib import version from dktasklib import upversion from dktasklib import publish from dktasklib import docs from dktasklib.package import Package, package @task def freeze(ctx): "pip freeze, but without -e installed packages" ctx.run("pip list --exclude-editable --format freeze") @task def outdated(ctx): "list all outdated requirements" ctx.run("pip list --outdated") ns = Collection( 'pydeps', freeze, outdated, version, upversion, publish, docs, package ) ns.configure({ 'pkg': Package(), 'run': { 'echo': True } })
659
16.368421
58
py
pydeps
pydeps-master/pydeps/target.py
# -*- coding: utf-8 -*- """ Abstracting the target for pydeps to work on. """ from __future__ import print_function import json import os import re import shutil import sys import tempfile from contextlib import contextmanager import logging log = logging.getLogger(__name__) class Target(object): """The compilation target. """ def __init__(self, path): # log.debug("CURDIR: %s, path: %s, exists: %s", os.getcwd(), path, os.path.exists(path)) # print("Target::CURDIR: %s, path: %s, exists: %s" % (os.getcwd(), path, os.path.exists(path))) self.calling_fname = path self.calling_dir = os.getcwd() self.exists = os.path.exists(path) if self.exists: self.path = os.path.realpath(path) else: # pragma: nocover print("No such file or directory:", repr(path), file=sys.stderr) if os.path.exists(path + '.py'): print("..did you mean:", path + '.py', '?', file=sys.stderr) sys.exit(1) self.is_dir = os.path.isdir(self.path) self.is_module = self.is_dir and '__init__.py' in os.listdir(self.path) self.is_pysource = os.path.splitext(self.path)[1] in ('.py', '.pyc', '.pyo', '.pyw') self.fname = os.path.basename(self.path) if self.is_dir: self.dirname = self.fname self.modname = self.fname else: self.dirname = os.path.dirname(self.path) self.modname = os.path.splitext(self.fname)[0] if self.is_pysource: # we will work directly on the file (in-situ) self.workdir = os.path.dirname(self.path) else: self.workdir = os.path.realpath(tempfile.mkdtemp()) self.syspath_dir = self.get_package_root() # split path such that syspath_dir + relpath == path self.relpath = self.path[len(self.syspath_dir):].lstrip(os.path.sep) if self.is_dir: self.modpath = self.relpath.replace(os.path.sep, '.') else: self.modpath = os.path.splitext(self.relpath)[0].replace(os.path.sep, '.') self.package_root = os.path.join( self.syspath_dir, self._path_parts(self.relpath)[0] ) @contextmanager def chdir_work(self): try: os.chdir(self.workdir) sys.path.insert(0, self.syspath_dir) yield finally: os.chdir(self.calling_dir) if sys.path[0] == self.syspath_dir: sys.path = sys.path[1:] self.close() def get_package_root(self): for d in self.get_parents(): if '__init__.py' not in os.listdir(d): return d raise Exception( "do you have an __init__.py file at the " "root of the drive..?") # pragma: nocover def get_parents(self): def _parent_iter(): parts = self._path_parts(self.path) for i in range(1, len(parts)): yield os.path.join(*parts[:-i]) return list(_parent_iter()) def _path_parts(self, pth): """Return a list of all directories in the path ``pth``. """ res = re.split(r"[\\/]", pth) if res and os.path.splitdrive(res[0]) == (res[0], ''): res[0] += os.path.sep return res def __del__(self): self.close() def close(self): """Clean up after ourselves. """ try: # make sure we don't delete the user's source file if we're working # on it in-situ. if not self.is_pysource and hasattr(self, 'workdir'): shutil.rmtree(self.workdir) except OSError: pass def __repr__(self): # pragma: nocover return json.dumps( {k: v for k, v in self.__dict__.items() if not k.startswith('_')}, indent=4, sort_keys=True )
3,933
32.058824
103
py
pydeps
pydeps-master/pydeps/dot.py
# -*- coding: utf-8 -*- """ Graphviz interface. """ import os import platform import sys from subprocess import Popen import subprocess import shlex from . import cli win32 = sys.platform == 'win32' def is_unicode(s): # pragma: nocover """Test unicode with py3 support. """ try: return isinstance(s, unicode) except NameError: return False def to_bytes(s): # pragma: nocover """Convert an item into bytes. """ if isinstance(s, bytes): return s if isinstance(s, str) or is_unicode(s): return s.encode("utf-8") try: return unicode(s).encode("utf-8") except NameError: return str(s).encode("utf-8") def cmd2args(cmd): """Prepare a command line for execution by Popen. """ if isinstance(cmd, str): return cmd if win32 else shlex.split(cmd) return cmd def pipe(cmd, txt): """Pipe `txt` into the command `cmd` and return the output. """ return Popen( cmd2args(cmd), stdout=subprocess.PIPE, stdin=subprocess.PIPE, shell=win32 ).communicate(txt)[0] def dot(src, **kw): """Execute the dot command to create an svg output. """ cmd = "dot -Gstart=1 -T%s" % kw.pop('T', 'svg') for k, v in list(kw.items()): if v is True: cmd += " -%s" % k else: cmd += " -%s%s" % (k, v) return pipe(cmd, to_bytes(src)) def call_graphviz_dot(src, fmt): """Call dot command, and provide helpful error message if we cannot find it. """ try: svg = dot(src, T=fmt) except OSError as e: # pragma: nocover if e.errno == 2: cli.error(""" cannot find 'dot' pydeps calls dot (from graphviz) to create svg diagrams, please make sure that the dot executable is available on your path. """) raise return svg def in_wsl(): """Are we running under wsl? """ return 'microsoft-standard' in platform.uname().release def display_svg(kw, fname): # pragma: nocover """Try to display the svg file on this platform. Note that this is also used to display PNG files, despite the name. """ display = kw['display'] if not display: display = os.getenv('PYDEPS_DISPLAY', os.getenv('BROWSER', None)) if not display: cli.verbose("Displaying:", fname) if sys.platform == 'win32': os.startfile(fname) else: if sys.platform == "darwin": display = "open" elif in_wsl(): # this is still borked... display = "/usr/bin/wslview" else: display = "xdg-open" subprocess.check_call([display, fname]) else: cli.verbose(display + " " + fname) subprocess.check_call([display, fname])
2,906
23.024793
74
py
pydeps
pydeps-master/pydeps/__main__.py
from .pydeps import pydeps pydeps()
36
11.333333
26
py
pydeps
pydeps-master/pydeps/pystdlib.py
# -*- coding: utf-8 -*- import sys import stdlib_list import warnings def pystdlib(): """Return a set of all module-names in the Python standard library. """ if sys.version_info[:2] >= (3, 10): # Python 3.10 has this functionality built-in. return list(sys.stdlib_module_names | set(sys.builtin_module_names)) curver = '.'.join(str(x) for x in sys.version_info[:2]) if curver not in stdlib_list.short_versions: # if stdlib_list doesn't know about our version, then use the last # version that stdlib_list knows about (not perfect, but it will # allow downstream packages to test on pre-release versions of # Python - which are difficult for stdlib_list to support). warnings.warn( ("stdlib_list does't support Python %s yet, " "pydeps will use symbols from %s for now.") % ( curver, stdlib_list.long_versions[-1] ) ) curver = stdlib_list.long_versions[-1] return (set(stdlib_list.stdlib_list(curver)) | { '_LWPCookieJar', '_MozillaCookieJar', '_abcoll', 'email._parseaddr', 'email.base64mime', 'email.feedparser', 'email.quoprimime', 'encodings', 'genericpath', 'ntpath', 'nturl2path', 'os2emxpath', 'posixpath', 'sre_compile', 'sre_parse', 'unittest.case', 'unittest.loader', 'unittest.main', 'unittest.result', 'unittest.runner', 'unittest.signals', 'unittest.suite', 'unittest.util', '_threading_local', 'sre_constants', 'strop', 'repr', 'opcode', 'nt', 'encodings.aliases', '_bisect', '_codecs', '_collections', '_functools', '_hashlib', '_heapq', '_io', '_locale', '_LWPCookieJar', '_md5', '_MozillaCookieJar', '_random', '_sha', '_sha256', '_sha512', '_socket', '_sre', '_ssl', '_struct', '_subprocess', '_threading_local', '_warnings', '_weakref', '_weakrefset', '_winreg' }) - {'__main__'}
1,969
43.772727
76
py
pydeps
pydeps-master/pydeps/depgraph2dot.py
# Based on original code, Copyright 2004 Toby Dickenson, # with changes 2014 (c) Bjorn Pettersen # # Permission is hereby granted, free of charge, to any person obtaining # a copy of this software and associated documentation files (the # "Software"), to deal in the Software without restriction, including # without limitation the rights to use, copy, modify, merge, publish, # distribute, sublicense, and/or sell copies of the Software, and to # permit persons to whom the Software is furnished to do so, subject # to the following conditions: # # The above copyright notice and this permission notice shall be included # in all copies or substantial portions of the Software. # # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. # IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY # CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE # SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. from .render_context import RenderBuffer from . import colors class PyDepGraphDot(object): def __init__(self, **kw): self.kw = kw def render(self, depgraph, ctx): with ctx.graph(): visited = set() drawn = set() for aname, bname in depgraph.cyclerelations: try: a = depgraph.sources[aname] b = depgraph.sources[bname] except KeyError: continue drawn.add((bname, aname)) ctx.write_rule( aname, bname, weight=depgraph.proximity_metric(a, b), minlen=depgraph.dissimilarity_metric(a, b), ) for a, b in sorted(depgraph): # b imports a aname = a.name bname = b.name if (bname, aname) in drawn: continue drawn.add((bname, aname)) ctx.write_rule( aname, bname, weight=depgraph.proximity_metric(a, b), minlen=depgraph.dissimilarity_metric(a, b)) visited.add(a) visited.add(b) space = colors.ColorSpace(visited) for src in sorted(visited): bg, fg = depgraph.get_colors(src, space) kwargs = {} if src.name in depgraph.cyclenodes: kwargs['shape'] = 'octagon' ctx.write_node( src.name, label=src.get_label(splitlength=14, rmprefix=self.kw.get('rmprefix')), fillcolor=colors.rgb2css(bg), fontcolor=colors.rgb2css(fg), **kwargs ) return ctx.text() class CycleGraphDot(object): def __init__(self, **kw): self.kw = kw def render(self, depgraph, ctx): with ctx.graph(concentrate=False): visited = set() drawn = set() for aname, bname in depgraph.cyclerelations: try: a = depgraph.sources[aname] b = depgraph.sources[bname] except KeyError: continue drawn.add((bname, aname)) ctx.write_rule( bname, aname, weight=depgraph.proximity_metric(a, b), minlen=depgraph.dissimilarity_metric(a, b), ) visited.add(a) visited.add(b) space = colors.ColorSpace(visited) for src in visited: bg, fg = depgraph.get_colors(src, space) kwargs = {} if src.name in depgraph.cyclenodes: kwargs['shape'] = 'octagon' ctx.write_node( src.name, label=src.label, fillcolor=colors.rgb2css(bg), fontcolor=colors.rgb2css(fg), **kwargs ) return ctx.text() def dep2dot(target, depgraph, **kw): dotter = PyDepGraphDot(**kw) ctx = RenderBuffer(target, **kw) return dotter.render(depgraph, ctx) def cycles2dot(target, depgraph, **kw): dotter = CycleGraphDot(**kw) ctx = RenderBuffer(target, **kw) return dotter.render(depgraph, ctx)
4,638
33.110294
74
py
pydeps
pydeps-master/pydeps/arguments.py
from __future__ import print_function, unicode_literals from io import StringIO import textwrap import json import argparse # from devtools import debug from .configs import Config, typefns, identity DEFAULT_NONE = '____' class Argument(object): def __init__(self, *flags, **args): if 'choices' in args and args['choices'] is None: del args['choices'] if 'container' in args: del args['container'] self._args = args self._flags = flags # self.__dict__.update(args) def __json__(self): return self.__dict__ def typename(self): if 'kind' in self._args: return self._args['kind'] if self._args.get('action') in {'store_true', 'store_false'}: return 'BOOL' tp = self._args.get('type') if tp is not None: return tp.__name__.upper() v = self._args.get('default') if v is None: return DEFAULT_NONE return v.__class__.__name__.upper() def typefn(self): return typefns.get(self.typename(), identity) def pytype(self): argname = self.argname() if argname == 'fname': return 'str' if 'choices' in self._args: return 'Literal[%s]' % ', '.join(repr(x) for x in self._args['choices']) if 'type' in self._args: res = self._args['type'] elif self._args.get('action') in {'store_true', 'store_false'}: res = bool elif self._args.get('kind', 'unknown') is None: return 'None' elif self._args.get('kind', '').startswith("FNAME"): res = str else: res = self._args.get('default').__class__ if res is None: return 'None' return res.__name__ def argname(self): if 'dest' in self._args: return self._args['dest'] return self._flags[0].lstrip('-').replace('-', '_') # return self._flags[-1].lstrip('-').replace('-', '_') def help(self): if 'help' in self._args: return self._args['help'] return '' def default(self): if 'default' in self._args: return self._args['default'] if self._args.get('action') == 'store_true': return False if self._args.get('action') == 'store_false': return True return DEFAULT_NONE def add_to_parser(self, parser): args = self._args args.pop('default', None) # remove default value args.pop('kind', None) # remove our attributes if args.get('action') in {'store_true', 'store_false'}: args['default'] = None # debug("ADD_TO_PARSER:", self._flags, args) parser.add_argument(*self._flags, **args) class Namespace(object): def __init__(self, ns): self.ns = ns def __repr__(self): return json.dumps(vars(self.ns), indent=4, sort_keys=True) def items(self): return list(dict(vars(self.ns)).items()) def __getitem__(self, key): return getattr(self.ns, key) def __setitem__(self, key, val): setattr(self.ns, key, val) def __delitem__(self, key): delattr(self.ns, key) def __getattr__(self, key): return getattr(self.ns, key) class Arguments(object): def __init__(self, config_files=None, debug=False, *posargs, parents=None, **kwargs): if config_files is None: config_files = [] # passthrough to argparse self.posargs = posargs self.kwargs = kwargs # print("CONFIG_FILES:", config_files) self.debug = debug self.arglist = [] self.args = {} self.config_files = config_files self.argtypes = {} self.defaults = {} self.parents = parents def load_config_files(self): for filename in self.config_files: if filename == '.pydeps' or filename.endswith('.pydeps'): self.load_pydeps_config(filename) def write_default_config(self): """Utility function to create .configs.Config # XXX: a more general utility to create configs from argparse # would be nice... """ fp = StringIO() # fp = sys.stdout print("class Config(object):", file=fp) arglist = self.arglist if self.parents: parent_actions = [] for parent in self.parents: for action in parent._actions: # if isinstance(action, argparse._StoreAction): parent_actions.append(Argument( *action.option_strings, **action.__dict__ )) arglist = parent_actions + arglist for arg in arglist: debug(arg._args) default = arg.default() if default == DEFAULT_NONE: default = None elif isinstance(default, str): default = repr(default) help = textwrap.wrap(arg.help(), 80 - 7) # 7 = len(" #: ") for helpline in help: print("", file=fp) print(" #: {help}".format(help=helpline), end="", file=fp) print("", file=fp) typename = arg.pytype() if typename == 'list' and default is None: typename = 'Optional[List[str]]' elif default is None: typename = f'Optional[{typename}]' elif default == [] and typename == 'list': typename = 'List[str]' # add py3 type annotations # print(f" {arg.argname()}: {typename} = {default}", file=fp) # witout type annotations print(" {argname} = {default}".format(argname=arg.argname(), default=default), file=fp) print("", file=fp) print(" def set_field(self, field, value):", file=fp) for arg in self.arglist: print(" if field == '{argname}':".format(argname=arg.argname()), file=fp) print(" self.{argname} = {typefn}(value)".format(argname=arg.argname(), typefn=arg.typefn().__name__), file=fp) print("", file=fp) # debug(res) res = fp.getvalue() print(res) return res def parse_args(self, argv): self.kwargs['parents'] = self.parents p = argparse.ArgumentParser(*self.posargs, **self.kwargs) for arg in self.arglist: arg.add_to_parser(p) args = Namespace(p.parse_args(argv)) config = Config.load(self.config_files) config.update({k : v for k, v in args.items() if v is not None}) # print('---- yaml ---------------') # print(config.write_yaml()) # print('---- json ---------------') # print(config.write_json()) # print('---- ini ---------------') # print(config.write_ini()) # print('---- toml ---------------') # print(config.write_toml()) # print('-------------------') return config def add(self, *flags, **kwargs): # debug(flags, kwargs) arg = Argument(*flags, **kwargs) self.arglist.append(arg) argname = arg.argname() self.args[argname] = arg self.argtypes[argname] = arg.typename() self.defaults[argname] = arg.default() def __repr__(self): return json.dumps(dict( types=self.argtypes, defaults=self.defaults, ), indent=4, sort_keys=True)
7,599
31.067511
134
py
pydeps
pydeps-master/pydeps/mfimp.py
""" Python's modulefinder._find_module has a bug that breaks a number of popular packages. This is a copy of the standard lib's imp._find_module which sort of does the right thing, ie. ignores namespace packages instead of crashing. This is vendorized/copied here to prevent the warning error that the regular imp module causes. """ from _imp import is_builtin, is_frozen from importlib import machinery import sys import os import tokenize PY_SOURCE = 1 PY_COMPILED = 2 C_EXTENSION = 3 PKG_DIRECTORY = 5 C_BUILTIN = 6 PY_FROZEN = 7 def _get_suffixes(): extensions = [(s, 'rb', C_EXTENSION) for s in machinery.EXTENSION_SUFFIXES] source = [(s, 'r', PY_SOURCE) for s in machinery.SOURCE_SUFFIXES] bytecode = [(s, 'rb', PY_COMPILED) for s in machinery.BYTECODE_SUFFIXES] return extensions + source + bytecode def find_module(name, path=None): if not isinstance(name, str): raise TypeError("'name' must be a str, not {}".format(type(name))) elif not isinstance(path, (type(None), list)): # Backwards-compatibility raise RuntimeError("'path' must be None or a list, " "not {}".format(type(path))) if path is None: if is_builtin(name): return None, None, ('', '', C_BUILTIN) elif is_frozen(name): return None, None, ('', '', PY_FROZEN) else: path = sys.path for entry in path: package_directory = os.path.join(entry, name) for suffix in ['.py', machinery.BYTECODE_SUFFIXES[0]]: package_file_name = '__init__' + suffix file_path = os.path.join(package_directory, package_file_name) if os.path.isfile(file_path): return None, package_directory, ('', '', PKG_DIRECTORY) for suffix, mode, type_ in _get_suffixes(): file_name = name + suffix file_path = os.path.join(entry, file_name) if os.path.isfile(file_path): break else: continue break # Break out of outer loop when breaking out of inner loop. else: raise ImportError('No module named {!r}'.format(name), name=name) encoding = None if 'b' not in mode: with open(file_path, 'rb') as file: encoding = tokenize.detect_encoding(file.readline)[0] file = open(file_path, mode, encoding=encoding) return file, file_path, (suffix, mode, type_)
2,445
31.613333
79
py
pydeps
pydeps-master/pydeps/render_context.py
# -*- coding: utf-8 -*- from collections import defaultdict from io import StringIO from contextlib import contextmanager import textwrap import enum def to_unicode(s): try: return unicode(s) except NameError: return s class Rankdir(enum.Enum): BOTTOM_TOP = 'BT' TOP_BOTTOM = 'TB' LEFT_RIGHT = 'LR' RIGHT_LEFT = 'RL' def reverse(self): return Rankdir(self.value[::-1]) class RenderContext(object): def __init__(self, out=None, reverse=False, rankdir=Rankdir.TOP_BOTTOM): self.out = out self.fp = StringIO() self.fillcolor = '#ffffff' self.fontcolor = '#000000' self.name = None self.concentrate = None self.compound = None self.width = 0.75 self.reverse = reverse self.rankdir = rankdir @contextmanager def graph(self, **kw): """Set up a graphviz graph context. """ self.name = kw.get('name', 'G') self.fillcolor = kw.get('fillcolor', '#ffffff') self.fontcolor = kw.get('fontcolor', '#000000') if kw.get('concentrate', True): self.concentrate = 'concentrate = true;' else: self.concentrate = '' self.compound = 'compound = true;' if kw.get('compound') else '' self.dedent(""" digraph {self.name} {{ {self.concentrate} {self.compound} rankdir = {self.rankdir.value}; node [style=filled,fillcolor="{self.fillcolor}",fontcolor="{self.fontcolor}",fontname=Helvetica,fontsize=10]; """.format(self=self)) yield self.writeln('}') def text(self): """Get value of output stream (StringIO). """ if self.out: self.out.close() # pragma: nocover return self.fp.getvalue() def write_rule(self, a, b, **attrs): """a -> b [a1=x,a2=y]; """ if self.reverse: a, b = b, a with self.rule(): self.write('%s -> %s' % (self._nodename(a), self._nodename(b))) # remove default values from output self._delattr(attrs, 'weight', 1) self._delattr(attrs, 'minlen', 1) self._delattr(attrs, 'len', 1) self.write_attributes(attrs) def write_node(self, a, **attrs): """a [a1=x,a2=y]; """ with self.rule(): nodename = self._nodename(a) self.write(nodename) # remove default values from output self._delattr(attrs, 'label', nodename) self._delattr(attrs, 'fillcolor', self.fillcolor) self._delattr(attrs, 'fontcolor', self.fontcolor) self._delattr(attrs, 'width', self.width) self.write_attributes(attrs) # -- end of external/public interface -- def write(self, txt): """Write ``txt`` to file and output stream (StringIO). """ self.fp.write(to_unicode(txt)) if self.out: self.out.write(txt) # pragma: nocover def writeln(self, txt): """Write ``txt`` and add newline. """ self.write(txt + '\n') def dedent(self, txt): """Write ``txt`` dedented. """ self.write(textwrap.dedent(txt)) def write_attributes(self, attrs): """Write comma separated attribute values (if exists). """ if attrs: self.write( ' [' + ','.join('%s="%s"' % kv for kv in sorted(attrs.items())) + ']' ) else: # pragma: nocover pass def _nodename(self, x): """Return a valid node name. """ return x.replace('.', '_') def _delattr(self, attr, key, value): if attr.get(key) == value: del attr[key] @contextmanager def rule(self): """Write indented rule. """ self.write(' ') yield self.writeln(';') class RenderBuffer(object): def __init__(self, target, reverse=False, rankdir=Rankdir.TOP_BOTTOM, cluster=False, min_cluster_size=0, max_cluster_size=1, keep_target_cluster=False, collapse_target_cluster=False, **kw): self.target = target self.nodes = [] self.clusters = defaultdict(list) self.rules = {} self.reverse = reverse self.rankdir = Rankdir(rankdir) if self.reverse: self.rankdir = self.rankdir.reverse() self.cluster = cluster self.min_cluster_size = min_cluster_size self.max_cluster_size = max_cluster_size self.graph_attrs = {} self.keep_target_cluster = keep_target_cluster self.collapse_target_cluster = collapse_target_cluster def _nodecolor(self, n): for node, attrs in self.nodes: if node == n: return attrs['fillcolor'] return '#000000' def cluster_stats(self): maxnodes = max(len(v) for v in self.clusters.values()) minnodes = min(len(v) for v in self.clusters.values()) return minnodes, maxnodes def _remove_small_clusters(self): # remove clusters that are too small target_cluster = self._target_clusterid() _remove = [] for clusterid, nodes in sorted(self.clusters.items()): if clusterid == target_cluster: # Target cluster must always be there, don't remove it even if it's small. We can get here # when --collapse-target-cluster flag is used. continue if len(nodes) < self.min_cluster_size: # print("REMOVING:CLUSTER:", clusterid, nodes) self.nodes += nodes _remove.append(clusterid) for _r in _remove: del self.clusters[_r] def _collapse_cluster(self, clusterid, nodes): """Add a single cluster node (with a label listing contents?) and change all rules to reference this node instead. """ first_node, first_attrs = nodes[0] first_attrs['shape'] = 'folder' first_attrs['label'] = clusterid self.nodes.append((clusterid, first_attrs)) for node, attrs in nodes: # for each node in this cluster # check all rules for in/out relations rules = list(self.rules.items()) self.rules = {} for (a, b), rule_attrs in rules: # orig = (a, b) if a == node: a = clusterid if b == node: b = clusterid # if orig != (a, b): # print("CHANGED[{}|{}]: {} TO {}".format(clusterid, node, orig, (a, b))) self.rules[(a, b)] = rule_attrs del self.clusters[clusterid] def triage_clusters(self): target_cluster = self._target_clusterid() if not self.collapse_target_cluster and not self.keep_target_cluster: # don't put nodes from the target into a cluster self.nodes += self.clusters[target_cluster] del self.clusters[target_cluster] self._remove_small_clusters() # collapse target cluster if requested if self.collapse_target_cluster: self._collapse_cluster(target_cluster, self.clusters[target_cluster]) # collapse clusters that are too big for clusterid, nodes in sorted(self.clusters.items()): if len(nodes) > self.max_cluster_size and clusterid != target_cluster: self._collapse_cluster(clusterid, nodes) def text(self): ctx = RenderContext(reverse=self.reverse, rankdir=self.rankdir) if self.cluster: self.triage_clusters() if self.clusters: # are there any clusters left after triage? self.graph_attrs['compound'] = True self.graph_attrs['concentrate'] = False with ctx.graph(**self.graph_attrs): clusters = set() for clusterid, nodes in sorted(self.clusters.items()): clusters.add(clusterid) ctx.writeln('subgraph cluster_%s {' % clusterid) ctx.writeln(' label = %s;' % clusterid) for n, attrs in nodes: ctx.write_node(n, **attrs) ctx.writeln('}') # non-clustered nodes for n, attrs in self.nodes: ctx.write_node(n, **attrs) intercluster = set() for (a, b), attrs in sorted(self.rules.items()): if a == b: continue cida = self._clusterid(a) cidb = self._clusterid(b) if cida == cidb: if self.reverse: attrs['fillcolor'] = self._nodecolor(b) else: attrs['fillcolor'] = self._nodecolor(a) ctx.write_rule(a, b, **attrs) elif cida in clusters and cidb in clusters: if (cida, cidb) not in intercluster: intercluster.add((cida, cidb)) if self.reverse: attrs['lhead'] = 'cluster_' + cida attrs['ltail'] = 'cluster_' + cidb attrs['fillcolor'] = self._nodecolor(b) else: attrs['ltail'] = 'cluster_' + cida attrs['lhead'] = 'cluster_' + cidb attrs['fillcolor'] = self._nodecolor(a) ctx.write_rule(a, b, **attrs) else: if cida in clusters: if self.reverse: attrs['lhead'] = 'cluster_' + cida else: attrs['ltail'] = 'cluster_' + cida if cidb in clusters: if self.reverse: attrs['ltail'] = 'cluster_' + cidb else: attrs['lhead'] = 'cluster_' + cidb if self.reverse: attrs['fillcolor'] = self._nodecolor(b) else: attrs['fillcolor'] = self._nodecolor(a) ctx.write_rule(a, b, **attrs) return ctx.text() @contextmanager def graph(self, **kw): self.graph_attrs.update(kw) yield def _clusterid(self, n): return n.split('.')[0] def write_node(self, n, **attrs): clusterid = self._clusterid(n) if self.cluster: self.clusters[clusterid].append((n, attrs)) else: self.nodes.append((n, attrs)) def write_rule(self, a, b, **attrs): self.rules[(a, b)] = attrs def _target_clusterid(self): return self._clusterid(self.target.fname)
11,072
33.388199
125
py
pydeps
pydeps-master/pydeps/dummymodule.py
# -*- coding: utf-8 -*- from __future__ import print_function import os import textwrap import logging from . import cli log = logging.getLogger(__name__) def is_module(directory): """A directory is a module if it contains an ``__init__.py`` file. """ return os.path.isdir(directory) and '__init__.py' in os.listdir(directory) def is_pysource(fname): """A file name is a python source file iff it ends with '.py(w?)' and doesn't start with a dot. """ return not fname.startswith('.') and fname.endswith(('.py', '.pyw')) def fname2modname(fname, package_root): subpath = os.path.splitext(fname)[0][len(package_root):] modname = subpath.lstrip(os.path.sep).replace(os.path.sep, '.') return modname def python_sources_below(directory, package=True): for root, dirs, files in os.walk(directory): if package and '__init__.py' not in files: continue dotdirs = [d for d in dirs if d.startswith('.')] for d in dotdirs: dirs.remove(d) if 'migrations' in dirs: dirs.remove('migrations') for fname in files: if is_pysource(fname): # and fname not in args['exclude']: if fname == '__init__.py': yield os.path.abspath(root) else: yield os.path.abspath(os.path.join(root, fname)) class DummyModule(object): """We create a file that imports the module to be investigated. """ def __init__(self, target, **args): self._legal_mnames = {} self.target = target self.fname = '_dummy_' + target.modpath.replace('.', '_') + '.py' self.absname = os.path.join(target.workdir, self.fname) if target.is_module: cli.verbose(1, "target is a PACKAGE") with open(self.fname, 'w') as fp: for fname in python_sources_below(target.package_root): modname = fname2modname(fname, target.syspath_dir) self.print_import(fp, modname) elif target.is_dir: # FIXME?: not sure what the intended semantics was here, as it is # this will almost certainly not do the right thing... cli.verbose(1, "target is a DIRECTORY") log.debug('curdir: %r', os.getcwd()) log.debug('fname: %r', self.fname) log.debug('target.dirname: %r', target.dirname) with open(self.fname, 'w') as fp: dirname = os.path.abspath(os.path.join(target.calling_dir, target.calling_fname)) for fname in os.listdir(dirname): fname = os.path.join(dirname, fname) log.debug("fname: %r", fname) if is_pysource(fname): self.print_import(fp, fname2modname(fname, '')) elif is_module(fname): log.debug("fname is a module: %r", fname) for fnamea in python_sources_below(fname): modname = fname2modname(fnamea, target.syspath_dir) self.print_import(fp, modname) else: assert target.is_pysource cli.verbose(1, "target is a FILE") # if working on a single file, we don't need to create a dummy # module, this also avoids problems with file names that are # not importable (e.g. `foo.bar.py) # self.fname = target.calling_fname self.fname = target.fname self.absname = target.package_root # with open(self.fname, 'w') as fp: # self.print_import(fp, target.modpath) log.debug("dummy-filename: %r (%s)[module=%s, dir=%s, file=%s]", self.fname, self.absname, target.is_module, target.is_dir, target.is_pysource) def text(self): """Return the content of the dummy module. """ log.debug("Getting text from %r", self.fname) # log.debug("sys.path: %r", sys.path) with open(self.fname) as fp: return fp.read() def legal_module_name(self, name): """Legal module names are dotted strings where each part is a valid Python identifier. (and not a keyword, and support unicode identifiers in Python3, ..) """ if name in self._legal_mnames: return self._legal_mnames[name] for part in name.split('.'): try: exec("%s = 42" % part, {}, {}) except: # pragma: nocover # noqa self._legal_mnames[name] = False return False self._legal_mnames[name] = True return True def print_header(self, fp): # pragma: nocover # we're not executing the file in fp, so really not necessary to # catch import errors print(textwrap.dedent(""" import sys import traceback """), file=fp) def print_import(self, fp, module): if not self.legal_module_name(module): log.warning("SKIPPING ILLEGAL MODULE_NAME: %s", module) return mparts = module.rsplit('.', 1) # we're not executing the file in fp, so really not necessary to # catch import errors if len(mparts) == 1: print(textwrap.dedent("""\ import {module} """).format(module=module), file=fp) else: print(textwrap.dedent("""\ from {prefix} import {mname} """).format(prefix=mparts[0], mname=mparts[1]), file=fp)
5,638
36.344371
97
py
pydeps
pydeps-master/pydeps/cli.py
# -*- coding: utf-8 -*- """ command line interface (cli) code. """ # pylint: disable=line-too-long from __future__ import print_function import argparse from pydeps.configs import Config from .arguments import Arguments # import json # from .pycompat import configparser import logging import os import sys import subprocess import textwrap from . import __version__ def error(*args, **kwargs): # pragma: nocover """Print an error message and exit. """ kwargs['file'] = sys.stderr print("\n\tERROR:", *args, **kwargs) if args and args[0].startswith("[Errno 2] No such file or directory"): print("\t(Did you forget to include an __init__.py?)") sys.exit(1) #: the (will become) verbose function verbose = None def _not_verbose(*args, **kwargs): # pragma: nocover pass verbose = _not_verbose def _mkverbose(level): def _verbose(n, *args, **kwargs): if not isinstance(n, int): # we're only interested in small integers # this allows the simpler usage cli.verbose(msg) args = (n,) + args n = 1 if 0 < level <= n: print(*args, **kwargs) return _verbose def _find_current_package(): cwd = os.getcwd() while 'setup.py' not in os.listdir(cwd) and cwd != os.path.dirname(cwd): cwd = os.path.dirname(cwd) if 'setup.py' not in os.listdir(cwd): raise Exception("--find-package didn't find setup.py in current, or any parent, directory") os.chdir(cwd) package_name = subprocess.check_output("python setup.py --name", shell=True).decode('u8').strip() return package_name def base_argparser(argv=()): """Initial parser that can set values for the rest of the parsing process. """ global verbose verbose = _not_verbose _p = argparse.ArgumentParser(add_help=False) _p.add_argument('--debug', action='store_true', dest='debug', help="turn on all the show and verbose options (mainly for debugging pydeps itself)") _p.add_argument('--config', help="specify config file", metavar="FILE") _p.add_argument('--no-config', help="disable processing of config files", action='store_true') _p.add_argument('--version', action='store_true', help='print pydeps version') _p.add_argument('-L', '--log', help=textwrap.dedent(''' set log-level to one of CRITICAL, ERROR, WARNING, INFO, DEBUG, NOTSET. ''')) _p.add_argument('--find-package', action='store_true', help="tries to automatically find the name of the current package.") _args, argv = _p.parse_known_args(argv) if _args.log: loglevels = "CRITICAL DEBUG ERROR FATAL INFO WARN" if _args.log not in loglevels: # pragma: nocover error('legal values for the -L parameter are:', loglevels) loglevel = getattr(logging, _args.log) else: loglevel = None logging.basicConfig( level=loglevel, format='%(filename)s:%(lineno)d: %(levelname)s: %(message)s' ) if _args.version: # pragma: nocover print("pydeps v" + __version__) sys.exit(0) return _p, _args, argv # return parsed and remaining args def parse_args(argv=()): """Parse command line arguments, and return a dict. """ _p, _args, argv = base_argparser(argv) find_package = _args.find_package config_files = [] if not _args.no_config: # process config files # extra config file specified with --config <fname> has highest precedence if _args.config: config_files.append(_args.config) # .pydeps file specified in current directory is next local_pydeps = os.path.join(os.getcwd(), '.pydeps') if os.path.exists(local_pydeps): config_files.append(local_pydeps) # pydeps.yml file specified in current directory is next local_pydeps = os.path.join(os.getcwd(), 'pydeps.yml') if os.path.exists(local_pydeps): config_files.append(local_pydeps) # is there a pyproject.toml file? pyproj = os.path.join(os.getcwd(), 'pyproject.toml') if os.path.exists(pyproj): config_files.append(pyproj) # is there a setup.cfg file? pyproj = os.path.join(os.getcwd(), 'setup.cfg') if os.path.exists(pyproj): config_files.append(pyproj) # finally the .pydeps file in the the user's homedir home = os.environ['USERPROFILE' if sys.platform == 'win32' else 'HOME'] home_pydeps = os.path.join(home, '.pydeps') if os.path.exists(home_pydeps): config_files.append(home_pydeps) args = Arguments(config_files, debug=True, parents=[_p]) if not find_package: args.add('fname', kind="FNAME:input", help='filename') else: args.add('--fname', kind="FNAME:input", help='filename') args.add('-v', '--verbose', default=0, dest='verbose', action='count', help="be more verbose (-vv, -vvv for more verbosity)") args.add('-o', default=None, kind="FNAME:output", dest='output', metavar="file", help="write output to 'file'") args.add('-T', default='svg', dest='format', help="output format (svg|png)") args.add('--display', kind="FNAME:exe", default=None, help="program to use to display the graph (png or svg file depending on the T parameter)", metavar="PROGRAM") args.add('--noshow', '--no-show', action='store_true', default=False, dest='no_show', help="don't call external program to display graph") args.add('--show-deps', action='store_true', help="show output of dependency analysis") args.add('--show-raw-deps', action='store_true', help="show output of dependency analysis before removing skips") args.add('--deps-output', dest='deps_out', default=None, kind="FNAME:output", help="write output of dependency analysis to 'file'") args.add('--show-dot', action='store_true', help="show output of dot conversion") args.add('--dot-output', dest='dot_out', default=None, kind="FNAME:output", help="write dot code to 'file'") args.add('--nodot', '--no-dot', action='store_true', help="skip dot conversion") args.add('--no-output', action='store_true', help="don't create .svg/.png file, implies --no-show (-t/-o will be ignored)") args.add('--show-cycles', action='store_true', help="show only import cycles") args.add('--debug-mf', default=0, type=int, metavar="INT", help="set the ModuleFinder.debug flag to this value") args.add('--noise-level', default=200, type=int, metavar="INT", help="exclude sources or sinks with degree greater than noise-level") args.add('--max-bacon', default=2, type=int, metavar="INT", help="exclude nodes that are more than n hops away (default=2, 0 -> infinite)") args.add('--max-module-depth', default=0, type=int, metavar="INT", help="coalesce deep modules to at most n levels") args.add('--pylib', action='store_true', help="include python std lib modules") args.add('--pylib-all', action='store_true', help="include python all std lib modules (incl. C modules)") args.add('--include-missing', action='store_true', help="include modules that are not installed (or can't be found on sys.path)") args.add('-x', '--exclude', default=[], nargs="+", metavar="PATTERN", help="input files to skip (e.g. `foo.*`), multiple file names can be provided") args.add('-xx', '--exclude-exact', default=[], nargs="+", metavar="MODULE", help="same as --exclude, except requires the full match. `-xx foo.bar` will exclude foo.bar, but not foo.bar.blob") args.add('--only', default=[], nargs="+", metavar="MODULE_PATH", help="only include modules that start with MODULE_PATH") args.add('--externals', action='store_true', help='create list of direct external dependencies') args.add('--reverse', action='store_true', help="draw arrows to (instead of from) imported modules") args.add('--rankdir', default='TB', type=str, choices=['TB', 'BT', 'LR', 'RL'], help="set the direction of the graph, legal values are TB (default, imported modules above importing modules), " "BT (opposite direction of TB), BT (opposite direction of TB), LR (left-to-right) and RL (right-to-left)") args.add('--cluster', action='store_true', help="draw external dependencies as separate clusters") args.add('--min-cluster-size', default=0, type=int, metavar="INT", help="the minimum number of nodes a dependency must have before being clustered (default=0)") args.add('--max-cluster-size', default=0, type=int, metavar="INT", help="the maximum number of nodes a dependency can have before the cluster is collapsed to a single node (default=0)") args.add('--keep-target-cluster', action='store_true', help="draw target module as a cluster") args.add('--collapse-target-cluster', action='store_true', help="collapse target module (--keep-target-cluster will be ignored)") args.add('--rmprefix', default=[], nargs="+", metavar="PREFIX", help="remove PREFIX from the displayed name of the nodes") args.add('--start-color', default=0, type=int, metavar="INT", help="starting value for hue from 0 (red/default) to 360.") # args.write_default_config() _args = args.parse_args(argv) if _args.externals: return Config(externals=True, fname=_args.fname, max_bacon=10, include_missing=True, no_show=True) if _args.no_output: _args.no_show = True _args.show = not _args.no_show if _args.no_dot and _args.show_cycles: error("Can't use --no=dot and --show-cycles together") # pragma: nocover if _args.no_dot: _args.show_dot = False if _args.max_bacon == 0: _args.max_bacon = sys.maxsize if ( _args.keep_target_cluster or _args.min_cluster_size > 0 or _args.max_cluster_size > 0 or _args.collapse_target_cluster ): _args.cluster = True if find_package: _args.fname = _find_current_package() _args.format = getattr(_args, 'format', 'svg') verbose = _mkverbose(max(_args.verbose, int(_args.debug))) verbose(2, _args, '\n') if _args.debug: # pragma: nocover _args.verbose = 1 _args.show = True _args.show_deps = True _args.show_dot = True return vars(_args)
10,311
46.302752
196
py
pydeps
pydeps-master/pydeps/pycompat.py
# -*- coding: utf-8 -*- """ Compatibility imports between py2/py3 """ # pragma: nocover try: from itertools import zip_longest # noqa except ImportError: from itertools import izip_longest as zip_longest # noqa try: import configparser # noqa except ImportError: import ConfigParser as configparser # noqa
391
25.133333
62
py
pydeps
pydeps-master/pydeps/package_names.py
# -*- coding: utf-8 -*- from __future__ import print_function import os import site def _find_top_level_file(site_pkg_dir, pth): if pth.endswith('.dist-info') or pth.endswith('.egg-info'): top_level_fname = os.path.join(site_pkg_dir, pth, 'top_level.txt') elif pth.endswith('.egg'): top_level_fname = os.path.join(site_pkg_dir, pth, 'EGG-INFO', 'top_level.txt') else: top_level_fname = None return top_level_fname def _extract_pkg_name(pth): name = pth.rsplit('.', 1)[0] name_no_version = name.split('-', 1)[0] return name_no_version.replace('_', '-') def find_package_names(): # initialize with well-known packages that don't seem to have a top_level.txt res = { 'yaml': 'PyYAML', 'Crypto': 'pycrypto', } site_package_dirs = [site.getusersitepackages()] site_package_dirs += site.getsitepackages() for site_packages in reversed(site_package_dirs): if not os.path.isdir(site_packages): continue for pth in os.listdir(site_packages): top_level_fname = _find_top_level_file(site_packages, pth) if top_level_fname is None: continue pkgname = _extract_pkg_name(pth) if not os.path.exists(top_level_fname): if pkgname not in res.values(): print("ERR:", pth, 'has not top_level.txt') continue with open(top_level_fname) as fp: modnames = fp.read().split() for modname in modnames: modname = modname.replace('/', '.') if modname.startswith(r'win32\lib'): modname = modname.rsplit('\\')[1] res[modname] = pkgname return res if __name__ == "__main__": import json print(json.dumps(find_package_names(), indent=4, sort_keys=True))
1,892
29.047619
86
py
pydeps
pydeps-master/pydeps/__init__.py
# -*- coding: utf-8 -*- """ Python module dependency visualization. This package installs the ``pydeps`` command, and normal usage will be to use it from the command line. """ __version__ = "1.12.12"
200
27.714286
76
py
pydeps
pydeps-master/pydeps/mf27.py
# from .mf.mf_next import * # for debugging next version import modulefinder from modulefinder import ( ModuleFinder as NativeModuleFinder ) from importlib.util import MAGIC_NUMBER import marshal import dis from . import mfimp HAVE_ARGUMENT = dis.HAVE_ARGUMENT # from stdlib's modulefinder _PY_SOURCE = mfimp.PY_SOURCE _PY_COMPILED = mfimp.PY_COMPILED _PKG_DIRECTORY = mfimp.PKG_DIRECTORY # monkey-patch broken modulefinder._find_module # (https://github.com/python/cpython/issues/84530) # in Python 3.8-3.10 if hasattr(modulefinder, '_find_module'): modulefinder._find_module = mfimp.find_module class ModuleFinder(NativeModuleFinder): def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook: name(%s) caller(%s) fromlist(%s) level(%s)" % (name, caller, fromlist, level)) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) if q.shortname in ('__future__', 'future'): # [pydeps] the future package causes recursion overflow return None m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def load_module(self, fqname, fp, pathname, file_info): # fqname = dotted module name we're loading suffix, mode, kind = file_info kstr = { _PKG_DIRECTORY: 'PKG_DIRECTORY', _PY_SOURCE: 'PY_SOURCE', _PY_COMPILED: 'PY_COMPILED', }.get(kind, 'unknown-kind') self.msgin(2, "load_module(%s) fqname=%s, fp=%s, pathname=%s" % (kstr, fqname, fp and "fp", pathname)) if kind == _PKG_DIRECTORY: module = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", module) return module if kind == _PY_SOURCE: txt = fp.read() txt += b'\n' if isinstance(txt, bytes) else '\n' co = compile( txt, pathname, 'exec', # compile code block dont_inherit=True # [pydeps] don't inherit future statements from current environment ) elif kind == _PY_COMPILED: # a .pyc file is a binary file containing only three things: # 1. a four-byte magic number # 2. a four byte modification timestamp, and # 3. a Marshalled code object # from: https://nedbatchelder.com/blog/200804/the_structure_of_pyc_files.html if fp.read(4) != MAGIC_NUMBER: self.msgout(2, "raise ImportError: Bad magic number", pathname) raise ImportError("Bad magic number in %s" % pathname) fp.read(4) # skip modification timestamp co = marshal.load(fp) # load marshalled code object. else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def scan_code(self, co, m): code = co.co_code # noqa # if sys.version_info >= (3, 4): # scanner = self.scan_opcodes # elif sys.version_info >= (2, 5): # scanner = self.scan_opcodes_25 # else: # scanner = self.scan_opcodes_24 scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what in ("import", "absolute_import"): fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] if what == "absolute_import": level = 0 else: level = -1 self._safe_import_hook(name, m, fromlist, level=level) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) # m is still the caller here... [bp] self._safe_import_hook(parent.__name__, m, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m)
5,842
38.47973
112
py
pydeps
pydeps-master/pydeps/pydeps.py
# -*- coding: utf-8 -*- """cli entrypoints. """ from __future__ import print_function import json import os import sys from pydeps.configs import Config from . import py2depgraph, cli, dot, target from .depgraph2dot import dep2dot, cycles2dot import logging from . import colors log = logging.getLogger(__name__) def _pydeps(trgt, **kw): # Pass args as a **kw dict since we need to pass it down to functions # called, but extract locally relevant parameters first to make the # code prettier (and more fault tolerant). # print("KW:", kw, '\n', os.getcwd()) # print('abspath:', os.path.abspath(kw.get('deps_out'))) # print('target', trgt.workdir) # print('target', trgt) colors.START_COLOR = kw.get('start_color') # show_cycles = kw.get('show_cycles') nodot = kw.get('no_dot') no_output = kw.get('no_output') output = kw.get('output') fmt = kw['format'] show_svg = kw.get('show') deps_out = kw.get('deps_out') dot_out = kw.get('dot_out') # reverse = kw.get('reverse') if os.getcwd() != trgt.workdir: # the tests are calling _pydeps directoy os.chdir(trgt.workdir) dep_graph = py2depgraph.py2dep(trgt, **kw) if kw.get('show_deps'): cli.verbose("DEPS:") if deps_out: # make sure output files are written to sensible directories directory, _fname = os.path.split(deps_out) if not directory: deps_out = os.path.join(trgt.calling_dir, deps_out) with open(deps_out, 'w') as fp: fp.write(dep_graph.__json__()) else: print(dep_graph.__json__()) dotsrc = depgraph_to_dotsrc(trgt, dep_graph, **kw) if not nodot: if kw.get('show_dot'): cli.verbose("DOTSRC:") if dot_out: # make sure output files are written to sensible directories directory, _fname = os.path.split(dot_out) if not directory: dot_out = os.path.join(trgt.calling_dir, dot_out) with open(dot_out, 'w') as fp: fp.write(dotsrc) else: print(dotsrc) if not no_output: try: svg = dot.call_graphviz_dot(dotsrc, fmt) except OSError as cause: raise RuntimeError("While rendering {!r}: {}".format(output, cause)) if fmt == 'svg': svg = svg.replace(b'</title>', b'</title><style>.edge>path:hover{stroke-width:8}</style>') try: with open(output, 'wb') as fp: cli.verbose("Writing output to:", output) fp.write(svg) except OSError as cause: raise RuntimeError("While writing {!r}: {}".format(output, cause)) if show_svg: try: dot.display_svg(kw, output) except OSError as cause: helpful = "" if cause.errno == 2: helpful = " (can be caused by not finding the program to open this file)" raise RuntimeError("While opening {!r}: {}{}".format(output, cause, helpful)) def depgraph_to_dotsrc(target, dep_graph, **kw): """Convert the dependency graph (DepGraph class) to dot source code. """ if kw.get('show_cycles'): dotsrc = cycles2dot(target, dep_graph, **kw) elif not kw.get('no_dot'): dotsrc = dep2dot(target, dep_graph, **kw) else: dotsrc = None return dotsrc def externals(trgt, **kwargs): """Return a list of direct external dependencies of ``pkgname``. Called for the ``pydeps --externals`` command. """ kw = dict( T='svg', config=None, debug=False, display=None, exclude=[], exclude_exact=[], externals=True, format='svg', max_bacon=2**65, no_config=True, nodot=False, noise_level=2**65, no_show=True, output=None, pylib=True, pylib_all=True, show=False, show_cycles=False, show_deps=False, show_dot=False, show_raw_deps=False, verbose=0, include_missing=True, start_color=0 ) kw.update(kwargs) depgraph = py2depgraph.py2dep(trgt, **kw) pkgname = trgt.fname log.info("DEPGRAPH: %s", depgraph) pkgname = os.path.splitext(pkgname)[0] res = {} ext = set() for k, src in list(depgraph.sources.items()): if k.startswith('_'): continue if not k.startswith(pkgname): continue if src.imports: imps = [imp for imp in src.imports if not imp.startswith(pkgname)] if imps: for imp in imps: ext.add(imp.split('.')[0]) res[k] = imps # return res # debug return list(sorted(ext)) def pydeps(**args): """Entry point for the ``pydeps`` command. This function should do all the initial parameter and environment munging before calling ``_pydeps`` (so that function has a clean execution path). """ sys.setrecursionlimit(10000) _args = dict(iter(Config(**args))) if args else cli.parse_args(sys.argv[1:]) _args['curdir'] = os.getcwd() inp = target.Target(_args['fname']) log.debug("Target: %r", inp) if _args.get('output'): _args['output'] = os.path.abspath(_args['output']) else: _args['output'] = os.path.join( inp.calling_dir, inp.modpath.replace('.', '_') + '.' + _args.get('format', 'svg') ) with inp.chdir_work(): # log.debug("Current directory: %s", os.getcwd()) _args['fname'] = inp.fname _args['isdir'] = inp.is_dir if _args.get('externals'): del _args['fname'] exts = externals(inp, **_args) print(json.dumps(exts, indent=4)) # return exts # so the tests can assert else: # this is the call you're looking for :-) try: return _pydeps(inp, **_args) except (OSError, RuntimeError) as cause: if log.isEnabledFor(logging.DEBUG): # we only want to log the exception if we're in debug mode log.exception("While running pydeps:") cli.error(str(cause)) def call_pydeps(file_or_dir, **kwargs): """Programatic entry point for pydeps. See :class:`pydeps.configs.Config` class for the available options. """ sys.setrecursionlimit(10000) inp = target.Target(file_or_dir) log.debug("Target: %r", inp) config = Config(**kwargs) if config.output: config.output = os.path.abspath(config.output) else: config.output = os.path.join( inp.calling_dir, inp.modpath.replace('.', '_') + '.' + config.format ) ctx = dict(iter(config)) with inp.chdir_work(): ctx['fname'] = inp.fname ctx['isdir'] = inp.is_dir if config.externals: del ctx['fname'] return externals(inp, **ctx) return _pydeps(inp, **ctx) if __name__ == '__main__': # pragma: nocover pydeps()
7,175
32.376744
106
py
pydeps
pydeps-master/pydeps/py2depgraph.py
# Copyright 2004,2009 Toby Dickenson # Changes 2014 (c) Bjorn Pettersen # # Permission is hereby granted, free of charge, to any person obtaining # a copy of this software and associated documentation files (the # "Software"), to deal in the Software without restriction, including # without limitation the rights to use, copy, modify, merge, publish, # distribute, sublicense, and/or sell copies of the Software, and to # permit persons to whom the Software is furnished to do so, subject # to the following conditions: # # The above copyright notice and this permission notice shall be included # in all copies or substantial portions of the Software. # # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, # EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF # MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. # IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY # CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, # TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE # SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. from __future__ import print_function import json import os import sys from collections import defaultdict import enum from .dummymodule import DummyModule from .pystdlib import pystdlib from . import depgraph from . import mf27 import logging log = logging.getLogger(__name__) PYLIB_PATH = depgraph.PYLIB_PATH class imp(enum.Enum): C_BUILTIN = 6 C_EXTENSION = 3 IMP_HOOK = 9 PKG_DIRECTORY = 5 PY_CODERESOURCE = 8 PY_COMPILED = 2 PY_FROZEN = 7 PY_RESOURCE = 4 PY_SOURCE = 1 class Module(object): def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} @property def shortname(self): if self.__name__ == "__main__": return self.__file__[:-3].replace('\\', '.') return self.__name__ def __repr__(self): return "Module(name=%s, file=%r, path=%r)" % ( self.__name__, (self.__file__ or "").replace('\\', '/'), self.__path__ ) class MyModuleFinder(mf27.ModuleFinder): def __init__(self, syspath, *args, **kwargs): self.args = kwargs self.verbose = kwargs.get('verbose', 0) # include all of python std lib (incl. C modules) self.include_pylib_all = kwargs.pop('pylib_all', False) # include python std lib modules. # self.include_pylib = kwargs.pop('pylib', self.include_pylib_all) self.include_pylib = kwargs.pop('pylib', self.include_pylib_all) self._depgraph = defaultdict(dict) self._types = {} self._last_caller = None # path=None, debug=0, excludes=[], replace_paths=[] debug = 5 if self.verbose >= 4 else 0 mf27.ModuleFinder.__init__(self, path=syspath, debug=debug, # debug=3, excludes=kwargs.get('excludes', [])) def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def import_hook(self, name, caller=None, fromlist=None, level=-1): old_last_caller = self._last_caller try: self._last_caller = caller # print " last_CALLER:", caller, "OLD-lastcaller:", old_last_caller return mf27.ModuleFinder.import_hook(self, name, caller, fromlist, level) finally: self._last_caller = old_last_caller def _add_import(self, module): if module is not None: if self._last_caller: # self._depgraph[self._last_caller.__name__][module.__name__] = module.__file__ if hasattr(module, '__file__') or self.include_pylib_all: pylib_p = [] if not module.__file__: pass else: rpath = os.path.split(module.__file__)[0].lower() if 'site-packages' not in rpath: pylib_p = [rpath.startswith(pp) for pp in PYLIB_PATH] if self.include_pylib or not any(pylib_p): # if self._last_caller.__name__ != module.__name__: # self._depgraph[self._last_caller.__name__][module.__name__] = module.__file__ self._depgraph[self._last_caller.__name__][module.__name__] = module.__file__ def import_module(self, partnam, fqname, parent): module = mf27.ModuleFinder.import_module(self, partnam, fqname, parent) self._add_import(module) return module def load_module(self, fqname, fp, pathname, suffix_mode_kind): (suffix, mode, kind) = suffix_mode_kind try: module = mf27.ModuleFinder.load_module( self, fqname, fp, pathname, (suffix, mode, kind) ) except SyntaxError: # this happened first when pyinvoke tried to load yaml3/2 based on # an `if six.PY3` # print "SYNTAX_ERROR" module = None except AttributeError as e: # See issues #139 and #140... print("ERROR trying to load {} from {} (py38+ _find_module reimplementation of imp.find_module called .is_package on None...)\n{}".format( fqname, pathname, e )) module = None if module is not None: self._types[module.__name__] = kind return module def ensure_fromlist(self, module, fromlist, recursive=0): self.msg(4, "ensure_fromlist", module, fromlist, recursive) for sub in fromlist: # print " for sub:", sub, "in fromlisst:", fromlist, "hasattr(module, sub):", hasattr(module, sub) if sub == "*": # print "STAR" if not recursive: submodules = self.find_all_submodules(module) if submodules: self.ensure_fromlist(module, submodules, 1) elif not hasattr(module, sub): # print "ELIF......" subname = "%s.%s" % (module.__name__, sub) submod = self.import_module(sub, subname, module) if not submod: raise ImportError("No module named " + subname) else: self._add_import(getattr(module, sub)) # print " SUB:", sub, "lastcaller:", self._last_caller class RawDependencies(object): def __init__(self, fname, **kw): path = sys.path[:] exclude = [] mf = MyModuleFinder(path, exclude, **kw) mf.run_script(fname) self.depgraph = mf._depgraph self.types = mf._types def py2dep(target, **kw) -> depgraph.DepGraph: """"Calculate dependencies for ``pattern`` and return a DepGraph. """ log.info("py2dep(%r)", target) dummy = DummyModule(target, **kw) kw['dummyname'] = dummy.fname syspath = sys.path[:] syspath.insert(0, target.syspath_dir) # remove exclude so we don't pass it twice to modulefinder exclude = ['migrations'] + kw.pop('exclude', []) log.debug("Exclude: %r", exclude) log.debug("KW: %r", kw) if 'fname' in kw: del kw['fname'] mf = MyModuleFinder( syspath, # module search path for this module finder excludes=exclude, # folders to exclude **kw ) mf.debug = max(mf.debug, kw.get('debug_mf', 0)) if log.isEnabledFor(logging.DEBUG): log.debug("CURDIR: %s", os.getcwd()) log.debug("FNAME: %r, CONTENT:\n%s\n", dummy.fname, dummy.text()) mf.run_script(dummy.fname) log.info("mf._depgraph:\n%s", json.dumps(dict(mf._depgraph), indent=4)) log.info("mf.badmodules:\n%s", json.dumps(mf.badmodules, indent=4)) if kw.get('include_missing'): for k, vdict in list(mf.badmodules.items()): if k not in mf._depgraph: mf._depgraph[k] = {} for v in vdict: if not target.is_pysource and v not in mf._depgraph['__main__']: mf._depgraph['__main__'][v] = None if v in mf._depgraph: mf._depgraph[v][k] = None else: mf._depgraph[v] = {k: None} log.info("mf._depgraph:\n%s", json.dumps(dict(mf._depgraph), indent=4)) kw['exclude'] = exclude if kw.get('pylib'): mf_depgraph = mf._depgraph for k, v in list(mf._depgraph.items()): log.debug('depgraph item: %r %r', k, v) # mf_modules = {k: os.syspath.abspath(v.__file__) # for k, v in mf.modules.items()} else: pylib = pystdlib() mf_depgraph = {} for k, v in list(mf._depgraph.items()): log.debug('depgraph item: %r %r', k, v) if k in pylib: continue vals = {vk: vv for vk, vv in v.items() if vk not in pylib} mf_depgraph[k] = vals # mf_modules = {k: os.syspath.abspath(v.__file__) # for k, v in mf.modules.items() # if k not in pylib} try: import yaml log.info("mf_depgraph:\n%s", yaml.dump(dict(mf_depgraph), default_flow_style=False)) # log.error("mf._types:\n%s", yaml.dump(mf._types, default_flow_style=False)) # log.debug("mf_modules:\n%s", yaml.dump(mf_modules, default_flow_style=False)) except ImportError: log.info("mf_depgraph:\n%s", json.dumps(dict(mf_depgraph), indent=4)) return depgraph.DepGraph(mf_depgraph, mf._types, target, **kw) def py2depgraph(): _fname = sys.argv[1] _graph = RawDependencies(_fname) sys.stdout.write( json.dumps(_graph.__dict__, indent=4) ) if __name__ == '__main__': py2depgraph()
10,490
35.681818
150
py
pydeps
pydeps-master/pydeps/depgraph.py
# -*- coding: utf-8 -*- from __future__ import print_function from collections import defaultdict import fnmatch from .pycompat import zip_longest import json import os import pprint import re import enum from . import colors, cli import sys import logging log = logging.getLogger(__name__) # we're normally not interested in imports of std python packages. PYLIB_PATH = { # in virtualenvs that see the system libs, these will be different. os.path.split(os.path.split(pprint.__file__)[0])[0].lower(), os.path.split(os.__file__)[0].lower() } class imp(enum.Enum): C_BUILTIN = 6 C_EXTENSION = 3 IMP_HOOK = 9 PKG_DIRECTORY = 5 PY_CODERESOURCE = 8 PY_COMPILED = 2 PY_FROZEN = 7 PY_RESOURCE = 4 PY_SOURCE = 1 UNKNOWN = 0 class Source(object): """A node (contained) in the dependency graph. It contains info about which modules are imported by this source, and which modules import this source. """ def __init__(self, name, path=None, imports=(), exclude=False, args=None): self.args = args or {} self.name = name # self.kind = kind self.path = path # needed here..? self.imports = set(imports) # modules we import self.imported_by = set() # modules that import us self.bacon = sys.maxsize # bacon distance self.excluded = exclude @property def name_parts(self): return self.name.split('.') @property def module_depth(self): return self.name.count('.') @property def path_parts(self): p = self.path or "" return p.replace('\\', '/').lower().split('/') @property def in_degree(self): """Number of incoming arrows. """ return len(self.imports) @property def out_degree(self): """Number of outgoing arrows. """ return len(self.imported_by) @property def degree(self): return self.in_degree + self.out_degree def is_noise(self): """Is this module just noise? (too common either at top or bottom of the graph). """ noise = self.args['noise_level'] if not (self.in_degree and self.out_degree): return self.degree > noise return False def __json__(self): res = dict( name=self.name, path=self.path, # kind=str(self.kind), bacon=self.bacon, ) if self.excluded: res['excluded'] = 'EXCLUDED' if self.imports: res['imports'] = list(sorted(self.imports)) if self.imported_by: res['imported_by'] = list(sorted(self.imported_by)) return res def __str__(self): return "%s (%s)" % (self.name, self.path) def __hash__(self): return hash(self.name) def __eq__(self, other): return self.name == other.name def __lt__(self, other): return self.name < other.name def __le__(self, other): return self.name <= other.name def __gt__(self, other): return self.name > other.name def __ge__(self, other): return self.name >= other.name def __repr__(self): return json.dumps(self.__json__(), indent=4) def __iadd__(self, other): """Merge other into self. """ if self.name == other.name and self.imports == other.imports and self.bacon == other.bacon: return self log.debug("iadd lhs: %r", self) log.debug("iadd rhs: %r", other) assert self.name == other.name self.path = self.path or other.path self.imports |= other.imports self.imported_by |= other.imported_by self.bacon = min(self.bacon, other.bacon) self.excluded = self.excluded or other.excluded log.debug("iadd result: %r", self) return self def get_label(self, splitlength=0, rmprefix=None): name = self.name if rmprefix: for prefix in rmprefix: if self.name.startswith(prefix): name = self.name[len(prefix):] break if splitlength and len(name) > splitlength and '.' in name: return '\\.\\n'.join(name.split('.')) # pragma: nocover return name @property def label(self): """Convert a module name to a formatted node label. This is a default policy - please override. """ return self.get_label(splitlength=14) class DepGraph(object): """The dependency graph. It is the output of :func:`pydeps.py2depgraph.py2dep` """ skip_modules = """ os sys qt time __future__ types re string bdb pdb __main__ south """.split() def __init__(self, depgraf, types, target, **args): # depgraph is py2depgraph.MyModulefinder._depgraph log.debug("DepGraph: depgraf=%r", depgraf) self.curhue = 150 # start with a green-ish color self.colors = {} self.cycles = [] self.cyclenodes = set() self.cyclerelations = set() self.max_module_depth = args.get('max_module_depth', 0) self.target = target self.args = args #: dict[module_name] -> Source object self.sources = {} self.skiplist = [re.compile(fnmatch.translate(arg)) for arg in args['exclude']] self.skiplist += [re.compile('^%s$' % fnmatch.translate(arg)) for arg in args['exclude_exact']] # depgraf = {name: imports for (name, imports) in depgraf.items()} for name, imports in depgraf.items(): log.debug("depgraph name=%r imports=%r", name, imports) src = Source( name=self.source_name(name), imports=[self.source_name(n) for n in imports.keys()], # values handled below args=args, exclude=self._exclude(name), ) log.debug("depgraph src=%r", src) self.add_source(src) for iname, path in imports.items(): src = Source( name=self.source_name(iname, path), path=path, args=args, exclude=self._exclude(iname) ) self.add_source(src) self.module_count = len(self.sources) cli.verbose(1, "there are", self.module_count, "total modules") self.connect_generations() if self.args['show_cycles']: self.find_import_cycles() self.calculate_bacon() if self.args['show_raw_deps']: print(self) self.exclude_noise() self.exclude_bacon(self.args['max_bacon']) self.only_filter(self.args.get('only')) excluded = [v for v in list(self.sources.values()) if v.excluded] # print "EXCLUDED:", excluded self.skip_count = len(excluded) cli.verbose(1, "skipping", self.skip_count, "modules") for module in excluded: # print 'exclude:', module.name cli.verbose(2, " ", module.name) self.remove_excluded() if not self.args['show_deps']: cli.verbose(3, self) def source_name(self, name, path=None): """Returns the module name, possibly limited by --max-module-depth. """ res = name if name == "__main__" and self.target.is_pysource: # use the target file name directly if we're working on a # single file return self.target.fname if name == "__main__" and path: # use the path to the main module if we're working on a module. res = path.replace('\\', '/').replace('/', '.') if self.args.get('verbose', 0) >= 2: # pragma: nocover print("changing __main__ =>", res) if self.max_module_depth > 0: res = '.'.join(res.split('.')[:self.max_module_depth]) return res def __json__(self): return json.dumps(self.sources, indent=4, sort_keys=True, default=lambda obj: obj.__json__() if hasattr(obj, '__json__') else obj) def levelcounts(self): pass def get_colors(self, src, colorspace=None): if colorspace is None: if src.basename not in self.colors: h = self.curhue # self.curhue += 7 # relative prime with 360 self.curhue += 37 # relative prime with 360 self.curhue %= 360 # print "NAME:", src.name, "BASENAME:", src.basename bg = colors.name2rgb(h) black = (0, 0, 0) white = (255, 255, 255) fg = colors.foreground(bg, black, white) self.colors[src.basename] = bg, fg return self.colors[src.basename] else: return colorspace.color(src) def _is_pylib(self, path): log.info('path %r in PYLIB_PATH %r => %s', path, PYLIB_PATH, path in PYLIB_PATH) return path in PYLIB_PATH def proximity_metric(self, a, b): """Return the weight of the dependency from a to b. Higher weights usually have shorter straighter edges. Return 1 if it has normal weight. A value of 4 is usually good for ensuring that a related pair of modules are drawn next to each other. Returns an int between 1 (unknown, default), and 4 (very related). """ res = 0 for ap, bp, n in zip(a.name_parts, b.name_parts, list(range(4))): res += ap == bp if n >= 3: break if res == 0: res = 1 # noqa return 4 if res > 4 else res def dissimilarity_metric(self, a, b): """Return non-zero if references to this module are strange, and should be drawn extra-long. The value defines the length, in rank. This is also good for putting some vertical space between seperate subsystems. Returns an int between 1 (default) and 4 (highly unrelated). """ res = 4 for an, bn, n in zip_longest(a.name_parts, b.name_parts, list(range(4))): res -= an == bn if n >= 3: break return 4 if res > 4 else res def _exclude(self, name): return any(skip.match(name) for skip in self.skiplist) def add_source(self, src): if src.name in self.sources: log.debug("ADD-SOURCE[+=]\n%r", src) self.sources[src.name] += src # merge else: log.debug("ADD-SOURCE[=]\n%r", src) self.sources[src.name] = src def __getitem__(self, item): return self.sources[item] def __iter__(self): visited = set(self.skip_modules) | set(self.args['exclude']) def visit(src): if src.name in visited: return visited.add(src.name) for name in src.imports: impmod = self.sources[name] # FIXME: why do we want to exclude **/*/__init__.py? This line # causes `collections` package in py3 to be excluded. # if impmod.path and not impmod.path.endswith('__init__.py'): if not src.name.startswith(impmod.name + "."): yield impmod, src visit(impmod) for _src in self.sources.values(): for source in visit(_src): cli.verbose(4, "Yielding", source[0], source[1]) yield source def __repr__(self): return json.dumps(self.sources, indent=4, sort_keys=True, default=lambda obj: obj.__json__() if hasattr(obj, '__json__') else obj) def find_import_cycles(self): def traverse(node, path): if node.name in self.cyclenodes: return if node.name in path: # found cycle cycle = path[path.index(node.name):] + [node.name] self.cycles.append(cycle) for nodename in cycle: self.cyclenodes.add(nodename) for i in range(len(cycle) - 1): self.cyclerelations.add( (cycle[i], cycle[i + 1]) ) return for impmod in node.imports: traverse(self.sources[impmod], path + [node.name]) for src in list(self.sources.values()): traverse(src, []) def connect_generations(self): """Traverse depth-first adding imported_by. """ for src in self.sources.values(): for _child in src.imports: if _child in self.sources: child = self.sources[_child] child.imported_by.add(src.name) def calculate_bacon(self): count = defaultdict(int) def bacon(src, n): count[src.name] += 1 if src.bacon <= n: return src.bacon = min(src.bacon, n) for imp in src.imports: bacon(self.sources[imp], n + 1) if '__main__' in self.sources: bacon(self.sources['__main__'], 0) elif self.args['dummyname'] in self.sources: bacon(self.sources[self.args['dummyname']], 0) def exclude_noise(self): for src in list(self.sources.values()): if src.excluded: continue if src.is_noise(): cli.verbose(2, "excluding", src, "because it is noisy:", src.degree) src.excluded = True self._add_skip(src.name) def exclude_bacon(self, limit): """Exclude modules that are more than `limit` hops away from __main__. """ for src in list(self.sources.values()): if src.bacon > limit: src.excluded = True self._add_skip(src.name) def only_filter(self, paths): """Exclude nodes that have a prefix in paths. """ if not paths: return paths = set(paths) def should_include(node): for p in paths: if node.name.startswith(p): return True return False for src in list(self.sources.values()): if not should_include(src): src.excluded = True # print "Excluding bacon:", src.name self._add_skip(src.name) def remove_excluded(self): """Remove all sources marked as excluded. """ sources = list(self.sources.values()) for src in sources: if src.excluded: del self.sources[src.name] src.imports = [m for m in src.imports if not self._exclude(m)] src.imported_by = [m for m in src.imported_by if not self._exclude(m)] def _add_skip(self, name): # print 'add skip:', name self.skiplist.append(re.compile(fnmatch.translate(name)))
15,089
32.019694
103
py
pydeps
pydeps-master/pydeps/configs.py
from io import StringIO import json import warnings import logging log = logging.getLogger(__name__) # from devtools import debug HAVE_TOML = False try: import tomllib as toml HAVE_TOML = True except ImportError: try: import tomlkit as toml HAVE_TOML = True except ImportError: try: import toml HAVE_TOML = True except ImportError: pass def is_string(v): return isinstance(v, str) def boolval(v): if isinstance(v, bool): return v if isinstance(v, int): return bool(v) if is_string(v): v = v.lower() if v in {'j', 'y', 'ja', 'yes', '1', 'true'}: return True if v in {'n', 'nei', 'no', '0', 'false'}: return False raise ValueError("Don't know how to convert %r to bool" % v) def listval(v): if is_string(v): return [x for x in v.split() if x.strip()] if isinstance(v, (list, tuple)): return v raise ValueError("Don't know how to convert %r to list" % v) def identity(v): return v typefns = { 'BOOL': boolval, 'INT': int, 'LIST': listval, 'STR': str, } def load_toml(filename): if not HAVE_TOML: return {} try: with open(filename) as fp: res = toml.loads(fp.read()) return res['tool']['pydeps'] except Exception as e: log.debug("Couldn't load toml file %s: %s", filename, e) return {} def load_json(filename): try: with open(filename) as fp: res = json.loads(fp.read()) return res['pydeps'] except (json.JSONDecodeError, KeyError): return {} def load_yaml(filename): try: import yaml from yaml import Loader with open(filename) as fp: res = yaml.load(fp.read(), Loader=Loader) return res['pydeps'] except (yaml.YAMLError, KeyError, ImportError): return {} def load_ini(filename): import configparser conf = configparser.ConfigParser() conf.read(filename) try: config = conf.items("pydeps") return dict(config) except (configparser.NoOptionError, configparser.NoSectionError): warnings.warn(' '.join(""" Couldn't find a [pydeps] section in your config files %r -- or it was empty """.split()) % filename) return {} def load_config(filename): if filename.endswith('.toml'): return load_toml(filename) if filename.endswith('.json'): return load_json(filename) if filename.endswith('.yaml'): return load_yaml(filename) raise ValueError("Unknown config file type: %s" % filename) # this class is generated from pydeps.arguments.Arguments.write_default_config() class Config(object): #: turn on all the show and verbose options (mainly for debugging pydeps #: itself) debug = False #: specify config file config = None #: disable processing of config files no_config = False #: print pydeps version version = False #: set log-level to one of CRITICAL, ERROR, WARNING, INFO, DEBUG, NOTSET. log = None #: tries to automatically find the name of the current package. find_package = False #: filename fname = None #: be more verbose (-vv, -vvv for more verbosity) verbose = 0 #: write output to 'file' output = None #: output format (svg|png) format = 'svg' #: program to use to display the graph (png or svg file depending on the T #: parameter) display = None #: don't call external program to display graph no_show = False #: show output of dependency analysis show_deps = False #: show output of dependency analysis before removing skips show_raw_deps = False #: write output of dependency analysis to 'file' deps_out = None #: show output of dot conversion show_dot = False #: write dot code to 'file' dot_out = None #: skip dot conversion no_dot = False #: don't create .svg/.png file, implies --no-show (-t/-o will be ignored) no_output = False #: show only import cycles show_cycles = False #: set the ModuleFinder.debug flag to this value debug_mf = 0 #: exclude sources or sinks with degree greater than noise-level noise_level = 200 #: exclude nodes that are more than n hops away (default=2, 0 -> infinite) max_bacon = 2 #: coalesce deep modules to at most n levels max_module_depth = 0 #: include python std lib modules pylib = False #: include python all std lib modules (incl. C modules) pylib_all = False #: include modules that are not installed (or can't be found on sys.path) include_missing = False #: input files to skip (e.g. `foo.*`), multiple file names can be provided exclude = [] #: same as --exclude, except requires the full match. `-xx foo.bar` will #: exclude foo.bar, but not foo.bar.blob exclude_exact = [] #: only include modules that start with MODULE_PATH only = [] #: create list of direct external dependencies externals = False #: draw arrows to (instead of from) imported modules reverse = False #: set the direction of the graph, legal values are TB (default, imported #: modules above importing modules), BT (opposite direction of TB), BT #: (opposite direction of TB), LR (left-to-right) and RL (right-to-left) rankdir = 'TB' #: draw external dependencies as separate clusters cluster = False #: the minimum number of nodes a dependency must have before being clustered #: (default=0) min_cluster_size = 0 #: the maximum number of nodes a dependency can have before the cluster is #: collapsed to a single node (default=0) max_cluster_size = 0 #: draw target module as a cluster keep_target_cluster = False #: collapse target module (--keep-target-cluster will be ignored) collapse_target_cluster = False #: remove PREFIX from the displayed name of the nodes rmprefix = [] #: starting value for hue from 0 (red/default) to 360. start_color = 0 def __init__(self, **kwargs): for key in dir(self.__class__): if not key.startswith('_'): val = getattr(self.__class__, key) if not callable(val): setattr(self, key, val) self.__dict__.update(kwargs) def __eq__(self, other): res = True for (lkey, lval), (rkey, fval) in zip(iter(self), iter(other)): if lkey != rkey: print(f"key mismatch: {lkey} != {rkey}") res = False if lval != fval: print(f"val mismatch {lkey}: {lval} != {fval}") res = False return res def set_field(self, field, value): if field == 'fname': self.fname = identity(value) if field == 'verbose': self.verbose = int(value) if field == 'output': self.output = identity(value) if field == 'format': self.format = str(value) if field == 'display': self.display = identity(value) if field == 'no_show': self.no_show = boolval(value) if field == 'show_deps': self.show_deps = boolval(value) if field == 'show_raw_deps': self.show_raw_deps = boolval(value) if field == 'deps_out': self.deps_out = identity(value) if field == 'show_dot': self.show_dot = boolval(value) if field == 'dot_out': self.dot_out = identity(value) if field == 'no_dot': self.no_dot = boolval(value) if field == 'no_output': self.no_output = boolval(value) if field == 'show_cycles': self.show_cycles = boolval(value) if field == 'debug_mf': self.debug_mf = int(value) if field == 'noise_level': self.noise_level = int(value) if field == 'max_bacon': self.max_bacon = int(value) if field == 'max_module_depth': self.max_module_depth = int(value) if field == 'pylib': self.pylib = boolval(value) if field == 'pylib_all': self.pylib_all = boolval(value) if field == 'include_missing': self.include_missing = boolval(value) if field == 'exclude': self.exclude = listval(value) if field == 'exclude_exact': self.exclude_exact = listval(value) if field == 'only': self.only = listval(value) if field == 'externals': self.externals = boolval(value) if field == 'reverse': self.reverse = boolval(value) if field == 'rankdir': self.rankdir = str(value) if field == 'cluster': self.cluster = boolval(value) if field == 'min_cluster_size': self.min_cluster_size = int(value) if field == 'max_cluster_size': self.max_cluster_size = int(value) if field == 'keep_target_cluster': self.keep_target_cluster = boolval(value) if field == 'collapse_target_cluster': self.collapse_target_cluster = boolval(value) if field == 'rmprefix': self.rmprefix = listval(value) if field == 'start_color': self.start_color = int(value) def __iter__(self): return iter(self.__dict__.items()) def __getattr__(self, name): return self.__dict__[name] def update(self, data): if set(data.keys()) - set(self.__dict__.keys()): warnings.warn("Unknown config keys: %s" % (set(data.keys()) - set(self.__dict__.keys()))) for k, v in data.items(): if k in self.__dict__: self.set_field(k, v) def write_ini(self): import configparser cfg = configparser.ConfigParser() cfg['pydeps'] = {k : str(v) for k, v in self} with StringIO() as fp: cfg.write(fp) return fp.getvalue() def write_toml(self): import tomlkit return tomlkit.dumps({ 'tools.pydeps': {k : v for k, v in self if v is not None} }) def write_json(self): data = {'pydeps': self.__dict__} return json.dumps(data, indent=4) def write_yaml(self): import yaml data = {'pydeps': self.__dict__} return yaml.dump(data, default_flow_style=False) @classmethod def load(cls, fnames): """Load config from file. """ conf = cls() for fname in fnames: ftype = filetype(fname) if ftype == 'toml': conf.update(load_toml(fname)) if ftype == 'json': conf.update(load_json(fname)) if ftype == 'yaml': conf.update(load_yaml(fname)) if ftype == 'ini': conf.update(load_ini(fname)) return conf def filetype(fname): if fname.endswith('.toml'): return 'toml' elif fname.endswith('.json'): return 'json' elif fname.endswith('.yaml'): return 'yaml' elif fname.endswith('.yml'): return 'yaml' elif fname.endswith('.ini'): return 'ini' else: return 'ini'
11,463
26.757869
101
py
pydeps
pydeps-master/pydeps/colors.py
# -*- coding: utf-8 -*- """Color calculations. """ import colorsys # noinspection PyAugmentAssignment # import hashlib START_COLOR = 0 # Value can be changed from command-line argument def frange(start, end, step): """Like range(), but with floats. """ val = start while val < end: yield val val += step def distinct_hues(count): """Return ``count`` hues, equidistantly spaced. """ for i in frange(0., 360., 360. / count): hue = ((i + START_COLOR) % 360) / 360. yield hue class ColorSpace(object): def __init__(self, nodes): self.nodes = {} for node in nodes: parts = node.name.split('.') self.add_to_tree(parts, self.nodes) self.basecolors = distinct_hues(len(self.nodes)) self.colors = dict(zip(sorted(self.nodes.keys()), self.basecolors)) def add_to_tree(self, parts, tree): if not parts: return first, rest = parts[0], parts[1:] if first not in tree: tree[first] = {} self.add_to_tree(rest, tree[first]) def color(self, src): nodename = src.name parts = nodename.split('.') hue = self.colors[parts[0]] saturation = min(0.95, 0.4 + 0.1 * (src.out_degree - 1)) lightness = max(0.3, 0.5 - 0.02 * (src.in_degree - 1)) bg = rgb2eightbit(colorsys.hls_to_rgb(hue, lightness, saturation)) black = (0, 0, 0) white = (255, 255, 255) fg = foreground(bg, black, white) return bg, fg def __str__(self): # pragma: nocover import pprint return pprint.pformat(self.colors) def rgb2eightbit(rgb): """Convert floats in [0..1] to integers in [0..256) """ return tuple(int(x * 256) for x in rgb) def name2rgb(hue): """Originally used to calculate color based on module name. """ r, g, b = colorsys.hsv_to_rgb(hue / 360.0, .8, .7) return tuple(int(x * 256) for x in [r, g, b]) def brightness(r, g, b): """From w3c (range 0..255). """ return (r * 299 + g * 587 + b * 114) / 1000 def brightnessdiff(a, b): """greater than 125 is good. """ return abs(brightness(*a) - brightness(*b)) def colordiff(rgb1, rgb2): """From w3c (greater than 500 is good). (range [0..765]) """ (r, g, b) = rgb1 (r2, g2, b2) = rgb2 return ( max(r, r2) - min(r, r2) + max(g, g2) - min(g, g2) + max(b, b2) - min(b, b2) ) def foreground(background, *options): """Find the best foreground color from `options` based on `background` color. """ def absdiff(a, b): return brightnessdiff(a, b) # return 3 * brightnessdiff(a, b) + colordiff(a, b) diffs = [(absdiff(background, color), color) for color in options] diffs.sort(reverse=True) return diffs[0][1] def rgb2css(rgb): """Convert rgb to hex. """ return '#%02x%02x%02x' % rgb
2,957
24.282051
75
py
pydeps
pydeps-master/pydeps/tools/pydeps2requirements.py
# -*- coding: utf-8 -*- """ Generate requirements.txt from pydeps output... Usage:: pydeps <packagename> --max-bacon=0 \ --show-raw-deps --nodot \ --noshow | python pydeps2requirements.py """ import json import os import site import sys from collections import defaultdict from pydeps.package_names import find_package_names WIDTH = 80 # packages that are difficult to eliminate but shouldn't ever be part # of a package's requirements. skiplist = { '_markerlib', 'pkg_resources' } def dep2req(name, package, imported_by): """Convert dependency to requirement. """ lst = [item for item in sorted(imported_by) if not item.startswith(name)] res = '%-15s # from: ' % package imps = ', '.join(lst) if len(imps) < WIDTH - 24: return res + imps return res + imps[:WIDTH - 24 - 3] + '...' _SITE_PACKAGE_DIRS = None def site_packages(): global _SITE_PACKAGE_DIRS if _SITE_PACKAGE_DIRS is None: site_package_dirs = [site.getusersitepackages()] site_package_dirs += site.getsitepackages() _SITE_PACKAGE_DIRS = [] for pth in reversed(site_package_dirs): if os.path.isdir(pth): _SITE_PACKAGE_DIRS.append(pth) return _SITE_PACKAGE_DIRS def is_site_package(p): for sp in site_packages(): if p.startswith(sp): return True return False def pydeps2reqs(deps): """Convert a deps instance into requirements. """ reqs = defaultdict(set) baseprefix = sys.real_prefix if hasattr(sys, 'real_prefix') else sys.base_prefix pkgnames = find_package_names() for k, v in list(deps.items()): # not a built-in p = v['path'] if p and not p.startswith(baseprefix): if is_site_package(p): if not p.endswith('.pyd'): if '/win32/' in p.replace('\\', '/'): reqs['win32'] |= set(v['imported_by']) else: name = k.split('.', 1)[0] if name not in skiplist: reqs[name] |= set(v['imported_by']) if '_dummy' in reqs: del reqs['_dummy'] return '\n'.join(dep2req(name, pkgnames[name], reqs[name]) for name in sorted(reqs)) def main(): """Cli entrypoint. """ if len(sys.argv) == 2: fname = sys.argv[1] with open(fname, 'rb') as fp: data = json.load(fp) else: data = json.loads(sys.stdin.read()) print(pydeps2reqs(data)) if __name__ == "__main__": main()
2,587
24.623762
88
py
pydeps
pydeps-master/tests/test_dep2dot.py
# -*- coding: utf-8 -*- import os import sys import pydeps.cli from pydeps import pydeps from pydeps.target import Target from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_dep2dot(): files = """ foo: - __init__.py - a.py: | from . import b - b.py """ with create_files(files) as workdir: assert simpledeps('foo', '-LDEBUG -vv') == { 'foo.b -> foo.a' } args = pydeps.cli.parse_args(["foo", "--noshow"]) pydeps.pydeps(**args) assert os.path.exists(os.path.join(workdir, 'foo.svg'))
648
22.178571
63
py
pydeps
pydeps-master/tests/test_externals.py
# -*- coding: utf-8 -*- import ast from pydeps.pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_relative_imports(capsys): files = """ foo: - __init__.py - a.py: | from bar import b bar: - __init__.py - b.py """ with create_files(files) as workdir: assert simpledeps('foo') == { 'bar -> foo.a', 'bar.b -> foo.a' } pydeps(fname='foo', externals=True) io = capsys.readouterr() assert ast.literal_eval(io.out) == ['bar']
638
22.666667
50
py
pydeps
pydeps-master/tests/test_dot.py
# -*- coding: utf-8 -*- from pydeps.dot import dot, cmd2args def test_svg(tmpdir): tmpdir.chdir() ab = tmpdir.join('ab.svg') dot(u""" digraph G { a -> b } """, o=ab.basename) assert ab.exists() def test_svg_str(tmpdir): tmpdir.chdir() ab = tmpdir.join('ab.svg') dot(""" digraph G { a -> b } """, o=ab.basename) assert ab.exists() def test_boolopt(tmpdir): tmpdir.chdir() ab = tmpdir.join('ab.svg') dot(""" digraph G { a -> b } """, x=True, o=ab.basename) print(tmpdir.listdir()) assert ab.exists() def test_obj(tmpdir): # pragma: nocover GRAPH = u""" digraph G { a -> b } """ import sys if sys.version_info >= (3,): class MyClass: def __str__(self): return GRAPH else: class MyClass(object): def __unicode__(self): return GRAPH tmpdir.chdir() ab = tmpdir.join('ab.svg') dot(MyClass(), x=True, o=ab.basename) print(tmpdir.listdir()) assert ab.exists() def test_cmd2args(): assert cmd2args([1, 2]) == [1, 2]
1,161
17.444444
41
py
pydeps
pydeps-master/tests/test_skinny_package.py
# -*- coding: utf-8 -*- from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_from_html5lib(): files = """ foo: - __init__.py - a.py: | from bar import py bar: - __init__.py """ with create_files(files) as workdir: assert simpledeps('foo', '--show-deps -LINFO -vv') == { 'bar -> foo.a', }
435
21.947368
63
py
pydeps
pydeps-master/tests/test_cli.py
# -*- coding: utf-8 -*- import os from pydeps.cli import error from pydeps.pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps, empty def test_output(tmpdir): files = """ unrelated: [] foo: - __init__.py - a.py: | from bar import b bar: - __init__.py - b.py """ with create_files(files) as workdir: assert os.getcwd() == workdir outname = os.path.join('unrelated', 'foo.svg') assert not os.path.exists(outname) pydeps(fname='foo', **empty('--noshow --show-dot', output=outname)) assert os.path.exists(outname) def test_rankdir_default(tmpdir, capsys): files = """ unrelated: [] foo: - __init__.py - a.py: | from bar import b bar: - __init__.py - b.py """ with create_files(files) as workdir: assert os.getcwd() == workdir outname = os.path.join('unrelated', 'foo.svg') pydeps(fname='foo', **empty('--noshow --show-dot', output=outname)) captured_stdout = capsys.readouterr().out assert 'rankdir = TB' in captured_stdout def test_error(capsys): """Test that error function prints reminder about missing inits on FileNotFoundErrors.""" try: error("[Errno 2] No such file or directory: 'foo'") except SystemExit: # because error invokes sys.exit(1), we have to catch it here, otherwise the test would always fail. pass else: # test should fail if error function doesn't raise assert False captured_stdout = capsys.readouterr().out assert "(Did you forget to include an __init__.py?)" in captured_stdout def test_rankdir_BT(tmpdir, capsys): files = """ unrelated: [] foo: - __init__.py - a.py: | from bar import b bar: - __init__.py - b.py """ with create_files(files) as workdir: assert os.getcwd() == workdir outname = os.path.join('unrelated', 'foo.svg') pydeps(fname='foo', **empty('--noshow --show-dot --rankdir=BT', output=outname)) captured_stdout = capsys.readouterr().out assert 'rankdir = BT' in captured_stdout
2,357
28.111111
108
py
pydeps
pydeps-master/tests/test_dirtree.py
# -*- coding: utf-8 -*- from pydeps.pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps import pytest @pytest.mark.skip(reason="TODO: fix this (issue #174)") def test_dirtree(): files = """ foo: - a: - __init__.py: '' - a.py: | from b.b import bval - b: - __init__.py: '' - b.py: | bval = 42 """ with create_files(files) as workdir: assert simpledeps('foo', '--show-deps -LINFO -vv') == { 'b.b -> a.a', }
633
24.36
63
py
pydeps
pydeps-master/tests/test_json.py
# -*- coding: utf-8 -*- import json import os from pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps, depgrf def test_dep2dot(): files = """ foo: - __init__.py - a.py: | from . import b - b.py """ with create_files(files) as workdir: g = depgrf("foo") d = json.loads(repr(g)) print(d) assert '__main__' in d['foo']['imported_by'] assert g.sources['foo.a'] == g.sources['foo.a'] assert str(g.sources['foo.a']).startswith('foo.a') assert 'foo.b' in repr(g.sources['foo.a'])
654
25.2
58
py
pydeps
pydeps-master/tests/test_config.py
# -*- coding: utf-8 -*- # from devtools import debug from pydeps.cli import parse_args from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_pydeps_config_ini(): files = """ config.ini: | [pydeps] rankdir = BT exclude = a c relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp', '--config=config.ini']) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_setupcfg_ini(): files = """ setup.cfg: | [pydeps] rankdir = BT exclude = a c relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp']) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_json(): files = """ pydeps.json: | { "pydeps": { "rankdir": "BT", "exclude": ["a", "c"] } } relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp', '--config=pydeps.json']) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_pydeps_yaml(): files = """ pydeps.yml: | pydeps: rankdir: BT exclude: - a - c relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp']) # debug(args) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_yaml(): files = """ config.yml: | pydeps: rankdir: BT exclude: - a - c relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp', '--config=config.yml']) # debug(args) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_pyproject_toml(): files = """ pyproject.toml: | [tool.pydeps] rankdir = "BT" exclude = ["a", "c"] relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp']) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c'] def test_pydeps_config_toml(): files = """ config.toml: | [tool.pydeps] rankdir = "BT" exclude = ["a", "c"] relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: args = parse_args(['relimp', '--config=config.toml']) assert args['rankdir'] == "BT" assert args['exclude'] == ['a', 'c']
4,035
23.760736
61
py
pydeps
pydeps-master/tests/filemaker.py
# -*- coding: utf-8 -*- from contextlib import contextmanager import os import shutil import tempfile import yaml class FilemakerBase(object): # pragma: nocover """Override marked methods to do something useful. Base class serves as a dry-run step generator. """ def __init__(self, root, fdef): self.fdef = fdef self.goto_root(root) self._makefiles(fdef) def goto_root(self, dirname): """override """ print("pushd", dirname) def makedir(self, dirname, content): """override, but call self.make_list(content) """ print("mkdir " + dirname) print("pushd " + dirname) self.make_list(content) print("popd") def make_file(self, filename, content): """override """ print("create file: %s %r" % (filename, content)) def make_empty_file(self, fname): """override """ print("touch", fname) def _make_empty_file(self, fname): if fname != 'empty': self.make_empty_file(fname) def make_list(self, lst): for item in lst: self._makefiles(item) def _makefiles(self, f): if isinstance(f, dict): for k, v in list(f.items()): if isinstance(v, list): self.makedir(dirname=k, content=v) elif isinstance(v, str): self.make_file(filename=k, content=v) else: # pragma: nocover raise ValueError("Unexpected:", k, v) elif isinstance(f, str): self._make_empty_file(f) elif isinstance(f, list): self.make_list(f) else: # pragma: nocover raise ValueError("Unknown type:", f) class Filemaker(FilemakerBase): def goto_root(self, dirname): os.chdir(dirname) def makedir(self, dirname, content): cwd = os.getcwd() os.mkdir(dirname) os.chdir(dirname) self.make_list(content) os.chdir(cwd) def make_file(self, filename, content): open(filename, 'w').write(content) def make_empty_file(self, fname): open(fname, 'w').close() @contextmanager def create_files(filedef, cleanup=True): fdef = yaml.safe_load(filedef) cwd = os.getcwd() tmpdir = os.path.realpath(tempfile.mkdtemp()) try: Filemaker(tmpdir, fdef) if not cleanup: print("TMPDIR =", tmpdir) yield tmpdir finally: os.chdir(cwd) if cleanup: shutil.rmtree(tmpdir, ignore_errors=True)
2,606
25.333333
76
py
pydeps
pydeps-master/tests/test_skip.py
# -*- coding: utf-8 -*- from __future__ import print_function from tests.filemaker import create_files from tests.simpledeps import simpledeps, depgrf def test_no_skip(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: print("plain", simpledeps('relimp')) assert simpledeps('relimp') == { 'relimp.b -> relimp.a', 'relimp.c -> relimp.b' } def test_skip_module_pattern(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: print("-x", simpledeps('relimp', '-x relimp.*')) assert simpledeps('relimp', '-x relimp.*') == set() def test_skip_exact_pattern(): files = """ relimp: - __init__.py - a.py: | from . import b from . import c - b.py: | from .c import d - c: - __init__.py - d.py """ with create_files(files) as workdir: print('plain', simpledeps('relimp')) assert simpledeps('relimp') == { 'relimp.b -> relimp.a', 'relimp.c.d -> relimp.b', 'relimp.c -> relimp.b', 'relimp.c -> relimp.a', } print('plain', simpledeps('relimp', '-xx relimp.c')) assert simpledeps('relimp', '-xx relimp.c') == { 'relimp.b -> relimp.a', 'relimp.c.d -> relimp.b', # this should remain # 'relimp.c -> relimp.b', # these should be filtered away.. # 'relimp.c -> relimp.a', } def test_skip_exact(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: print('-xx', simpledeps('relimp', '-xx relimp.c')) assert simpledeps('relimp', '-xx relimp.c') == { 'relimp.b -> relimp.a' } def test_skip_modules(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: assert simpledeps('relimp', '-x relimp.c') == { 'relimp.b -> relimp.a' } # g = depgrf('relimp', '-x relimp.c relimp.b') # print g # print simpledeps('relimp', '-x relimp.c relimp.b') # assert 1 # assert g == 42 def test_rawdeps(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: assert simpledeps('relimp', '--show-raw-deps -x relimp.c') == { 'relimp.b -> relimp.a' }
3,232
25.284553
72
py
pydeps
pydeps-master/tests/test_pyw.py
# -*- coding: utf-8 -*- import sys from tests.filemaker import create_files from tests.simpledeps import simpledeps import pytest @pytest.mark.skipif(sys.platform != 'win32', reason=".pyw files only exist on windows") def test_from_pyw(): files = """ baz.pyw: | import foo.a import bar foo: - __init__.py - a.py: | from bar import py bar: - __init__.py """ with create_files(files) as workdir: assert simpledeps('baz.pyw', '--show-deps -LINFO -vv') == { 'foo.a -> baz.pyw', 'foo -> baz.pyw', 'bar -> baz.pyw', 'bar -> foo.a', }
706
24.25
87
py
pydeps
pydeps-master/tests/test_cycles.py
# -*- coding: utf-8 -*- from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_cycle(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import a """ with create_files(files, cleanup=False) as workdir: print("WORKDIR:", workdir) deps = simpledeps('relimp') assert 'relimp.a -> relimp.b' in deps assert 'relimp.b -> relimp.a' in deps
518
24.95
55
py
pydeps
pydeps-master/tests/test_colors.py
# -*- coding: utf-8 -*- import os from pydeps.colors import rgb2css, brightness, brightnessdiff, colordiff, name2rgb, foreground red = (255, 0, 0) green = (0, 255, 0) yellow = (0, 255, 255) blue = (0, 0, 255) black = (0, 0, 0) white = (255, 255, 255) def test_rgb2css(): assert rgb2css(red) == '#ff0000' assert rgb2css(green) == '#00ff00' assert rgb2css(yellow) == '#00ffff' assert rgb2css(blue) == '#0000ff' assert rgb2css(black) == '#000000' assert rgb2css(white) == '#ffffff' def test_brightness(): assert brightnessdiff(yellow, white) < brightnessdiff(yellow, black) def test_colordiff(): assert colordiff(blue, yellow) < colordiff(blue, red) def test_foreground(): assert foreground(black, red, green, yellow, blue, black, white) == white assert foreground(black, red, green, yellow, blue, black) == yellow assert foreground(black, red, green, blue, black) == green assert foreground(black, red, blue, black) == red assert foreground(black, blue, black) == blue def test_name2rgb(): def fg(name): return foreground( name2rgb(13), black, white) assert fg('hello') == fg('hello.world')
1,194
24.978261
94
py
pydeps
pydeps-master/tests/test_package_names.py
from pydeps.package_names import find_package_names def test_find_package_names(): packages = find_package_names() assert 'pip' in packages assert 'pytest' in packages
184
17.5
51
py
pydeps
pydeps-master/tests/test_relative_imports.py
# -*- coding: utf-8 -*- import os import sys from tests.filemaker import create_files from tests.simpledeps import simpledeps import pytest def test_relative_imports(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py """ with create_files(files) as workdir: assert simpledeps('relimp') == {'relimp.b -> relimp.a'} def test_relative_imports2(): files = """ relimp: - __init__.py - a.py: | from . import b - b.py: | from . import c - c.py """ with create_files(files) as workdir: deps = simpledeps('relimp') assert 'relimp.c -> relimp.b' in deps assert 'relimp.b -> relimp.a' in deps def test_relative_imports3(): files = """ relimp: - __init__.py - a.py: | from .b import c - b.py """ with create_files(files) as workdir: assert simpledeps('relimp') == {'relimp.b -> relimp.a'} def test_relative_imports_same_name_with_std(): files = """ relimp: - __init__.py - io.py: | import io """ with create_files(files) as workdir: if sys.version_info < (3,): # pragma: nocover deps = {'relimp.io -> relimp.io'} else: # pragma: nocover deps = {'io -> relimp.io'} assert simpledeps('relimp', '--pylib') == deps def test_relative_imports_same_name_with_std_future(): files = """ relimp: - __init__.py - io.py: | from __future__ import absolute_import import io """ with create_files(files) as workdir: deps = { '__future__ -> relimp.io', 'io -> relimp.io' } assert simpledeps('relimp', '--pylib') == deps def test_pydeps_colors(): files = """ pdeps: - __init__.py - colors.py: | import colorsys - depgraph.py: | import json import pprint import enum from . import colors """ with create_files(files, cleanup=False) as workdir: assert simpledeps('pdeps', '-x enum') == { 'pdeps.colors -> pdeps.depgraph', } def test_hierarchy(): files = """ relimp: - __init__.py - a: - __init__.py - amodule.py: | from ..b import bmodule - b: - __init__.py - bmodule.py """ with create_files(files, cleanup=True) as workdir: os.system("tree " + workdir) deps = simpledeps('relimp') assert 'relimp.b -> relimp.a.amodule' in deps assert 'relimp.b.bmodule -> relimp.a.amodule' in deps
2,990
24.784483
68
py
pydeps
pydeps-master/tests/simpledeps.py
import pydeps.cli from pydeps import pydeps from pydeps.py2depgraph import py2dep from pydeps.target import Target def empty(args="", **kw): args = pydeps.cli.parse_args(['foo', '--no-config'] + args.split()) args.pop('fname') args.update(kw) return args def depgrf(item, args=""): t = Target(item) # print("TARGET:", t) with t.chdir_work(): res = py2dep(t, **empty(args)) # print("DEPGRPH:", res) return res def simpledeps(item, args=""): return {"%s -> %s" % (a.name, b.name) for a, b in depgrf(item, args)}
572
21.92
73
py
pydeps
pydeps-master/tests/__init__.py
# -*- coding: utf-8 -*-
26
5.75
23
py
pydeps
pydeps-master/tests/test_file.py
# -*- coding: utf-8 -*- import os from pydeps.py2depgraph import py2dep from pydeps.pydeps import _pydeps from pydeps.target import Target from tests.filemaker import create_files from tests.simpledeps import empty, simpledeps def test_file(): files = """ a.py: | import collections """ with create_files(files) as workdir: assert simpledeps('a.py') == set() def test_file_in_sub_directory(): files = """ foo: - a: - b.py: | import c - c.py: "" """ with create_files(files) as workdir: # os.chdir('foo') # t = Target('a/b.py') # with t.chdir_work(): # print("CURDIR:", os.getcwd(), os.listdir('.')) # # deps = py2dep(t, **empty(log="DEBUG", no_dot=True, show_deps=True)) # deps = py2dep(t, **empty()) # print("DEPSx:", deps) # rels = {f'{a.name} -> {b.name}' for a, b in deps} # print("RELS:", rels) # assert 'c -> b.py' in rels assert 'c -> b.py' in simpledeps('foo/a/b.py') # , '-L DEBUG') def test_file_in_directory(): files = """ - a: - b.py: | import c - c.py: "" """ with create_files(files) as workdir: assert 'c -> b.py' in simpledeps('a/b.py') def test_file_pylib(): files = """ a.py: | import collections """ with create_files(files) as workdir: assert 'collections -> a.py' in simpledeps('a.py', '--pylib') def test_file_pyliball(): files = """ a.py: | import collections """ with create_files(files) as workdir: assert 'collections -> a.py' in simpledeps('a.py', '--pylib --pylib-all')
1,816
24.591549
83
py
pydeps
pydeps-master/tests/test_py2dep.py
# -*- coding: utf-8 -*- import os # from devtools import debug from pydeps.pydeps import call_pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_py2depgraph(capsys): files = """ - a.py: | import b - b.py """ with create_files(files) as workdir: assert simpledeps('a.py') == {'b -> a.py'} def test_pydeps_api(): files = """ - a.py: | import b - b.py """ with create_files(files) as workdir: call_pydeps('a.py', no_show=True, show_dot=True, dot_out='a.dot') # debug(os.getcwd()) # debug(os.listdir('.')) with open('a.dot') as f: dot = f.read() # debug(dot) assert 'b -> a_py' in dot
783
22.757576
73
py
pydeps
pydeps-master/tests/test_cluster.py
import os from pydeps.cli import parse_args from pydeps.pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_cluster(): files = """ - bar_module: - __init__.py: '' - bar_a: - __init__.py: '' - a.py: | from bar_module.bar_a import aa - aa.py: '' - bar_b: - __init__.py: '' - b.py: from bar_module.bar_a import aa - bar_c: - __init__.py: '' - c.py: | from bar_module.bar_a import a - foo_module: - __init__.py: '' - foo_a: - __init__.py: '' - a.py: | from . import aa - aa.py: '' - foo_c: - __init__.py: '' - c.py: |- from foo_module.foo_a import aa from bar_module.bar_c import c """ with create_files(files) as workdir: args = parse_args(['foo_module', '--no-config', '--show-deps', '--cluster', '--max-cluster-size=100', '--show-dot', '--dot-output', 'output.dot', '--no-show', '-LINFO', '-vv']) pydeps(**args) assert 'output.dot' in os.listdir(workdir) dot_output = open('output.dot').read() assert 'subgraph cluster_bar_module' in dot_output assert 'ltail="cluster_bar_module"' in dot_output
1,535
32.391304
109
py
pydeps
pydeps-master/tests/test_funny_names.py
# -*- coding: utf-8 -*- from pydeps.pydeps import pydeps from tests.filemaker import create_files from tests.simpledeps import simpledeps def test_from_html5lib(): files = """ foo: - __init__.py - a.py: | from bar import py bar: - __init__.py - py.py: | barpy = 42 """ with create_files(files) as workdir: assert simpledeps('foo', '--show-deps -LINFO -vv') == { 'bar -> foo.a', 'bar.py -> foo.a' } def test_multidot(): files = """ foo.bar.py: | from math import pi """ with create_files(files) as workdir: assert simpledeps('foo.bar.py', '--show-deps --pylib -LINFO -vv') == { 'math -> foo.bar.py', }
813
22.941176
78
py
pydeps
pydeps-master/tests/test_render_context.py
# -*- coding: utf-8 -*- from pydeps.render_context import RenderContext, Rankdir def test_render_context(): ctx = RenderContext() with ctx.graph(): ctx.write_rule('a', 'b') assert 'a -> b' in ctx.text() assert 'b -> a' not in ctx.text() assert 'rankdir = TB' in ctx.text() def test_render_context_reverse(): # verify that rankdir is reversed in RenderBuffer, not RenderContext. ctx = RenderContext(reverse=True, rankdir=Rankdir.BOTTOM_TOP) with ctx.graph(): ctx.write_rule('a', 'b') assert 'b -> a' in ctx.text() assert 'a -> b' not in ctx.text() assert 'rankdir = BT' in ctx.text() def test_render_context_rankdir(): ctx = RenderContext(rankdir=Rankdir.LEFT_RIGHT) with ctx.graph(): pass text = ctx.text() assert 'rankdir = LR' in text
827
26.6
73
py
pydeps
pydeps-master/docs/conf.py
# -*- coding: utf-8 -*- # # pydeps documentation build configuration file, created by # sphinx-quickstart on Thu Feb 13 18:59:19 2014. # # This file is execfile()d with the current directory set to its containing dir. # # Note that not all possible configuration values are present in this # autogenerated file. # # All configuration values have a default; values that are commented out # serve to show the default. import sys, os, datetime # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. sys.path.insert(0, os.path.abspath('..')) # -- General configuration ----------------------------------------------------- # If your documentation needs a minimal Sphinx version, state it here. #needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be extensions # coming with Sphinx (named 'sphinx.ext.*') or your custom ones. extensions = [ 'sphinx.ext.autodoc', 'sphinx.ext.intersphinx', 'sphinx.ext.todo', 'sphinx.ext.coverage', 'sphinx.ext.viewcode' ] # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix of source filenames. source_suffix = '.rst' # The encoding of source files. #source_encoding = 'utf-8-sig' # The master toctree document. master_doc = 'index' # General information about the project. project = u'pydeps' copyright = u'2014 - %s Bjorn Pettersen' % datetime.date.today().year # The version info for the project you're documenting, acts as replacement for # |version| and |release|, also used in various other places throughout the # built documents. # # The short X.Y version. version = '1.12.12' # The full version, including alpha/beta/rc tags. release = '1.12.12' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. #language = None # There are two options for replacing |today|: either, you set today to some # non-false value, then it is used: #today = '' # Else, today_fmt is used as the format for a strftime call. #today_fmt = '%B %d, %Y' # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. exclude_patterns = ['_build'] # The reST default role (used for this markup: `text`) to use for all documents. #default_role = None # If true, '()' will be appended to :func: etc. cross-reference text. #add_function_parentheses = True # If true, the current module name will be prepended to all description # unit titles (such as .. function::). #add_module_names = True # If true, sectionauthor and moduleauthor directives will be shown in the # output. They are ignored by default. #show_authors = False # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # A list of ignored prefixes for module index sorting. #modindex_common_prefix = [] # If true, keep warnings as "system message" paragraphs in the built documents. keep_warnings = False # -- Options for HTML output --------------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. html_theme = 'default' # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. #html_theme_options = {} # Add any paths that contain custom themes here, relative to this directory. #html_theme_path = [] # The name for this set of Sphinx documents. If None, it defaults to # "<project> v<release> documentation". #html_title = None # A shorter title for the navigation bar. Default is the same as html_title. #html_short_title = None # The name of an image file (relative to this directory) to place at the top # of the sidebar. #html_logo = None # The name of an image file (within the static path) to use as favicon of the # docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32 # pixels large. #html_favicon = None # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] # If not '', a 'Last updated on:' timestamp is inserted at every page bottom, # using the given strftime format. #html_last_updated_fmt = '%b %d, %Y' # If true, SmartyPants will be used to convert quotes and dashes to # typographically correct entities. #html_use_smartypants = True # Custom sidebar templates, maps document names to template names. #html_sidebars = {} # Additional templates that should be rendered to pages, maps page names to # template names. #html_additional_pages = {} # If false, no module index is generated. #html_domain_indices = True # If false, no index is generated. #html_use_index = True # If true, the index is split into individual pages for each letter. #html_split_index = False # If true, links to the reST sources are added to the pages. #html_show_sourcelink = True # If true, "Created using Sphinx" is shown in the HTML footer. Default is True. #html_show_sphinx = True # If true, "(C) Copyright ..." is shown in the HTML footer. Default is True. #html_show_copyright = True # If true, an OpenSearch description file will be output, and all pages will # contain a <link> tag referring to it. The value of this option must be the # base URL from which the finished HTML is served. #html_use_opensearch = '' # This is the file name suffix for HTML files (e.g. ".xhtml"). #html_file_suffix = None # Output file base name for HTML help builder. htmlhelp_basename = 'pydepsdoc' # -- Options for LaTeX output -------------------------------------------------- latex_elements = { # The paper size ('letterpaper' or 'a4paper'). #'papersize': 'letterpaper', # The font size ('10pt', '11pt' or '12pt'). #'pointsize': '10pt', # Additional stuff for the LaTeX preamble. #'preamble': '', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, author, documentclass [howto/manual]). latex_documents = [ ('index', 'pydeps.tex', u'pydeps Documentation', u'', 'manual'), ] # The name of an image file (relative to this directory) to place at the top of # the title page. #latex_logo = None # For "manual" documents, if this is true, then toplevel headings are parts, # not chapters. #latex_use_parts = False # If true, show page references after internal links. #latex_show_pagerefs = False # If true, show URL addresses after external links. #latex_show_urls = False # Documents to append as an appendix to all manuals. #latex_appendices = [] # If false, no module index is generated. #latex_domain_indices = True # -- Options for manual page output -------------------------------------------- # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ ('index', 'pydeps', u'pydeps Documentation', [u''], 1) ] # If true, show URL addresses after external links. #man_show_urls = False # -- Options for Texinfo output ------------------------------------------------ # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ ('index', 'pydeps', u'pydeps Documentation', u'', 'pydeps', 'command line access to ConfigParser.', 'Miscellaneous'), ] # Documents to append as an appendix to all manuals. #texinfo_appendices = [] # If false, no module index is generated. #texinfo_domain_indices = True # How to display URL addresses: 'footnote', 'no', or 'inline'. #texinfo_show_urls = 'footnote' # If true, do not generate a @detailmenu in the "Top" node's menu. #texinfo_no_detailmenu = False # Example configuration for intersphinx: refer to the Python standard library. intersphinx_mapping = {'python': ('https://docs.python.org/3/', None)}
8,225
30.760618
80
py
pydeps
pydeps-master/docs/module-finder-archive/mf_36.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import sys import types import struct import warnings with warnings.catch_warnings(): warnings.simplefilter('ignore', DeprecationWarning) import imp LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=[], replace_paths=[]): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes self.replace_paths = replace_paths self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with open(pathname) as fp: stuff = ("", "r", imp.PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with open(pathname) as fp: stuff = (ext, "r", imp.PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == imp.PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == imp.PY_SOURCE: co = compile(fp.read()+'\n', pathname, 'exec') elif type == imp.PY_COMPILED: try: marshal_data = importlib._bootstrap_external._validate_bytecode_header(fp.read()) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(marshal_data) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: if sub in self.badmodules: self._add_badmodule(sub, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) fullname = name + "." + sub self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", imp.C_BUILTIN)) path = self.path return imp.find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return types.CodeType(co.co_argcount, co.co_kwonlyargcount, co.co_nlocals, co.co_stacksize, co.co_flags, co.co_code, tuple(consts), co.co_names, co.co_varnames, new_filename, co.co_name, co.co_firstlineno, co.co_lnotab, co.co_freevars, co.co_cellvars) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
23,027
35.321767
97
py
pydeps
pydeps-master/docs/module-finder-archive/mf_310.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import io import sys LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Old imp constants: _SEARCH_ERROR = 0 _PY_SOURCE = 1 _PY_COMPILED = 2 _C_EXTENSION = 3 _PKG_DIRECTORY = 5 _C_BUILTIN = 6 _PY_FROZEN = 7 # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname def _find_module(name, path=None): """An importlib reimplementation of imp.find_module (for our purposes).""" # It's necessary to clear the caches for our Finder first, in case any # modules are being added/deleted/modified at runtime. In particular, # test_modulefinder.py changes file tree contents in a cache-breaking way: importlib.machinery.PathFinder.invalidate_caches() spec = importlib.machinery.PathFinder.find_spec(name, path) if spec is None: raise ImportError("No module named {name!r}".format(name=name), name=name) # Some special cases: if spec.loader is importlib.machinery.BuiltinImporter: return None, None, ("", "", _C_BUILTIN) if spec.loader is importlib.machinery.FrozenImporter: return None, None, ("", "", _PY_FROZEN) file_path = spec.origin if spec.loader.is_package(name): return None, os.path.dirname(file_path), ("", "", _PKG_DIRECTORY) if isinstance(spec.loader, importlib.machinery.SourceFileLoader): kind = _PY_SOURCE elif isinstance(spec.loader, importlib.machinery.ExtensionFileLoader): kind = _C_EXTENSION elif isinstance(spec.loader, importlib.machinery.SourcelessFileLoader): kind = _PY_COMPILED else: # Should never happen. return None, None, ("", "", _SEARCH_ERROR) file = io.open_code(file_path) suffix = os.path.splitext(file_path)[-1] return file, file_path, (suffix, "rb", kind) class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=None, replace_paths=None): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes if excludes is not None else [] self.replace_paths = replace_paths if replace_paths is not None else [] self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with io.open_code(pathname) as fp: stuff = ("", "rb", _PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with io.open_code(pathname) as fp: stuff = (ext, "rb", _PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == _PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == _PY_SOURCE: co = compile(fp.read(), pathname, 'exec') elif type == _PY_COMPILED: try: data = fp.read() importlib._bootstrap_external._classify_pyc(data, fqname, {}) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(memoryview(data)[16:]) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) except SyntaxError as msg: self.msg(2, "SyntaxError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: fullname = name + "." + sub if fullname in self.badmodules: self._add_badmodule(fullname, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", _C_BUILTIN)) path = self.path return _find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return co.replace(co_consts=tuple(consts), co_filename=new_filename) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
24,401
34.571429
86
py
pydeps
pydeps-master/docs/module-finder-archive/mf_27.py
"""Find modules used by a script, using introspection.""" from __future__ import generators import dis import imp import marshal import os import sys import types import struct if hasattr(sys.__stdout__, "newlines"): READ_MODE = "U" # universal line endings else: # Python < 2.3 compatibility, no longer strictly required READ_MODE = "r" LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL HAVE_ARGUMENT = dis.HAVE_ARGUMENT EXTENDED_ARG = dis.EXTENDED_ARG def _unpack_opargs(code): # enumerate() is not an option, since we sometimes process # multiple elements on a single pass through the loop extended_arg = 0 n = len(code) i = 0 while i < n: op = ord(code[i]) offset = i i = i+1 arg = None if op >= HAVE_ARGUMENT: arg = ord(code[i]) + ord(code[i+1])*256 + extended_arg extended_arg = 0 i = i+2 if op == EXTENDED_ARG: extended_arg = arg*65536 yield (offset, op, arg) # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): paths = packagePathMap.get(packagename, []) paths.append(path) packagePathMap[packagename] = paths replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around the # way the _xmlplus package injects itself under the name "xml" into # sys.modules at runtime by calling ReplacePackage("_xmlplus", "xml") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=[], replace_paths=[]): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes self.replace_paths = replace_paths self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print " ", print str, for arg in args: print repr(arg), print def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with open(pathname, READ_MODE) as fp: stuff = ("", "r", imp.PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with open(pathname, READ_MODE) as fp: stuff = (ext, "r", imp.PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError, "relative importpath too deep" pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError, "No module named " + qname def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError, "No module named " + mname self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError, "No module named " + subname def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc", ".pyo"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] for triple in imp.get_suffixes(): suffixes.append(triple[0]) for dir in m.__path__: try: names = os.listdir(dir) except os.error: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == imp.PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == imp.PY_SOURCE: co = compile(fp.read()+'\n', pathname, 'exec') elif type == imp.PY_COMPILED: if fp.read(4) != imp.get_magic(): self.msgout(2, "raise ImportError: Bad magic number", pathname) raise ImportError, "Bad magic number in %s" % pathname fp.read(4) co = marshal.load(fp) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError, msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: if sub in self.badmodules: self._add_badmodule(sub, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError, msg: self.msg(2, "ImportError:", str(msg)) fullname = name + "." + sub self._add_badmodule(fullname, caller) def scan_opcodes(self, co, unpack = struct.unpack): # Scan the code, and yield 'interesting' opcode combinations # Version for Python 2.4 and older code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in _unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if c in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 1 and opargs[i-1][0] == LOAD_CONST): fromlist = consts[opargs[i-1][1]] yield "import", (fromlist, names[oparg]) continue def scan_opcodes_25(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in _unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == -1: # normal import yield "import", (fromlist, names[oparg]) elif level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code if sys.version_info >= (2, 5): scanner = self.scan_opcodes_25 else: scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what in ("import", "absolute_import"): fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] if what == "absolute_import": level = 0 else: level = -1 self._safe_import_hook(name, m, fromlist, level=level) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) if fp: fp.close() return m def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError, name if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", imp.C_BUILTIN)) path = self.path return imp.find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print print " %-25s %s" % ("Name", "File") print " %-25s %s" % ("----", "----") # Print modules found keys = self.modules.keys() keys.sort() for key in keys: m = self.modules[key] if m.__path__: print "P", else: print "m", print "%-25s" % key, m.__file__ or "" # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print print "Missing modules:" for name in missing: mods = self.badmodules[name].keys() mods.sort() print "?", name, "imported from", ', '.join(mods) # Print modules that may be missing, but then again, maybe not... if maybe: print print "Submodules that appear to be missing, but could also be", print "global names in the parent package:" for name in maybe: mods = self.badmodules[name].keys() mods.sort() print "?", name, "imported from", ', '.join(mods) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return types.CodeType(co.co_argcount, co.co_nlocals, co.co_stacksize, co.co_flags, co.co_code, tuple(consts), co.co_names, co.co_varnames, new_filename, co.co_name, co.co_firstlineno, co.co_lnotab, co.co_freevars, co.co_cellvars) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error, msg: print msg return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print "path:" for item in path: print " ", repr(item) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print "\n[interrupt]"
24,461
34.973529
86
py
pydeps
pydeps-master/docs/module-finder-archive/mf_39.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import io import sys LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Old imp constants: _SEARCH_ERROR = 0 _PY_SOURCE = 1 _PY_COMPILED = 2 _C_EXTENSION = 3 _PKG_DIRECTORY = 5 _C_BUILTIN = 6 _PY_FROZEN = 7 # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname def _find_module(name, path=None): """An importlib reimplementation of imp.find_module (for our purposes).""" # It's necessary to clear the caches for our Finder first, in case any # modules are being added/deleted/modified at runtime. In particular, # test_modulefinder.py changes file tree contents in a cache-breaking way: importlib.machinery.PathFinder.invalidate_caches() spec = importlib.machinery.PathFinder.find_spec(name, path) if spec is None: raise ImportError("No module named {name!r}".format(name=name), name=name) # Some special cases: if spec.loader is importlib.machinery.BuiltinImporter: return None, None, ("", "", _C_BUILTIN) if spec.loader is importlib.machinery.FrozenImporter: return None, None, ("", "", _PY_FROZEN) file_path = spec.origin if spec.loader.is_package(name): return None, os.path.dirname(file_path), ("", "", _PKG_DIRECTORY) if isinstance(spec.loader, importlib.machinery.SourceFileLoader): kind = _PY_SOURCE elif isinstance(spec.loader, importlib.machinery.ExtensionFileLoader): kind = _C_EXTENSION elif isinstance(spec.loader, importlib.machinery.SourcelessFileLoader): kind = _PY_COMPILED else: # Should never happen. return None, None, ("", "", _SEARCH_ERROR) file = io.open_code(file_path) suffix = os.path.splitext(file_path)[-1] return file, file_path, (suffix, "rb", kind) class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=None, replace_paths=None): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes if excludes is not None else [] self.replace_paths = replace_paths if replace_paths is not None else [] self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with io.open_code(pathname) as fp: stuff = ("", "rb", _PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with io.open_code(pathname) as fp: stuff = (ext, "rb", _PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == _PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == _PY_SOURCE: co = compile(fp.read(), pathname, 'exec') elif type == _PY_COMPILED: try: data = fp.read() importlib._bootstrap_external._classify_pyc(data, fqname, {}) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(memoryview(data)[16:]) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) except SyntaxError as msg: self.msg(2, "SyntaxError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: fullname = name + "." + sub if fullname in self.badmodules: self._add_badmodule(fullname, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", _C_BUILTIN)) path = self.path return _find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return co.replace(co_consts=tuple(consts), co_filename=new_filename) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
24,401
34.571429
86
py
pydeps
pydeps-master/docs/module-finder-archive/mf_pydeps_orig.py
"""Find modules used by a script, using introspection.""" # This module should be kept compatible with Python 2.2, see PEP 291. from __future__ import print_function from __future__ import generators import dis import imp import marshal import os import sys import types import struct READ_MODE = "r" LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG HAVE_ARGUMENT = dis.HAVE_ARGUMENT # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=[], replace_paths=[]): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes self.replace_paths = replace_paths self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with open(pathname) as fp: stuff = ("", "r", imp.PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with open(pathname) as fp: stuff = (ext, "r", imp.PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook: name(%s) caller(%s) fromlist(%s) level(%s)" % (name, caller, fromlist, level)) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) if q.shortname in ('__future__', 'future'): # the future package causes recursion overflow return None m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] for triple in imp.get_suffixes(): suffixes.append(triple[0]) for dir in m.__path__: try: names = os.listdir(dir) except os.error: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module: partname(%s) fqname(%s) parent(%s)" % (partname, fqname, parent)) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): # fqname = dotted module name we're loading suffix, mode, kind = file_info kstr = { imp.PKG_DIRECTORY: 'PKG_DIRECTORY', imp.PY_SOURCE: 'PY_SOURCE', imp.PY_COMPILED: 'PY_COMPILED', }.get(kind, 'unknown-kind') self.msgin(2, "load_module(%s) fqname=%s, fp=%s, pathname=%s" % (kstr, fqname, fp and "fp", pathname)) if kind == imp.PKG_DIRECTORY: module = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", module) return module if kind == imp.PY_SOURCE: co = compile( fp.read() + '\n', pathname, 'exec', # compile code block dont_inherit=True # don't inherit future statements from current environment ) elif kind == imp.PY_COMPILED: # a .pyc file is a binary file containing only thee things: # 1. a four-byte magic number # 2. a four byte modification timestamp, and # 3. a Marshalled code object # from: https://nedbatchelder.com/blog/200804/the_structure_of_pyc_files.html if fp.read(4) != imp.get_magic(): self.msgout(2, "raise ImportError: Bad magic number", pathname) raise ImportError("Bad magic number in %s" % pathname) fp.read(4) # skip modification timestamp co = marshal.load(fp) # load marshalled code object. else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: if sub in self.badmodules: self._add_badmodule(sub, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) fullname = name + "." + sub self._add_badmodule(fullname, caller) def scan_opcodes(self, co, unpack=struct.unpack): # pragma: nocover # Scan the code, and yield 'interesting' opcode combinations # Version for Python 2.4 and older code = co.co_code names = co.co_names consts = co.co_consts while code: c = code[0] if c in STORE_OPS: oparg, = unpack('<H', code[1:3]) yield "store", (names[oparg],) code = code[3:] continue if c == LOAD_CONST and code[3] == IMPORT_NAME: oparg_1, oparg_2 = unpack('<xHxH', code[:6]) yield "import", (consts[oparg_1], names[oparg_2]) code = code[6:] continue if c >= HAVE_ARGUMENT: code = code[3:] else: code = code[1:] def scan_opcodes_25(self, co, unpack = struct.unpack): # Scan the code, and yield 'interesting' opcode combinations # Python 2.5 version (has absolute and relative imports) code = co.co_code names = co.co_names consts = co.co_consts LOAD_LOAD_AND_IMPORT = LOAD_CONST + LOAD_CONST + IMPORT_NAME while code: c = code[0] if c in STORE_OPS: oparg, = unpack('<H', code[1:3]) yield "store", (names[oparg],) code = code[3:] continue if code[:9:3] == LOAD_LOAD_AND_IMPORT: oparg_1, oparg_2, oparg_3 = unpack('<xHxHxH', code[:9]) level = consts[oparg_1] if level == -1: # normal import yield "import", (consts[oparg_2], names[oparg_3]) elif level == 0: # absolute import yield "absolute_import", (consts[oparg_2], names[oparg_3]) else: # relative import yield "relative_import", (level, consts[oparg_2], names[oparg_3]) code = code[9:] continue if c >= HAVE_ARGUMENT: code = code[3:] else: code = code[1:] def scan_opcodes_34(self, co): i = 0 bytecode = list(dis.Bytecode(co)) while i < len(bytecode): if ( bytecode[i].opname == "LOAD_CONST" and bytecode[i + 1].opname == "LOAD_CONST" and bytecode[i + 2].opname == "IMPORT_NAME" ): level = bytecode[i].argval fromlist = bytecode[i + 1].argval import_name = bytecode[i + 2].argval if level == 0: yield "absolute_import", (fromlist, import_name) else: yield "relative_import", (level, fromlist, import_name) i += 2 i += 1 def scan_code(self, co, m): code = co.co_code if sys.version_info >= (3, 4): scanner = self.scan_opcodes_34 elif sys.version_info >= (2, 5): scanner = self.scan_opcodes_25 else: scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what in ("import", "absolute_import"): fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] if what == "absolute_import": level = 0 else: level = -1 self._safe_import_hook(name, m, fromlist, level=level) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) # m is still the caller here... [bp] self._safe_import_hook(parent.__name__, m, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", imp.C_BUILTIN)) path = self.path return imp.find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return types.CodeType(co.co_argcount, co.co_nlocals, co.co_stacksize, co.co_flags, co.co_code, tuple(consts), co.co_names, co.co_varnames, new_filename, co.co_name, co.co_firstlineno, co.co_lnotab, co.co_freevars, co.co_cellvars) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
26,278
35.651325
112
py
pydeps
pydeps-master/docs/module-finder-archive/mf_next.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import io import sys # Old imp constants: _SEARCH_ERROR = 0 _PY_SOURCE = 1 _PY_COMPILED = 2 _C_EXTENSION = 3 _PKG_DIRECTORY = 5 _C_BUILTIN = 6 _PY_FROZEN = 7 # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname def _find_module(name, path=None): """An importlib reimplementation of imp.find_module (for our purposes).""" # It's necessary to clear the caches for our Finder first, in case any # modules are being added/deleted/modified at runtime. In particular, # test_modulefinder.py changes file tree contents in a cache-breaking way: importlib.machinery.PathFinder.invalidate_caches() spec = importlib.machinery.PathFinder.find_spec(name, path) if spec is None: raise ImportError("No module named {name!r}".format(name=name), name=name) # Some special cases: if spec.loader is importlib.machinery.BuiltinImporter: return None, None, ("", "", _C_BUILTIN) if spec.loader is importlib.machinery.FrozenImporter: return None, None, ("", "", _PY_FROZEN) file_path = spec.origin if spec.loader.is_package(name): return None, os.path.dirname(file_path), ("", "", _PKG_DIRECTORY) if isinstance(spec.loader, importlib.machinery.SourceFileLoader): kind = _PY_SOURCE elif isinstance(spec.loader, importlib.machinery.ExtensionFileLoader): kind = _C_EXTENSION elif isinstance(spec.loader, importlib.machinery.SourcelessFileLoader): kind = _PY_COMPILED else: # Should never happen. return None, None, ("", "", _SEARCH_ERROR) file = io.open_code(file_path) suffix = os.path.splitext(file_path)[-1] return file, file_path, (suffix, "rb", kind) class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=None, replace_paths=None): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes if excludes is not None else [] self.replace_paths = replace_paths if replace_paths is not None else [] self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with io.open_code(pathname) as fp: stuff = ("", "rb", _PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with io.open_code(pathname) as fp: stuff = (ext, "rb", _PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == _PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == _PY_SOURCE: co = compile(fp.read(), pathname, 'exec') elif type == _PY_COMPILED: try: data = fp.read() importlib._bootstrap_external._classify_pyc(data, fqname, {}) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(memoryview(data)[16:]) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) except SyntaxError as msg: self.msg(2, "SyntaxError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: fullname = name + "." + sub if fullname in self.badmodules: self._add_badmodule(fullname, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations for name in dis._find_store_names(co): yield "store", (name,) for name, level, fromlist in dis._find_imports(co): if level == 0: # absolute import yield "absolute_import", (fromlist, name) else: # relative import yield "relative_import", (level, fromlist, name) def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", _C_BUILTIN)) path = self.path return _find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return co.replace(co_consts=tuple(consts), co_filename=new_filename) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
23,699
34.532234
86
py
pydeps
pydeps-master/docs/module-finder-archive/mf_35.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import sys import types import struct import warnings with warnings.catch_warnings(): warnings.simplefilter('ignore', PendingDeprecationWarning) import imp LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=[], replace_paths=[]): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes self.replace_paths = replace_paths self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with open(pathname) as fp: stuff = ("", "r", imp.PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with open(pathname) as fp: stuff = (ext, "r", imp.PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == imp.PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == imp.PY_SOURCE: co = compile(fp.read()+'\n', pathname, 'exec') elif type == imp.PY_COMPILED: try: marshal_data = importlib._bootstrap_external._validate_bytecode_header(fp.read()) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(marshal_data) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: if sub in self.badmodules: self._add_badmodule(sub, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) fullname = name + "." + sub self._add_badmodule(fullname, caller) def scan_opcodes_25(self, co, unpack = struct.unpack): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes_25 for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", imp.C_BUILTIN)) path = self.path return imp.find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return types.CodeType(co.co_argcount, co.co_kwonlyargcount, co.co_nlocals, co.co_stacksize, co.co_flags, co.co_code, tuple(consts), co.co_names, co.co_varnames, new_filename, co.co_name, co.co_firstlineno, co.co_lnotab, co.co_freevars, co.co_cellvars) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
23,085
35.355906
97
py
pydeps
pydeps-master/docs/module-finder-archive/__init__.py
0
0
0
py
pydeps
pydeps-master/docs/module-finder-archive/mf_38.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import io import sys import types import warnings LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Old imp constants: _SEARCH_ERROR = 0 _PY_SOURCE = 1 _PY_COMPILED = 2 _C_EXTENSION = 3 _PKG_DIRECTORY = 5 _C_BUILTIN = 6 _PY_FROZEN = 7 # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname def _find_module(name, path=None): """An importlib reimplementation of imp.find_module (for our purposes).""" # It's necessary to clear the caches for our Finder first, in case any # modules are being added/deleted/modified at runtime. In particular, # test_modulefinder.py changes file tree contents in a cache-breaking way: importlib.machinery.PathFinder.invalidate_caches() spec = importlib.machinery.PathFinder.find_spec(name, path) if spec is None: raise ImportError("No module named {name!r}".format(name=name), name=name) # Some special cases: if spec.loader is importlib.machinery.BuiltinImporter: return None, None, ("", "", _C_BUILTIN) if spec.loader is importlib.machinery.FrozenImporter: return None, None, ("", "", _PY_FROZEN) file_path = spec.origin if spec.loader.is_package(name): return None, os.path.dirname(file_path), ("", "", _PKG_DIRECTORY) if isinstance(spec.loader, importlib.machinery.SourceFileLoader): kind = _PY_SOURCE elif isinstance(spec.loader, importlib.machinery.ExtensionFileLoader): kind = _C_EXTENSION elif isinstance(spec.loader, importlib.machinery.SourcelessFileLoader): kind = _PY_COMPILED else: # Should never happen. return None, None, ("", "", _SEARCH_ERROR) file = io.open_code(file_path) suffix = os.path.splitext(file_path)[-1] return file, file_path, (suffix, "rb", kind) class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=None, replace_paths=None): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes if excludes is not None else [] self.replace_paths = replace_paths if replace_paths is not None else [] self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with io.open_code(pathname) as fp: stuff = ("", "rb", _PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with io.open_code(pathname) as fp: stuff = (ext, "rb", _PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == _PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == _PY_SOURCE: co = compile(fp.read(), pathname, 'exec') elif type == _PY_COMPILED: try: data = fp.read() importlib._bootstrap_external._classify_pyc(data, fqname, {}) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(memoryview(data)[16:]) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) except SyntaxError as msg: self.msg(2, "SyntaxError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: fullname = name + "." + sub if fullname in self.badmodules: self._add_badmodule(fullname, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", _C_BUILTIN)) path = self.path return _find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return co.replace(co_consts=tuple(consts), co_filename=new_filename) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
24,430
34.510174
86
py
pydeps
pydeps-master/docs/module-finder-archive/mf_37.py
"""Find modules used by a script, using introspection.""" import dis import importlib._bootstrap_external import importlib.machinery import marshal import os import sys import types import warnings with warnings.catch_warnings(): warnings.simplefilter('ignore', DeprecationWarning) import imp LOAD_CONST = dis.opmap['LOAD_CONST'] IMPORT_NAME = dis.opmap['IMPORT_NAME'] STORE_NAME = dis.opmap['STORE_NAME'] STORE_GLOBAL = dis.opmap['STORE_GLOBAL'] STORE_OPS = STORE_NAME, STORE_GLOBAL EXTENDED_ARG = dis.EXTENDED_ARG # Modulefinder does a good job at simulating Python's, but it can not # handle __path__ modifications packages make at runtime. Therefore there # is a mechanism whereby you can register extra paths in this map for a # package, and it will be honored. # Note this is a mapping is lists of paths. packagePathMap = {} # A Public interface def AddPackagePath(packagename, path): packagePathMap.setdefault(packagename, []).append(path) replacePackageMap = {} # This ReplacePackage mechanism allows modulefinder to work around # situations in which a package injects itself under the name # of another package into sys.modules at runtime by calling # ReplacePackage("real_package_name", "faked_package_name") # before running ModuleFinder. def ReplacePackage(oldname, newname): replacePackageMap[oldname] = newname class Module: def __init__(self, name, file=None, path=None): self.__name__ = name self.__file__ = file self.__path__ = path self.__code__ = None # The set of global names that are assigned to in the module. # This includes those names imported through starimports of # Python modules. self.globalnames = {} # The set of starimports this module did that could not be # resolved, ie. a starimport from a non-Python module. self.starimports = {} def __repr__(self): s = "Module(%r" % (self.__name__,) if self.__file__ is not None: s = s + ", %r" % (self.__file__,) if self.__path__ is not None: s = s + ", %r" % (self.__path__,) s = s + ")" return s class ModuleFinder: def __init__(self, path=None, debug=0, excludes=[], replace_paths=[]): if path is None: path = sys.path self.path = path self.modules = {} self.badmodules = {} self.debug = debug self.indent = 0 self.excludes = excludes self.replace_paths = replace_paths self.processed_paths = [] # Used in debugging only def msg(self, level, str, *args): if level <= self.debug: for i in range(self.indent): print(" ", end=' ') print(str, end=' ') for arg in args: print(repr(arg), end=' ') print() def msgin(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent + 1 self.msg(*args) def msgout(self, *args): level = args[0] if level <= self.debug: self.indent = self.indent - 1 self.msg(*args) def run_script(self, pathname): self.msg(2, "run_script", pathname) with open(pathname) as fp: stuff = ("", "r", imp.PY_SOURCE) self.load_module('__main__', fp, pathname, stuff) def load_file(self, pathname): dir, name = os.path.split(pathname) name, ext = os.path.splitext(name) with open(pathname) as fp: stuff = (ext, "r", imp.PY_SOURCE) self.load_module(name, fp, pathname, stuff) def import_hook(self, name, caller=None, fromlist=None, level=-1): self.msg(3, "import_hook", name, caller, fromlist, level) parent = self.determine_parent(caller, level=level) q, tail = self.find_head_package(parent, name) m = self.load_tail(q, tail) if not fromlist: return q if m.__path__: self.ensure_fromlist(m, fromlist) return None def determine_parent(self, caller, level=-1): self.msgin(4, "determine_parent", caller, level) if not caller or level == 0: self.msgout(4, "determine_parent -> None") return None pname = caller.__name__ if level >= 1: # relative import if caller.__path__: level -= 1 if level == 0: parent = self.modules[pname] assert parent is caller self.msgout(4, "determine_parent ->", parent) return parent if pname.count(".") < level: raise ImportError("relative importpath too deep") pname = ".".join(pname.split(".")[:-level]) parent = self.modules[pname] self.msgout(4, "determine_parent ->", parent) return parent if caller.__path__: parent = self.modules[pname] assert caller is parent self.msgout(4, "determine_parent ->", parent) return parent if '.' in pname: i = pname.rfind('.') pname = pname[:i] parent = self.modules[pname] assert parent.__name__ == pname self.msgout(4, "determine_parent ->", parent) return parent self.msgout(4, "determine_parent -> None") return None def find_head_package(self, parent, name): self.msgin(4, "find_head_package", parent, name) if '.' in name: i = name.find('.') head = name[:i] tail = name[i+1:] else: head = name tail = "" if parent: qname = "%s.%s" % (parent.__name__, head) else: qname = head q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail if parent: qname = head parent = None q = self.import_module(head, qname, parent) if q: self.msgout(4, "find_head_package ->", (q, tail)) return q, tail self.msgout(4, "raise ImportError: No module named", qname) raise ImportError("No module named " + qname) def load_tail(self, q, tail): self.msgin(4, "load_tail", q, tail) m = q while tail: i = tail.find('.') if i < 0: i = len(tail) head, tail = tail[:i], tail[i+1:] mname = "%s.%s" % (m.__name__, head) m = self.import_module(head, mname, m) if not m: self.msgout(4, "raise ImportError: No module named", mname) raise ImportError("No module named " + mname) self.msgout(4, "load_tail ->", m) return m def ensure_fromlist(self, m, fromlist, recursive=0): self.msg(4, "ensure_fromlist", m, fromlist, recursive) for sub in fromlist: if sub == "*": if not recursive: all = self.find_all_submodules(m) if all: self.ensure_fromlist(m, all, 1) elif not hasattr(m, sub): subname = "%s.%s" % (m.__name__, sub) submod = self.import_module(sub, subname, m) if not submod: raise ImportError("No module named " + subname) def find_all_submodules(self, m): if not m.__path__: return modules = {} # 'suffixes' used to be a list hardcoded to [".py", ".pyc"]. # But we must also collect Python extension modules - although # we cannot separate normal dlls from Python extensions. suffixes = [] suffixes += importlib.machinery.EXTENSION_SUFFIXES[:] suffixes += importlib.machinery.SOURCE_SUFFIXES[:] suffixes += importlib.machinery.BYTECODE_SUFFIXES[:] for dir in m.__path__: try: names = os.listdir(dir) except OSError: self.msg(2, "can't list directory", dir) continue for name in names: mod = None for suff in suffixes: n = len(suff) if name[-n:] == suff: mod = name[:-n] break if mod and mod != "__init__": modules[mod] = mod return modules.keys() def import_module(self, partname, fqname, parent): self.msgin(3, "import_module", partname, fqname, parent) try: m = self.modules[fqname] except KeyError: pass else: self.msgout(3, "import_module ->", m) return m if fqname in self.badmodules: self.msgout(3, "import_module -> None") return None if parent and parent.__path__ is None: self.msgout(3, "import_module -> None") return None try: fp, pathname, stuff = self.find_module(partname, parent and parent.__path__, parent) except ImportError: self.msgout(3, "import_module ->", None) return None try: m = self.load_module(fqname, fp, pathname, stuff) finally: if fp: fp.close() if parent: setattr(parent, partname, m) self.msgout(3, "import_module ->", m) return m def load_module(self, fqname, fp, pathname, file_info): suffix, mode, type = file_info self.msgin(2, "load_module", fqname, fp and "fp", pathname) if type == imp.PKG_DIRECTORY: m = self.load_package(fqname, pathname) self.msgout(2, "load_module ->", m) return m if type == imp.PY_SOURCE: co = compile(fp.read()+'\n', pathname, 'exec') elif type == imp.PY_COMPILED: try: data = fp.read() importlib._bootstrap_external._classify_pyc(data, fqname, {}) except ImportError as exc: self.msgout(2, "raise ImportError: " + str(exc), pathname) raise co = marshal.loads(memoryview(data)[16:]) else: co = None m = self.add_module(fqname) m.__file__ = pathname if co: if self.replace_paths: co = self.replace_paths_in_code(co) m.__code__ = co self.scan_code(co, m) self.msgout(2, "load_module ->", m) return m def _add_badmodule(self, name, caller): if name not in self.badmodules: self.badmodules[name] = {} if caller: self.badmodules[name][caller.__name__] = 1 else: self.badmodules[name]["-"] = 1 def _safe_import_hook(self, name, caller, fromlist, level=-1): # wrapper for self.import_hook() that won't raise ImportError if name in self.badmodules: self._add_badmodule(name, caller) return try: self.import_hook(name, caller, level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) self._add_badmodule(name, caller) else: if fromlist: for sub in fromlist: if sub in self.badmodules: self._add_badmodule(sub, caller) continue try: self.import_hook(name, caller, [sub], level=level) except ImportError as msg: self.msg(2, "ImportError:", str(msg)) fullname = name + "." + sub self._add_badmodule(fullname, caller) def scan_opcodes(self, co): # Scan the code, and yield 'interesting' opcode combinations code = co.co_code names = co.co_names consts = co.co_consts opargs = [(op, arg) for _, op, arg in dis._unpack_opargs(code) if op != EXTENDED_ARG] for i, (op, oparg) in enumerate(opargs): if op in STORE_OPS: yield "store", (names[oparg],) continue if (op == IMPORT_NAME and i >= 2 and opargs[i-1][0] == opargs[i-2][0] == LOAD_CONST): level = consts[opargs[i-2][1]] fromlist = consts[opargs[i-1][1]] if level == 0: # absolute import yield "absolute_import", (fromlist, names[oparg]) else: # relative import yield "relative_import", (level, fromlist, names[oparg]) continue def scan_code(self, co, m): code = co.co_code scanner = self.scan_opcodes for what, args in scanner(co): if what == "store": name, = args m.globalnames[name] = 1 elif what == "absolute_import": fromlist, name = args have_star = 0 if fromlist is not None: if "*" in fromlist: have_star = 1 fromlist = [f for f in fromlist if f != "*"] self._safe_import_hook(name, m, fromlist, level=0) if have_star: # We've encountered an "import *". If it is a Python module, # the code has already been parsed and we can suck out the # global names. mm = None if m.__path__: # At this point we don't know whether 'name' is a # submodule of 'm' or a global module. Let's just try # the full name first. mm = self.modules.get(m.__name__ + "." + name) if mm is None: mm = self.modules.get(name) if mm is not None: m.globalnames.update(mm.globalnames) m.starimports.update(mm.starimports) if mm.__code__ is None: m.starimports[name] = 1 else: m.starimports[name] = 1 elif what == "relative_import": level, fromlist, name = args if name: self._safe_import_hook(name, m, fromlist, level=level) else: parent = self.determine_parent(m, level=level) self._safe_import_hook(parent.__name__, None, fromlist, level=0) else: # We don't expect anything else from the generator. raise RuntimeError(what) for c in co.co_consts: if isinstance(c, type(co)): self.scan_code(c, m) def load_package(self, fqname, pathname): self.msgin(2, "load_package", fqname, pathname) newname = replacePackageMap.get(fqname) if newname: fqname = newname m = self.add_module(fqname) m.__file__ = pathname m.__path__ = [pathname] # As per comment at top of file, simulate runtime __path__ additions. m.__path__ = m.__path__ + packagePathMap.get(fqname, []) fp, buf, stuff = self.find_module("__init__", m.__path__) try: self.load_module(fqname, fp, buf, stuff) self.msgout(2, "load_package ->", m) return m finally: if fp: fp.close() def add_module(self, fqname): if fqname in self.modules: return self.modules[fqname] self.modules[fqname] = m = Module(fqname) return m def find_module(self, name, path, parent=None): if parent is not None: # assert path is not None fullname = parent.__name__+'.'+name else: fullname = name if fullname in self.excludes: self.msgout(3, "find_module -> Excluded", fullname) raise ImportError(name) if path is None: if name in sys.builtin_module_names: return (None, None, ("", "", imp.C_BUILTIN)) path = self.path return imp.find_module(name, path) def report(self): """Print a report to stdout, listing the found modules with their paths, as well as modules that are missing, or seem to be missing. """ print() print(" %-25s %s" % ("Name", "File")) print(" %-25s %s" % ("----", "----")) # Print modules found keys = sorted(self.modules.keys()) for key in keys: m = self.modules[key] if m.__path__: print("P", end=' ') else: print("m", end=' ') print("%-25s" % key, m.__file__ or "") # Print missing modules missing, maybe = self.any_missing_maybe() if missing: print() print("Missing modules:") for name in missing: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) # Print modules that may be missing, but then again, maybe not... if maybe: print() print("Submodules that appear to be missing, but could also be", end=' ') print("global names in the parent package:") for name in maybe: mods = sorted(self.badmodules[name].keys()) print("?", name, "imported from", ', '.join(mods)) def any_missing(self): """Return a list of modules that appear to be missing. Use any_missing_maybe() if you want to know which modules are certain to be missing, and which *may* be missing. """ missing, maybe = self.any_missing_maybe() return missing + maybe def any_missing_maybe(self): """Return two lists, one with modules that are certainly missing and one with modules that *may* be missing. The latter names could either be submodules *or* just global names in the package. The reason it can't always be determined is that it's impossible to tell which names are imported when "from module import *" is done with an extension module, short of actually importing it. """ missing = [] maybe = [] for name in self.badmodules: if name in self.excludes: continue i = name.rfind(".") if i < 0: missing.append(name) continue subname = name[i+1:] pkgname = name[:i] pkg = self.modules.get(pkgname) if pkg is not None: if pkgname in self.badmodules[name]: # The package tried to import this module itself and # failed. It's definitely missing. missing.append(name) elif subname in pkg.globalnames: # It's a global in the package: definitely not missing. pass elif pkg.starimports: # It could be missing, but the package did an "import *" # from a non-Python module, so we simply can't be sure. maybe.append(name) else: # It's not a global in the package, the package didn't # do funny star imports, it's very likely to be missing. # The symbol could be inserted into the package from the # outside, but since that's not good style we simply list # it missing. missing.append(name) else: missing.append(name) missing.sort() maybe.sort() return missing, maybe def replace_paths_in_code(self, co): new_filename = original_filename = os.path.normpath(co.co_filename) for f, r in self.replace_paths: if original_filename.startswith(f): new_filename = r + original_filename[len(f):] break if self.debug and original_filename not in self.processed_paths: if new_filename != original_filename: self.msgout(2, "co_filename %r changed to %r" \ % (original_filename,new_filename,)) else: self.msgout(2, "co_filename %r remains unchanged" \ % (original_filename,)) self.processed_paths.append(original_filename) consts = list(co.co_consts) for i in range(len(consts)): if isinstance(consts[i], type(co)): consts[i] = self.replace_paths_in_code(consts[i]) return types.CodeType(co.co_argcount, co.co_kwonlyargcount, co.co_nlocals, co.co_stacksize, co.co_flags, co.co_code, tuple(consts), co.co_names, co.co_varnames, new_filename, co.co_name, co.co_firstlineno, co.co_lnotab, co.co_freevars, co.co_cellvars) def test(): # Parse command line import getopt try: opts, args = getopt.getopt(sys.argv[1:], "dmp:qx:") except getopt.error as msg: print(msg) return # Process options debug = 1 domods = 0 addpath = [] exclude = [] for o, a in opts: if o == '-d': debug = debug + 1 if o == '-m': domods = 1 if o == '-p': addpath = addpath + a.split(os.pathsep) if o == '-q': debug = 0 if o == '-x': exclude.append(a) # Provide default arguments if not args: script = "hello.py" else: script = args[0] # Set the path based on sys.path and the script directory path = sys.path[:] path[0] = os.path.dirname(script) path = addpath + path if debug > 1: print("path:") for item in path: print(" ", repr(item)) # Create the module finder and turn its crank mf = ModuleFinder(path, debug, exclude) for arg in args[1:]: if arg == '-m': domods = 1 continue if domods: if arg[-2:] == '.*': mf.import_hook(arg[:-2], None, ["*"]) else: mf.import_hook(arg) else: mf.load_file(arg) mf.run_script(script) mf.report() return mf # for -i debugging if __name__ == '__main__': try: mf = test() except KeyboardInterrupt: print("\n[interrupted]")
23,035
35.334385
86
py
MRE-ISE
MRE-ISE-main/run.py
import argparse import logging import sys sys.path.append("..") import torch import numpy as np import random from torchvision import transforms from torch.utils.data import DataLoader from cores.gene.model import MRE from transformers import CLIPProcessor, CLIPModel from transformers import CLIPConfig from processor.dataset import MREProcessor, MREDataset from cores.gene.model import Trainer import os os.environ["CUDA_VISIBLE_DEVICES"] = "0" import warnings warnings.filterwarnings("ignore", category=UserWarning) from torch.utils.tensorboard import SummaryWriter logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s', datefmt = '%m/%d/%Y %H:%M:%S', level = logging.INFO) logger = logging.getLogger(__name__) MODEL_CLASS = { 'bert': (MREProcessor, MREDataset), } DATA_PATH = { 'MRE': {'train': 'data/vsg_tsg/ours_train.json', 'dev': 'data/vsg_tsg/ours_valid.json', 'test': 'data/vsg_tsg/ours_test.json', 'vbow': 'data/vsg_tsg/vbow.pk', 'tbow': 'data/vsg_tsg/tbow.pk' } } IMG_PATH = { 'MRE': {'train': 'data/img_org/train/', 'dev': 'data/img_org/val/', 'test': 'data/img_org/test'}} def set_seed(seed=2021): """set random seed""" torch.manual_seed(seed) torch.cuda.manual_seed_all(seed) torch.backends.cudnn.deterministic = True torch.backends.cudnn.enabled = True torch.backends.cudnn.benchmark = True np.random.seed(seed) random.seed(seed) def main(): parser = argparse.ArgumentParser() parser.add_argument('--pretrain_name', default='openai/clip-vit-base-patch32', type=str, help="The name of pretrained model") parser.add_argument('--dataset_name', default='MRE', type=str, help="The name of example_dataset.") parser.add_argument('--num_epochs', default=30, type=int, help="Training epochs") parser.add_argument('--device', default='cuda', type=str, help="cuda or cpu") parser.add_argument('--batch_size', default=32, type=int, help="batch size") parser.add_argument('--lr_pretrained', default=2e-5, type=float, help="pre-trained learning rate") parser.add_argument('--lr_main', default=2e-4, type=float, help="learning rate") parser.add_argument('--warmup_ratio', default=0.01, type=float) parser.add_argument('--eval_begin_epoch', default=1, type=int) parser.add_argument('--seed', default=1234, type=int, help="random seed, default is 1") parser.add_argument('--load_path', default=None, type=str, help="Load model from load_path") parser.add_argument('--save_path', default='ckpt', type=str, help="save model at save_path") parser.add_argument('--write_path', default=None, type=str, help="do_test=True, predictions will be write in write_path") parser.add_argument('--notes', default="", type=str, help="input some remarks for making save path dir.") parser.add_argument('--do_train', action='store_true') # , action='store_true' parser.add_argument('--do_test', action='store_true') parser.add_argument('--do_predict', action='store_true') parser.add_argument('--max_seq', default=40, type=int) parser.add_argument('--max_obj', default=40, type=int) parser.add_argument('--hid_size', default=768, type=int, help="hidden state size") parser.add_argument('--num_layers', default=2, type=int, help="number of refine layers") parser.add_argument('--beta', default=0.01, type=float, help="Default is 1e-2") parser.add_argument("--num_per", type=int, default=16, help="Default is 16") parser.add_argument("--feature_denoise", type=bool, default=True, help="Default is False.") parser.add_argument("--top_k", type=int, default=10, help="Default is 10.") parser.add_argument("--epsilon", type=float, default=0.3, help="Default is 0.3.") parser.add_argument("--temperature", type=float, default=0.1, help="Default is 0.1.") parser.add_argument("--graph_skip_conn", type=float, default=0.0, help="Default is 0.0.") parser.add_argument("--graph_include_self", type=bool, default=True, help="Default is True.") parser.add_argument("--dropout", type=float, default=0.5, help="Default is 0.0") parser.add_argument("--graph_type", type=str, default="epsilonNN", help="epsilonNN, KNN, prob") parser.add_argument("--graph_metric_type", type=str, default="multi_mlp") parser.add_argument("--repar", type=bool, default=True, help="Default is True.") parser.add_argument("--threshold", type=float, default=0.2, help="Default is 0.2.") parser.add_argument("--prior_mode", type=str, default="Gaussian", help="Default is Gaussian.") parser.add_argument("--is_IB", type=bool, default=True, help="Default is True.") parser.add_argument("--eta1", type=float, default=0.7, help="Default is 0.7") parser.add_argument("--eta1", type=float, default=0.9, help="Default is 0.9") parser.add_argument("--text_bow_size", type=int, default=2000, help="Default is 2000") parser.add_argument("--visual_bow_size", type=int, default=2000, help="Default is 2000") parser.add_argument("--neighbor_num", type=int, default=2, help="Default is 2") parser.add_argument("--topic_keywords_number", type=int, default=10, help="Default is 10") parser.add_argument("--topic_number", type=int, default=10, help="Default is 10") args = parser.parse_args() data_path, img_path, aux_path = DATA_PATH[args.dataset_name], IMG_PATH[args.dataset_name], AUX_PATH[args.dataset_name] data_process, dataset_class = MODEL_CLASS[args.model_name] re_path = 'data/ours_rel2id.json' transform = transforms.Compose([ transforms.Resize(256), transforms.CenterCrop(224), transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]) set_seed(args.seed) # set seed, default is 1 if args.save_path is not None: # make save_path dir args.save_path = os.path.join(args.save_path, args.model_name, args.dataset_name+"_"+str(args.batch_size)+"_"+str(args.lr)+"_"+args.notes) if not os.path.exists(args.save_path): os.makedirs(args.save_path, exist_ok=True) print(args) logdir = "logs/" + args.model_name + "_"+args.dataset_name + "_"+str(args.batch_size) + "_" + str(args.lr) + args.notes writer = SummaryWriter(log_dir=logdir) if args.do_train: clip_vit, clip_processor, aux_processor, rcnn_processor = None, None, None, None clip_processor = CLIPProcessor.from_pretrained(args.vit_name) aux_processor = CLIPProcessor.from_pretrained(args.vit_name) aux_processor.feature_extractor.size, aux_processor.feature_extractor.crop_size = args.aux_size, args.aux_size rcnn_processor = CLIPProcessor.from_pretrained(args.vit_name) rcnn_processor.feature_extractor.size, rcnn_processor.feature_extractor.crop_size = args.rcnn_size, args.rcnn_size clip_model = CLIPModel.from_pretrained(args.vit_name) clip_vit = clip_model.vision_model clip_text = clip_model.text_model processor = data_process(data_path, re_path, args.bert_name, args.vit_name, clip_processor=clip_processor, aux_processor=aux_processor, rcnn_processor=rcnn_processor) train_dataset = dataset_class(processor, transform, img_path, aux_path, args.max_seq, aux_size=args.aux_size, rcnn_size=args.rcnn_size, mode='train', max_obj_num=args.max_obj_num) train_dataloader = DataLoader(train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=4, pin_memory=True) dev_dataset = dataset_class(processor, transform, img_path, aux_path, args.max_seq, aux_size=args.aux_size, rcnn_size=args.rcnn_size, mode='dev', max_obj_num=args.max_obj_num) dev_dataloader = DataLoader(dev_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4, pin_memory=True) test_dataset = dataset_class(processor, transform, img_path, aux_path, args.max_seq, aux_size=args.aux_size, rcnn_size=args.rcnn_size, mode='test', max_obj_num=args.max_obj_num) test_dataloader = DataLoader(test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=4, pin_memory=True) re_dict = processor.get_relation_dict() num_labels = len(re_dict) tokenizer = processor.tokenizer # test vision_config = CLIPConfig.from_pretrained(args.vit_name).vision_config text_config = CLIPConfig.from_pretrained(args.vit_name).text_config model = MRE(args, vision_config, text_config, clip_vit, clip_text, num_labels, args.text_bow_size, args.visual_bow_size, tokenizer, processor) trainer = Trainer(train_data=train_dataloader, dev_data=dev_dataloader, test_data=test_dataloader, re_dict=re_dict, model=model, args=args, logger=logger, writer=writer) trainer.train() torch.cuda.empty_cache() writer.close() if __name__ == "__main__": main()
9,056
52.276471
187
py
MRE-ISE
MRE-ISE-main/VSG/RelTR_parser/visual_scene_graph.py
import os import torch import torch.nn as nn import torch.nn.functional as F import torchvision.transforms as T from PIL import Image import requests import matplotlib.pyplot as plt import json import pickle import ast from tqdm import tqdm from transformers import CLIPProcessor, CLIPModel import cv2 from models.backbone import Backbone, Joiner from models.position_encoding import PositionEmbeddingSine from models.transformer import Transformer from models.reltr import RelTR os.environ["CUDA_VISIBLE_DEVICES"] = "1" CLASSES = ['N/A', 'airplane', 'animal', 'arm', 'bag', 'banana', 'basket', 'beach', 'bear', 'bed', 'bench', 'bike', 'bird', 'board', 'boat', 'book', 'boot', 'bottle', 'bowl', 'box', 'boy', 'branch', 'building', 'bus', 'cabinet', 'cap', 'car', 'cat', 'chair', 'child', 'clock', 'coat', 'counter', 'cow', 'cup', 'curtain', 'desk', 'dog', 'door', 'drawer', 'ear', 'elephant', 'engine', 'eye', 'face', 'fence', 'finger', 'flag', 'flower', 'food', 'fork', 'fruit', 'giraffe', 'girl', 'glass', 'glove', 'guy', 'hair', 'hand', 'handle', 'hat', 'head', 'helmet', 'hill', 'horse', 'house', 'jacket', 'jean', 'kid', 'kite', 'lady', 'lamp', 'laptop', 'leaf', 'leg', 'letter', 'light', 'logo', 'man', 'men', 'motorcycle', 'mountain', 'mouth', 'neck', 'nose', 'number', 'orange', 'pant', 'paper', 'paw', 'people', 'person', 'phone', 'pillow', 'pizza', 'plane', 'plant', 'plate', 'player', 'pole', 'post', 'pot', 'racket', 'railing', 'rock', 'roof', 'room', 'screen', 'seat', 'sheep', 'shelf', 'shirt', 'shoe', 'short', 'sidewalk', 'sign', 'sink', 'skateboard', 'ski', 'skier', 'sneaker', 'snow', 'sock', 'stand', 'street', 'surfboard', 'table', 'tail', 'tie', 'tile', 'tire', 'toilet', 'towel', 'tower', 'track', 'train', 'tree', 'truck', 'trunk', 'umbrella', 'vase', 'vegetable', 'vehicle', 'wave', 'wheel', 'window', 'windshield', 'wing', 'wire', 'woman', 'zebra'] REL_CLASSES = ['__background__', 'above', 'across', 'against', 'along', 'and', 'at', 'attached to', 'behind', 'belonging to', 'between', 'carrying', 'covered in', 'covering', 'eating', 'flying in', 'for', 'from', 'growing on', 'hanging from', 'has', 'holding', 'in', 'in front of', 'laying on', 'looking at', 'lying on', 'made of', 'mounted on', 'near', 'of', 'on', 'on back of', 'over', 'painted on', 'parked on', 'part of', 'playing', 'riding', 'says', 'sitting on', 'standing on', 'to', 'under', 'using', 'walking in', 'walking on', 'watching', 'wearing', 'wears', 'with'] # for output bounding box post-processing def box_cxcywh_to_xyxy(x): x_c, y_c, w, h = x.unbind(1) b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)] return torch.stack(b, dim=1) def rescale_bboxes(out_bbox, size): img_w, img_h = size b = box_cxcywh_to_xyxy(out_bbox) x = torch.tensor([img_w, img_h, img_w, img_h], dtype=out_bbox.dtype).to(torch.get_device(out_bbox)) b = b * x return b def find_repeat(new_bbox, bbox): flag = 0 index = 0 def get_list(x): return [i for i in range(max(0, x-5), x+5)] for idx, b in enumerate(bbox): if (new_bbox[0] in get_list(b[0])) and (new_bbox[1] in get_list(b[1])) and (new_bbox[2] in get_list(b[2])) and (new_bbox[3] in get_list(b[3])): flag = 1 index = idx break return flag, index def construct_scene_graph(data, original_img_dir, target_file, mode='train'): position_embedding = PositionEmbeddingSine(128, normalize=True) backbone = Backbone('resnet50', False, False, False) backbone = Joiner(backbone, position_embedding) backbone.num_channels = 2048 transformer = Transformer(d_model=256, dropout=0.1, nhead=8, dim_feedforward=2048, num_encoder_layers=6, num_decoder_layers=6, normalize_before=False, return_intermediate_dec=True) model = RelTR(backbone, transformer, num_classes=151, num_rel_classes=51, num_entities=100, num_triplets=200) # The checkpoint is pretrained on Visual Genome ckpt = torch.hub.load_state_dict_from_url( url='https://cloud.tnt.uni-hannover.de/index.php/s/PB8xTKspKZF7fyK/download/checkpoint0149.pth', # map_location=lambda storage, loc: storage.cuda(0), check_hash=True) # map_location='cpu' model.load_state_dict(ckpt['model']) model.to(torch.device('cuda')) # Some transformation functions transform = T.Compose([ T.Resize(800), T.ToTensor(), T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) ]) # img_list = os.listdir(dirname) # img_list = [] # for d in data: # img_list.append(d['img_id']) # img_list = ['twitter_stream_2018_10_10_25_0_2_7.jpg'] res_list = [] with torch.no_grad(): model.eval() for d in tqdm(data, total=len(data)): # print("1:{}".format(torch.cuda.memory_allocated(0))) res = dict() img_id = d['img_id'] # res['img'] = i im = Image.open(os.path.join(original_img_dir, mode, img_id)) img = transform(im).unsqueeze(0) img = img.to(torch.device('cuda')) # propagate through the model outputs = model(img) # outputs = dict() # for k, v in res.items(): # outputs[k] = v.to(torch.device('cpu')) keep_thresh = 0.35 # keep only predictions with >0.3 confidence probas = outputs['rel_logits'].softmax(-1)[0, :, :-1] probas_sub = outputs['sub_logits'].softmax(-1)[0, :, :-1] probas_obj = outputs['obj_logits'].softmax(-1)[0, :, :-1] keep = torch.logical_and(probas.max(-1).values > keep_thresh, torch.logical_and(probas_sub.max(-1).values > keep_thresh, probas_obj.max(-1).values > keep_thresh)) # convert boxes from [0; 1] to image scales sub_bboxes_scaled = rescale_bboxes(outputs['sub_boxes'][0, keep], im.size) obj_bboxes_scaled = rescale_bboxes(outputs['obj_boxes'][0, keep], im.size) # topk = 10 # display up to 10 images keep_queries = torch.nonzero(keep, as_tuple=True)[0] topk = keep_queries.size(0) # display up to 10 images indices = torch.argsort( -probas[keep_queries].max(-1)[0] * probas_sub[keep_queries].max(-1)[0] * probas_obj[keep_queries].max(-1)[ 0])[ :topk] keep_queries = keep_queries[indices] bbox = [[0, 0, im.width, im.height]] bbox_attri = ['img'] # s_bbox = [] # s_bbox_attri = [] # o_bbox = [] # o_bbox_attri = [] rel = [{'s_index': 0, 'o_index': 0, 'name': 'self'}] for idx, s, o in zip(keep_queries, sub_bboxes_scaled[indices], obj_bboxes_scaled[indices]): (sxmin, symin, sxmax, symax) = [max(0, round(x)) for x in s.tolist()] sxmax, symax = min(im.width, sxmax), min(im.height, symax) (oxmin, oymin, oxmax, oymax) = [max(0, round(x)) for x in o.tolist()] oxmax, oymax = min(im.width, oxmax), min(im.height, oymax) # new_img = im.new('RGB', size=s) _flag, _idx = find_repeat((sxmin, symin, sxmax, symax), bbox) if not _flag: bbox.append((sxmin, symin, sxmax, symax)) bbox_attri.append(CLASSES[probas_sub[idx].argmax()]) s_index = len(bbox)-1 else: s_index = _idx _flag, _idx = find_repeat((oxmin, oymin, oxmax, oymax), bbox) if not _flag: bbox.append((oxmin, oymin, oxmax, oymax)) bbox_attri.append(CLASSES[probas_sub[idx].argmax()]) o_index = len(bbox) - 1 else: o_index = _idx # if (sxmin, symin, sxmax, symax) not in bbox: # bbox.append((sxmin, symin, sxmax, symax)) # bbox_attri.append(CLASSES[probas_sub[idx].argmax()]) # if (oxmin, oymin, oxmax, oymax) not in bbox: # bbox.append((oxmin, oymin, oxmax, oymax)) # bbox_attri.append(CLASSES[probas_obj[idx].argmax()]) # s_index = bbox.index((sxmin, symin, sxmax, symax)) # o_index = bbox.index((oxmin, oymin, oxmax, oymax)) rel_attri = REL_CLASSES[probas[idx].argmax()] rel.append({'s_index': s_index, 'o_index': o_index, 'name': rel_attri}) # res['s_bbox'] = s_bbox # res['s_bbox_attri'] = s_bbox_attri # res['o_bbox'] = o_bbox # res['o_bbox_attri'] = o_bbox_attri res['bbox'] = bbox res['bbox_attri'] = bbox_attri res['rel'] = rel d['VSG'] = res res_list.append(d) # torch.empty_cache() assert len(data) == len(res_list) with open(target_file, 'w', encoding='utf-8') as f: json.dump(res_list, f) def get_obj_features(dirname): vision_model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32") processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32") if torch.cuda.is_available(): vision_model.to(torch.device('cuda')) with open(os.path.join(os.path.dirname(dirname), os.path.basename(dirname).split('.')[0] + '.json')) as f: data = json.load(f) mode = os.path.basename(dirname).split('.')[0].split('_')[1] with torch.no_grad(): vision_model.eval() for d in tqdm(data, total=len(data)): imgid = d['img'] img_path = os.path.join('../../data/img_org', mode, imgid) bbox = d['bbox'] features = [] for b in bbox: crop_img = cv2.imread(img_path) # print(crop_img.shape) try: crop_region = crop_img[b[1]:b[3], b[0]:b[2]] except TypeError as e: print(e) print(bbox) print(b) print(imgid) exit(0) im = Image.fromarray(crop_region, mode="RGB") # print(im.size) images = processor(images=im, return_tensors="pt") images = images.to(torch.device('cuda')) image_features = vision_model.get_image_features(**images).squeeze() features.append(image_features.tolist()) d['features'] = features with open(os.path.join(os.path.dirname(dirname), os.path.basename(dirname).split('.')[0] + '.pk'), 'wb') as fout: pickle.dump(data, fout) if __name__ == '__main__': FILE_DIR = '../data/tsg/' IMG_DIR = '../data/img_org/' DIST_DIR = '../data/vsg_tsg/' for i in ['ours_train.json', 'ours_val.json', 'ours_test.json']: print(f'parsing {i} ... ') with open(os.path.join(FILE_DIR, i)) as f: data = json.load(f) base_name = os.path.basename(i).split('.')[0] # dirname = os.path.join(FILE_DIR, i) mode = i.split('.')[0].split('_')[1] print(mode) construct_scene_graph(data, original_img_dir=IMG_DIR, target_file=os.path.join(DIST_DIR, f'{base_name}.json'), mode=mode) # get_obj_features(dirname) # data = [1, 2, 3] # output_file = os.path.join(os.path.dirname(dirname), os.path.basename(dirname).split('.')[0] + '.json') # print(output_file) # with open(output_file, 'w', encoding='utf-8') as f: # json.dump(data, f) # with open(os.path.join(os.path.dirname(dirname), os.path.basename(dirname).split('.')[0] + '.pk'), 'wb') as fout: # pickle.dump(data, fout) # x = [{'s_index': 0, 'o_index': 0, 'name': 'self'}, {'s_index': 1, 'o_index': 2, 'name': 'wearing'}, {'s_index': 1, 'o_index': 3, 'name': 'wearing'}, {'s_index': 4, 'o_index': 5, 'name': 'wearing'}, {'s_index': 4, 'o_index': 6, 'name': 'has'}, {'s_index': 4, 'o_index': 7, 'name': 'wearing'}, {'s_index': 8, 'o_index': 9, 'name': 'on'}] # b = [[i['s_index'], i['o_index']] for i in x] # print(b) # with open(os.path.join(os.path.dirname(dirname), os.path.basename(dirname) + '.pk'), 'wb') as fout: # pickle.dump([1, 2], fout)
12,857
45.086022
341
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/config.py
""" Configuration file! """ import os from argparse import ArgumentParser import numpy as np ROOT_PATH = os.path.dirname(os.path.realpath(__file__)) DATA_PATH = os.path.join(ROOT_PATH, 'data') def path(fn): return os.path.join(DATA_PATH, fn) def stanford_path(fn): return os.path.join(DATA_PATH, 'stanford_filtered', fn) # ============================================================================= # Update these with where your data is stored ~~~~~~~~~~~~~~~~~~~~~~~~~ VG_IMAGES = '/datasets2/VG_100K_2/VG_100K' RCNN_CHECKPOINT_FN = path('faster_rcnn_500k.h5') IM_DATA_FN = stanford_path('image_data.json') VG_SGG_FN = stanford_path('VG-SGG.h5') VG_SGG_DICT_FN = stanford_path('VG-SGG-dicts.json') PROPOSAL_FN = stanford_path('proposals.h5') COCO_PATH = 'datasets/mscoco' # ============================================================================= # ============================================================================= MODES = ('sgdet', 'sgcls', 'predcls') BOX_SCALE = 1024 # Scale at which we have the boxes IM_SCALE = 592 # Our images will be resized to this res without padding # Proposal assignments BG_THRESH_HI = 0.5 BG_THRESH_LO = 0.0 RPN_POSITIVE_OVERLAP = 0.7 # IOU < thresh: negative example RPN_NEGATIVE_OVERLAP = 0.3 # Max number of foreground examples RPN_FG_FRACTION = 0.5 FG_FRACTION = 0.25 # Total number of examples RPN_BATCHSIZE = 256 ROIS_PER_IMG = 256 REL_FG_FRACTION = 0.25 RELS_PER_IMG = 256 RELS_PER_IMG_REFINE = 64 BATCHNORM_MOMENTUM = 0.01 ANCHOR_SIZE = 16 ANCHOR_RATIOS = (0.23232838, 0.63365731, 1.28478321, 3.15089189) #(0.5, 1, 2) ANCHOR_SCALES = (2.22152954, 4.12315647, 7.21692515, 12.60263013, 22.7102731) #(4, 8, 16, 32) class ModelConfig(object): """Wrapper class for model hyperparameters.""" def __init__(self): """ Defaults """ self.coco = None self.ckpt = None self.save_dir = None self.lr = None self.batch_size = None self.val_size = None self.l2 = None self.clip = None self.num_gpus = None self.num_workers = None self.print_interval = None self.gt_box = None self.mode = None self.refine = None self.ad3 = False self.test = False self.adam = False self.multi_pred=False self.cache = None self.model = None self.use_proposals=False self.use_resnet=False self.use_tanh=False self.use_bias = False self.limit_vision=False self.num_epochs=None self.old_feats=False self.order=None self.det_ckpt=None self.nl_edge=None self.nl_obj=None self.hidden_dim=None self.pass_in_obj_feats_to_decoder = None self.pass_in_obj_feats_to_edge = None self.pooling_dim = None self.rec_dropout = None self.parser = self.setup_parser() self.args = vars(self.parser.parse_args()) print("~~~~~~~~ Hyperparameters used: ~~~~~~~") for x, y in self.args.items(): print("{} : {}".format(x, y)) self.__dict__.update(self.args) if len(self.ckpt) != 0: self.ckpt = os.path.join(ROOT_PATH, self.ckpt) else: self.ckpt = None if len(self.cache) != 0: self.cache = os.path.join(ROOT_PATH, self.cache) else: self.cache = None if len(self.save_dir) == 0: self.save_dir = None else: self.save_dir = os.path.join(ROOT_PATH, self.save_dir) if not os.path.exists(self.save_dir): os.mkdir(self.save_dir) assert self.val_size >= 0 if self.mode not in MODES: raise ValueError("Invalid mode: mode must be in {}".format(MODES)) if self.model not in ('motifnet', 'stanford'): raise ValueError("Invalid model {}".format(self.model)) if self.ckpt is not None and not os.path.exists(self.ckpt): raise ValueError("Ckpt file ({}) doesnt exist".format(self.ckpt)) def setup_parser(self): """ Sets up an argument parser :return: """ parser = ArgumentParser(description='training code') # Options to deprecate parser.add_argument('-coco', dest='coco', help='Use COCO (default to VG)', action='store_true') parser.add_argument('-ckpt', dest='ckpt', help='Filename to load from', type=str, default='') parser.add_argument('-det_ckpt', dest='det_ckpt', help='Filename to load detection parameters from', type=str, default='') parser.add_argument('-save_dir', dest='save_dir', help='Directory to save things to, such as checkpoints/save', default='', type=str) parser.add_argument('-ngpu', dest='num_gpus', help='cuantos GPUs tienes', type=int, default=3) parser.add_argument('-nwork', dest='num_workers', help='num processes to use as workers', type=int, default=1) parser.add_argument('-lr', dest='lr', help='learning rate', type=float, default=1e-3) parser.add_argument('-b', dest='batch_size', help='batch size per GPU',type=int, default=2) parser.add_argument('-val_size', dest='val_size', help='val size to use (if 0 we wont use val)', type=int, default=5000) parser.add_argument('-l2', dest='l2', help='weight decay', type=float, default=1e-4) parser.add_argument('-clip', dest='clip', help='gradients will be clipped to have norm less than this', type=float, default=5.0) parser.add_argument('-p', dest='print_interval', help='print during training', type=int, default=100) parser.add_argument('-m', dest='mode', help='mode \in {sgdet, sgcls, predcls}', type=str, default='sgdet') parser.add_argument('-model', dest='model', help='which model to use? (motifnet, stanford). If you want to use the baseline (NoContext) model, then pass in motifnet here, and nl_obj, nl_edge=0', type=str, default='motifnet') parser.add_argument('-old_feats', dest='old_feats', help='Use the original image features for the edges', action='store_true') parser.add_argument('-order', dest='order', help='Linearization order for Rois (confidence -default, size, random)', type=str, default='confidence') parser.add_argument('-cache', dest='cache', help='where should we cache predictions', type=str, default='') parser.add_argument('-gt_box', dest='gt_box', help='use gt boxes during training', action='store_true') parser.add_argument('-adam', dest='adam', help='use adam. Not recommended', action='store_true') parser.add_argument('-test', dest='test', help='test set', action='store_true') parser.add_argument('-multipred', dest='multi_pred', help='Allow multiple predicates per pair of box0, box1.', action='store_true') parser.add_argument('-nepoch', dest='num_epochs', help='Number of epochs to train the model for',type=int, default=25) parser.add_argument('-resnet', dest='use_resnet', help='use resnet instead of VGG', action='store_true') parser.add_argument('-proposals', dest='use_proposals', help='Use Xu et als proposals', action='store_true') parser.add_argument('-nl_obj', dest='nl_obj', help='Num object layers', type=int, default=1) parser.add_argument('-nl_edge', dest='nl_edge', help='Num edge layers', type=int, default=2) parser.add_argument('-hidden_dim', dest='hidden_dim', help='Num edge layers', type=int, default=256) parser.add_argument('-pooling_dim', dest='pooling_dim', help='Dimension of pooling', type=int, default=4096) parser.add_argument('-pass_in_obj_feats_to_decoder', dest='pass_in_obj_feats_to_decoder', action='store_true') parser.add_argument('-pass_in_obj_feats_to_edge', dest='pass_in_obj_feats_to_edge', action='store_true') parser.add_argument('-rec_dropout', dest='rec_dropout', help='recurrent dropout to add', type=float, default=0.1) parser.add_argument('-use_bias', dest='use_bias', action='store_true') parser.add_argument('-use_tanh', dest='use_tanh', action='store_true') parser.add_argument('-limit_vision', dest='limit_vision', action='store_true') return parser
8,445
41.656566
212
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/models/eval_rels.py
from dataloaders.visual_genome import VGDataLoader, VG import numpy as np import torch from config import ModelConfig from lib.pytorch_misc import optimistic_restore from lib.evaluation.sg_eval import BasicSceneGraphEvaluator from tqdm import tqdm from config import BOX_SCALE, IM_SCALE import dill as pkl import os conf = ModelConfig() if conf.model == 'motifnet': from lib.rel_model import RelModel elif conf.model == 'stanford': from lib.rel_model_stanford import RelModelStanford as RelModel else: raise ValueError() train, val, test = VG.splits(num_val_im=conf.val_size, filter_duplicate_rels=True, use_proposals=conf.use_proposals, filter_non_overlap=conf.mode == 'sgdet') if conf.test: val = test train_loader, val_loader = VGDataLoader.splits(train, val, mode='rel', batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) detector = RelModel(classes=train.ind_to_classes, rel_classes=train.ind_to_predicates, num_gpus=conf.num_gpus, mode=conf.mode, require_overlap_det=True, use_resnet=conf.use_resnet, order=conf.order, nl_edge=conf.nl_edge, nl_obj=conf.nl_obj, hidden_dim=conf.hidden_dim, use_proposals=conf.use_proposals, pass_in_obj_feats_to_decoder=conf.pass_in_obj_feats_to_decoder, pass_in_obj_feats_to_edge=conf.pass_in_obj_feats_to_edge, pooling_dim=conf.pooling_dim, rec_dropout=conf.rec_dropout, use_bias=conf.use_bias, use_tanh=conf.use_tanh, limit_vision=conf.limit_vision ) detector.cuda() ckpt = torch.load(conf.ckpt) optimistic_restore(detector, ckpt['state_dict']) # if conf.mode == 'sgdet': # det_ckpt = torch.load('checkpoints/new_vgdet/vg-19.tar')['state_dict'] # detector.detector.bbox_fc.weight.data.copy_(det_ckpt['bbox_fc.weight']) # detector.detector.bbox_fc.bias.data.copy_(det_ckpt['bbox_fc.bias']) # detector.detector.score_fc.weight.data.copy_(det_ckpt['score_fc.weight']) # detector.detector.score_fc.bias.data.copy_(det_ckpt['score_fc.bias']) all_pred_entries = [] def val_batch(batch_num, b, evaluator, thrs=(20, 50, 100)): det_res = detector[b] if conf.num_gpus == 1: det_res = [det_res] for i, (boxes_i, objs_i, obj_scores_i, rels_i, pred_scores_i) in enumerate(det_res): gt_entry = { 'gt_classes': val.gt_classes[batch_num + i].copy(), 'gt_relations': val.relationships[batch_num + i].copy(), 'gt_boxes': val.gt_boxes[batch_num + i].copy(), } assert np.all(objs_i[rels_i[:,0]] > 0) and np.all(objs_i[rels_i[:,1]] > 0) # assert np.all(rels_i[:,2] > 0) pred_entry = { 'pred_boxes': boxes_i * BOX_SCALE/IM_SCALE, 'pred_classes': objs_i, 'pred_rel_inds': rels_i, 'obj_scores': obj_scores_i, 'rel_scores': pred_scores_i, } all_pred_entries.append(pred_entry) evaluator[conf.mode].evaluate_scene_graph_entry( gt_entry, pred_entry, ) evaluator = BasicSceneGraphEvaluator.all_modes(multiple_preds=conf.multi_pred) if conf.cache is not None and os.path.exists(conf.cache): print("Found {}! Loading from it".format(conf.cache)) with open(conf.cache,'rb') as f: all_pred_entries = pkl.load(f) for i, pred_entry in enumerate(tqdm(all_pred_entries)): gt_entry = { 'gt_classes': val.gt_classes[i].copy(), 'gt_relations': val.relationships[i].copy(), 'gt_boxes': val.gt_boxes[i].copy(), } evaluator[conf.mode].evaluate_scene_graph_entry( gt_entry, pred_entry, ) evaluator[conf.mode].print_stats() else: detector.eval() for val_b, batch in enumerate(tqdm(val_loader)): val_batch(conf.num_gpus*val_b, batch, evaluator) evaluator[conf.mode].print_stats() if conf.cache is not None: with open(conf.cache,'wb') as f: pkl.dump(all_pred_entries, f)
4,353
37.530973
89
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/models/train_detector.py
""" Training script 4 Detection """ from dataloaders.mscoco import CocoDetection, CocoDataLoader from dataloaders.visual_genome import VGDataLoader, VG from lib.object_detector import ObjectDetector import numpy as np from torch import optim import torch import pandas as pd import time import os from config import ModelConfig, FG_FRACTION, RPN_FG_FRACTION, IM_SCALE, BOX_SCALE from torch.nn import functional as F from lib.fpn.box_utils import bbox_loss import torch.backends.cudnn as cudnn from pycocotools.cocoeval import COCOeval from lib.pytorch_misc import optimistic_restore, clip_grad_norm from torch.optim.lr_scheduler import ReduceLROnPlateau cudnn.benchmark = True conf = ModelConfig() if conf.coco: train, val = CocoDetection.splits() val.ids = val.ids[:conf.val_size] train.ids = train.ids train_loader, val_loader = CocoDataLoader.splits(train, val, batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) else: train, val, _ = VG.splits(num_val_im=conf.val_size, filter_non_overlap=False, filter_empty_rels=False, use_proposals=conf.use_proposals) train_loader, val_loader = VGDataLoader.splits(train, val, batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) detector = ObjectDetector(classes=train.ind_to_classes, num_gpus=conf.num_gpus, mode='rpntrain' if not conf.use_proposals else 'proposals', use_resnet=conf.use_resnet) detector.cuda() # Note: if you're doing the stanford setup, you'll need to change this to freeze the lower layers if conf.use_proposals: for n, param in detector.named_parameters(): if n.startswith('features'): param.requires_grad = False optimizer = optim.SGD([p for p in detector.parameters() if p.requires_grad], weight_decay=conf.l2, lr=conf.lr * conf.num_gpus * conf.batch_size, momentum=0.9) scheduler = ReduceLROnPlateau(optimizer, 'max', patience=3, factor=0.1, verbose=True, threshold=0.001, threshold_mode='abs', cooldown=1) start_epoch = -1 if conf.ckpt is not None: ckpt = torch.load(conf.ckpt) if optimistic_restore(detector, ckpt['state_dict']): start_epoch = ckpt['epoch'] def train_epoch(epoch_num): detector.train() tr = [] start = time.time() for b, batch in enumerate(train_loader): tr.append(train_batch(batch)) if b % conf.print_interval == 0 and b >= conf.print_interval: mn = pd.concat(tr[-conf.print_interval:], axis=1).mean(1) time_per_batch = (time.time() - start) / conf.print_interval print("\ne{:2d}b{:5d}/{:5d} {:.3f}s/batch, {:.1f}m/epoch".format( epoch_num, b, len(train_loader), time_per_batch, len(train_loader) * time_per_batch / 60)) print(mn) print('-----------', flush=True) start = time.time() return pd.concat(tr, axis=1) def train_batch(b): """ :param b: contains: :param imgs: the image, [batch_size, 3, IM_SIZE, IM_SIZE] :param all_anchors: [num_anchors, 4] the boxes of all anchors that we'll be using :param all_anchor_inds: [num_anchors, 2] array of the indices into the concatenated RPN feature vector that give us all_anchors, each one (img_ind, fpn_idx) :param im_sizes: a [batch_size, 4] numpy array of (h, w, scale, num_good_anchors) for each image. :param num_anchors_per_img: int, number of anchors in total over the feature pyramid per img Training parameters: :param train_anchor_inds: a [num_train, 5] array of indices for the anchors that will be used to compute the training loss (img_ind, fpn_idx) :param gt_boxes: [num_gt, 4] GT boxes over the batch. :param gt_classes: [num_gt, 2] gt boxes where each one is (img_id, class) :return: """ result = detector[b] scores = result.od_obj_dists box_deltas = result.od_box_deltas labels = result.od_obj_labels roi_boxes = result.od_box_priors bbox_targets = result.od_box_targets rpn_scores = result.rpn_scores rpn_box_deltas = result.rpn_box_deltas # detector loss valid_inds = (labels.data != 0).nonzero().squeeze(1) fg_cnt = valid_inds.size(0) bg_cnt = labels.size(0) - fg_cnt class_loss = F.cross_entropy(scores, labels) # No gather_nd in pytorch so instead convert first 2 dims of tensor to 1d box_reg_mult = 2 * (1. / FG_FRACTION) * fg_cnt / (fg_cnt + bg_cnt + 1e-4) twod_inds = valid_inds * box_deltas.size(1) + labels[valid_inds].data box_loss = bbox_loss(roi_boxes[valid_inds], box_deltas.view(-1, 4)[twod_inds], bbox_targets[valid_inds]) * box_reg_mult loss = class_loss + box_loss # RPN loss if not conf.use_proposals: train_anchor_labels = b.train_anchor_labels[:, -1] train_anchors = b.train_anchors[:, :4] train_anchor_targets = b.train_anchors[:, 4:] train_valid_inds = (train_anchor_labels.data == 1).nonzero().squeeze(1) rpn_class_loss = F.cross_entropy(rpn_scores, train_anchor_labels) # print("{} fg {} bg, ratio of {:.3f} vs {:.3f}. RPN {}fg {}bg ratio of {:.3f} vs {:.3f}".format( # fg_cnt, bg_cnt, fg_cnt / (fg_cnt + bg_cnt + 1e-4), FG_FRACTION, # train_valid_inds.size(0), train_anchor_labels.size(0)-train_valid_inds.size(0), # train_valid_inds.size(0) / (train_anchor_labels.size(0) + 1e-4), RPN_FG_FRACTION), flush=True) rpn_box_mult = 2 * (1. / RPN_FG_FRACTION) * train_valid_inds.size(0) / (train_anchor_labels.size(0) + 1e-4) rpn_box_loss = bbox_loss(train_anchors[train_valid_inds], rpn_box_deltas[train_valid_inds], train_anchor_targets[train_valid_inds]) * rpn_box_mult loss += rpn_class_loss + rpn_box_loss res = pd.Series([rpn_class_loss.data[0], rpn_box_loss.data[0], class_loss.data[0], box_loss.data[0], loss.data[0]], ['rpn_class_loss', 'rpn_box_loss', 'class_loss', 'box_loss', 'total']) else: res = pd.Series([class_loss.data[0], box_loss.data[0], loss.data[0]], ['class_loss', 'box_loss', 'total']) optimizer.zero_grad() loss.backward() clip_grad_norm( [(n, p) for n, p in detector.named_parameters() if p.grad is not None], max_norm=conf.clip, clip=True) optimizer.step() return res def val_epoch(): detector.eval() # all_boxes is a list of length number-of-classes. # Each list element is a list of length number-of-images. # Each of those list elements is either an empty list [] # or a numpy array of detection. vr = [] for val_b, batch in enumerate(val_loader): vr.append(val_batch(val_b, batch)) vr = np.concatenate(vr, 0) if vr.shape[0] == 0: print("No detections anywhere") return 0.0 val_coco = val.coco coco_dt = val_coco.loadRes(vr) coco_eval = COCOeval(val_coco, coco_dt, 'bbox') coco_eval.params.imgIds = val.ids if conf.coco else [x for x in range(len(val))] coco_eval.evaluate() coco_eval.accumulate() coco_eval.summarize() mAp = coco_eval.stats[1] return mAp def val_batch(batch_num, b): result = detector[b] if result is None: return np.zeros((0, 7)) scores_np = result.obj_scores.data.cpu().numpy() cls_preds_np = result.obj_preds.data.cpu().numpy() boxes_np = result.boxes_assigned.data.cpu().numpy() im_inds_np = result.im_inds.data.cpu().numpy() im_scales = b.im_sizes.reshape((-1, 3))[:, 2] if conf.coco: boxes_np /= im_scales[im_inds_np][:, None] boxes_np[:, 2:4] = boxes_np[:, 2:4] - boxes_np[:, 0:2] + 1 cls_preds_np[:] = [val.ind_to_id[c_ind] for c_ind in cls_preds_np] im_inds_np[:] = [val.ids[im_ind + batch_num * conf.batch_size * conf.num_gpus] for im_ind in im_inds_np] else: boxes_np *= BOX_SCALE / IM_SCALE boxes_np[:, 2:4] = boxes_np[:, 2:4] - boxes_np[:, 0:2] + 1 im_inds_np += batch_num * conf.batch_size * conf.num_gpus return np.column_stack((im_inds_np, boxes_np, scores_np, cls_preds_np)) print("Training starts now!") for epoch in range(start_epoch + 1, start_epoch + 1 + conf.num_epochs): rez = train_epoch(epoch) print("overall{:2d}: ({:.3f})\n{}".format(epoch, rez.mean(1)['total'], rez.mean(1)), flush=True) mAp = val_epoch() scheduler.step(mAp) torch.save({ 'epoch': epoch, 'state_dict': detector.state_dict(), 'optimizer': optimizer.state_dict(), }, os.path.join(conf.save_dir, '{}-{}.tar'.format('coco' if conf.coco else 'vg', epoch)))
9,155
40.808219
115
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/models/eval_rel_count.py
""" Baseline model that works by simply iterating through the training set to make a dictionary. Also, caches this (we can use this for training). The model is quite simple, so we don't use the base train/test code """ from dataloaders.visual_genome import VGDataLoader, VG from lib.object_detector import ObjectDetector import numpy as np import torch import os from lib.get_dataset_counts import get_counts, box_filter from config import ModelConfig, FG_FRACTION, RPN_FG_FRACTION, DATA_PATH, BOX_SCALE, IM_SCALE, PROPOSAL_FN import torch.backends.cudnn as cudnn from lib.pytorch_misc import optimistic_restore, nonintersecting_2d_inds from lib.evaluation.sg_eval import BasicSceneGraphEvaluator from tqdm import tqdm from copy import deepcopy import dill as pkl cudnn.benchmark = True conf = ModelConfig() MUST_OVERLAP=False train, val, test = VG.splits(num_val_im=conf.val_size, filter_non_overlap=MUST_OVERLAP, filter_duplicate_rels=True, use_proposals=conf.use_proposals) if conf.test: print("test data!") val = test train_loader, val_loader = VGDataLoader.splits(train, val, mode='rel', batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) fg_matrix, bg_matrix = get_counts(train_data=train, must_overlap=MUST_OVERLAP) detector = ObjectDetector(classes=train.ind_to_classes, num_gpus=conf.num_gpus, mode='rpntrain' if not conf.use_proposals else 'proposals', use_resnet=conf.use_resnet, nms_filter_duplicates=True, thresh=0.01) detector.eval() detector.cuda() classifier = ObjectDetector(classes=train.ind_to_classes, num_gpus=conf.num_gpus, mode='gtbox', use_resnet=conf.use_resnet, nms_filter_duplicates=True, thresh=0.01) classifier.eval() classifier.cuda() ckpt = torch.load(conf.ckpt) mismatch = optimistic_restore(detector, ckpt['state_dict']) mismatch = optimistic_restore(classifier, ckpt['state_dict']) MOST_COMMON_MODE = True if MOST_COMMON_MODE: prob_matrix = fg_matrix.astype(np.float32) prob_matrix[:,:,0] = bg_matrix # TRYING SOMETHING NEW. prob_matrix[:,:,0] += 1 prob_matrix /= np.sum(prob_matrix, 2)[:,:,None] # prob_matrix /= float(fg_matrix.max()) np.save(os.path.join(DATA_PATH, 'pred_stats.npy'), prob_matrix) prob_matrix[:,:,0] = 0 # Zero out BG else: prob_matrix = fg_matrix.astype(np.float64) prob_matrix = prob_matrix / prob_matrix.max(2)[:,:,None] np.save(os.path.join(DATA_PATH, 'pred_dist.npy'), prob_matrix) # It's test time! def predict(boxes, classes): relation_possibilities_ = np.array(box_filter(boxes, must_overlap=MUST_OVERLAP), dtype=int) full_preds = np.zeros((boxes.shape[0], boxes.shape[0], train.num_predicates)) for o1, o2 in relation_possibilities_: c1, c2 = classes[[o1, o2]] full_preds[o1, o2] = prob_matrix[c1, c2] full_preds[:,:,0] = 0.0 # Zero out BG. return full_preds # ########################################################################################## # ########################################################################################## # For visualizing / exploring c_to_ind = {c: i for i, c in enumerate(train.ind_to_classes)} def gimme_the_dist(c1name, c2name): c1 = c_to_ind[c1name] c2 = c_to_ind[c2name] dist = prob_matrix[c1, c2] argz = np.argsort(-dist) for i, a in enumerate(argz): if dist[a] > 0.0: print("{:3d}: {:10s} ({:.4f})".format(i, train.ind_to_predicates[a], dist[a])) counts = np.zeros((train.num_classes, train.num_classes, train.num_predicates), dtype=np.int64) for ex_ind in tqdm(range(len(val))): gt_relations = val.relationships[ex_ind].copy() gt_classes = val.gt_classes[ex_ind].copy() o1o2 = gt_classes[gt_relations[:, :2]].tolist() for (o1, o2), pred in zip(o1o2, gt_relations[:, 2]): counts[o1, o2, pred] += 1 zeroshot_case = counts[np.where(prob_matrix == 0)].sum() / float(counts.sum()) max_inds = prob_matrix.argmax(2).ravel() max_counts = counts.reshape(-1, 51)[np.arange(max_inds.shape[0]), max_inds] most_freq_port = max_counts.sum()/float(counts.sum()) print(" Rel acc={:.2f}%, {:.2f}% zsl".format( most_freq_port*100, zeroshot_case*100)) # ########################################################################################## # ########################################################################################## T = len(val) evaluator = BasicSceneGraphEvaluator.all_modes(multiple_preds=conf.multi_pred) # First do detection results img_offset = 0 all_pred_entries = {'sgdet':[], 'sgcls':[], 'predcls':[]} for val_b, b in enumerate(tqdm(val_loader)): det_result = detector[b] img_ids = b.gt_classes_primary.data.cpu().numpy()[:,0] scores_np = det_result.obj_scores.data.cpu().numpy() cls_preds_np = det_result.obj_preds.data.cpu().numpy() boxes_np = det_result.boxes_assigned.data.cpu().numpy()* BOX_SCALE/IM_SCALE # boxpriors_np = det_result.box_priors.data.cpu().numpy() im_inds_np = det_result.im_inds.data.cpu().numpy() + img_offset for img_i in np.unique(img_ids + img_offset): gt_entry = { 'gt_classes': val.gt_classes[img_i].copy(), 'gt_relations': val.relationships[img_i].copy(), 'gt_boxes': val.gt_boxes[img_i].copy(), } pred_boxes = boxes_np[im_inds_np == img_i] pred_classes = cls_preds_np[im_inds_np == img_i] obj_scores = scores_np[im_inds_np == img_i] all_rels = nonintersecting_2d_inds(pred_boxes.shape[0]) fp = predict(pred_boxes, pred_classes) fp_pred = fp[all_rels[:,0], all_rels[:,1]] scores = np.column_stack(( obj_scores[all_rels[:,0]], obj_scores[all_rels[:,1]], fp_pred.max(1) )).prod(1) sorted_inds = np.argsort(-scores) sorted_inds = sorted_inds[scores[sorted_inds] > 0] #[:100] pred_entry = { 'pred_boxes': pred_boxes, 'pred_classes': pred_classes, 'obj_scores': obj_scores, 'pred_rel_inds': all_rels[sorted_inds], 'rel_scores': fp_pred[sorted_inds], } all_pred_entries['sgdet'].append(pred_entry) evaluator['sgdet'].evaluate_scene_graph_entry( gt_entry, pred_entry, ) img_offset += img_ids.max() + 1 evaluator['sgdet'].print_stats() # ----------------------------------------------------------------------------------------- # EVAL CLS AND SG img_offset = 0 for val_b, b in enumerate(tqdm(val_loader)): det_result = classifier[b] scores, cls_preds = det_result.rm_obj_dists[:,1:].data.max(1) scores_np = scores.cpu().numpy() cls_preds_np = (cls_preds+1).cpu().numpy() img_ids = b.gt_classes_primary.data.cpu().numpy()[:,0] boxes_np = b.gt_boxes_primary.data.cpu().numpy() im_inds_np = det_result.im_inds.data.cpu().numpy() + img_offset for img_i in np.unique(img_ids + img_offset): gt_entry = { 'gt_classes': val.gt_classes[img_i].copy(), 'gt_relations': val.relationships[img_i].copy(), 'gt_boxes': val.gt_boxes[img_i].copy(), } pred_boxes = boxes_np[im_inds_np == img_i] pred_classes = cls_preds_np[im_inds_np == img_i] obj_scores = scores_np[im_inds_np == img_i] all_rels = nonintersecting_2d_inds(pred_boxes.shape[0]) fp = predict(pred_boxes, pred_classes) fp_pred = fp[all_rels[:,0], all_rels[:,1]] sg_cls_scores = np.column_stack(( obj_scores[all_rels[:,0]], obj_scores[all_rels[:,1]], fp_pred.max(1) )).prod(1) sg_cls_inds = np.argsort(-sg_cls_scores) sg_cls_inds = sg_cls_inds[sg_cls_scores[sg_cls_inds] > 0] #[:100] pred_entry = { 'pred_boxes': pred_boxes, 'pred_classes': pred_classes, 'obj_scores': obj_scores, 'pred_rel_inds': all_rels[sg_cls_inds], 'rel_scores': fp_pred[sg_cls_inds], } all_pred_entries['sgcls'].append(deepcopy(pred_entry)) evaluator['sgcls'].evaluate_scene_graph_entry( gt_entry, pred_entry, ) ######################################################## fp = predict(gt_entry['gt_boxes'], gt_entry['gt_classes']) fp_pred = fp[all_rels[:, 0], all_rels[:, 1]] pred_cls_scores = fp_pred.max(1) pred_cls_inds = np.argsort(-pred_cls_scores) pred_cls_inds = pred_cls_inds[pred_cls_scores[pred_cls_inds] > 0][:100] pred_entry['pred_rel_inds'] = all_rels[pred_cls_inds] pred_entry['rel_scores'] = fp_pred[pred_cls_inds] pred_entry['pred_classes'] = gt_entry['gt_classes'] pred_entry['obj_scores'] = np.ones(pred_entry['pred_classes'].shape[0]) all_pred_entries['predcls'].append(pred_entry) evaluator['predcls'].evaluate_scene_graph_entry( gt_entry, pred_entry, ) img_offset += img_ids.max() + 1 evaluator['predcls'].print_stats() evaluator['sgcls'].print_stats() for mode, entries in all_pred_entries.items(): with open('caches/freqbaseline-{}-{}.pkl'.format('overlap' if MUST_OVERLAP else 'nonoverlap', mode), 'wb') as f: pkl.dump(entries, f)
9,552
36.758893
116
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/models/_visualize.py
""" Visualization script. I used this to create the figures in the paper. WARNING: I haven't tested this in a while. It's possible that some later features I added break things here, but hopefully there should be easy fixes. I'm uploading this in the off chance it might help someone. If you get it to work, let me know (and also send a PR with bugs/etc) """ from dataloaders.visual_genome import VGDataLoader, VG from lib.rel_model import RelModel import numpy as np import torch from config import ModelConfig from lib.pytorch_misc import optimistic_restore from lib.evaluation.sg_eval import BasicSceneGraphEvaluator from tqdm import tqdm from config import BOX_SCALE, IM_SCALE from lib.fpn.box_utils import bbox_overlaps from collections import defaultdict from PIL import Image, ImageDraw, ImageFont import os from functools import reduce conf = ModelConfig() train, val, test = VG.splits(num_val_im=conf.val_size) if conf.test: val = test train_loader, val_loader = VGDataLoader.splits(train, val, mode='rel', batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) detector = RelModel(classes=train.ind_to_classes, rel_classes=train.ind_to_predicates, num_gpus=conf.num_gpus, mode=conf.mode, require_overlap_det=True, use_resnet=conf.use_resnet, order=conf.order, nl_edge=conf.nl_edge, nl_obj=conf.nl_obj, hidden_dim=conf.hidden_dim, use_proposals=conf.use_proposals, pass_in_obj_feats_to_decoder=conf.pass_in_obj_feats_to_decoder, pass_in_obj_feats_to_edge=conf.pass_in_obj_feats_to_edge, pooling_dim=conf.pooling_dim, rec_dropout=conf.rec_dropout, use_bias=conf.use_bias, use_tanh=conf.use_tanh, limit_vision=conf.limit_vision ) detector.cuda() ckpt = torch.load(conf.ckpt) optimistic_restore(detector, ckpt['state_dict']) ############################################ HELPER FUNCTIONS ################################### def get_cmap(N): import matplotlib.cm as cmx import matplotlib.colors as colors """Returns a function that maps each index in 0, 1, ... N-1 to a distinct RGB color.""" color_norm = colors.Normalize(vmin=0, vmax=N - 1) scalar_map = cmx.ScalarMappable(norm=color_norm, cmap='hsv') def map_index_to_rgb_color(index): pad = 40 return np.round(np.array(scalar_map.to_rgba(index)) * (255 - pad) + pad) return map_index_to_rgb_color cmap = get_cmap(len(train.ind_to_classes) + 1) def load_unscaled(fn): """ Loads and scales images so that it's 1024 max-dimension""" image_unpadded = Image.open(fn).convert('RGB') im_scale = 1024.0 / max(image_unpadded.size) image = image_unpadded.resize((int(im_scale * image_unpadded.size[0]), int(im_scale * image_unpadded.size[1])), resample=Image.BICUBIC) return image font = ImageFont.truetype('/usr/share/fonts/truetype/freefont/FreeMonoBold.ttf', 32) def draw_box(draw, boxx, cls_ind, text_str): box = tuple([float(b) for b in boxx]) if '-GT' in text_str: color = (255, 128, 0, 255) else: color = (0, 128, 0, 255) # color = tuple([int(x) for x in cmap(cls_ind)]) # draw the fucking box draw.line([(box[0], box[1]), (box[2], box[1])], fill=color, width=8) draw.line([(box[2], box[1]), (box[2], box[3])], fill=color, width=8) draw.line([(box[2], box[3]), (box[0], box[3])], fill=color, width=8) draw.line([(box[0], box[3]), (box[0], box[1])], fill=color, width=8) # draw.rectangle(box, outline=color) w, h = draw.textsize(text_str, font=font) x1text = box[0] y1text = max(box[1] - h, 0) x2text = min(x1text + w, draw.im.size[0]) y2text = y1text + h print("drawing {}x{} rectangle at {:.1f} {:.1f} {:.1f} {:.1f}".format( h, w, x1text, y1text, x2text, y2text)) draw.rectangle((x1text, y1text, x2text, y2text), fill=color) draw.text((x1text, y1text), text_str, fill='black', font=font) return draw def val_epoch(): detector.eval() evaluator = BasicSceneGraphEvaluator.all_modes() for val_b, batch in enumerate(tqdm(val_loader)): val_batch(conf.num_gpus * val_b, batch, evaluator) evaluator[conf.mode].print_stats() def val_batch(batch_num, b, evaluator, thrs=(20, 50, 100)): det_res = detector[b] # if conf.num_gpus == 1: # det_res = [det_res] assert conf.num_gpus == 1 boxes_i, objs_i, obj_scores_i, rels_i, pred_scores_i = det_res gt_entry = { 'gt_classes': val.gt_classes[batch_num].copy(), 'gt_relations': val.relationships[batch_num].copy(), 'gt_boxes': val.gt_boxes[batch_num].copy(), } # gt_entry = {'gt_classes': gtc[i], 'gt_relations': gtr[i], 'gt_boxes': gtb[i]} assert np.all(objs_i[rels_i[:, 0]] > 0) and np.all(objs_i[rels_i[:, 1]] > 0) # assert np.all(rels_i[:, 2] > 0) pred_entry = { 'pred_boxes': boxes_i * BOX_SCALE / IM_SCALE, 'pred_classes': objs_i, 'pred_rel_inds': rels_i, 'obj_scores': obj_scores_i, 'rel_scores': pred_scores_i, } pred_to_gt, pred_5ples, rel_scores = evaluator[conf.mode].evaluate_scene_graph_entry( gt_entry, pred_entry, ) # SET RECALL THRESHOLD HERE pred_to_gt = pred_to_gt[:20] pred_5ples = pred_5ples[:20] # Get a list of objects that match, and GT objects that dont objs_match = (bbox_overlaps(pred_entry['pred_boxes'], gt_entry['gt_boxes']) >= 0.5) & ( objs_i[:, None] == gt_entry['gt_classes'][None] ) objs_matched = objs_match.any(1) has_seen = defaultdict(int) has_seen_gt = defaultdict(int) pred_ind2name = {} gt_ind2name = {} edges = {} missededges = {} badedges = {} if val.filenames[batch_num].startswith('2343676'): import ipdb ipdb.set_trace() def query_pred(pred_ind): if pred_ind not in pred_ind2name: has_seen[objs_i[pred_ind]] += 1 pred_ind2name[pred_ind] = '{}-{}'.format(train.ind_to_classes[objs_i[pred_ind]], has_seen[objs_i[pred_ind]]) return pred_ind2name[pred_ind] def query_gt(gt_ind): gt_cls = gt_entry['gt_classes'][gt_ind] if gt_ind not in gt_ind2name: has_seen_gt[gt_cls] += 1 gt_ind2name[gt_ind] = '{}-GT{}'.format(train.ind_to_classes[gt_cls], has_seen_gt[gt_cls]) return gt_ind2name[gt_ind] matching_pred5ples = pred_5ples[np.array([len(x) > 0 for x in pred_to_gt])] for fiveple in matching_pred5ples: head_name = query_pred(fiveple[0]) tail_name = query_pred(fiveple[1]) edges[(head_name, tail_name)] = train.ind_to_predicates[fiveple[4]] gt_5ples = np.column_stack((gt_entry['gt_relations'][:, :2], gt_entry['gt_classes'][gt_entry['gt_relations'][:, 0]], gt_entry['gt_classes'][gt_entry['gt_relations'][:, 1]], gt_entry['gt_relations'][:, 2], )) has_match = reduce(np.union1d, pred_to_gt) for gt in gt_5ples[np.setdiff1d(np.arange(gt_5ples.shape[0]), has_match)]: # Head and tail namez = [] for i in range(2): matching_obj = np.where(objs_match[:, gt[i]])[0] if matching_obj.size > 0: name = query_pred(matching_obj[0]) else: name = query_gt(gt[i]) namez.append(name) missededges[tuple(namez)] = train.ind_to_predicates[gt[4]] for fiveple in pred_5ples[np.setdiff1d(np.arange(pred_5ples.shape[0]), matching_pred5ples)]: if fiveple[0] in pred_ind2name: if fiveple[1] in pred_ind2name: badedges[(pred_ind2name[fiveple[0]], pred_ind2name[fiveple[1]])] = train.ind_to_predicates[fiveple[4]] theimg = load_unscaled(val.filenames[batch_num]) theimg2 = theimg.copy() draw2 = ImageDraw.Draw(theimg2) # Fix the names for pred_ind in pred_ind2name.keys(): draw2 = draw_box(draw2, pred_entry['pred_boxes'][pred_ind], cls_ind=objs_i[pred_ind], text_str=pred_ind2name[pred_ind]) for gt_ind in gt_ind2name.keys(): draw2 = draw_box(draw2, gt_entry['gt_boxes'][gt_ind], cls_ind=gt_entry['gt_classes'][gt_ind], text_str=gt_ind2name[gt_ind]) recall = int(100 * len(reduce(np.union1d, pred_to_gt)) / gt_entry['gt_relations'].shape[0]) id = '{}-{}'.format(val.filenames[batch_num].split('/')[-1][:-4], recall) pathname = os.path.join('qualitative', id) if not os.path.exists(pathname): os.mkdir(pathname) theimg.save(os.path.join(pathname, 'img.jpg'), quality=100, subsampling=0) theimg2.save(os.path.join(pathname, 'imgbox.jpg'), quality=100, subsampling=0) with open(os.path.join(pathname, 'shit.txt'), 'w') as f: f.write('good:\n') for (o1, o2), p in edges.items(): f.write('{} - {} - {}\n'.format(o1, p, o2)) f.write('fn:\n') for (o1, o2), p in missededges.items(): f.write('{} - {} - {}\n'.format(o1, p, o2)) f.write('shit:\n') for (o1, o2), p in badedges.items(): f.write('{} - {} - {}\n'.format(o1, p, o2)) mAp = val_epoch()
9,693
36.867188
280
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/models/train_rels.py
""" Training script for scene graph detection. Integrated with my faster rcnn setup """ from dataloaders.visual_genome import VGDataLoader, VG import numpy as np from torch import optim import torch import pandas as pd import time import os from config import ModelConfig, BOX_SCALE, IM_SCALE from torch.nn import functional as F from lib.pytorch_misc import optimistic_restore, de_chunkize, clip_grad_norm from lib.evaluation.sg_eval import BasicSceneGraphEvaluator from lib.pytorch_misc import print_para from torch.optim.lr_scheduler import ReduceLROnPlateau conf = ModelConfig() if conf.model == 'motifnet': from lib.rel_model import RelModel elif conf.model == 'stanford': from lib.rel_model_stanford import RelModelStanford as RelModel else: raise ValueError() train, val, _ = VG.splits(num_val_im=conf.val_size, filter_duplicate_rels=True, use_proposals=conf.use_proposals, filter_non_overlap=conf.mode == 'sgdet') train_loader, val_loader = VGDataLoader.splits(train, val, mode='rel', batch_size=conf.batch_size, num_workers=conf.num_workers, num_gpus=conf.num_gpus) detector = RelModel(classes=train.ind_to_classes, rel_classes=train.ind_to_predicates, num_gpus=conf.num_gpus, mode=conf.mode, require_overlap_det=True, use_resnet=conf.use_resnet, order=conf.order, nl_edge=conf.nl_edge, nl_obj=conf.nl_obj, hidden_dim=conf.hidden_dim, use_proposals=conf.use_proposals, pass_in_obj_feats_to_decoder=conf.pass_in_obj_feats_to_decoder, pass_in_obj_feats_to_edge=conf.pass_in_obj_feats_to_edge, pooling_dim=conf.pooling_dim, rec_dropout=conf.rec_dropout, use_bias=conf.use_bias, use_tanh=conf.use_tanh, limit_vision=conf.limit_vision ) # Freeze the detector for n, param in detector.detector.named_parameters(): param.requires_grad = False print(print_para(detector), flush=True) def get_optim(lr): # Lower the learning rate on the VGG fully connected layers by 1/10th. It's a hack, but it helps # stabilize the cores. fc_params = [p for n,p in detector.named_parameters() if n.startswith('roi_fmap') and p.requires_grad] non_fc_params = [p for n,p in detector.named_parameters() if not n.startswith('roi_fmap') and p.requires_grad] params = [{'params': fc_params, 'lr': lr / 10.0}, {'params': non_fc_params}] # params = [p for n,p in detector.named_parameters() if p.requires_grad] if conf.adam: optimizer = optim.Adam(params, weight_decay=conf.l2, lr=lr, eps=1e-3) else: optimizer = optim.SGD(params, weight_decay=conf.l2, lr=lr, momentum=0.9) scheduler = ReduceLROnPlateau(optimizer, 'max', patience=3, factor=0.1, verbose=True, threshold=0.0001, threshold_mode='abs', cooldown=1) return optimizer, scheduler ckpt = torch.load(conf.ckpt) if conf.ckpt.split('-')[-2].split('/')[-1] == 'vgrel': print("Loading EVERYTHING") start_epoch = ckpt['epoch'] if not optimistic_restore(detector, ckpt['state_dict']): start_epoch = -1 # optimistic_restore(detector.detector, torch.load('checkpoints/vgdet/vg-28.tar')['state_dict']) else: start_epoch = -1 optimistic_restore(detector.detector, ckpt['state_dict']) detector.roi_fmap[1][0].weight.data.copy_(ckpt['state_dict']['roi_fmap.0.weight']) detector.roi_fmap[1][3].weight.data.copy_(ckpt['state_dict']['roi_fmap.3.weight']) detector.roi_fmap[1][0].bias.data.copy_(ckpt['state_dict']['roi_fmap.0.bias']) detector.roi_fmap[1][3].bias.data.copy_(ckpt['state_dict']['roi_fmap.3.bias']) detector.roi_fmap_obj[0].weight.data.copy_(ckpt['state_dict']['roi_fmap.0.weight']) detector.roi_fmap_obj[3].weight.data.copy_(ckpt['state_dict']['roi_fmap.3.weight']) detector.roi_fmap_obj[0].bias.data.copy_(ckpt['state_dict']['roi_fmap.0.bias']) detector.roi_fmap_obj[3].bias.data.copy_(ckpt['state_dict']['roi_fmap.3.bias']) detector.cuda() def train_epoch(epoch_num): detector.train() tr = [] start = time.time() for b, batch in enumerate(train_loader): tr.append(train_batch(batch, verbose=b % (conf.print_interval*10) == 0)) #b == 0)) if b % conf.print_interval == 0 and b >= conf.print_interval: mn = pd.concat(tr[-conf.print_interval:], axis=1).mean(1) time_per_batch = (time.time() - start) / conf.print_interval print("\ne{:2d}b{:5d}/{:5d} {:.3f}s/batch, {:.1f}m/epoch".format( epoch_num, b, len(train_loader), time_per_batch, len(train_loader) * time_per_batch / 60)) print(mn) print('-----------', flush=True) start = time.time() return pd.concat(tr, axis=1) def train_batch(b, verbose=False): """ :param b: contains: :param imgs: the image, [batch_size, 3, IM_SIZE, IM_SIZE] :param all_anchors: [num_anchors, 4] the boxes of all anchors that we'll be using :param all_anchor_inds: [num_anchors, 2] array of the indices into the concatenated RPN feature vector that give us all_anchors, each one (img_ind, fpn_idx) :param im_sizes: a [batch_size, 4] numpy array of (h, w, scale, num_good_anchors) for each image. :param num_anchors_per_img: int, number of anchors in total over the feature pyramid per img Training parameters: :param train_anchor_inds: a [num_train, 5] array of indices for the anchors that will be used to compute the training loss (img_ind, fpn_idx) :param gt_boxes: [num_gt, 4] GT boxes over the batch. :param gt_classes: [num_gt, 2] gt boxes where each one is (img_id, class) :return: """ result = detector[b] losses = {} losses['class_loss'] = F.cross_entropy(result.rm_obj_dists, result.rm_obj_labels) losses['rel_loss'] = F.cross_entropy(result.rel_dists, result.rel_labels[:, -1]) loss = sum(losses.values()) optimizer.zero_grad() loss.backward() clip_grad_norm( [(n, p) for n, p in detector.named_parameters() if p.grad is not None], max_norm=conf.clip, verbose=verbose, clip=True) losses['total'] = loss optimizer.step() res = pd.Series({x: y.data[0] for x, y in losses.items()}) return res def val_epoch(): detector.eval() evaluator = BasicSceneGraphEvaluator.all_modes() for val_b, batch in enumerate(val_loader): val_batch(conf.num_gpus * val_b, batch, evaluator) evaluator[conf.mode].print_stats() return np.mean(evaluator[conf.mode].result_dict[conf.mode + '_recall'][100]) def val_batch(batch_num, b, evaluator): det_res = detector[b] if conf.num_gpus == 1: det_res = [det_res] for i, (boxes_i, objs_i, obj_scores_i, rels_i, pred_scores_i) in enumerate(det_res): gt_entry = { 'gt_classes': val.gt_classes[batch_num + i].copy(), 'gt_relations': val.relationships[batch_num + i].copy(), 'gt_boxes': val.gt_boxes[batch_num + i].copy(), } assert np.all(objs_i[rels_i[:, 0]] > 0) and np.all(objs_i[rels_i[:, 1]] > 0) pred_entry = { 'pred_boxes': boxes_i * BOX_SCALE/IM_SCALE, 'pred_classes': objs_i, 'pred_rel_inds': rels_i, 'obj_scores': obj_scores_i, 'rel_scores': pred_scores_i, # hack for now. } evaluator[conf.mode].evaluate_scene_graph_entry( gt_entry, pred_entry, ) print("Training starts now!") optimizer, scheduler = get_optim(conf.lr * conf.num_gpus * conf.batch_size) for epoch in range(start_epoch + 1, start_epoch + 1 + conf.num_epochs): rez = train_epoch(epoch) print("overall{:2d}: ({:.3f})\n{}".format(epoch, rez.mean(1)['total'], rez.mean(1)), flush=True) if conf.save_dir is not None: torch.save({ 'epoch': epoch, 'state_dict': detector.state_dict(), #{k:v for k,v in detector.state_dict().items() if not k.startswith('detector.')}, # 'optimizer': optimizer.state_dict(), }, os.path.join(conf.save_dir, '{}-{}.tar'.format('vgrel', epoch))) mAp = val_epoch() scheduler.step(mAp) if any([pg['lr'] <= (conf.lr * conf.num_gpus * conf.batch_size)/99.0 for pg in optimizer.param_groups]): print("exiting training early", flush=True) break
8,782
41.225962
130
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/dataloaders/visual_genome.py
""" File that involves dataloaders for the Visual Genome example_dataset. """ import json import os import h5py import numpy as np import torch from PIL import Image from torch.utils.data import Dataset from torchvision.transforms import Resize, Compose, ToTensor, Normalize from dataloaders.blob import Blob from lib.fpn.box_intersections_cpu.bbox import bbox_overlaps from config import VG_IMAGES, IM_DATA_FN, VG_SGG_FN, VG_SGG_DICT_FN, BOX_SCALE, IM_SCALE, PROPOSAL_FN from dataloaders.image_transforms import SquarePad, Grayscale, Brightness, Sharpness, Contrast, \ RandomOrder, Hue, random_crop from collections import defaultdict from pycocotools.coco import COCO class VG(Dataset): def __init__(self, mode, roidb_file=VG_SGG_FN, dict_file=VG_SGG_DICT_FN, image_file=IM_DATA_FN, filter_empty_rels=True, num_im=-1, num_val_im=5000, filter_duplicate_rels=True, filter_non_overlap=True, use_proposals=False): """ Torch example_dataset for VisualGenome :param mode: Must be train, test, or val :param roidb_file: HDF5 containing the GT boxes, classes, and relationships :param dict_file: JSON Contains mapping of classes/relationships to words :param image_file: HDF5 containing image filenames :param filter_empty_rels: True if we filter out images without relationships between boxes. One might want to set this to false if training a detector. :param filter_duplicate_rels: Whenever we see a duplicate relationship we'll sample instead :param num_im: Number of images in the entire example_dataset. -1 for all images. :param num_val_im: Number of images in the validation set (must be less than num_im unless num_im is -1.) :param proposal_file: If None, we don't provide proposals. Otherwise file for where we get RPN proposals """ if mode not in ('test', 'train', 'val'): raise ValueError("Mode must be in test, train, or val. Supplied {}".format(mode)) self.mode = mode # Initialize self.roidb_file = roidb_file self.dict_file = dict_file self.image_file = image_file self.filter_non_overlap = filter_non_overlap self.filter_duplicate_rels = filter_duplicate_rels and self.mode == 'train' self.split_mask, self.gt_boxes, self.gt_classes, self.relationships = load_graphs( self.roidb_file, self.mode, num_im, num_val_im=num_val_im, filter_empty_rels=filter_empty_rels, filter_non_overlap=self.filter_non_overlap and self.is_train, ) self.filenames = load_image_filenames(image_file) self.filenames = [self.filenames[i] for i in np.where(self.split_mask)[0]] self.ind_to_classes, self.ind_to_predicates = load_info(dict_file) if use_proposals: print("Loading proposals", flush=True) p_h5 = h5py.File(PROPOSAL_FN, 'r') rpn_rois = p_h5['rpn_rois'] rpn_scores = p_h5['rpn_scores'] rpn_im_to_roi_idx = np.array(p_h5['im_to_roi_idx'][self.split_mask]) rpn_num_rois = np.array(p_h5['num_rois'][self.split_mask]) self.rpn_rois = [] for i in range(len(self.filenames)): rpn_i = np.column_stack(( rpn_scores[rpn_im_to_roi_idx[i]:rpn_im_to_roi_idx[i] + rpn_num_rois[i]], rpn_rois[rpn_im_to_roi_idx[i]:rpn_im_to_roi_idx[i] + rpn_num_rois[i]], )) self.rpn_rois.append(rpn_i) else: self.rpn_rois = None # You could add data augmentation here. But we didn't. # tform = [] # if self.is_train: # tform.append(RandomOrder([ # Grayscale(), # Brightness(), # Contrast(), # Sharpness(), # Hue(), # ])) tform = [ SquarePad(), Resize(IM_SCALE), ToTensor(), Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), ] self.transform_pipeline = Compose(tform) @property def coco(self): """ :return: a Coco-like object that we can use to evaluate detection! """ anns = [] for i, (cls_array, box_array) in enumerate(zip(self.gt_classes, self.gt_boxes)): for cls, box in zip(cls_array.tolist(), box_array.tolist()): anns.append({ 'area': (box[3] - box[1] + 1) * (box[2] - box[0] + 1), 'bbox': [box[0], box[1], box[2] - box[0] + 1, box[3] - box[1] + 1], 'category_id': cls, 'id': len(anns), 'image_id': i, 'iscrowd': 0, }) fauxcoco = COCO() fauxcoco.dataset = { 'info': {'description': 'ayy lmao'}, 'images': [{'id': i} for i in range(self.__len__())], 'categories': [{'supercategory': 'person', 'id': i, 'name': name} for i, name in enumerate(self.ind_to_classes) if name != '__background__'], 'annotations': anns, } fauxcoco.createIndex() return fauxcoco @property def is_train(self): return self.mode.startswith('train') @classmethod def splits(cls, *args, **kwargs): """ Helper method to generate splits of the example_dataset""" train = cls('train', *args, **kwargs) val = cls('val', *args, **kwargs) test = cls('test', *args, **kwargs) return train, val, test def __getitem__(self, index): image_unpadded = Image.open(self.filenames[index]).convert('RGB') # Optionally flip the image if we're doing training flipped = self.is_train and np.random.random() > 0.5 gt_boxes = self.gt_boxes[index].copy() # Boxes are already at BOX_SCALE if self.is_train: # crop boxes that are too large. This seems to be only a problem for image heights, but whatevs gt_boxes[:, [1, 3]] = gt_boxes[:, [1, 3]].clip( None, BOX_SCALE / max(image_unpadded.size) * image_unpadded.size[1]) gt_boxes[:, [0, 2]] = gt_boxes[:, [0, 2]].clip( None, BOX_SCALE / max(image_unpadded.size) * image_unpadded.size[0]) # # crop the image for data augmentation # image_unpadded, gt_boxes = random_crop(image_unpadded, gt_boxes, BOX_SCALE, round_boxes=True) w, h = image_unpadded.size box_scale_factor = BOX_SCALE / max(w, h) if flipped: scaled_w = int(box_scale_factor * float(w)) # print("Scaled w is {}".format(scaled_w)) image_unpadded = image_unpadded.transpose(Image.FLIP_LEFT_RIGHT) gt_boxes[:, [0, 2]] = scaled_w - gt_boxes[:, [2, 0]] img_scale_factor = IM_SCALE / max(w, h) if h > w: im_size = (IM_SCALE, int(w * img_scale_factor), img_scale_factor) elif h < w: im_size = (int(h * img_scale_factor), IM_SCALE, img_scale_factor) else: im_size = (IM_SCALE, IM_SCALE, img_scale_factor) gt_rels = self.relationships[index].copy() if self.filter_duplicate_rels: # Filter out dupes! assert self.mode == 'train' old_size = gt_rels.shape[0] all_rel_sets = defaultdict(list) for (o0, o1, r) in gt_rels: all_rel_sets[(o0, o1)].append(r) gt_rels = [(k[0], k[1], np.random.choice(v)) for k,v in all_rel_sets.items()] gt_rels = np.array(gt_rels) entry = { 'img': self.transform_pipeline(image_unpadded), 'img_size': im_size, 'gt_boxes': gt_boxes, 'gt_classes': self.gt_classes[index].copy(), 'gt_relations': gt_rels, 'scale': IM_SCALE / BOX_SCALE, # Multiply the boxes by this. 'index': index, 'flipped': flipped, 'fn': self.filenames[index], } if self.rpn_rois is not None: entry['proposals'] = self.rpn_rois[index] assertion_checks(entry) return entry def __len__(self): return len(self.filenames) @property def num_predicates(self): return len(self.ind_to_predicates) @property def num_classes(self): return len(self.ind_to_classes) # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ # MISC. HELPER FUNCTIONS ~~~~~~~~~~~~~~~~~~~~~ # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ def assertion_checks(entry): im_size = tuple(entry['img'].size()) if len(im_size) != 3: raise ValueError("Img must be dim-3") c, h, w = entry['img'].size() if c != 3: raise ValueError("Must have 3 color channels") num_gt = entry['gt_boxes'].shape[0] if entry['gt_classes'].shape[0] != num_gt: raise ValueError("GT classes and GT boxes must have same number of examples") assert (entry['gt_boxes'][:, 2] >= entry['gt_boxes'][:, 0]).all() assert (entry['gt_boxes'] >= -1).all() def load_image_filenames(image_file, image_dir=VG_IMAGES): """ Loads the image filenames from visual genome from the JSON file that contains them. This matches the preprocessing in scene-graph-TF-release/data_tools/vg_to_imdb.py. :param image_file: JSON file. Elements contain the param "image_id". :param image_dir: directory where the VisualGenome images are located :return: List of filenames corresponding to the good images """ with open(image_file, 'r') as f: im_data = json.load(f) corrupted_ims = ['1592.jpg', '1722.jpg', '4616.jpg', '4617.jpg'] fns = [] for i, img in enumerate(im_data): basename = '{}.jpg'.format(img['image_id']) if basename in corrupted_ims: continue filename = os.path.join(image_dir, basename) if os.path.exists(filename): fns.append(filename) assert len(fns) == 108073 return fns def load_graphs(graphs_file, mode='train', num_im=-1, num_val_im=0, filter_empty_rels=True, filter_non_overlap=False): """ Load the file containing the GT boxes and relations, as well as the example_dataset split :param graphs_file: HDF5 :param mode: (train, val, or test) :param num_im: Number of images we want :param num_val_im: Number of validation images :param filter_empty_rels: (will be filtered otherwise.) :param filter_non_overlap: If training, filter images that dont overlap. :return: image_index: numpy array corresponding to the index of images we're using boxes: List where each element is a [num_gt, 4] array of ground truth boxes (x1, y1, x2, y2) gt_classes: List where each element is a [num_gt] array of classes relationships: List where each element is a [num_r, 3] array of (box_ind_1, box_ind_2, predicate) relationships """ if mode not in ('train', 'val', 'test'): raise ValueError('{} invalid'.format(mode)) roi_h5 = h5py.File(graphs_file, 'r') data_split = roi_h5['split'][:] split = 2 if mode == 'test' else 0 split_mask = data_split == split # Filter out images without bounding boxes split_mask &= roi_h5['img_to_first_box'][:] >= 0 if filter_empty_rels: split_mask &= roi_h5['img_to_first_rel'][:] >= 0 image_index = np.where(split_mask)[0] if num_im > -1: image_index = image_index[:num_im] if num_val_im > 0: if mode == 'val': image_index = image_index[:num_val_im] elif mode == 'train': image_index = image_index[num_val_im:] split_mask = np.zeros_like(data_split).astype(bool) split_mask[image_index] = True # Get box information all_labels = roi_h5['labels'][:, 0] all_boxes = roi_h5['boxes_{}'.format(BOX_SCALE)][:] # will index later assert np.all(all_boxes[:, :2] >= 0) # sanity check assert np.all(all_boxes[:, 2:] > 0) # no empty box # convert from xc, yc, w, h to x1, y1, x2, y2 all_boxes[:, :2] = all_boxes[:, :2] - all_boxes[:, 2:] / 2 all_boxes[:, 2:] = all_boxes[:, :2] + all_boxes[:, 2:] im_to_first_box = roi_h5['img_to_first_box'][split_mask] im_to_last_box = roi_h5['img_to_last_box'][split_mask] im_to_first_rel = roi_h5['img_to_first_rel'][split_mask] im_to_last_rel = roi_h5['img_to_last_rel'][split_mask] # load relation labels _relations = roi_h5['relationships'][:] _relation_predicates = roi_h5['predicates'][:, 0] assert (im_to_first_rel.shape[0] == im_to_last_rel.shape[0]) assert (_relations.shape[0] == _relation_predicates.shape[0]) # sanity check # Get everything by image. boxes = [] gt_classes = [] relationships = [] for i in range(len(image_index)): boxes_i = all_boxes[im_to_first_box[i]:im_to_last_box[i] + 1, :] gt_classes_i = all_labels[im_to_first_box[i]:im_to_last_box[i] + 1] if im_to_first_rel[i] >= 0: predicates = _relation_predicates[im_to_first_rel[i]:im_to_last_rel[i] + 1] obj_idx = _relations[im_to_first_rel[i]:im_to_last_rel[i] + 1] - im_to_first_box[i] assert np.all(obj_idx >= 0) assert np.all(obj_idx < boxes_i.shape[0]) rels = np.column_stack((obj_idx, predicates)) else: assert not filter_empty_rels rels = np.zeros((0, 3), dtype=np.int32) if filter_non_overlap: assert mode == 'train' inters = bbox_overlaps(boxes_i, boxes_i) rel_overs = inters[rels[:, 0], rels[:, 1]] inc = np.where(rel_overs > 0.0)[0] if inc.size > 0: rels = rels[inc] else: split_mask[image_index[i]] = 0 continue boxes.append(boxes_i) gt_classes.append(gt_classes_i) relationships.append(rels) return split_mask, boxes, gt_classes, relationships def load_info(info_file): """ Loads the file containing the visual genome label meanings :param info_file: JSON :return: ind_to_classes: sorted list of classes ind_to_predicates: sorted list of predicates """ info = json.load(open(info_file, 'r')) info['label_to_idx']['__background__'] = 0 info['predicate_to_idx']['__background__'] = 0 class_to_ind = info['label_to_idx'] predicate_to_ind = info['predicate_to_idx'] ind_to_classes = sorted(class_to_ind, key=lambda k: class_to_ind[k]) ind_to_predicates = sorted(predicate_to_ind, key=lambda k: predicate_to_ind[k]) return ind_to_classes, ind_to_predicates def vg_collate(data, num_gpus=3, is_train=False, mode='det'): assert mode in ('det', 'rel') blob = Blob(mode=mode, is_train=is_train, num_gpus=num_gpus, batch_size_per_gpu=len(data) // num_gpus) for d in data: blob.append(d) blob.reduce() return blob class VGDataLoader(torch.utils.data.DataLoader): """ Iterates through the data, filtering out None, but also loads everything as a (cuda) variable """ @classmethod def splits(cls, train_data, val_data, batch_size=3, num_workers=1, num_gpus=3, mode='det', **kwargs): assert mode in ('det', 'rel') train_load = cls( dataset=train_data, batch_size=batch_size * num_gpus, shuffle=True, num_workers=num_workers, collate_fn=lambda x: vg_collate(x, mode=mode, num_gpus=num_gpus, is_train=True), drop_last=True, # pin_memory=True, **kwargs, ) val_load = cls( dataset=val_data, batch_size=batch_size * num_gpus if mode=='det' else num_gpus, shuffle=False, num_workers=num_workers, collate_fn=lambda x: vg_collate(x, mode=mode, num_gpus=num_gpus, is_train=False), drop_last=True, # pin_memory=True, **kwargs, ) return train_load, val_load
16,373
37.527059
129
py
MRE-ISE
MRE-ISE-main/VSG/VG_parser/dataloaders/blob.py
""" Data blob, hopefully to make collating less painful and MGPU training possible """ from lib.fpn.anchor_targets import anchor_target_layer import numpy as np import torch from torch.autograd import Variable class Blob(object): def __init__(self, mode='det', is_train=False, num_gpus=1, primary_gpu=0, batch_size_per_gpu=3): """ Initializes an empty Blob object. :param mode: 'det' for detection and 'rel' for det+relationship :param is_train: True if it's training """ assert mode in ('det', 'rel') assert num_gpus >= 1 self.mode = mode self.is_train = is_train self.num_gpus = num_gpus self.batch_size_per_gpu = batch_size_per_gpu self.primary_gpu = primary_gpu self.imgs = [] # [num_images, 3, IM_SCALE, IM_SCALE] array self.im_sizes = [] # [num_images, 4] array of (h, w, scale, num_valid_anchors) self.all_anchor_inds = [] # [all_anchors, 2] array of (img_ind, anchor_idx). Only has valid # boxes (meaning some are gonna get cut out) self.all_anchors = [] # [num_im, IM_SCALE/4, IM_SCALE/4, num_anchors, 4] shapes. Anchors outside get squashed # to 0 self.gt_boxes = [] # [num_gt, 4] boxes self.gt_classes = [] # [num_gt,2] array of img_ind, class self.gt_rels = [] # [num_rels, 3]. Each row is (gtbox0, gtbox1, rel). self.gt_sents = [] self.gt_nodes = [] self.sent_lengths = [] self.train_anchor_labels = [] # [train_anchors, 5] array of (img_ind, h, w, A, labels) self.train_anchors = [] # [train_anchors, 8] shapes with anchor, target self.train_anchor_inds = None # This will be split into GPUs, just (img_ind, h, w, A). self.batch_size = None self.gt_box_chunks = None self.anchor_chunks = None self.train_chunks = None self.proposal_chunks = None self.proposals = [] @property def is_flickr(self): return self.mode == 'flickr' @property def is_rel(self): return self.mode == 'rel' @property def volatile(self): return not self.is_train def append(self, d): """ Adds a single image to the blob :param datom: :return: """ i = len(self.imgs) self.imgs.append(d['img']) h, w, scale = d['img_size'] # all anchors self.im_sizes.append((h, w, scale)) gt_boxes_ = d['gt_boxes'].astype(np.float32) * d['scale'] self.gt_boxes.append(gt_boxes_) self.gt_classes.append(np.column_stack(( i * np.ones(d['gt_classes'].shape[0], dtype=np.int64), d['gt_classes'], ))) # Add relationship info if self.is_rel: self.gt_rels.append(np.column_stack(( i * np.ones(d['gt_relations'].shape[0], dtype=np.int64), d['gt_relations']))) # Augment with anchor targets if self.is_train: train_anchors_, train_anchor_inds_, train_anchor_targets_, train_anchor_labels_ = \ anchor_target_layer(gt_boxes_, (h, w)) self.train_anchors.append(np.hstack((train_anchors_, train_anchor_targets_))) self.train_anchor_labels.append(np.column_stack(( i * np.ones(train_anchor_inds_.shape[0], dtype=np.int64), train_anchor_inds_, train_anchor_labels_, ))) if 'proposals' in d: self.proposals.append(np.column_stack((i * np.ones(d['proposals'].shape[0], dtype=np.float32), d['scale'] * d['proposals'].astype(np.float32)))) def _chunkize(self, datom, tensor=torch.LongTensor): """ Turn data list into chunks, one per GPU :param datom: List of lists of numpy arrays that will be concatenated. :return: """ chunk_sizes = [0] * self.num_gpus for i in range(self.num_gpus): for j in range(self.batch_size_per_gpu): chunk_sizes[i] += datom[i * self.batch_size_per_gpu + j].shape[0] return Variable(tensor(np.concatenate(datom, 0)), volatile=self.volatile), chunk_sizes def reduce(self): """ Merges all the detections into flat lists + numbers of how many are in each""" if len(self.imgs) != self.batch_size_per_gpu * self.num_gpus: raise ValueError("Wrong batch size? imgs len {} bsize/gpu {} numgpus {}".format( len(self.imgs), self.batch_size_per_gpu, self.num_gpus )) self.imgs = Variable(torch.stack(self.imgs, 0), volatile=self.volatile) self.im_sizes = np.stack(self.im_sizes).reshape( (self.num_gpus, self.batch_size_per_gpu, 3)) if self.is_rel: self.gt_rels, self.gt_rel_chunks = self._chunkize(self.gt_rels) self.gt_boxes, self.gt_box_chunks = self._chunkize(self.gt_boxes, tensor=torch.FloatTensor) self.gt_classes, _ = self._chunkize(self.gt_classes) if self.is_train: self.train_anchor_labels, self.train_chunks = self._chunkize(self.train_anchor_labels) self.train_anchors, _ = self._chunkize(self.train_anchors, tensor=torch.FloatTensor) self.train_anchor_inds = self.train_anchor_labels[:, :-1].contiguous() if len(self.proposals) != 0: self.proposals, self.proposal_chunks = self._chunkize(self.proposals, tensor=torch.FloatTensor) def _scatter(self, x, chunk_sizes, dim=0): """ Helper function""" if self.num_gpus == 1: return x.cuda(self.primary_gpu, async=True) return torch.nn.parallel.scatter_gather.Scatter.apply( list(range(self.num_gpus)), chunk_sizes, dim, x) def scatter(self): """ Assigns everything to the GPUs""" self.imgs = self._scatter(self.imgs, [self.batch_size_per_gpu] * self.num_gpus) self.gt_classes_primary = self.gt_classes.cuda(self.primary_gpu, async=True) self.gt_boxes_primary = self.gt_boxes.cuda(self.primary_gpu, async=True) # Predcls might need these self.gt_classes = self._scatter(self.gt_classes, self.gt_box_chunks) self.gt_boxes = self._scatter(self.gt_boxes, self.gt_box_chunks) if self.is_train: self.train_anchor_inds = self._scatter(self.train_anchor_inds, self.train_chunks) self.train_anchor_labels = self.train_anchor_labels.cuda(self.primary_gpu, async=True) self.train_anchors = self.train_anchors.cuda(self.primary_gpu, async=True) if self.is_rel: self.gt_rels = self._scatter(self.gt_rels, self.gt_rel_chunks) else: if self.is_rel: self.gt_rels = self.gt_rels.cuda(self.primary_gpu, async=True) if self.proposal_chunks is not None: self.proposals = self._scatter(self.proposals, self.proposal_chunks) def __getitem__(self, index): """ Returns a tuple containing data :param index: Which GPU we're on, or 0 if no GPUs :return: If training: (image, im_size, img_start_ind, anchor_inds, anchors, gt_boxes, gt_classes, train_anchor_inds) test: (image, im_size, img_start_ind, anchor_inds, anchors) """ if index not in list(range(self.num_gpus)): raise ValueError("Out of bounds with index {} and {} gpus".format(index, self.num_gpus)) if self.is_rel: rels = self.gt_rels if index > 0 or self.num_gpus != 1: rels_i = rels[index] if self.is_rel else None elif self.is_flickr: rels = (self.gt_sents, self.gt_nodes) if index > 0 or self.num_gpus != 1: rels_i = (self.gt_sents[index], self.gt_nodes[index]) else: rels = None rels_i = None if self.proposal_chunks is None: proposals = None else: proposals = self.proposals if index == 0 and self.num_gpus == 1: image_offset = 0 if self.is_train: return (self.imgs, self.im_sizes[0], image_offset, self.gt_boxes, self.gt_classes, rels, proposals, self.train_anchor_inds) return self.imgs, self.im_sizes[0], image_offset, self.gt_boxes, self.gt_classes, rels, proposals # Otherwise proposals is None assert proposals is None image_offset = self.batch_size_per_gpu * index # TODO: Return a namedtuple if self.is_train: return ( self.imgs[index], self.im_sizes[index], image_offset, self.gt_boxes[index], self.gt_classes[index], rels_i, None, self.train_anchor_inds[index]) return (self.imgs[index], self.im_sizes[index], image_offset, self.gt_boxes[index], self.gt_classes[index], rels_i, None)
9,073
38.281385
118
py