repo_name
stringlengths
6
97
path
stringlengths
3
341
text
stringlengths
8
1.02M
anandakelvin/instagrump
instagrump/__init__.py
from instagrump.Profile import Profile from instagrump.Content import Content
anandakelvin/instagrump
instagrump/Content.py
import requests class Content: def __init__(self, url): api_endp = '__a=1' if not api_endp in url: if url[-1] != '/': url += '/' if '?' in url: url += '&%s' % api_endp else: url += '?%s' % api_endp self.data = requests.get(url).json()['graphql']['shortcode_media'] self.type = self.data['__typename'] self.id = self.data['id'] self.caption = self.data['edge_media_to_caption']['edges'] self.caption_is_edited = self.data['caption_is_edited'] self.tagged_user = self.data['edge_media_to_tagged_user']['edges'] self.is_video = self.data['is_video'] self.dimensions = self.data['dimensions'] self.location = self.data['location'] self.timestamp = self.data['taken_at_timestamp'] self.likes = self.data['edge_media_preview_like']['count'] self.comments_disabled = self.data['comments_disabled'] self.viewer_can_reshare = self.data['viewer_can_reshare'] self.has_ranked_comments = self.data['has_ranked_comments'] try: self.title = self.data['title'] except: self.title = None def get_content(self): if self.type == 'GraphSidecar': data = self.data['edge_sidecar_to_children']['edges'] else: data = [{'node':self.data}] result = [] for node in data: item = {} if node['node']['is_video']: item['content'] = node['node']['video_url'] else: item['content'] = node['node']['display_url'] item['display_url'] = node['node']['display_url'] result.append(item) return result
anurag-ks/check-pep8
plugin.py
<reponame>anurag-ks/check-pep8 import sublime import sublime_plugin import subprocess import os class Pep8Command(sublime_plugin.TextCommand): """pep8 command""" def run(self, view): if self.view.is_dirty(): sublime.message_dialog("Please save the file") else: filename = self.view.file_name() if filename is None: sublime.message_dialog('Wirte some code, save it and then run \ this command') return name, ext = os.path.splitext(filename) if ext == ".py": try: p = subprocess.Popen(['pep8', str(filename)], stdout=subprocess.PIPE) (output, _) = p.communicate() output = output.decode('utf-8') if output == "" or output is None: sublime.message_dialog('No pep8 errors found.') else: sublime.message_dialog(output) except subprocess.CalledProcessError as e: sublime.message_dialog(str(e.output)) else: sublime.message_dialog("This is not a Python file")
kayru/librush
Scripts/embed_shaders.py
<gh_stars>10-100 import codecs from collections import namedtuple import os import subprocess import sys import textwrap BinText = namedtuple('BinText', 'declaration, definition') def bin2cpp(filename, variableName): s = open(filename, 'rb').read() s = codecs.encode(s, "hex").decode().upper() t=",".join(["0x"+x+y for (x,y) in zip(s[0::2], s[1::2])]) declarationText = "extern const unsigned char " + variableName + "data[];" declarationText += "\n"; declarationText += "extern const size_t " + variableName + "size;" definitionText = "const unsigned char " + variableName + "data[] = {\n\t%s\n};"%" \n\t".join(textwrap.wrap(t,80)) definitionText += "\n"; definitionText += "const size_t " + variableName + "size = sizeof(" + variableName + "data);" return BinText( declaration = declarationText, definition = definitionText ) # Compile shaders and generate .h / .cpp code header = '#pragma once\n// clang-format off\nnamespace Rush\n{\n' source = '#include "GfxEmbeddedShaders.h"\n// clang-format off\nnamespace Rush\n{\n' Shader = namedtuple('Shader', 'filename, entry, target') hlslShaders = [ Shader( filename = "..\\Shaders\\Primitive.hlsl", entry = "psMain", target = "ps_5_0", ), Shader( filename = "..\\Shaders\\Primitive.hlsl", entry = "psMainTextured", target = "ps_5_0", ), Shader( filename = "..\\Shaders\\Primitive.hlsl", entry = "vsMain3D", target = "vs_5_0", ), Shader( filename = "..\\Shaders\\Primitive.hlsl", entry = "vsMain2D", target = "vs_5_0", ), ] # SPIR-V shaders for shader in hlslShaders: # TODO: find glslc.exe and report error if not found hlslCompiler = "glslc.exe" inputFilename = os.path.normpath(shader.filename) shaderDirectory = os.path.dirname(shader.filename) outputFilename = os.path.join(shaderDirectory, shader.entry+".spv") stageMap = dict( vs_5_0="vertex", ps_5_0="fragment" ) command = [ hlslCompiler, "-x", "hlsl", "-o", outputFilename, "-fentry-point=" + shader.entry, "-fshader-stage=" + stageMap[shader.target], #"-DVULKAN=1", inputFilename ] print(" ".join(command)) compileResult = subprocess.call(command, stdout=subprocess.DEVNULL) if compileResult != 0: exit(1) variableNamePrefix = "SPV_" + shader.entry + "_" text = bin2cpp(outputFilename, variableNamePrefix) header += text.declaration + "\n"; source += text.definition + "\n"; os.remove(outputFilename) # DXBC shaders for shader in hlslShaders: # TODO: find fxc.exe and report error if not found hlslCompiler = "C:\\Program Files (x86)\\Windows Kits\\10\\bin\\x64\\fxc.exe" inputFilename = os.path.normpath(shader.filename) shaderDirectory = os.path.dirname(shader.filename) outputFilename = os.path.join(shaderDirectory, shader.entry+".dxbc") command = [ hlslCompiler, "/nologo", "/Fo", outputFilename, "/E", shader.entry, "/T", shader.target, inputFilename ] print(" ".join(command)) compileResult = subprocess.call(command, stdout=subprocess.DEVNULL) if compileResult != 0: exit(1) variableNamePrefix = "DXBC_" + shader.entry + "_" text = bin2cpp(outputFilename, variableNamePrefix) header += text.declaration + "\n"; source += text.definition + "\n"; os.remove(outputFilename) # Output results header += "}" source += "}" with open("..\\Rush\\GfxEmbeddedShaders.h", "w") as fileOut: fileOut.write(header) with open("..\\Rush\\GfxEmbeddedShaders.cpp", "w") as fileOut: fileOut.write(source)
jakobjassmann/ecodes-dk-lidar
scripts/fill_processing_gaps.py
# fix_processing_gaps.py # short script to fix missing tiles for each variable # to be run post completion of processing with process_tiles.py # <NAME> <EMAIL> 11 May 2021 # NB: uses scandir for speed, on OPALS Shell an install might be required with # python -m pip install scandir --user # Dependencies import os import glob import pandas import re import scandir import copy import sys import subprocess import shutil from dklidar import settings ## 1) Determine output folder structure # Status print('#' * 80 + 'Check EcoDes-DK processing outputs for completness' + '\n\n') print('Preparing environment...'), # Initiate list folders = [] # Check for subfolders present (max depth = 1) for folder in scandir.scandir(settings.output_folder): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # Remove tile_footprints folder if present folders = [folder for folder in folders if not bool(re.match('.*tile_footprints.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*point_source_proportion.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*point_source_counts.*', folder))] ## Get reference set of tiles based on dtm_10m dtm_10m = [folder for folder in folders if bool(re.match('.*dtm_10m.*', folder))][0] dtm_10m_tiles = [re.sub('.*_(\d*_\d*).tif', '\g<1>', file_name) for file_name in glob.glob(dtm_10m + '/*.tif')] dtm_10m_tiles = set(dtm_10m_tiles) print(' done.') ## 2) Check completeness of tiles for all variables # Status print('Scanning tiles for...') # Initiate empty dictionary missing_tiles = {} # Scan folders for missing tiles for folder in folders: variable_name = re.sub('.*[\\\\\/]', '', folder) print('\t' + variable_name) tiles = [re.sub('.*_(\d*_\d*).tif', '\g<1>', file_name) for file_name in glob.glob(folder + '/*.tif')] tiles = set(tiles) tiles_missing = dtm_10m_tiles - tiles missing_tiles.update({variable_name: tiles_missing}) # Status print('Scan complete.\n') print('Exporting missing tile_ids to csv...'), # Save missing tiles for each variable to csv missing_tiles_df_list = [] for variable in missing_tiles.keys(): missing_diles_df_local = pandas.DataFrame(missing_tiles[variable], columns = ['tile_id']) missing_diles_df_local['variable'] = variable missing_tiles_df_list.append(missing_diles_df_local) # Concatenate list of dfs into one df and export to csv missing_tiles_df = pandas.concat(missing_tiles_df_list) missing_tiles_df.to_csv(settings.wd + '/documentation/empty_tile_ids.csv', index = False) # Status print(' done.') ## 3) Generate empty tiles for all missing tile ids # Plenty of overlap means this is faster to do once then just copy # Make temp dir os.mkdir(settings.scratch_folder + '/fill_temp') # Get unique tile ids unique_missing_tiles = [tile_id for tile_list in missing_tiles.values() for tile_id in tile_list] unique_missing_tiles = set(unique_missing_tiles) # Status print('Generating ' + str(len(unique_missing_tiles)) + ' empty tiles in temp folder...') # Generate empty tiles using gdal for tile_id in unique_missing_tiles: in_file = dtm_10m + '/dtm_10m_' + tile_id + '.tif' out_file = settings.scratch_folder + '/fill_temp/' + "empty_" + tile_id + '.tif' cmd = settings.gdal_calc_bin + ' -A ' + in_file + ' --outfile ' + out_file + ' --calc=(-9999*greater(A,-9999)) --type=Int16 --NoDataValue=-9999' call_return = subprocess.check_output(cmd, shell=False) sys.stdout.write('.') sys.stdout.flush() # Status print(' done.') ## 4) Perpare summary oputput and fill missing rasters # Status print('Generating stummary stats...'), # Duplicate missing_tiles dict for summary stats summary_stats = copy.copy(missing_tiles) # Loop through key value pairs, get stats and write missing NA rasters for variable, tiles_missing in missing_tiles.items(): # Status print(variable + ' with ' + str(len(tiles_missing)) + ' tiles.') # Determine output folder output_folder = [folder for folder in folders if bool(re.match('.*[^_]' + variable + '.*', folder))][0] # Set summary stats summary_stats[variable] = len(tiles_missing) # Fill missing NA rasters if needed if len(tiles_missing)>0: for tile_id in tiles_missing: source = settings.scratch_folder + '/fill_temp/' + "empty_" + tile_id + '.tif' destination = output_folder + '/' + variable + "_" + tile_id + '.tif' if not os.path.exists(destination): shutil.copy(source, destination) else: print('Warning: file already exists') # Write summary stats to text file stats_file = open(output_folder + '/empty_tiles_' + variable + '.txt', 'w') stats_file.write(variable + '\n' + '' + str(len(tiles_missing)) + ' tiles did not complete processing and were replaced with empty tiles (all values = NA).\nThe affected tile_ids are:\n' + '\n'.join(list(tiles_missing)) + '\n') stats_file.close() # Status print(' done.') print('Exporting summary stats to csv...'), # Export stats as csv summary_stats_df = pandas.DataFrame() summary_stats_df['variable'] = summary_stats.keys() summary_stats_df['n_missing_tiles'] = summary_stats.values() summary_stats_df = summary_stats_df.sort_values('variable') summary_stats_df.to_csv(settings.wd + '/documentation/empty_tiles_summary.csv', index=False) # Status print(' done.') print('Cleaning up temp files...'), # Remove temp files for tile_id in unique_missing_tiles: temp_file = settings.scratch_folder + '/fill_temp/' + "empty_" + tile_id + '.tif' os.remove(temp_file) os.rmdir(settings.scratch_folder + '/fill_temp/') # Status print(' done.') print('Script complete.' + '\n' + 80 * '#') ## End of File
jakobjassmann/ecodes-dk-lidar
scripts/generate_tile_footprints.py
### DK Nationwide LiDAR - A short R script to generate the tile footprints file. ### <NAME> <EMAIL> 27 October 2021 import os import subprocess import re import ogr from dklidar import settings # Status update print('#' * 80) print(' EcoDes-DK15 generating tile_footprint variable\n') print(' 1. Creating output folder ...'), # Check whether output folder exists if not create if not os.path.exists(settings.output_folder + '/tile_footprints/'): os.mkdir(settings.output_folder + '/tile_footprints/') print('Done.') # Set variable to base the footprints on base_var = 'dtm_10m' # Export tile footprints using gdaltlindex print(' 2. Exporting tile footprints using gdaltindex based on "' + base_var + '" ...'), cmd = settings.gdaltlindex_bin + \ settings.output_folder + '/tile_footprints/tile_footprints.shp ' + \ settings.output_folder + '/' + base_var + '/*.tif' subprocess.check_output(cmd, stderr=subprocess.STDOUT) print('Done.') # Status print(' 3. Extracting tile_ids from file name attributes ...'), # Load shapefile shp_driver = ogr.GetDriverByName('ESRI Shapefile') tile_footprints = shp_driver.Open(settings.output_folder + '/tile_footprints/tile_footprints.shp', 1) # Open layer and get attribute name tile_layer = tile_footprints.GetLayer() tile_attribute_name = tile_layer.GetLayerDefn().GetFieldDefn(0).name # Set new attribute name new_attribute = ogr.FieldDefn('tile_id', ogr.OFTString) new_attribute.SetWidth(8) #16 char string width tile_layer.CreateField(new_attribute) # Convert file names to tile_id for feature in tile_layer: tile_id = feature.GetField(tile_attribute_name) tile_id = re.sub('.*(\d{4}_\d{3}).tif', '\g<1>', tile_id) feature.SetField('tile_id', tile_id) tile_layer.SetFeature(feature) feature = None tile_id = None # Delete file name field tile_layer.DeleteField(0) # Close file connection and tidy up tile_footprints.Destroy() tile_layer = None new_attribute = None tile_attribute_name = None # Status print('Done.\n') print(' Export of "tile_footprints" complete.\n') print('#' * 80),
jakobjassmann/ecodes-dk-lidar
dklidar/dtm.py
### Functions for DTM raster file handling for DK Lidar project ### <NAME> <EMAIL> 29 January 2019 import os import subprocess import re import pandas import opals import glob import time import shutil from dklidar import settings from dklidar import common #### Function definitions ## Generate tile footprint def dtm_generate_footprint(tile_id): """ Generates a footprint file using gdal. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return execution status """ # Initiate return value and log output return_value = '' log_file = open('log.txt', 'a') try: cmd = settings.gdaltlindex_bin + \ settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ settings.dtm_folder + '/DTM_1km_' + tile_id + '.tif' cmd_return = subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) log_file.write( '\n' + tile_id + ' footprint generation... \n' + cmd_return + \ '\n' + tile_id + ' successful.\n\n') return_value = 'success' except: log_file.write('\n' + tile_id + ' footprint generation failed. \n') return_value = 'gdalError' # Close log file log_file.close() return return_value ## Create neighbourhood mosaic def dtm_neighbourhood_mosaic(tile_id): """ Generates a mosaic of the dem for a given tile and it's 8 neighbourings. Incoplete mosaics are generated should a given neighbour be missing. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initate return value and open log file return_value = '' log_file = open('log.txt', 'a+') # get current (temporary) work directory temp_wd = os.getcwd() # Retrieve row and col numbers for the current tile_id center_row = int(re.sub('(\d+)_\d+', '\g<1>', tile_id)) center_col = int(re.sub('\d+_(\d+)', '\g<1>', tile_id)) # Determine row and column numbers for tiles in the 3 x 3 window rows_to_load = [center_row - 1, center_row, center_row + 1] cols_to_load = [center_col - 1, center_col, center_col + 1] # Generate list of tile_ids for tiles to load tiles_to_load = [] for row in rows_to_load: for col in cols_to_load: tile_to_load = str(row) + '_' + str(col) tiles_to_load.extend([tile_to_load]) # Prep filenames and check if files exists: tile_file_names = [] for tile_to_load in tiles_to_load: tile_file_name = settings.dtm_folder + '/DTM_1km_' + tile_to_load + '.tif' if os.path.exists(tile_file_name): tile_file_names.append(tile_file_name) n_neighbours = len(tile_file_names) tile_file_names = ' '.join(tile_file_names) if n_neighbours == 9: log_file.write(tile_id + ' mosaicing...\n' + 'Number of neighbours = ' + str(n_neighbours) + '. Complete!\n') else: log_file.write(tile_id + ' mosaicing...\n' + 'Warning! Number of neighbours = ' + str(n_neighbours) + '. Incomplete. Edge effects possible!\n') # Construct command: cmd = settings.gdalwarp_bin + ' ' + tile_file_names + ' ' + \ settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif ' # Execute command as subprocess and return message: try: log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) log_file.write('\n' + tile_id + ' successful.\n\n') return_value = 'success' except: log_file.write(tile_id + ' failed.\n\n') return_value = "gdalError" # Close log file log_file.close() return return_value ## Validate crs def dtm_validate_crs(tile_id, mosaic = True): """ Function to validate the crs for dtm files (single tile and mosaic) :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param mosaic: if Ture validates crs for mosaic also, default: True :return: execution status """ # Initiate return value return_value = '' # Generate odm files path names dtm_file = settings.dtm_folder + '/DTM_1km_' + tile_id + '.tif' dtm_mosaic = settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif' # Retrieve CRS string for single tile try: crs_str = subprocess.check_output(settings.gdalsrsinfo_bin + '-o proj4 ' + dtm_file, shell=False, stderr=subprocess.STDOUT) # Clean up string by removing first line all white space before and after just in case crs_str = re.sub('^.*?\n', '', crs_str) # Check whether CRS exists, if different issue warning. if crs_str.strip() == settings.crs_proj4_gdal.strip(): return_value = 'Tile: match' else: return_value = 'Tile: warning - no match' except: return_value = 'Single: error' # Retrieve CRS string for mosaic if mosaic == True: try: crs_str = subprocess.check_output(settings.gdalsrsinfo_bin + '-o proj4 ' + dtm_mosaic, shell=False, stderr=subprocess.STDOUT) # Clean up string by removing first line all white space before and after just in case crs_str = re.sub('^.*?\n', '', crs_str) # Check whether CRS exists, if not assign, if different throw error. if crs_str.strip() == settings.crs_proj4_gdal.strip(): return_value = return_value + '; Mosaic: match' else: return_value = return_value + '; Mosaic: warning - no match' except: return_value = return_value + '; Mosaic: error' return return_value ## Aggregate dem to 10 m def dtm_aggregate_tile(tile_id): """ Aggregates the 0.4 m DTM to 10 m size for final output and further calculations. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return valule and open log file return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/dtm_10m' if not os.path.exists(out_folder): os.mkdir(out_folder) try: ## Aggregate dtm to temporary file: # Specify gdal command cmd = settings.gdalwarp_bin + \ '-tr 10 10 -r average ' + \ settings.dtm_folder + '/DTM_1km_' + tile_id + '.tif ' + \ wd + '/dtm_10m_' + tile_id + '_float.tif ' # Execute gdal command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' aggregating dtm_10m successful.\n\n') out_file = out_folder + '/dtm_10m_' + tile_id + '.tif' # Stretch by 100, round and store as int16 # Specify gdal command cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/dtm_10m_' + tile_id + '_float.tif ' + \ ' --outfile=' + out_file + \ ' --calc=rint(100*A)' + ' --type=Int16' + ' --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' converting dtm_10m to int16... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file) return_value = 'success' except: log_file.write('\n' + tile_id + ' dtm_10m aggregation failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() # Remove temporary files try: os.remove(wd + '/dtm_10m_' + tile_id + '_float.tif') except: pass return return_value ## Aggregate dem mosaic to 10 m def dtm_aggregate_mosaic(tile_id): """ Aggregates the 0.4 m DTM mosaic to 10 m size for final output and other calculations. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return valule and open log file return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder out_folder = settings.dtm_mosaics_10m_folder if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Specify gdal command cmd = settings.gdalwarp_bin + \ '-tr 10 10 -r average -overwrite ' + \ settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif ' + \ settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' # Execute gdal command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' aggregating dtm_10m mosaic successful.\n\n') return_value = 'success' except: log_file.write('\n' + tile_id + ' dtm_10m mosaic aggregation failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() return return_value ## Calculate slope for tile def dtm_calc_slope(tile_id): """ Calculates the slope parameter for a DTM neighbourhood mosaic and crops to original tile_size. Requires dtm_generate_mosaic() to be executed. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: stdout and stderr command line output of gdal command execution. """ # Initiate return value and log output return_value = '' log_file = open('log.txt', 'a+') # Get current wd wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/slope' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Calculate slope parameter cmd = settings.gdaldem_bin + ' slope ' + \ settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' + \ wd + '/slope_' + tile_id + '_mosaic.tif ' log_file.write(tile_id + ' slope calculation... \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop slope mosaic output to original tile size cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/slope_' + tile_id + '_mosaic.tif ' + \ wd + '/slope_' + tile_id + '_mosaic_cropped.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropped slope.\n\n') # Round and store slope as int16 cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/slope_' + tile_id + '_mosaic_cropped.tif ' + \ ' --outfile=' + out_folder + '/slope_' + tile_id + '.tif ' + \ ' --calc=rint(10*A) --type=Int16 --NoDataValue=-9999' log_file.write('\n' + tile_id + ' rounding slope and calculation successful. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_folder + '/slope_' + tile_id + '.tif ') return_value = 'success' except: log_file.write('\n' + tile_id + ' slope calculation failed. \n\n') return_value = 'gdalError' # Close log file log_file.close() # Remove temporary file try: os.remove(wd + '/slope_' + tile_id + '_mosaic.tif ') os.remove(wd + '/slope_' + tile_id + '_mosaic_cropped.tif ') except: pass # Return return value return return_value ## Calculate aspect for a tile def dtm_calc_aspect(tile_id, slope_zero = 'nodata'): """ Calculates the aspect for all 10 m cells in a DTM neighbourhood mosaic and crops to original tile_size. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param slope_zero: integer value or 'nodata', sets the value for cells were slope = 0. :return: execution status """ # Initiate return valule and log output return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/aspect' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Calculate aspect cmd = settings.gdaldem_bin + ' aspect ' + \ settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' + \ wd + '/aspect_' + tile_id + '_mosaic.tif ' log_file.write(tile_id + ' aspect calculation... \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop aspect output to original tile size cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/aspect_' + tile_id + '_mosaic.tif ' + \ wd + '/aspect_' + tile_id + '_mosaic_cropped.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' aspect mosaic cropped.\n\n') # Set aspect for cells slope = 0 if the value is not nodata if slope_zero != 'nodata': # Calculate slope cmd = settings.gdaldem_bin + ' slope ' + \ settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' + \ wd + '/slope_' + tile_id + '_mosaic.tif ' log_file.write(tile_id + ' slope calculated. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop slope mosaic cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/slope_' + tile_id + '_mosaic.tif ' + \ wd + '/slope_' + tile_id + '_mosaic_cropped.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' slope mosaic cropped.\n\n') # Prepare mask from slope raster for slope = 0 cmd = settings.gdal_calc_bin + \ '-S ' + wd + '/slope_' + tile_id + '_mosaic_cropped.tif ' + \ ' --outfile=' + wd + '/slope_mask_' + tile_id + '.tif ' + \ ' --calc=' + str(slope_zero) + '*(S==0)-9999*(S!=0)' log_file.write('\n' + tile_id + ' generated slope mask successfuly. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Merge mask over aspect to set value for aspect where slope = 0 cmd = settings.gdal_merge_bin + \ ' -o ' + wd + '/aspect_' + tile_id + '_mosaic_cropped_masked.tif '+ \ ' -a_nodata -9999 -init -9999 -ot Float32 ' + \ wd + '/slope_mask_' + tile_id + '.tif ' + \ wd + '/aspect_' + tile_id + '_mosaic_cropped.tif ' log_file.write('\n' + tile_id + ' set aspect for slope = 0 successfuly. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) else: # Rename cropped mosaic os.rename(wd + '/aspect_' + tile_id + '_mosaic_cropped.tif ', wd + '/aspect_' + tile_id + '_mosaic_cropped_masked.tif ') # Round and store as int16 cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/aspect_' + tile_id + '_mosaic_cropped_masked.tif ' + \ ' --outfile=' + out_folder + '/aspect_' + tile_id + '.tif ' + \ ' --calc=rint(10*A) --type=Int16 --NoDataValue=-9999' log_file.write('\n' + tile_id + ' rounding aspect to int16 and calculation success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_folder + '/aspect_' + tile_id + '.tif ') return_value = 'success' except: log_file.write('\n' + tile_id + ' aspect calculation failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() # Remove temporary files try: os.remove(wd + '/aspect_' + tile_id + '_mosaic.tif ') os.remove(wd + '/aspect_' + tile_id + '_mosaic_cropped.tif ') os.remove(wd + '/aspect_' + tile_id + '_mosaic_cropped_masked.tif ') os.remove(wd + '/slope_' + tile_id + '_mosaic.tif ') os.remove(wd + '/slope_' + tile_id + '_mosaic_cropped.tif ') os.remove(wd + '/slope_mask_' + tile_id + '.tif ') except: pass return return_value ## Calculcate heat index def dtm_calc_heat_index(tile_id, slope_zero = 'nodata'): """ Calculates the heat index from McCune and Keon (2002) based on the aspect only. Aspect must have been calculated using dtm_calc_aspect(). :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param zero_slope: zero_slope - inteeger value assigned to the aspect when slope = 0 or 'nodata' :return: execution status """ # Intialise return value and log return_value = '' log_file = open('log.txt', 'a+') # Get current wd wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/heat_load_index' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Specify path to aspect raster A aspect_file = '-A ' + settings.output_folder + '/aspect/aspect_' + tile_id + '.tif ' # Construct numpy equation, stretch by 10k and round heat_index = 'rint(10000*((1-cos(radians((A/10)-45)))/2))' # Specify temp_file path temp_file = wd + '/heat_load_index_' + tile_id + '.tif ' # Specify output path out_file = out_folder + '/heat_load_index_' + tile_id + '.tif ' # Construct gdal command, save file as Int16: cmd = settings.gdal_calc_bin + \ aspect_file + \ '--outfile=' + temp_file + \ ' --calc=' + heat_index + \ ' --type=Int16 --NoDataValue=-9999 --overwrite' # Execute gdal command log_file.write('\n' + tile_id + ' calculating heat index success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # maks index for zero slope value if specified if slope_zero != 'nodata': cmd = settings.gdal_calc_bin + \ aspect_file + \ ' -B ' + temp_file + \ ' --outfile=' + out_file + \ ' --calc=-9999*(A==' + str(slope_zero) + ')+B*(A!=' + str(slope_zero) + ')' + \ ' --type=Int16 --NoDataValue=-9999 --overwrite ' log_file.write('\n' + tile_id + ' applied aspect mask successfuly. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) else: # Move file try: os.remove(out_file) except: pass os.rename(temp_file, out_file) # Apply mask(s) common.apply_mask(out_file) # remove temp file try: os.remove(temp_file) except: pass return_value = 'success' except: log_file.write(tile_id + ' calculating heat index failed. \n ') return_value = 'gdal_error' # Close log file log_file.close() return return_value ## Calculate solar radiation def dtm_calc_solar_radiation(tile_id): """ Returns cell by cell solar radiation following McCune and Keon 2002. Slope and aspect must have been calculated beforehand using dtm_calc_slope() and dtm_calc_aspect(). :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # The calculation of the solar radiation is a two step process # 1) Obtain a raster with the latitude of the centre of the cell in radians # 2) Calculate the solar radiation using the formula form McCune and Keon 2002 # initiate return value and log ouptut return_value = '' log_file = open('log.txt', 'a') # Get current wd wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/solar_radiation' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # 1) Create raster with latitude of the centre of a cell # Construct gdal command to export xyz file in utm dtm_file = settings.output_folder + '/slope/slope_' + tile_id + '.tif' out_file = wd + '/xyz_' + tile_id + '.xyz' cmd = settings.gdal_translate_bin + \ ' -of xyz -co COLUMN_SEPARATOR="," -co ADD_HEADER_LINE=YES ' + \ dtm_file + ' ' + \ out_file log_file.write('\n converting slope raster to xyz with gdal command: ' + cmd) # Execute gdal command and log log_file.write('\n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' generated xyz. \n\n ') log_file.write('\n conversion completed ') # Read in xyz as a pandas dataframe xyz = pandas.read_csv(wd + '/xyz_' + tile_id + '.xyz') xy = xyz[["X", "Y"]] xy.to_csv(wd + '/xy_' + tile_id + '.csv', index=False, header=False, sep=' ') log_file.write('\n removed z coordinate. ') # Construct gdal commands to transform cell coordinates from utm to lat long in_file = wd + '\\xy_' + tile_id + '.csv' # Script used to break here insert a pause time.sleep(1) cmd = settings.gdaltransform_bin + \ ' -s_srs EPSG:25832 -t_srs WGS84 ' + \ ' < ' + in_file log_file.write('\n gdal transform command: ' + cmd) # And execute the gdal command xy_transformed = subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) log_file.write(xy_transformed[0:200] + '\n') # Write to file out_file = wd + '/xy_' + tile_id + '_latlong.csv' xy_trans_file = open(out_file, 'w') xy_trans_file.write(xy_transformed) xy_trans_file.close() log_file.write('\n gdal transform completed. ') # Load lat long file as pandas df skip_rows = 1 if settings.gdal_version == '3.3.3': skip_rows = 0 xy_latlong = pandas.read_csv(wd + '/xy_' + tile_id + '_latlong.csv', sep='\s+', names=['X', 'Y', 'return_status'], skiprows=skip_rows) log_file.write('\n check whether lenght matches. ') # check data frames are of the same length if len(xyz.index) != len(xy_latlong.index): log_file.write('\n lenght of dataframes did not match \n') raise Exception("") log_file.write('\n added latitude utm as z to latlong file. ') # Assign lat (deg) to UTM z coordinate xyz["Z"] = xy_latlong["Y"] xyz.to_csv(wd + '/xyz_' + tile_id + '.xyz', index=False, header=False, sep=' ') # Convert back to geotiff, prepare gdal translate command in_file = wd + '/xyz_' + tile_id + '.xyz' out_file = wd + '/lat_' + tile_id + '.tif' cmd = settings.gdal_translate_bin + \ ' -of GTiff -a_srs EPSG:25832 ' + \ in_file + ' ' + \ out_file log_file.write('\n converting to geotiff using gdal command:' + cmd) # Execute command and log log_file.write('\n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' generated lat tif. \n\n') # Intermediate clean up os.remove(wd + '/xyz_' + tile_id + '.xyz') os.remove(wd + '/xy_' + tile_id + '.csv') os.remove(wd + '/xy_' + tile_id + '_latlong.csv') del (xy) del (xy_latlong) del (xyz) log_file.write('\n finished preparations. ') ## 2) Calculate Solar radiation # The equation from McCune and Keon goes as follows: # solar radiation = 0.339 + # 0.808 x cos(lat) x cos(slope) + # -0.196 x sin(lat) x sin(slope) + # -0.482 x cos(asp) x sin(slope) # Aspect must be foldered around the S-N line: # asp = 180 - |180 - asp| # and all values mus be in radians: # rad = deg * pi / 180 or using numpy simply: rad = radians(deg) # Finally, the result needs to be stretched by 1000 and rounded for storage as an Int16 # Specify path to latitude raster L lat_file = '-L ' + wd + '/lat_' + tile_id + '.tif ' # Specify path to slope raster as raster S slope_file = '-S ' + settings.output_folder + '/slope/slope_' + tile_id + '.tif ' # Specify path to aspect raster A aspect_file = '-A ' + settings.output_folder + '/aspect/aspect_' + tile_id + '.tif ' # Construct numpy equation (based on McCune and Keon 2002) and stretch by 1000 and round to nearest int. #solar_rad_eq = 'rint(1000*(0.339+0.808*cos(radians(L))*cos(radians((S/10)))-0.196*sin(radians(L))*sin(radians((S/10)))-0.482*cos(radians(180-absolute(180-(A/10))))*sin(radians((S/10)))))' # Construct numpy equation (based on McCune and Keon 2002), convert to MJ/yr/100m2 (cell) and round to nearest int. solar_rad_eq = 'rint(1000000*(numpy.exp(0.339+0.808*cos(radians(L))*cos(radians((S/10)))-0.196*sin(radians(L))*sin(radians((S/10)))-0.482*cos(radians(180-absolute(180-(A/10))))*sin(radians((S/10))))))' # Specify output path out_file = out_folder + '/solar_radiation_' + tile_id + '.tif ' # Construct gdal command: cmd = settings.gdal_calc_bin + \ lat_file + \ slope_file + \ aspect_file + \ '--outfile=' + out_file + \ '--calc=' + solar_rad_eq + ' ' + \ '--type=Int32 --NoDataValue=-9999 --overwrite' #Int16 if original equation is used. log_file.write('\n calculating solar radiaiton with gdal calc: ' + cmd) # Execute gdal command log_file.write('\n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' calculated solar radiation. \n\n') # Apply mask(s) common.apply_mask(out_file) # Remove latitude tif os.remove(wd + '/lat_' + tile_id + '.tif') return_value = 'success' log_file.write('\n done. ') except: log_file.write('\n\n' + tile_id + ' calculating solar radiation failed. \n ') return_value = 'gdal_error' # Write log output to log file log_file.close() return return_value ## Calculate landscape openness mean def dtm_openness_mean(tile_id): """ Exports the mean landscape openness for all eight cardinal directions with a 150 m search radius using the OPALS Openness module. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Get working directory wd = os.getcwd() # Generate folder paths out_folder = settings.output_folder + '/openness_mean' if not os.path.exists(out_folder): os.mkdir(out_folder) # Attempt calculation of mean openness try: # Initialise Opals Openness Module export_openness = opals.Openness.Openness() # Export positive openness for a given cell cell with a search radius of 150 m (15 cells) export_openness.inFile = settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' export_openness.outFile = wd + '/openness_150m_' + tile_id + '_mosaic.tif ' export_openness.feature = opals.Types.OpennessFeature.positive export_openness.kernelSize = 15 # 15 x 10 m = 150 m export_openness.selMode = 0 export_openness.noData = -9999 export_openness.commons.screenLogLevel = opals.Types.LogLevel.none export_openness.commons.nbThreads = settings.nbThreads export_openness.run() # Convert to degrees, round and store as int16 # Specify gdal command cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/openness_150m_' + tile_id + '_mosaic.tif ' + \ ' --outfile=' + wd + '/landscape_openness_' + tile_id + '_mosaic.tif ' + \ ' --calc=rint(degrees(A))' + \ ' --type=Int16 --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' converted and rounded to degrees. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Obtain file extent for cropping then remove outer 150 m of mosaic to avoid edge effects cmd = settings.gdalinfo_bin + wd + '/landscape_openness_' + tile_id + '_mosaic.tif ' mosaic_info = subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) upper_left = re.search("Upper *Left *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) lower_right = re.search("Lower *Right *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) xmin = float(upper_left.group(1)) + 150 ymax = float(upper_left.group(2)) - 150 xmax = float(lower_right.group(1)) - 150 ymin = float(lower_right.group(2)) + 150 # remove 150 m on outer edge using gdalwarp cmd = settings.gdalwarp_bin + \ '-te ' + str(xmin) + ' ' + str(ymin) + ' ' + str(xmax) + ' ' + str(ymax) + ' -overwrite ' + \ wd + '/landscape_openness_' + tile_id + '_mosaic.tif ' + \ wd + '/landscape_openness_' + tile_id + '_mosaic_cropped.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropping landscape openness mosaic finished.\n\n') # Crop openness mosaic to original tile size (this will set all edges removed earlier to NA) cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/landscape_openness_' + tile_id + '_mosaic_cropped.tif ' + \ out_folder + '/openness_mean_' + tile_id + '.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' landscape openness calculation successful.\n\n') # Apply mask(s) common.apply_mask(out_folder + '/openness_mean_' + tile_id + '.tif ') return_value = 'success' except: return_value = 'opals/gdal/Error' # Remove temporary files try: os.remove(wd + '/openness_150m_' + tile_id + '_mosaic.tif ') os.remove(wd + '/landscape_openness_' + tile_id + '_mosaic.tif ') os.remove(wd + '/landscape_openness_' + tile_id + '_mosaic_cropped.tif ') # and are super random files created by OPALS for temp_file in glob.glob(wd + '/../*' + tile_id + '_mosaic_dz._dz.tif'): os.remove(temp_file) except: pass # Close log file log_file.close() return return_value ## Calculate landscape openness difference def dtm_openness_difference(tile_id): """ Exports the difference between the minimum and maximum positive openness within a 50 m search radius using the Opals Openness module. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Get working directory wd = os.getcwd() # Generate folder paths out_folder = settings.output_folder + '/openness_difference' if not os.path.exists(out_folder): os.mkdir(out_folder) # Attempt openness difference calculation try: # Initialise Opals Openness Module export_openness = opals.Openness.Openness() # Export minimum positive openness for a given cell cell with a kernel size of 50 m x 50 m export_openness.inFile = settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' export_openness.outFile = wd + '/openness_50m_min_' + tile_id + '_mosaic.tif ' export_openness.feature = opals.Types.OpennessFeature.positive export_openness.kernelSize = 5 # 5 x 10 m = 50 m export_openness.selMode = 1 export_openness.noData = -9999 export_openness.commons.screenLogLevel = opals.Types.LogLevel.none export_openness.commons.nbThreads = settings.nbThreads export_openness.run() export_openness.reset() # Export maximum positive openness for a given cell with a kernel size of 50 m x 50 m export_openness.inFile = settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' export_openness.outFile = wd + '/openness_50m_max_' + tile_id + '_mosaic.tif ' export_openness.feature = opals.Types.OpennessFeature.positive export_openness.kernelSize = 5 # 5 x 10 m = 50 m export_openness.selMode = 2 export_openness.noData = -9999 export_openness.commons.screenLogLevel = opals.Types.LogLevel.none export_openness.commons.nbThreads = settings.nbThreads export_openness.run() # Calculate difference, round and store as int16 # Specify gdal command cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/openness_50m_min_' + tile_id + '_mosaic.tif ' + \ '-B ' + wd + '/openness_50m_max_' + tile_id + '_mosaic.tif ' + \ ' --outfile=' + wd + '/diff_openness_' + tile_id + '_mosaic.tif ' + \ ' --calc=rint(degrees(B)-degrees(A))' + \ ' --type=Int16 --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' calculated difference openness. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Obtain file extent for cropping (remove outer 50 m of mosaic) cmd = settings.gdalinfo_bin + wd + '/diff_openness_' + tile_id + '_mosaic.tif ' mosaic_info = subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) upper_left = re.search("Upper *Left *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) lower_right = re.search("Lower *Right *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) xmin = float(upper_left.group(1)) + 50 ymax = float(upper_left.group(2)) - 50 xmax = float(lower_right.group(1)) - 50 ymin = float(lower_right.group(2)) + 50 # Remove 50 m on outer edge using gdalwarp cmd = settings.gdalwarp_bin + \ '-te ' + str(xmin) + ' ' + str(ymin) + ' ' + str(xmax) + ' ' + str(ymax) + ' -overwrite ' + \ wd + '/diff_openness_' + tile_id + '_mosaic.tif ' + \ wd + '/diff_openness_' + tile_id + '_mosaic_cropped.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropped openness difference mosaic.\n\n') # Crop diff openness to original tile size (this will set all edges removed earlier to NA) cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/diff_openness_' + tile_id + '_mosaic_cropped.tif ' + \ out_folder + '/openness_difference_' + tile_id + '.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' openness calculation successful.\n\n') # Apply mask(s) common.apply_mask(out_folder + '/openness_difference_' + tile_id + '.tif ') return_value = 'success' except: return_value = 'opals/gdal/Error' # Remove temporary files try: os.remove(wd + '/openness_50m_min_' + tile_id + '_mosaic.tif ') os.remove(wd + '/openness_50m_max_' + tile_id + '_mosaic.tif ') os.remove(wd + '/diff_openness_' + tile_id + '_mosaic.tif ') os.remove(wd + '/diff_openness_' + tile_id + '_mosaic_cropped.tif ') # and are super random files created by OPALS for temp_file in glob.glob(wd + '/../*' + tile_id + '_mosaic_dz._dz.tif'): os.remove(temp_file) except: pass # Close log file log_file.close() # Return exist status return return_value ## Calculate TWI following Kopecky et al. 2020 def dtm_kopecky_twi(tile_id): """ Calculates the topographic wetness indec (TWI) following Kopecky et al. 2020. Requires SAGA GIS 7.8.2 or later to be specified in settings.py. Calculations are done on the aggregated 10 m tile neighbourhood mosaic. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return valule and log output return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Replace backslash with forward slash for SAGA GIS wd = re.sub('\\\\', '/', wd) # Prepare output folder out_folder = settings.output_folder + '/twi' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Fill in sinks on mosaic cmd = settings.saga_bin + 'ta_preprocessor 5 ' + \ '-ELEV ' + settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif ' + \ '-FILLED ' + wd + '/' + tile_id + '_mosaic_10m_filled.sdat ' + \ '-MINSLOPE 0.01' log_file.write(tile_id + ' finished filling sinks. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Calculate Flow Accumulation on filled mosaic cmd = settings.saga_bin + 'ta_hydrology 0 ' + \ '-ELEVATION ' + wd + '/' + tile_id + '_mosaic_10m_filled.sdat ' + \ '-METHOD 4 -CONVERGENCE 1.0 ' + \ '-FLOW ' + wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd.sdat ' log_file.write(tile_id + ' finished flow claculation. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Calculate Flow Width and Catchment Area cmd = settings.saga_bin + 'ta_hydrology 19 ' + \ '-DEM ' + wd + '/' + tile_id + '_mosaic_10m_filled.sdat ' + \ '-TCA ' + wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd.sdat ' + \ '-WIDTH ' + wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd_width.sdat ' + \ '-SCA ' + wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd_sca.sdat ' log_file.write(tile_id + ' finished flow width and catchment area calculation. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Calculate slope on mosaic cmd = settings.saga_bin + 'ta_morphometry 0 ' + \ '-ELEVATION ' + wd + '/' + tile_id + '_mosaic_10m_filled.sdat ' + \ '-METHOD 7 ' + \ '-SLOPE ' + wd + '/' + tile_id + '_mosaic_10m_filled_slope.sdat ' log_file.write(tile_id + ' finished slope claculation. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Calculate TWI on mosaic cmd = settings.saga_bin + 'ta_hydrology 20 ' + \ '-SLOPE ' + wd + '/' + tile_id + '_mosaic_10m_filled_slope.sdat ' + \ '-AREA ' +wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd_sca.sdat ' + \ '-TWI '+ wd + '/' + tile_id + '_mosaic_10m_filled_twi.sdat ' log_file.write(tile_id + ' finished twi claculation. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop output to original tile size and convert to tif: cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ ' -crop_to_cutline -overwrite ' + \ wd + '/' + tile_id + '_mosaic_10m_filled_twi.sdat ' + \ wd + '/twi_' + tile_id + '_float.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropping wetness index mosaic successful.\n\n') return_value = 'success' # Stretch to by 1000, round and convert to int 16 # Construct gdal command: cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/twi_' + tile_id + '_float.tif ' + \ '--outfile=' + out_folder + '/twi_' + tile_id + '.tif ' + \ ' --calc=rint(1000*A) --type=Int16 --NoDataValue=-9999 --overwrite' # Execute gdal command log_file.write('\n' + tile_id + ' rounding and conversion finished. ' \ 'TWI calculation successful. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) except: log_file.write('\n' + tile_id + ' wetness index calculation failed.\n\n') return_value = 'SAGA/GDAL Error' # Close log file log_file.close() # Remove temporary file try: for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled.*'): os.remove(temp_file) for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd.*'): os.remove(temp_file) for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd_width.*'): os.remove(temp_file) for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled_flow_mfd_sca.*'): os.remove(temp_file) for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled_slope.*'): os.remove(temp_file) for temp_file in glob.glob(wd + '/' + tile_id + '_mosaic_10m_filled_twi.*'): os.remove(temp_file) os.remove(wd + '/twi_' + tile_id + '_float.tif ') except: pass return return_value ## Calculate SAGA Wetness Index for a tile def dtm_saga_wetness(tile_id): """ Calculates the saga wetness index for a tile mosaic then crops to the original tile. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return valule and log output return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/saga_wetness_index' if not os.path.exists(out_folder): os.mkdir(out_folder) try: # Calculate wetness index at DTM scale cmd = settings.saga_wetness_bin + '-DEM ' + \ settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif ' + \ '-TWI ' + wd + '/wetness_index_' + tile_id + '_mosaic.tif' log_file.write(tile_id + ' wetness index calculation finished. \n ' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop output to original tile size: cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-tr 10 10 -r med -crop_to_cutline -overwrite ' + \ wd + '/wetness_index_' + tile_id + '_mosaic.sdat ' + \ wd + '/wetness_index_' + tile_id + '.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropping wetness index mosaic successful.\n\n') return_value = 'success' # Set input file path in_file = wd + '/wetness_index_' + tile_id + '.tif ' # Set output file path out_file = out_folder + '/wetness_index_' + tile_id + '.tif ' # Stretch to by 1000, round and convert to int 16 # Construct gdal command: cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/wetness_index_' + tile_id + '.tif ' + \ '--outfile=' + out_file + \ ' --calc=rint(1000*A) --type=Int16 --NoDataValue=-9999 --overwrite' # Execute gdal command log_file.write('\n' + tile_id + ' rounding and conversion finished. ' \ 'Wetness index calculation successful. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) except: log_file.write('\n' + tile_id + ' wetness index calculation failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() # Remove temporary file try: os.remove(wd + '/wetness_index_' + tile_id + '_mosaic.sdat ') os.remove(wd + '/wetness_index_' + tile_id + '_mosaic.prj ') os.remove(wd + '/wetness_index_' + tile_id + '_mosaic.sgrd ') os.remove(wd + '/wetness_index_' + tile_id + '_mosaic.mgrd ') os.remove(wd + '/wetness_index_' + tile_id + '_mosaic.tif ') os.remove(wd + '/wetness_index_' + tile_id + '.tif ') except: pass return return_value ## Calculate landscape openness (mean) using SAGA def dtm_saga_landscape_openness(tile_id): """ Calculates landscape opennes following Yokoyama et al. 2002 based on an aggregated 10 m dtm and a 150 m search radius. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return valule and log output return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder out_folder = settings.output_folder + '/landscape_openness' if not os.path.exists(out_folder): os.mkdir(out_folder) try: ## Aggregate dtm mosaic to temporary file: # Specify gdal command cmd = settings.gdalwarp_bin + \ '-tr 10 10 -r average ' + \ settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif ' + \ wd + '/dtm_10m_' + tile_id + '_mosaic_float.tif ' + ' -overwrite' # Execute gdal command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' aggregating dtm_10m mosaic successful.\n\n') # Use saga gis openness module for calculating the openness in 150 m cmd = settings.saga_openness_bin + \ '-DEM ' + wd + '/dtm_10m_' + tile_id + '_mosaic_float.tif ' + \ '-POS ' + wd + '/openness_10m_' + tile_id + '_mosaic.sdat ' + \ '-RADIUS 150 -METHOD 1' # Execute saga command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' openness from mosaic successful.\n\n') # Obtain file extent for cropping (remove outer 150 m of mosaic) cmd = settings.gdalinfo_bin + wd + '/openness_10m_' + tile_id + '_mosaic.sdat ' mosaic_info = subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) upper_left = re.search("Upper *Left *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) lower_right = re.search("Lower *Right *\( *(\d+.\d+), *(\d+.\d+)\)", mosaic_info) xmin = float(upper_left.group(1)) + 150 ymax = float(upper_left.group(2)) - 150 xmax = float(lower_right.group(1)) - 150 ymin = float(lower_right.group(2)) + 150 # remove 150 m on outer edge using gdal warp cmd = settings.gdalwarp_bin + \ '-te ' + str(xmin) + ' ' + str(ymin) + ' ' + str(xmax) + ' ' + str(ymax) + ' -overwrite ' + \ wd + '/openness_10m_' + tile_id + '_mosaic.sdat ' + \ wd + '/openness_10m_' + tile_id + '_mosaic.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' cropping wetness index mosaic.\n\n') # Convert to degrees, round and store as int16 # Specify gdal command cmd = settings.gdal_calc_bin + \ '-A ' + wd + '/openness_10m_' + tile_id + '_mosaic.tif ' + \ ' --outfile=' + wd + '/openness_10m_' + tile_id + '_mosaic_deg.tif ' + \ ' --calc=rint(degrees(A))' + ' --type=Int16' + ' --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' converting dtm_10m to int16 successful. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Crop slope output to original tile size: cmd = settings.gdalwarp_bin + \ ' -cutline ' + settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.shp ' + \ '-crop_to_cutline -overwrite ' + \ wd + '/openness_10m_' + tile_id + '_mosaic_deg.tif ' + \ out_folder + '/openness_10m_' + tile_id + '.tif ' log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' openness calculation successful.\n\n') return_value = 'success' except: log_file.write('\n' + tile_id + ' opennes calculation failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() # Remove temporary files try: os.remove(wd + '/dtm_10m_' + tile_id + '_mosaic_float.tif') os.remove(wd + '/openness_10m_' + tile_id + '_mosaic.sdat') os.remove(wd + '/openness_10m_' + tile_id + '_mosaic.sgrd') os.remove(wd + '/openness_10m_' + tile_id + '_mosaic.prj') os.remove(wd + '/openness_10m_' + tile_id + '_mosaic.mgrd') os.remove(wd + wd + '/openness_10m_' + tile_id + '_mosaic.tif') os.remove(wd + '/openness_10m_' + tile_id + '_mosaic_deg.tif') except: pass return return_value def dtm_remove_temp_files(tile_id): """ Removes footprint and mosaic files for the dtm to clear up space for subsequent processing. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # initiate return value return_value = '' dtm_mosaic = settings.dtm_mosaics_folder + '/dtm_' + tile_id + '_mosaic.tif' dtm_mosaic_10m = settings.dtm_mosaics_10m_folder + '/dtm_' + tile_id + '_float_mosaic_10m.tif' dtm_footprint_files = glob.glob(settings.dtm_footprint_folder + '/DTM_1km_' + tile_id + '_footprint.*') try: os.remove(dtm_mosaic) return_value = 'success' except: return_value = 'unable to delete dtm mosaic file' try: os.remove(dtm_mosaic_10m) return_value = 'success' except: return_value = 'unable to delete dtm mosaic 10 m file' try: for file in dtm_footprint_files: os.remove(file) return_value = 'success' except: return_value = return_value + 'unable to delete dtm footprint file' # return execution status return return_value
jakobjassmann/ecodes-dk-lidar
scripts/archive_outputs.py
<reponame>jakobjassmann/ecodes-dk-lidar # Quick helper script to archive all main variable folders # <NAME> j.assmann # Zip all folders import shutil import glob import re import multiprocessing import os from dklidar import settings # Function to archive folders using shutil def zip_folder(folder_path): # Set folder for zips zip_output = 'D:/Jakob/dk_nationwide_lidar/data/zipped_outputs' folder_name = zip_output + '/' + re.sub('.*\\\\(.*)\\\\', '\g<1>', folder_path) shutil.make_archive(folder_name, 'bztar', folder_path) # Main body of script if __name__ == '__main__': # Get list of folders in output directories and their names folders = glob.glob(settings.output_folder + '*/') # archive in parallel multiprocessing.set_executable(settings.python_exec_path) pool = multiprocessing.Pool(processes=len(folders)) zipping = pool.map_async(zip_folder, folders) zipping.wait() pool.close()
jakobjassmann/ecodes-dk-lidar
scripts/check_vrt_completeness.py
# EcoDes-DK15 check validity of outputs based on whether gdal can open the file. # <NAME> <EMAIL> 2 December 2021 # Dependencies import scandir import pandas import re import os import tqdm import itertools import multiprocessing from osgeo import gdal from dklidar import settings # Switch on gdal Python exceptions gdal.UseExceptions() ## 1) Function definitions # Check file function def check_vrt_completeness(folder): var_name = re.sub('.*/(.*)$', '\g<1>', folder) vrt_file_name = folder + '/' + var_name + '.vrt' file_names = [] status = [] # Check presence of vrt file if os.path.exists(vrt_file_name): # Open file vrt_file = open(vrt_file_name, 'r') # Read content vrt_file_contents = ''.join(vrt_file.readlines()) # List folder content file_list = list_files(folder) # drop non tif files from list file_list = list(itertools.compress( file_list, [bool(re.search('tif', x)) for x in file_list])) # Search tif files in vrt #progress = 0 for i in range(0, len(file_list)): if not bool(re.search( re.sub('.*/(.*)$', '\g<1>', file_list[i]), vrt_file_contents)): file_names.append(re.sub('.*/(.*)$', '\g<1>', file_list[i])) status.append('missing') ## # Update progress ## progress = float(i + 1) / float(len(file_list)) ## # Update progress bar ## print('\r|' + ## '#' * int(round(progress * 54)) + ## '-' * int(round((1 - progress) * 54)) + ## '| ' + ## str(int(round(progress * 100))) + '%'), else: file_names.append('') status.append('vrt file missing') # Combined outputs to df status_check_df = pandas.DataFrame(zip(*[file_names, status]), columns = ['file_name','error']) status_check_df['variable'] = var_name return(status_check_df) def list_files(folder_path): files = [] # Scan directory for file_name in scandir.scandir(folder_path): files.append(folder_path + '/' + file_name.name) return(files) if __name__ == '__main__': ## 2) Prepare environment # Status print('#' * 80 + '\n') print('Checking EcoDes-DK15 vrts for completeness\n\n') print('Preparing environment... '), # determine output folder structure based on original processing folders = [] for folder in scandir.scandir(settings.output_folder): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # Clean up folder paths folders = map(lambda folder: re.sub('\\\\', '/', folder), folders) # Status print('done.'), print('Checking vrts... '), # Run file checks in parallel multiprocessing.set_executable(settings.python_exec_path) pool = multiprocessing.Pool(processes=62) file_checks = list(tqdm.tqdm(pool.imap_unordered(check_vrt_completeness, folders), total = len(folders))) # Concatenate into one dataframe status_df = pandas.concat(file_checks) # Write to file status_df.to_csv(settings.log_folder + '/missing_files_in_vrts.csv', index = False) # Status print('done.'), print('\nFound ' + str(len(status_df.index)) + ' missing files.') print('Check log file for names of missing files:\n\t' + settings.log_folder + '/missing_files_in_vrts.csv') # End of File
jakobjassmann/ecodes-dk-lidar
documentation/source_data/merger_scripts/merge_outputs.py
<reponame>jakobjassmann/ecodes-dk-lidar ## EcoDes-DK output merger ## This script is used to merge the outputs of the various EcoDes processing ## runs to create the merged dataset based on the DHM_201415 merger. ## <NAME> <EMAIL> # Dependencies import os import shutil import pandas import glob import re import scandir # Status print('#' * 80) print('Merging EcoDes-DK outputs from the different reprocessing batches.') print('\nPreparing environment...'), ## 1) Set global variables # tile_ids to source tiles_to_source_original_processing = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_from_DHM2018.csv') tiles_to_source_reprocessing_1 = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_from_DHM2015.csv') tiles_to_source_reprocessing_2 = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_to_process_dhm201415_merger.csv') tiles_incomplete = pandas.read_csv('D:/Jakob/dhm201415_merger/incomplete_tile_pairs.csv') # Remove incomplete tiles from tile_ids tiles_to_source_original_processing = pandas.DataFrame( set(tiles_to_source_original_processing['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_1 = pandas.DataFrame( set(tiles_to_source_reprocessing_1['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_2 = pandas.DataFrame( set(tiles_to_source_reprocessing_2['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) # Remove redundancies tiles_to_source_original_processing = pandas.DataFrame( set(tiles_to_source_original_processing['tile_id'].tolist()) - set(tiles_to_source_reprocessing_2['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_1 = pandas.DataFrame( set(tiles_to_source_reprocessing_1['tile_id'].tolist()) - set(tiles_to_source_reprocessing_2['tile_id'].tolist()), columns = ['tile_id']) # source folders folder_original_processing = 'D:/Jakob/dk_nationwide_lidar/data/outputs' folder_reprocessing_1 = 'D:/Jakob/ecodes-dk-lidar/data/outputs' folder_reprocessing_2 = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/outputs' # destination folder dest_folder = 'D:/Jakob/ecodes-dk-lidar-rev1/data/outputs' # base folders dtm_files = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/dtm' laz_files = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/laz' # determine output folder structure based on original processing folders = [] for folder in scandir.scandir(folder_original_processing): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # remove variables that were / will be separately reprocessed (if present) folders = [folder for folder in folders if not bool(re.match('.*tile_footprints.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*solar_radiation.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*date_stamp.*', folder))] # Clean up file paths folders = map(lambda folder: re.sub('\\\\', '/', folder), folders) # Keep only relative paths folders = map(lambda folder: '/' + os.path.relpath(folder, folder_original_processing), folders) ###!!! Break(s) for debugging !!! ##folders = [folders[0]] ##folders = ['/point_source_info/point_source_counts'] ## 2) Function definitons def list_files(folder_path): files = [] # Scan directory for file_name in scandir.scandir(folder_path): files.append(file_name.name) return(files) def get_tile_ids(file_names): # initiate empty list for tile_ids tile_ids = [] # clean up files names for i in range(0, len(file_names)): file_names[i] = re.sub('\\\\', '/', file_names[i]) # fill list with tile_id for file_name in file_names: tile_id = re.sub('.*(\d{4}_\d{3}).*', '\g<1>', file_name) tile_ids.append(tile_id) # combine to data frame files_df = pandas.DataFrame(zip(*[tile_ids, file_names]), columns = ['tile_id', 'file_name']) # return files_df return(files_df) def copy_tiles(tiles_to_copy, files_df, source_folder, out_folder): # Subset files to copy files_to_copy = files_df[files_df['tile_id']. isin(tiles_to_copy['tile_id'].tolist())]['file_name'].tolist() # Set counter to 0: progress = 0 # Copy files for i in range(0, len(files_to_copy)): # Check whether file exists, if not copy if not os.path.isfile(out_folder + '/' + files_to_copy[i]): # Copy file shutil.copy(source_folder + '/' + files_to_copy[i], out_folder + '/' + files_to_copy[i]) # Update progress progress = float(i + 1) / float(len(files_to_copy)) # Update progress bar print('\r\t|' + '-' * int(round(progress * 54)) + ' ' * int(round((1 - progress) * 54)) + '| ' + str(int(round(progress * 100))) + '%'), def check_dir(folder_path): # Check if dir exists if os.path.exists(folder_path): return(0) # Determine file path components and check filpath from root to end subfolders = [folder_path] current_dir = folder_path while not current_dir == (os.path.splitdrive(folder_path)[0] + '/'): parent_dir = os.path.dirname(current_dir) subfolders.append(parent_dir) current_dir = parent_dir subfolders = subfolders[::-1][1:len(subfolders)] for folder in subfolders: if not os.path.exists(folder): os.mkdir(folder) return(0) def compare_tile_dfs(df1, df2): if(set(df1['tile_id'].tolist()) == set(df2['tile_id'].tolist())): return(pandas.DataFrame([], columns = ['tile_id'])) else: diff = pandas.DataFrame( set(df1['tile_id'].tolist()) - set(df2['tile_id'].tolist()), columns = ['tile_id']) return(diff) def get_var_from_source(var_folder, tiles_df, source_folder): # Get all files for one variable from one source # Set dest directory based on global variable global dest_folder out_folder = dest_folder + var_folder # and create if needed check_dir(out_folder) # Set source folder source_folder = source_folder + var_folder # Generate df of tile_ids and file names files_to_copy = get_tile_ids(list_files(source_folder)) # Copy files copy_tiles(tiles_df, files_to_copy, source_folder, out_folder) return(0) def get_var(var_folder): # Get global variables global tiles_to_source_original_processing, tiles_to_source_reprocessing_1, tiles_to_source_reprocessing_2, folder_original_processing, folder_reprocessing_1, folder_reprocessing_2 # Get all files for one variable from the three data sources print('Sourcing: ' + var_folder) print('\tOriginal processing') get_var_from_source(var_folder, tiles_to_source_original_processing, folder_original_processing) print('\n\tReprocessing #1') get_var_from_source(var_folder, tiles_to_source_reprocessing_1, folder_reprocessing_1) print('\n\tReprocessing #2') get_var_from_source(var_folder, tiles_to_source_reprocessing_2, folder_reprocessing_2) # Return df of copied tiles and files_copied = get_tile_ids(list_files(dest_folder + var_folder)) files_copied['var_folder'] = var_folder print('\tDone.\n') return(files_copied) # Status print(' done.\n') ## 3) Main body of script ## Check completeness of tile lists # Status print('Checking completness of tile list to copy...'), # Get all tiles in data set dhm_merged_tiles = get_tile_ids(list_files(laz_files)) # Compare with tiles to merge missing_tiles = len(compare_tile_dfs(dhm_merged_tiles, pandas.concat([tiles_to_source_original_processing, tiles_to_source_reprocessing_1, tiles_to_source_reprocessing_2]))) if missing_tiles > 0: # Prompt for choice to stop del_choice = raw_input('\n' + str(missing_tiles) + ' are missing! Continue anyways?' + '[y/n]') if not del_choice == 'y': print('Aborting on request or invalid choice!') quit() else: print('Okay, continuing merger.') else: print('\n=> Sets are complete, proceeding as planned.\n') ## Copy tiles for all variables # Status print('Starting merger:\n\n') files_copied_dfs = [] for var_folder in folders: files_copied_df = get_var(var_folder) files_copied_dfs.append(files_copied_df) # Stauts print('Merger complete!\n') print('#' * 80 + '\n') ## Quality control # Status print('Quality control:\n') # Quality control: files_missing_dfs = [] for files_copied_df in files_copied_dfs: # Get tiles missed in copying (e.g. due to absence in source) files_missing_df = compare_tile_dfs(dhm_merged_tiles, files_copied_df) if len(files_missing_df) > 0: # Status print('\t' + files_copied_df['var_folder'][1] + ' is missing: ' + str(len(files_missing_df)) + ' tiles.\n') # Add missing tiles to list files_missing_df['var_folder'] = files_copied_df['var_folder'][1] files_missing_dfs.append(files_missing_df) else: # Status print('\t' + files_copied_df['var_folder'][1] + ' complete.\n') files_missing_dfs = pandas.concat(files_missing_dfs) files_missing_dfs.to_csv('final_merger_missing_files.csv', index = False) # Status print('A total of ' + str(len(files_missing_dfs)) + ' files are missing.\n') print('Quality control complete.\n') print('Merger complete.\n') print('#' * 80 + '\n') # EOF
jakobjassmann/ecodes-dk-lidar
scripts/generate_dems.py
import multiprocessing import pandas import subprocess import re import datetime import os import opals from dklidar import settings # initate opals opals.loadAllModules() # make folder for temp odms os.mkdir('temp_odms') os.chdir('temp_odms') # load missing tiles missing_dtms = pandas.read_csv(settings.dtm_folder + '../missing_dtm_tile_ids.csv')['tile_id'].tolist() print('Generating dems for ' + str(len(missing_dtms)) + ' tiles...') for tile_id in missing_dtms: os.mkdir(tile_id) os.chdir(tile_id) laz_file = settings.laz_folder + '/PUNKTSKY_1km_' + tile_id + '.laz' odm_file = tile_id + '_temp.odm' dtm_file = settings.dtm_folder + '/DTM_1km_' + tile_id # 1) Import tile into temporary odm try: import_tile = opals.Import.Import() import_tile.inFile = laz_file import_tile.outFile = odm_file import_tile.commons.screenLogLevel = opals.Types.LogLevel.none import_tile.run() print('Imported ' + tile_id + '. '), except: print('Unable to import ' + tile_id + '.'), #1) Validate CRS try: odm_dm = opals.pyDM.Datamanager.load(odm_file) crs_str = odm_dm.getCRS() # Check whether CRS exists, if not assign, if different throw error. if crs_str == settings.crs_wkt_opals: print ('crs: match. '), elif crs_str == '': odm_dm.setCRS(settings.crs_wkt_opals) print('crs: empty - set. '), else: print('crs: warning - no match. '), odm_dm = None # This is needed as opals locks the file connection otherwise. except: print('crs validation error. '), #2) Generate DEM # Section adapted from <NAME> code in ALS Calculator.py # --- try: dtm_export = opals.DTM.DTM() dtm_export.inFile = odm_file dtm_export.feature = opals.Types.GridFeature.sigmaz dtm_export.outFile = dtm_file dtm_export.gridSize = 0.4 dtm_export.filter = "Generic[Classification == 2]" dtm_export.multiBand = False dtm_export.neighbours = 8 # recommendation by TU Wien (<NAME>) dtm_export.searchRadius = 6 # recommendation by TU Wien (<NAME>) dtm_export.commons.screenLogLevel = opals.Types.LogLevel.none dtm_export.run() print('dtm export success.'), except: print('dtm export failed. '), # --- # tidy up try: os.rename((dtm_file + '_dtm.tif'), (dtm_file + '.tif')) os.remove(dtm_file + '_sigmaz.tif') print('tidy up successfull') except: print('tidy up failed.') out_file = open(settings.laz_folder + '../missing_dtms_generated.txt', 'a+') out_file.write(dtm_file + '.tif\n') out_file.close() os.chdir('..')
jakobjassmann/ecodes-dk-lidar
scripts/check_outputs_integrity.py
# EcoDes-DK15 check validity of outputs based on whether gdal can open the file. # <NAME> <EMAIL> 2 December 2021 # Dependencies import scandir import pandas import re import tqdm from osgeo import gdal import multiprocessing from dklidar import settings # Switch on gdal Python exceptions gdal.UseExceptions() ## 1) Function definitions # Check file function def check_file(file_name): file_path = [] error = [] try: gtif = gdal.Open(file_name) gtif = None except RuntimeError, e: file_path.append(file_name) error.append(e) return(pandas.DataFrame(zip(*[file_path, error]), columns = ['file_name','error'])) def list_files(folder_path): files = [] # Scan directory for file_name in scandir.scandir(folder_path): files.append(folder_path + '/' + file_name.name) return(files) if __name__ == '__main__': ## 2) Prepare environment # Status print('#' * 80 + '\n') print('Validating EcoDes-DK15 outputs through a gdal load\n\n') print('Preparing environment...'), # determine output folder structure based on original processing folders = [] for folder in scandir.scandir(settings.output_folder): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # Scan files file_lists = [list_files(folder) for folder in folders] file_list = [file_name for file_list in file_lists for file_name in file_list] # break for debug: file_list = file_list # Status print check - if needed print(' done.\n') #print('\n') #print(''.join(map(lambda folder: folder + '\n', folders))) print('Checking ' + str(len(folders)) + ' output folders.') print('Containing: ' + str(len(file_list)) + ' files.\n') # Run file checks in parallel multiprocessing.set_executable(settings.python_exec_path) pool = multiprocessing.Pool(processes=62) file_checks = list(tqdm.tqdm(pool.imap_unordered(check_file, file_list), total = len(file_list))) # Concatenate into one dataframe errors_df = pandas.concat(file_checks) # Write to file errors_df.to_csv(settings.log_folder + '/output_file_erros.csv', index = False) # Status print('\nFound ' + str(len(errors_df.index)) + ' errors.') print('Check log file for errors:\n\t' + settings.log_folder + '/output_file_erros.csv') # End of File
jakobjassmann/ecodes-dk-lidar
scripts/download_files.py
<reponame>jakobjassmann/ecodes-dk-lidar import multiprocessing import pandas import subprocess import re import datetime import os from dklidar import settings remote_url_laz = "https://download.kortforsyningen.dk/system/files/Statsaftalen/DANMARK/3_HOJDEDATA/PUNKTSKY/" remote_url_dtm = "https://download.kortforsyningen.dk/system/files/Statsaftalen/DANMARK/3_HOJDEDATA/DTM/" file_list_laz = 'D:/Jakob/dk_nationwide_lidar/data/file_lists_kortforsyningen/all_laz_files.csv' file_list_dtm = 'D:/Jakob/dk_nationwide_lidar/data/file_lists_kortforsyningen/all_dtm_files.csv' # A cookie is needd to access the data. To generate the following curl command you can use Google Chrome. # First, log into the kortforsyningen website then select one tile from the laz files in the catalogues and check out # Open the "Mine downloads" section of the website and activate the Chrome Developer Mode by pressing "F12". # In the newly opened panel, select the "Network" tap and pres "crtl + R" to activate the monitoring for the website. # Then start the download of the file that you checked out and right-click on the associated item in the network tab. # Click "copy" and select "curl CMD". Then replace the following cookie string with the cookie string in # the copied curl command. You may have to adapt the curl command in download_file() accordingly if you change system. cookie = 'Cookie: nmstat=1571735317171; _1d3fe=http://10.0.0.177:80; MY_SESSION=rd4o00000000000000000000ffffac1105c7o80; has_js=1; SESS1fef01291fb7205e2d46065129ba3bc3=jrFlh_HehTZC8-4x7l0loZ92TuhL0ikAewDwXNCpKiQ; downloadticket=1853700203f2d9b03cce33e4cb0c8c37' def download_file(file_name): """ Wee helper function to download file from Kortforsyningen server :param file_name - file name on server and to be saved as :returns exit code of curl download command """ remote_url = 'empty_url' local_path = '/empty_path' if not(re.match('.*LAZ.*', file_name) is None): remote_url = remote_url_laz + file_name local_path = settings.laz_folder + file_name if not (re.match('.*DTM.*', file_name) is None): remote_url = remote_url_dtm + file_name local_path = settings.dtm_folder + file_name curl_cmd = '"C:/Program Files/Git/mingw64/bin/curl.exe" ' + \ '"' + remote_url + '" ' +\ '-H "Connection: keep-alive" ' +\ '-H "Upgrade-Insecure-Requests: 1" ' + \ '-H "User-Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/80.0.3987.163 Safari/537.36" ' + \ '-H "Sec-Fetch-Dest: iframe" ' + \ '-H "Accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9" ' + \ '-H "Sec-Fetch-Site: same-origin" ' + \ '-H "Sec-Fetch-Mode: navigate" ' +\ '-H "Sec-Fetch-User: ?1" ' + \ '-H "Referer: https://download.kortforsyningen.dk/mine/downloads" ' + \ '-H "Accept-Language: en,da-DK;q=0.9,da;q=0.8,en-US;q=0.7" ' + \ '-H ' + '"' + cookie + '" ' +\ '--compressed ' + \ '--output ' + '"' + local_path + '"' try: FNULL = open(os.devnull, 'w') subprocess.check_call(curl_cmd, stdout=FNULL, stderr=subprocess.STDOUT) return_value = 'success' except subprocess.CalledProcessError as error: return_value = 'curl non-zero exit status: ' + str(error.returncode) print '.', return return_value def extract_file(file_name): """ Wee helper function to download file from Kortforsyningen server :param file_name - file name on server and to be saved as :returns exit code of curl download command """ local_path = '/empty_path' dest_folder = '/empty_path' if not(re.match('.*LAZ.*', file_name) is None): local_path = settings.laz_folder + file_name dest_folder = settings.laz_folder if not (re.match('.*DTM.*', file_name) is None): local_path = settings.dtm_folder + file_name dest_folder = settings.dtm_folder extract_cmd = '"C:/Program Files/7-Zip/7z" ' + \ 'e ' + \ '-y ' + local_path + ' ' + \ '-o' + dest_folder try: FNULL = open(os.devnull, 'w') subprocess.check_call(extract_cmd, stdout=FNULL, stderr=subprocess.STDOUT) return_value = 'success' except subprocess.CalledProcessError as error: return_value = '7zip non-zero exit status: ' + str(error.returncode) print '.', return return_value #### Main body of script if __name__ == '__main__': # Start timer startTime = datetime.datetime.now() print('\n') print('-' * 80) print("DK nationwide LiDAR download script") print(str(startTime)) # Load file lists as data frames: laz_files_df = pandas.read_csv(file_list_laz) dtm_files_df = pandas.read_csv(file_list_dtm) if not ('download' in laz_files_df): laz_files_df['download'] = 'pending' if not ('extraction' in laz_files_df): laz_files_df['extraction'] = 'pending' if not ('download' in dtm_files_df): dtm_files_df['download'] = 'pending' if not ('extraction' in dtm_files_df): dtm_files_df['extraction'] = 'pending' # Prepare download lists laz_files_to_download = laz_files_df['file_name'][laz_files_df['download'] != 'success'].tolist() dtm_files_to_download = dtm_files_df['file_name'][laz_files_df['download'] != 'success'].tolist() # Prep processing pool multiprocessing.set_executable(settings.python_exec_path) pool = multiprocessing.Pool(processes=4) # Download LAZ files if len(laz_files_to_download) > 0: print(datetime.datetime.now().strftime('%X') + ' Downloading ' + str(len(laz_files_to_download)) + ' laz files: '), download_pool = pool.map_async(download_file, laz_files_to_download) download_pool.wait() laz_files_df['download'] = download_pool.get() else: print('No LAZ files to download.') print('\n') # Download DTM files if len(dtm_files_to_download) > 0: print(datetime.datetime.now().strftime('%X') + ' Downloading ' + str(len(dtm_files_to_download)) + ' dtm files: '), download_pool = pool.map_async(download_file, dtm_files_to_download) download_pool.wait() dtm_files_df['download'] = download_pool.get() else: print('No DTM files to download.') print('\n') # Prepare extraction lists laz_files_to_extract = laz_files_df['file_name'][laz_files_df['extraction'] != 'success'].tolist() dtm_files_to_extract = dtm_files_df['file_name'][dtm_files_df['extraction'] != 'success'].tolist() # # Extract LAZ files if len(laz_files_to_extract) > 0: print(datetime.datetime.now().strftime('%X') + ' Extracting ' + str(len(laz_files_to_extract)) + ' laz files: '), download_pool = pool.map_async(extract_file, laz_files_to_extract) download_pool.wait() laz_files_df['extraction'] = download_pool.get() else: print('No LAZ files to extract.') print('\n') # Extract DTM files if len(dtm_files_to_extract) > 0: print(datetime.datetime.now().strftime('%X') + ' Extracting ' + str(len(dtm_files_to_extract)) + ' dtm files: '), download_pool = pool.map_async(extract_file, dtm_files_to_extract) download_pool.wait() dtm_files_df['extraction'][dtm_files_df['extraction'] != 'success'] = download_pool.get() else: print('No LAZ files to extract.') print('\n') laz_files_df.to_csv(file_list_laz, index=False, header=True) dtm_files_df.to_csv(file_list_dtm, index=False, header=True) print('-' * 80 + '\nDone.\nTime elapsed: ' + str(datetime.datetime.now() - startTime))
jakobjassmann/ecodes-dk-lidar
scripts/process_tiles.py
<filename>scripts/process_tiles.py ### Script to rasterise point clouds for the DK nationwide lidar re-processing ### <NAME> <EMAIL> 29 January 2019 ## Imports import glob import re import os import shutil import datetime import time import multiprocessing import pandas import opals from dklidar import points from dklidar import dtm from dklidar import settings from dklidar import common #### Prepare the environment # Set working directory os.chdir(settings.wd) # Set number of parallel processes: n_processes = 62 # 54 # Confirm essential folders exist if not os.path.exists(settings.wd): print('Working Directory ' + settings.wd + ' does not exist. Exiting script...') exit() if not os.path.exists(settings.laz_folder): print('laz_folder ' + settings.laz_folder + ' does not exist. Exiting script...') exit() if not os.path.exists(settings.dtm_folder): print('dtm_folder ' + settings.dtm_folder + ' does not exist. Exiting script...') exit() # Conmfirm other folders exists and if not create them for folder in [settings.dtm_mosaics_folder, settings.dtm_footprint_folder, settings.odm_folder, settings.odm_footprint_folder, settings.output_folder]: if not os.path.exists(folder): os.mkdir(folder) # Load file names dtm_files = glob.glob(settings.dtm_folder + '/*.tif') laz_files = glob.glob(settings.laz_folder + '/*.laz') # initiate empty lists for tile_ids dtm_tile_ids = [] laz_tile_ids = [] # fill dictionaries with tile_id, as well as row number and column number for each file name: for file_name in dtm_files: tile_id = re.sub('.*DTM_1km_(\d*_\d*).tif', '\g<1>', file_name) dtm_tile_ids.append(tile_id) for file_name in laz_files: tile_id = re.sub('.*PUNKTSKY_1km_(\d*_\d*).laz', '\g<1>', file_name) laz_tile_ids.append(tile_id) ## Define processing steps for each tile to be carried out in parallel def process_tile(tile_id): ## Prepare environment # Generate temporary wd for parallel worker, this will allow for smooth logging and opals sessions to run in parallel wd = os.getcwd() current_pid = re.sub('[(),]', '', str(multiprocessing.current_process()._identity)) temp_wd = settings.scratch_folder + '/temp_' + current_pid if not os.path.exists(temp_wd): os.mkdir(temp_wd) os.chdir(temp_wd) # Create folder for logging tile_log_folder = settings.log_folder + '/process_tiles/' + tile_id if not os.path.exists(tile_log_folder): os.mkdir(tile_log_folder) # Stagger processing if this is the first turn during a processing id # This is a crucial step to reduce the chance of the functions using gdal_rasterize to run directly in parallel # (generate mask function), if more than 10 gdla_rasterize instances run in parallel there is a massive drop in # performance for some reason if not os.path.exists(temp_wd + '/first_go_complete.txt'): # sleep for 5 seconds. Providing the maximum performance gain for all full integers < 6s = sequential time.sleep(5 * int(re.sub('[(),]', '', str(multiprocessing.current_process()._identity)))) lock_file = open(temp_wd + '/first_go_complete.txt', 'w') lock_file.close() # opals loadModules opals.loadAllModules() ## Logging: Keep track of progress for each step and overall progress # Initiate progress variables steps = ['processing'] status_steps = [['complete']] ## Generate masks return_value = common.generate_water_masks(tile_id) # Update progress variables steps.append('generate_water_masks') status_steps.append([return_value]) # gather logs for step and tile common.gather_logs('process_tiles', 'generate_water_masks', tile_id) ## Import tile to ODM return_value = points.odm_import_single_tile(tile_id) # Update progress variables steps.append('odm_import_single_tile') status_steps.append([return_value]) # gather logs for step and tile common.gather_logs('process_tiles', 'odm_import_single_tile', tile_id) ## Validate CRS of odm files return_value = points.odm_validate_crs(tile_id) # Update progress variables steps.append('odm_validate_crs') status_steps.append([return_value]) ## Export footprint return_value = points.odm_generate_footprint(tile_id) # Update progress variables steps.append('odm_generate_footprint') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_generate_footprint', tile_id) ## Normalise height return_value = points.odm_add_normalized_z(tile_id) # Update progress variables steps.append('odm_add_normalized_z') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_add_normalized_z', tile_id) ## Export mean normalised height for 10 m x 10 m cell return_value = points.odm_export_normalized_z(tile_id) # Update progress variables steps.append('odm_export_normalized_z') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_normalized_z', tile_id) ## Export canopy height return_value = points.odm_export_canopy_height(tile_id) # Update progress variables steps.append('odm_export_canopy_height') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_canopy_height', tile_id) ## Export point counts for pre-defined intervals and classess return_value = points.odm_export_point_counts(tile_id) # Update progress variables steps.append('odm_export_point_counts') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_point_counts', tile_id) ## Export proportions based on point counts return_value = points.odm_export_proportions(tile_id) # Update progress variables steps.append('odm_export_proportions') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_proportions', tile_id) ## Export point source information return_value = points.odm_export_point_source_info(tile_id) # Update progress variables steps.append('odm_export_point_source_info') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_point_source_info', tile_id) ## Export amplitude mean and sd return_value = points.odm_export_amplitude(tile_id) # Update progress variables steps.append('odm_export_amplitude') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_amplitude', tile_id) ## Export date stamps return_value = points.odm_export_date_stamp(tile_id) # Update progress variables steps.append('odm_export_date_stamp') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_export_date_stamp', tile_id) ## Remove unneeded odm files return_value = points.odm_remove_temp_files(tile_id) # Update progress variables steps.append('odm_remove_temp_files') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'odm_remove_temp_files', tile_id) ## Terrain model derived variables ## Generate tile footprint return_value = dtm.dtm_generate_footprint(tile_id) # Update progress variables steps.append('dtm_generate_footprint') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_generate_footprint', tile_id) ## Generate neighbourhood mosaic return_value = dtm.dtm_neighbourhood_mosaic(tile_id) # Update progress variables steps.append('dtm_neighbourhood_mosaic') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_neighbourhood_mosaic', tile_id) ## Validate CRS return_value = dtm.dtm_validate_crs(tile_id) # Update progress variables steps.append('dtm_validate_crs') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_validate_crs', tile_id) ## Generate 10 m aggregate of DEM return_value = dtm.dtm_aggregate_tile(tile_id) # Update progress variables steps.append('dtm_aggregate_tile') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_aggregate_tile', tile_id) ## Generate 10 m aggregate of neighbourhood mosaic return_value = dtm.dtm_aggregate_mosaic(tile_id) # Update progress variables steps.append('dtm_aggregate_mosaic') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_aggregate_mosaic', tile_id) ## Calculate slope return_value = dtm.dtm_calc_slope(tile_id) # Update progress variables steps.append('dtm_calc_slope') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_calc_slope', tile_id) ## Calculate aspect return_value = dtm.dtm_calc_aspect(tile_id) # Update progress variables steps.append('dtm_calc_aspect') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_calc_aspect', tile_id) ## Calculate heat index return_value = dtm.dtm_calc_heat_index(tile_id) # Update progress variables steps.append('dtm_calc_heat_index') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_calc_heat_index', tile_id) ## Calculate solar radiation return_value = dtm.dtm_calc_solar_radiation(tile_id) # Update progress variables steps.append('dtm_calc_solar_radiation') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_calc_solar_radiation', tile_id) ## Calculate landscape openness mean return_value = dtm.dtm_openness_mean(tile_id) # Update progress variables steps.append('dtm_openness_mean') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_openness_mean', tile_id) ## Calculate landscape openness difference return_value = dtm.dtm_openness_difference(tile_id) # Update progress variables steps.append('dtm_openness_difference') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_openness_difference', tile_id) ## Calculate Kopecky TWI return_value = dtm.dtm_kopecky_twi(tile_id) # Update progress variables steps.append('dtm_kopecky_twi') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_kopecky_twi', tile_id) ## Remove unneeded dtm files return_value = dtm.dtm_remove_temp_files(tile_id) # Update progress variables steps.append('dtm_remove_temp_files') status_steps.append([return_value]) # gather logs for step and tile] common.gather_logs('process_tiles', 'dtm_remove_temp_files', tile_id) ## Logging: finalise log outputs # Zip into pandas data frame status_df = pandas.DataFrame(zip(*status_steps), index = [tile_id], columns=steps) status_df.index.name = 'tile_id' # Export as CSV status_df.to_csv(tile_log_folder + '/status.csv', index=True, header=True) # Change back to original working directory os.chdir(wd) # Print tile_id to console to update on status print(datetime.datetime.now().strftime('%X') + ' ' + tile_id + ' '), #### Main body of script if __name__ == '__main__': ## Start timer startTime = datetime.datetime.now() ## Status output to console print('\n' + '-' * 80 + 'Starting process_tiles.py at ' + str(startTime.strftime('%c')) + '\n') ## Prepare process managment and logging progress_df = common.init_log_folder('process_tiles', laz_tile_ids) ## Identify which tiles still require processing tiles_to_process = set(progress_df.index.values[progress_df['processing'] != 'complete'].tolist()) ## If processing of a specific subset of tiles is needed ## the following lines can be helpful in achieving the task ## Remove comments as needed. ## tiles_to_process_completed = set(progress_df.index.values[progress_df['processing'] == 'complete'].tolist()) ## tiles_to_process_dhm201415 = set(pandas.read_csv('auxillary_files/tiles_to_process_dhm201415_merger.csv')['tile_id'].tolist()) ## tiles_to_process = tiles_to_process_dhm201415 - tiles_to_process_completed ## tiles_to_process = tiles_to_process - (tiles_to_process - set(progress_df.index.values.tolist())) ## print('Processing ' + str(len(tiles_to_process)) + ' tiles. \n') ## time.sleep(60) # Set up processing pool multiprocessing.set_executable(settings.python_exec_path) pool = multiprocessing.Pool(processes=n_processes) # Execute processing of tiles print(datetime.datetime.now().strftime('%X') + ' Processing tiles: ... '), tile_processing = pool.map_async(process_tile, tiles_to_process) # Make sure all processes finish before carrying on. tile_processing.wait() print('... done.') # Clear scratch folder shutil.rmtree('scratch') os.mkdir('scratch') shutil.copy('data/empty_on_purpose.txt', 'scratch/empty_on_purpose.txt') # Update progress status progress_df = common.update_progress_df('process_tiles', progress_df) # Export progress_df as CSV progress_file = settings.log_folder + '/process_tiles/' + 'overall_progress.csv' progress_df.to_csv(progress_file, index=True, header=True) # Print out time elapsed: print('\nTime elapsed: ' + str(datetime.datetime.now() - startTime))
jakobjassmann/ecodes-dk-lidar
dklidar/settings.py
### Settings file for the DK Lidar project ### <NAME> <EMAIL> 29 January 2019 # This file is used to provide global variables to all scripts in the DK nationwide lidar project. # Path to python executable python_exec_path = 'C:/Program Files/opals_nightly_2.3.2/opals/python.exe' # Set paths to gdal executables / binaries (here we use OSGE4W64) as the OPALS gdal binaries do not work reliably # remember the trailing spaces at the end! # simply set to the gdal command e.g. 'gdalwarp ' if you want to use the OPALS gdal binaries (gdaldem slope won't work) gdalwarp_bin = 'C:/OSGeo4W/OSGeo4W.bat gdalwarp ' gdaldem_bin = 'C:/OSGeo4W/OSGeo4W.bat gdaldem ' gdaltlindex_bin = 'C:/OSGeo4W/OSGeo4W.bat gdaltindex ' gdal_translate_bin = 'C:/OSGeo4W/OSGeo4W.bat gdal_translate ' gdal_calc_bin = 'C:/OSGeo4W/OSGeo4W.bat gdal_calc ' gdal_merge_bin = 'C:/OSGeo4W/OSGeo4W.bat gdal_merge ' gdaltransform_bin = 'C:/OSGeo4W/OSGeo4W.bat gdaltransform ' gdalinfo_bin = 'C:/OSGeo4W/OSGeo4W.bat gdalinfo ' gdal_rasterize_bin = 'C:/OSGeo4W/OSGeo4W.bat gdal_rasterize ' gdalsrsinfo_bin = 'C:/OSGeo4W/OSGeo4W.bat gdalsrsinfo ' # set gdal version (this matters for the crs outputs) gdal_version = '3.3.3' # '2.2.4' # Saga binary commands saga_wetness_bin = 'C:/OSGeo4W/apps/saga-ltr/saga_cmd.exe --cores=1 ta_hydrology 15 ' saga_openness_bin = 'C:/OSGeo4W/apps/saga-ltr/saga_cmd.exe --cores=1 ta_lighting 5 ' saga_bin = 'D:/Jakob/saga-7.8.2_x64/saga_cmd.exe --cores=1 ' ### Set folder locations # Main working directory wd = 'D:/Jakob/ecodes-dk-lidar-rev1' # Point cloud folder laz_folder = wd + '/data/laz/' # DTM folder dtm_folder = wd + '/data/dtm/' # DTM mosaics and footprints dtm_mosaics_folder = wd + '/data/dtm_mosaics' dtm_mosaics_10m_folder = wd + '/data/dtm_mosaics_10m' dtm_footprint_folder = wd + '/data/dtm_footprints' # ODM folder odm_folder = wd + '/data/odm/' # ODM Mosaics and footprint odm_mosaics_folder = wd +'/data/odm_mosaics' odm_footprint_folder = wd + '/data/odm_footprint/' # Folder for ouptuts output_folder = wd + '/data/outputs/' # Scratch folder for temporary data storage scratch_folder = wd + '/scratch' # Log folder for global log ouptus log_folder = wd + '/log' # Mask files dk_coastline_poly = 'D:/Jakob/ecodes-dk-lidar-rev1/auxillary_files/dk_bounds_jesper_poly.shp' dk_lakes_poly = 'D:/Jakob/eecodes-dk-lidar-rev1/auxillary_files/lake_mask_jesper.shp' ## Spatial reference settings # common crs as WKT string crs_wkt_opals = r'PROJCS["ETRS89 / UTM 32N",GEOGCS["ETRS89",DATUM["European_Terrestrial_Reference_System_1989",SPHEROID["GRS 1980",6378137,298.257222101,AUTHORITY["EPSG","7019"]],AUTHORITY["EPSG","6258"]],PRIMEM["Greenwich",0,AUTHORITY["EPSG","8901"]],UNIT["degree",0.01745329251994328,AUTHORITY["EPSG","9122"]],AUTHORITY["EPSG","4258"]],PROJECTION["Transverse_Mercator"],PARAMETER["latitude_of_origin",0],PARAMETER["central_meridian",9],PARAMETER["scale_factor",0.9996],PARAMETER["false_easting",500000],PARAMETER["false_northing",0],UNIT["metre",1,AUTHORITY["EPSG","9001"]],AXIS["Easting",EAST],AXIS["Northing",NORTH],AUTHORITY["EPSG","25832"]]' crs_wkt_gdal = r'PROJCS["ETRS89 / UTM zone 32N",GEOGCS["ETRS89",DATUM["European_Terrestrial_Reference_System_1989",SPHEROID["GRS 1980",6378137,298.257222101,AUTHORITY["EPSG","7019"]],TOWGS84[0,0,0,0,0,0,0],AUTHORITY["EPSG","6258"]],PRIMEM["Greenwich",0,AUTHORITY["EPSG","8901"]],UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]],AUTHORITY["EPSG","4258"]],PROJECTION["Transverse_Mercator"],PARAMETER["latitude_of_origin",0],PARAMETER["central_meridian",9],PARAMETER["scale_factor",0.9996],PARAMETER["false_easting",500000],PARAMETER["false_northing",0],UNIT["metre",1,AUTHORITY["EPSG","9001"]],AXIS["Easting",EAST],AXIS["Northing",NORTH],AUTHORITY["EPSG","25832"]]' crs_proj4_gdal = r'+proj=utm +zone=32 +ellps=GRS80 +towgs84=0,0,0,0,0,0,0 +units=m +no_defs' ## Multiprocessing settings # common nbThreads parameter - a throttle limiter for OPALS, ensures Opals subprocesses use only a single core nbThreads = 1 ## Processing Options # Output cell size out_cell_size = 10 ## Filter Strings # point filter for all three vegetation classes as OPALS WKT veg_classes_filter = "Generic[Classification == 3 OR Classification == 4 OR Classification == 5]" ground_and_veg_classes_filter = "Generic[Classification == 2 OR Classification == 3 OR Classification == 4 OR Classification == 5]" all_classes = "Generic[Classification == 2 OR Classification == 3 OR Classification == 4 OR Classification == 5 OR Classification == 6 OR Classification == 9]"
jakobjassmann/ecodes-dk-lidar
scripts/debug.py
# Script for debugging of dklidar module functons # <NAME> <EMAIL> 18 February 2021 # This script requires the input files (dtms and laz) to be stored # in the folders specified in dklidar/settings.py # This also applies to the binaries specified in the same file ## Preparations print('Preparing environment\n') # Dependencies import os import sys import opals import datetime # Load dklidar module parts for direct access from dklidar import points from dklidar import dtm from dklidar import settings from dklidar import common # Set temporary workdirectory temp_wd = 'D:/Jakob/ecodes-dk-lidar-reprocessing/scratch/debug' # Set temporary out directory settings.output_folder = 'D:/Jakob/ecodes-dk-lidar-reprocessing/scratch/debug/output' # Create folders if needed for folder in [temp_wd, settings.output_folder]: if not os.path.exists(folder): os.mkdir(folder) # Change to temporary work dir os.chdir(temp_wd) # Set tile_id if not set as argument args = [arg for arg in sys.argv[1:] if not arg.startswith("-")] if len(args) == 0: tile_id = '6186_524' else: tile_id = args[0] print('Processing tile_id: ' + tile_id + '\n') ### Load opals modules opals.loadAllModules() # Confirm essential folders exist if not os.path.exists(settings.wd): print('Working Directory ' + settings.wd + ' does not exist. Exiting script...') exit() if not os.path.exists(settings.laz_folder): print('laz_folder ' + settings.laz_folder + ' does not exist. Exiting script...') exit() if not os.path.exists(settings.dtm_folder): print('dtm_folder ' + settings.dtm_folder + ' does not exist. Exiting script...') exit() # Conmfirm other folders exists and if not create them for folder in [settings.dtm_mosaics_folder, settings.dtm_footprint_folder, settings.odm_folder, settings.odm_footprint_folder, settings.output_folder]: if not os.path.exists(folder): os.mkdir(folder) ## ------------------- ## Start timer print('\nStarting Timer\n') startTime = datetime.datetime.now() ## ------------------- ## Test dtm functions print('#' * 60) print('\nTesting DTM Functions\n') # Generate tile footprint print('=> Generate Tile Footprint') print(dtm.dtm_generate_footprint(tile_id)) # Generate tile neighbourhood mosaic print('=> Generate Neighbourhood Mosaic') print(dtm.dtm_neighbourhood_mosaic(tile_id)) # Validate CRS print('=> Validate CRS') print(dtm.dtm_validate_crs(tile_id)) # Aggregate tile to 10 m print('=> Aggregate tile to 10 m') print(dtm.dtm_aggregate_tile(tile_id)) # Aggregate neighbourhood mosaic to 10 m print('=> Aggregate Neighbourhood Mosaic to 10 m') print(dtm.dtm_aggregate_mosaic(tile_id)) # Calculate slope print('=> Calculate Slope') print(dtm.dtm_calc_slope(tile_id)) # Calculate aspect print('=> Calculate Aspect') print(dtm.dtm_calc_aspect(tile_id, -1)) # Calculate heat index print('=> Calculate Heat Index') print(dtm.dtm_calc_heat_index(tile_id, -10)) # Calculate solar radiation print('=> Calculate Solar Radiation') print(dtm.dtm_calc_solar_radiation(tile_id)) # Calculate landscape openness mean print('=> Calculate Openness Mean') print(dtm.dtm_openness_mean(tile_id)) # Calculate landscape openness difference print('=> Calculate Openness Difference') print(dtm.dtm_openness_difference(tile_id)) ## Calculate Kopecky TWI print('=> Calculating Kopecky TWI') print(dtm.dtm_kopecky_twi(tile_id)) # Remove old temp dtm files print('=> Removing DTM temp files') print(dtm.dtm_remove_temp_files(tile_id)) ## ------------------- ## Test common functions print('\n') print('#' * 60) print('\nTesting Common Functions\n') ## Generate Masks print('=> Generating Masks') print(common.generate_water_masks(tile_id)) ## ------------------- ## Test points functions print('\n') print('#' * 60) print('\nTesting Point Cloud Functions\n') ## Import tile to ODM print('=> Importing ODM') print(points.odm_import_single_tile(tile_id)) ## Validate CRS of odm files print('=> Validate CRS of ODMs') print(points.odm_validate_crs(tile_id)) ## Export footprint print('=> Generating footprints from ODM') print(points.odm_generate_footprint(tile_id)) ## Normalise height print('=> Normalize Height') print(points.odm_add_normalized_z(tile_id)) ## Export mean normalised height for 10 m x 10 m cell print('=> Export Normalize Height') print(points.odm_export_normalized_z(tile_id)) ## Export canopy height print('=> Export Canopy Height') print(points.odm_export_canopy_height(tile_id)) ## Export point counts for pre-defined intervals and classess print('=> Export Point Counts') print(points.odm_export_point_counts(tile_id)) ## Export proportions based on point counts print('=> Export Proportions') print(points.odm_export_proportions(tile_id)) ## Export point source information print('=> Export Point Source Information') print(points.odm_export_point_source_info(tile_id)) ## Export amplitude mean and sd print('=> Export Amplitude Mean and SD') print(points.odm_export_amplitude(tile_id)) ## Export date stamp print('=> Exporting Date Stamps') print(points.odm_export_date_stamp(tile_id)) ## Additional export for selected point counts (to speed up testing) #### Export total point count ##print('=> Export Total Point Count') ##print(points.odm_export_point_count(tile_id, 'total_point_count', -1, 50, [2,3,4,5,6,9])) ## #### Export veg point count ##print('=> Export Total Point Count') ##print(points.odm_export_point_count(tile_id, 'vegetation_point_count', 0, 50, [3,4,5])) ## Remove unneeded odm files print('=> Remove ODM Temp Files') print(points.odm_remove_temp_files(tile_id)) ## ------------------- # Report time elapsed print('\n') print('#' * 60) print('Debug Complete\nTime elapsed: ' + str(datetime.datetime.now() - startTime)) print("Run debug.Rmd for visual analysis") print('\n')
jakobjassmann/ecodes-dk-lidar
scripts/fix_na_tile_data_type.py
<gh_stars>0 # EcoDes-DK - Fix files missing from VRTs # <NAME> <EMAIL> 2 December 2021 # Most files missing from the VRTs are missing because they are the wrong raster # file type (e.g. Int16 instead of float32) - the only tiles affected seem to # be NA tiles. # These originate from the processing workflow and OPALS output, but # also from the fill_processing_gaps.py script that does not accoutn for # differences in the file paths. This script is here to correct the raster # type of those files. # !!! This scripts requires check_vrt_completeness.py to be run beforehand !!! # Prep environment: # Dependencies import pandas import os import re import scandir import shutil import itertools import subprocess from osgeo import gdal from dklidar import settings # Function definitions def get_data_type(file_name): raster = gdal.Open(file_name) dataType = gdal.GetDataTypeName(raster.GetRasterBand(1).DataType) raster = None return(dataType) def translate_file(file_name, data_type): # Copy file to temp folder: temp_file = settings.scratch_folder + '/temp_raster.tif' shutil.copy(file_name, temp_file) # remove old file os.remove(file_name) # translate file os.system(settings.gdal_translate_bin + '-ot ' + data_type + ' ' + temp_file + ' ' + file_name) # Load missing tiles missing_files = pandas.read_csv(settings.log_folder + '/missing_files_in_vrts.csv') # Set DataTypes for non-int16 variables data_types_df = pandas.DataFrame( zip(*[ ['solar_radiation', 'amplitude_mean', 'amplitude_sd', 'date_stamp_min', 'date_stamp_max', 'date_stamp_mode'], ['Int32', 'Float32', 'Float32', 'Int32', 'Int32', 'Int32']]), columns = ['variable','data_type']) # determine output folder structure based on original processing folders = [] for folder in scandir.scandir(settings.output_folder): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # Clean up folder paths folders = map(lambda folder: re.sub('\\\\', '/', folder), folders) # Set up progres bar progress = 0 for i in range(0,len(missing_files.index)): # Grab folder name folder = list( itertools.compress( folders, [bool(re.search(missing_files.variable[i], folder)) for folder in folders])) folder = folder[0] # Grab data_type data_type = list( itertools.compress( data_types_df.data_type, [bool(re.search(missing_files.variable[i], variable)) for variable in data_types_df.variable])) data_type = data_type[0] # Set file path file_name = folder + '/' + missing_files.file_name[i] # Check wehether data types match if data_type == get_data_type(file_name): break # Copy to temp file temp_file = settings.scratch_folder + '/' + missing_files.file_name[i] shutil.copy(file_name, temp_file) # Break for debugging # file_name = settings.scratch_folder + '/test_out/' + missing_files.file_name[i] # Remove file from original folder os.remove(file_name) # Construct gdal command cmd = settings.gdal_translate_bin + '-ot ' + data_type + ' ' + temp_file + ' ' + file_name print(cmd) # Execute gdal commannd and swallow output os.system(cmd) # Remove temp_file os.remove(temp_file) # Update progress progress = float(i + 1) / float(len(missing_files.index)) # Update progress bar print('\n\r|' + '#' * int(round(progress * 54)) + '-' * int(round((1 - progress) * 54)) + '| ' + str(int(round(progress * 100))) + '%\n'),
jakobjassmann/ecodes-dk-lidar
documentation/source_data/merger_scripts/check_merged_outputs.py
## EcoDes-DK output merger ## This script is used to merge the outputs of the various EcoDes processing ## runs to create the merged dataset based on the DHM_201415 merger. ## <NAME> <EMAIL> # Dependencies import os import shutil import pandas import glob import re import scandir import hashlib # Status print('#' * 80) print('Check MD5 sums of merged EcoDes-DK outputs from the different reprocessing batches.') print('\nPreparing environment...'), ## 1) Set global variables # tile_ids to source tiles_to_source_original_processing = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_from_DHM2018.csv') tiles_to_source_reprocessing_1 = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_from_DHM2015.csv') tiles_to_source_reprocessing_2 = pandas.read_csv('D:/Jakob/dhm201415_merger/tiles_to_process_dhm201415_merger.csv') tiles_incomplete = pandas.read_csv('D:/Jakob/dhm201415_merger/incomplete_tile_pairs.csv') # Remove incomplete tiles from tile_ids tiles_to_source_original_processing = pandas.DataFrame( set(tiles_to_source_original_processing['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_1 = pandas.DataFrame( set(tiles_to_source_reprocessing_1['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_2 = pandas.DataFrame( set(tiles_to_source_reprocessing_2['tile_id'].tolist()) - set(tiles_incomplete['tile_id'].tolist()), columns = ['tile_id']) # Remove redundancies tiles_to_source_original_processing = pandas.DataFrame( set(tiles_to_source_original_processing['tile_id'].tolist()) - set(tiles_to_source_reprocessing_2['tile_id'].tolist()), columns = ['tile_id']) tiles_to_source_reprocessing_1 = pandas.DataFrame( set(tiles_to_source_reprocessing_1['tile_id'].tolist()) - set(tiles_to_source_reprocessing_2['tile_id'].tolist()), columns = ['tile_id']) # source folders folder_original_processing = 'D:/Jakob/dk_nationwide_lidar/data/outputs' folder_reprocessing_1 = 'D:/Jakob/ecodes-dk-lidar/data/outputs' folder_reprocessing_2 = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/outputs' # destination folder dest_folder = 'D:/Jakob/ecodes-dk-lidar-rev1/data/outputs' # base folders dtm_files = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/dtm' laz_files = 'D:/Jakob/ecodes-dk-lidar-reprocessing/data/laz' # determine output folder structure based on original processing folders = [] for folder in scandir.scandir(folder_original_processing): if folder.is_dir(): sub_folders = [sub_folder.path for sub_folder in scandir.scandir(folder.path) if sub_folder.is_dir()] if len(sub_folders) > 0: for sub_folder in sub_folders: folders.append(sub_folder) else: folders.append(folder.path) # remove variables that were / will be separately reprocessed (if present) folders = [folder for folder in folders if not bool(re.match('.*tile_footprints.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*solar_radiation.*', folder))] folders = [folder for folder in folders if not bool(re.match('.*date_stamp.*', folder))] # Clean up file paths folders = map(lambda folder: re.sub('\\\\', '/', folder), folders) # Keep only relative paths folders = map(lambda folder: '/' + os.path.relpath(folder, folder_original_processing), folders) ###!!! Break(s) for debugging !!! ##folders = [folders[1]] ##folders = ['/point_source_info/point_source_counts'] ## 2) Function definitons def list_files(folder_path): files = [] # Scan directory for file_name in scandir.scandir(folder_path): files.append(file_name.name) return(files) def get_tile_ids(file_names): # initiate empty list for tile_ids tile_ids = [] # clean up files names for i in range(0, len(file_names)): file_names[i] = re.sub('\\\\', '/', file_names[i]) # fill list with tile_id for file_name in file_names: tile_id = re.sub('.*(\d{4}_\d{3}).*', '\g<1>', file_name) tile_ids.append(tile_id) # combine to data frame files_df = pandas.DataFrame(zip(*[tile_ids, file_names]), columns = ['tile_id', 'file_name']) # return files_df return(files_df) def compare_tile_dfs(df1, df2): if(set(df1['tile_id'].tolist()) == set(df2['tile_id'].tolist())): return(pandas.DataFrame([], columns = ['tile_id'])) else: diff = pandas.DataFrame( set(df1['tile_id'].tolist()) - set(df2['tile_id'].tolist()), columns = ['tile_id']) return(diff) def check_var_from_source(var_folder, tiles_df, source_folder): # Get all files for one variable from one source # Set dest directory based on global variable global dest_folder out_folder = dest_folder + var_folder # Set source folder source_folder = source_folder + var_folder # Generate df of tile_ids and file names files_to_check = get_tile_ids(list_files(source_folder)) # check files check_df = check_tiles(tiles_df, files_to_check, source_folder, out_folder) return(check_df) def check_var(var_folder): # Get global variables global tiles_to_source_original_processing, tiles_to_source_reprocessing_1, tiles_to_source_reprocessing_2, folder_original_processing, folder_reprocessing_1, folder_reprocessing_2 # Get all files for one variable from the three data sources print('Checking: ' + var_folder) print('\tOriginal processing') original_check = check_var_from_source(var_folder, tiles_to_source_original_processing, folder_original_processing) print('\n\tReprocessing #1') reprocessing_1_check = check_var_from_source(var_folder, tiles_to_source_reprocessing_1, folder_reprocessing_1) print('\n\tReprocessing #2') reprocessing_2_check = check_var_from_source(var_folder, tiles_to_source_reprocessing_2, folder_reprocessing_2) # Return df of copied tiles and files_with_errors = pandas.concat([original_check, reprocessing_1_check, reprocessing_2_check]) files_with_errors['variable'] = var_folder print('\tDone.\n') return(files_with_errors) # The function below is thanks to stackoverflow usesrs quantumSoup and user2653663 # https://stackoverflow.com/questions/3431825/generating-an-md5-checksum-of-a-file def md5(fname): hash_md5 = hashlib.md5() with open(fname, "rb") as f: for chunk in iter(lambda: f.read(4096), b""): hash_md5.update(chunk) return hash_md5.hexdigest() def check_tiles(tiles_to_check, files_df, source_folder, out_folder): # Subset files to check files_to_check = files_df[files_df['tile_id']. isin(tiles_to_check['tile_id'].tolist())]['file_name'].tolist() # initate output lists files = [] status = [] # Set counter to 0: progress = 0 # Check files for i in range(0, len(files_to_check)): # Check whether files exists if os.path.isfile(source_folder + '/' + files_to_check[i]): if os.path.isfile(out_folder + '/' + files_to_check[i]): match = md5(source_folder + '/' + files_to_check[i]) == md5(out_folder + '/' + files_to_check[i]) if not match: files.append(files_to_check[i]) status.append('md5_mismatch') else: files.append(files_to_check[i]) status.append('out_file_missing') else: files.append(files_to_check[i]) status.append('source_file_missing') # Update progress progress = float(i + 1) / float(len(files_to_check)) # Update progress bar print('\r\t|' + '-' * int(round(progress * 54)) + ' ' * int(round((1 - progress) * 54)) + '| ' + str(int(round(progress * 100))) + '%'), # Compile outputs to dataframe and retun status_df = pandas.DataFrame(zip(*[files, status]), columns = ['file_name','status']) return(status_df) # Status print(' done.\n') ## 3) Main body of script ## Check completeness of tile lists # Status print('Checking completness of tile list to check...'), # Get all tiles in data set dhm_merged_tiles = get_tile_ids(list_files(laz_files)) # Compare with tiles to merge missing_tiles = len(compare_tile_dfs(dhm_merged_tiles, pandas.concat([tiles_to_source_original_processing, tiles_to_source_reprocessing_1, tiles_to_source_reprocessing_2]))) if missing_tiles > 0: # Prompt for choice to stop del_choice = raw_input('\n' + str(missing_tiles) + ' are missing! Continue anyways?' + '[y/n]') if not del_choice == 'y': print('Aborting on request or invalid choice!') quit() else: print('Okay, continuing merger.') else: print('\n=> Sets are complete, proceeding as planned.\n') ## Check tiles for all variables # Status print('Starting check:\n\n') files_checked_dfs = [] for var_folder in folders: files_copied_df = check_var(var_folder) files_checked_dfs.append(files_copied_df) files_checked_df = pandas.concat(files_checked_dfs) files_checked_df.to_csv('checkusm_errors.csv', index = False) # Stauts print('Check complete!\n') print('#' * 80 + '\n') # EOF
jakobjassmann/ecodes-dk-lidar
scripts/generate_list_of_vrts.py
# Short script to generate a list of all vrt files import glob import re from dklidar import settings # list vrts vrts = glob.glob(settings.output_folder + '*/*.vrt') more_vrts = glob.glob(settings.output_folder + '*/*/*.vrt') for vrt in more_vrts: vrts.append(vrt) # Clean up file paths vrts = [re.sub(settings.output_folder[0:len(settings.output_folder)-1], '', vrt) for vrt in vrts] vrts = [re.sub('\\\\', '/', vrt) for vrt in vrts] vrts = [re.sub('^/', '', vrt) for vrt in vrts] # write out to file out_file = open(settings.output_folder + '/list_of_vrts.txt', 'w') out_file.write('\n'.join(vrts)) out_file.close()
jakobjassmann/ecodes-dk-lidar
dklidar/points.py
<reponame>jakobjassmann/ecodes-dk-lidar ### Functions for point cloud handling for the DK Lidar project ### <NAME> <EMAIL> 29 January 2019 ## Imports import re import os import opals import subprocess import numpy import glob import shutil from osgeo import gdal_array from datetime import datetime, timedelta from dklidar import common from dklidar import settings ##### Function definitions ## Import a single tile into ODM def odm_import_single_tile(tile_id): """ Imports a single tile (specified by tile_id) into an ODM for subsequent processing. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: returns execution status """ # Initiate return value return_value = '' # Generate relevant file names: laz_file = settings.laz_folder + '/PUNKTSKY_1km_' + tile_id + '.laz' odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' # Try import try: # Import tile id import_tile = opals.Import.Import() import_tile.inFile = laz_file import_tile.outFile = odm_file import_tile.commons.screenLogLevel = opals.Types.LogLevel.none import_tile.run() return_value = 'success' except: return_value = 'opalsError' # return execution status return return_value ## Load neighbourhood of tiles into ODM (this is currently not neede by any of the funcitons below) def odm_import_mosaic(tile_id): """ Imports a tile (specified by tile_id) and it's 3 x 3 neighbourhood into a mosaiced ODM file for subsequent processing. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: returns execution status """ # Initiate return value, open log file return_value = '' log_file = open('log.txt', 'a+') # Retrieve row and col numbers for the current tile_id center_row = int(re.sub('(\d+)_\d+', '\g<1>', tile_id)) center_col = int(re.sub('\d+_(\d+)', '\g<1>', tile_id)) # Determine row and column numbers for tiles in the 3 x 3 window rows_to_load = [center_row - 1, center_row, center_row + 1] cols_to_load = [center_col - 1, center_col, center_col + 1] # Generate list of tile_ids for tiles to load tiles_to_load = [] for row in rows_to_load: for col in cols_to_load: tile_to_load = str(row) + '_' + str(col) tiles_to_load.extend([tile_to_load]) # Prep filenames and check if files exists: tile_file_names = [] for tile_to_load in tiles_to_load: tile_file_name = settings.laz_folder + '/PUNKTSKY_1km_' + tile_to_load + '.laz' if os.path.exists(tile_file_name): tile_file_names.append(tile_file_name) n_neighbours = len(tile_file_names) # Update log output depending of the number of valid neighbours if n_neighbours == 9: log_file.write(' importing point clouds into ODM mosaic...\n' + 'Number of neighbours = ' + str(n_neighbours) + '. Complete!\n') else: log_file.write(tile_id + ' importing point clouds into ODM mosaic...\n' + 'Warning! Number of neighbours = ' + str(n_neighbours) + '. Incomplete. Edge effects possible!\n') # Generate output file name string odm_file = settings.odm_mosaics_folder + '/odm_mosaic_' + tile_id + '.odm' # Execute command as subprocess and return message: try: # Import tiles into odm. import_tile = opals.Import.Import() import_tile.inFile = tile_file_names import_tile.outFile = odm_file import_tile.commons.screenLogLevel = opals.Types.LogLevel.none import_tile.run() log_file.write(tile_id + ' success.\n\n') return_value = return_value + 'success' if n_neighbours != 9: return_value = 'Warning: Incomplete Neighbourhood!' except: return_value = 'opalsError' log_file.write(tile_id + ' failed. OpalsError.\n\n') # Write log output to log file log_file.close() # return status output return return_value ## Def: Export tile footprint def odm_generate_footprint(tile_id): """ Exports footprint from an odm file based on the tile_id in the DK nationwide dataset :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: returns execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Generate relevant file names: odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' temp_tif_file = os.getcwd() + '/temp_' + tile_id + '.tif' footprint_file = settings.odm_footprint_folder + '/footprint_' + tile_id + '.shp' # Try token raster export to temp tif try: # Export temporary tif export_tif = opals.Cell.Cell() export_tif.inFile = odm_file export_tif.outFile = temp_tif_file export_tif.feature = 'min' export_tif.cellSize = 10 # This is also the default cell size, so technically not needed. export_tif.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_tif.commons.screenLogLevel = opals.Types.LogLevel.none export_tif.run() log_file.write('\n' + tile_id + ' temporary raster export successful.\n\n') except: return_value = 'opalsError' log_file.write('\n' + tile_id + ' temporary raster export failed.\n\n') # Try generating footprint from temp tif try: # Specify gdal command cmd = settings.gdaltlindex_bin + ' ' + footprint_file + ' ' + temp_tif_file # Execute gdal command log_file.write('\n' + tile_id + ' footprint generation... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ tile_id + ' successful.\n\n') # set exit status return_value = 'success' except: log_file.write('\n' + tile_id + ' footprint generation... \n' + tile_id + ' failed.\n\n') if return_value == 'opalsError': pass else: return_value = 'gdalError' # Close log file log_file.close() # Remove temp raster file try: os.remove(temp_tif_file) except: pass # return status output return return_value ## Def: Validiate CRS def odm_validate_crs(tile_id, mosaic = False): """ Function to validate the crs for odm files (single tile and mosaic) :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param mosaic: if True validates crs for mosaic also, default: False :return: execution status """ # Initiate return value return_value = '' # Generate odm files path names odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' odm_mosaic = settings.odm_mosaics_folder + '/odm_mosaic_' + tile_id + '.odm' # Retrieve CRS string for single tile try: odm_dm = opals.pyDM.Datamanager.load(odm_file) crs_str = odm_dm.getCRS() # Check whether CRS exists, if not assign, if different throw error. if crs_str == settings.crs_wkt_opals: return_value = 'Tile: match' elif crs_str == '': odm_dm.setCRS(settings.crs_wkt_opals) return_value = 'Tile: empty - set' else: return_value = 'Tile: warning - no match' odm_dm = None # This is needed as opals locks the file connection otherwise. except: return_value = 'Single: error; ' # Retrieve CRS string for mosaic if mosaic == True: try: odm_dm = opals.pyDM.Datamanager.load(odm_mosaic) crs_str = odm_dm.getCRS() # Check whether CRS exists, if not assign, if different throw error. if crs_str == settings.crs_wkt_opals: return_value = return_value + '; Mosaic: match' elif crs_str == '': odm_dm.setCRS(settings.crs_wkt_opals) return_value = return_value + '; Mosaic: empty - set' else: return_value = return_value + '; Mosaic: warning - no match' odm_dm = None # This is needed as opals locks the file connection otherwise. except: return_value = return_value + 'Mosaic: error;' return return_value ## Add height above ground (normalized z) to a tile odm def odm_add_normalized_z(tile_id, mosaic = False): """ Adds a "normalizedZ' variable to each point in an ODM file by normalising the height using the 0.4 m DTM. Can deal with either single tile odms or neighbourhood mosaics (option mosaic). If a mosaic is normalised, then the corresponding dtm mosaic will have to be generated first (use dtm_neighbourhood_mosaic()). :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param mosaic: boolean (true or false) specifies whether a single tile pointcloud or a neighbourhood mosaic should be normalised :return: execution status """ # Initiate return value return_value = '' # Generate file paths if(mosaic == False): odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' dtm_file = settings.dtm_folder + '/DTM_1km_' + tile_id + '.tif' else: odm_file = settings.odm_mosaics_folder + '/odm_mosaic_' + tile_id + '.odm' dtm_file = settings.dtm_folder + '/dtm_' + tile_id + '_mosaic.tif' # Normalise the point cloud data try: add_normalized_z = opals.AddInfo.AddInfo() add_normalized_z.inFile = odm_file add_normalized_z.gridFile = dtm_file add_normalized_z.attribute = 'normalizedZ = z - r[0]' add_normalized_z.commons.screenLogLevel = opals.Types.LogLevel.none add_normalized_z.commons.nbThreads = settings.nbThreads add_normalized_z.run() return_value = 'success' except: return_value = 'opalsError' # Return exist status return return_value ## Export mean and sd of height above ground for all 10 m cells in a tile def odm_export_normalized_z(tile_id): """ Exports mean and standard deviation of the normalisedZ variable for the 10 m x 10 m raster grid. This function requires the normalizedZ to be added first (use odm_add_normalized_z()). :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Set file and folder paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' temp_file_mean = os.getcwd() + '/temp_' + tile_id + '_mean.tif' temp_file_sd = os.getcwd() + '/temp_' + tile_id + '_sd.tif' out_folder = settings.output_folder + '/normalized_z' out_file_mean = out_folder + '/normalized_z_mean/normalized_z_mean_' + tile_id + '.tif' out_file_sd = out_folder + '/normalized_z_sd/normalized_z_sd_' + tile_id + '.tif' # Create folders if they do not already exists if not os.path.exists(out_folder): os.mkdir(out_folder) if not os.path.exists(out_folder + '/normalized_z_mean'): os.mkdir(out_folder + '/normalized_z_mean') if not os.path.exists(out_folder + '/normalized_z_sd'): os.mkdir(out_folder + '/normalized_z_sd') # Export normalized z raster mean and sd try: # Initialise exporter export_normalized_z = opals.Cell.Cell() # Export mean export_normalized_z.inFile = odm_file export_normalized_z.outFile = temp_file_mean export_normalized_z.attribute = 'normalizedZ' export_normalized_z.feature = 'mean' export_normalized_z.cellSize = settings.out_cell_size export_normalized_z.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_normalized_z.filter = settings.all_classes export_normalized_z.commons.screenLogLevel = opals.Types.LogLevel.none export_normalized_z.commons.nbThreads = settings.nbThreads export_normalized_z.run() # Reset exporter export_normalized_z.reset() # Export sd export_normalized_z = opals.Cell.Cell() export_normalized_z.inFile = odm_file export_normalized_z.outFile = temp_file_sd export_normalized_z.attribute = 'normalizedZ' export_normalized_z.feature = 'stdDev' export_normalized_z.cellSize = settings.out_cell_size export_normalized_z.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_normalized_z.filter = settings.all_classes export_normalized_z.commons.screenLogLevel = opals.Types.LogLevel.none export_normalized_z.commons.nbThreads = settings.nbThreads export_normalized_z.run() return_value = 'success' except: return_value = 'opalsError' # Stretch and convert to 16 bit integer try: # Construct gdal command for mean cmd = settings.gdal_calc_bin + \ '-A ' + temp_file_mean + ' ' + \ '--outfile=' + out_file_mean + ' ' + \ '--calc=rint(A*100) ' + \ '--type=Int16 --NoDataValue=-9999 ' # Execute and log command log_file.write('\n' + tile_id + ' rounding mean to int16 and calculation success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file_mean) # Construct gdal command for sd cmd = settings.gdal_calc_bin + \ '-A ' + temp_file_sd + ' ' + \ '--outfile=' + out_file_sd + ' ' + \ '--calc=rint(A*100) ' + \ '--type=Int16 --NoDataValue=-9999 ' # Execute and log command log_file.write('\n' + tile_id + ' rounding sd to int16 and calculation success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply masks common.apply_mask(out_file_sd) return_value = 'success' except: if return_value == 'opalsError': pass else: return_value = 'gdalError' log_file.write('\n' + tile_id + ' normalized_z export failed. \n') # Tidy up try: os.remove(temp_file_mean) os.remove(temp_file_sd) except: pass # Close log file log_file.close() # Return exist status return return_value ## Export canopy height for all 10 m cells in a tile def odm_export_canopy_height(tile_id): """ Exports the canopy height (95 percentile of normalised height only for points classified as vegetation) for the 10 m x 10 m raster grid. This function requires the normalizedZ to be added first (use odm_add_normalized_z()). :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Generate file paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' temp_file1 = os.getcwd() + '/' + tile_id + '_temp1.tif' temp_file2 = os.getcwd() + '/' + tile_id + '_temp2.tif' out_folder = settings.output_folder + '/canopy_height' out_file = out_folder + '/canopy_height_' + tile_id + '.tif' # Create folder if it does not exist if not os.path.exists(out_folder): os.mkdir(out_folder) # Export canopy height try: # Initialise exporter export_canopy_height = opals.Cell.Cell() # Export mean export_canopy_height.inFile = odm_file export_canopy_height.outFile = temp_file1 export_canopy_height.attribute = 'normalizedZ' export_canopy_height.feature = 'quantile:0.95' # Apply extraction only to points classified as vegetation: export_canopy_height.filter = settings.veg_classes_filter # Set no data value to zero. Later the no data value will be set to -9999 this will assure that there are no # holes in the dataset for pixels where there are no points classified as vegetation. export_canopy_height.noData = 0 export_canopy_height.cellSize = settings.out_cell_size export_canopy_height.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_canopy_height.commons.screenLogLevel = opals.Types.LogLevel.none export_canopy_height.commons.nbThreads = settings.nbThreads export_canopy_height.run() return_value = 'success' except: return_value = 'opalsError' # Stretch by 100, round to Int16 and set Nodata value to -9999 try: # Construct gdal command to set no data value (this is done first to keep no data point counts as 0) cmd = settings.gdal_translate_bin + ' -a_nodata -9999 ' + temp_file1 + ' ' + temp_file2 # Execute gdal commant and add to log output log_file.write('\n' + tile_id + ' setting no data value... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ tile_id + ' successful.\n\n') # Construct gdal command to stredtch and round to int 16 cmd = settings.gdal_calc_bin + \ '-A ' + temp_file2 + ' ' + \ '--outfile=' + out_file + ' ' + \ '--calc=rint(A*100) ' + \ '--type=Int16 ' + \ '--NoDataValue=-9999' # Execute command and log log_file.write('\n' + tile_id + ' stretching and rounding success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file) # set exit status return_value = 'success' except: log_file.write(tile_id + ' setting no data value for canopy height failed.\n\n') if return_value == 'opalsError': pass else: return_value = 'gdalError' # Close log file log_file.close() # Remove temp raster file try: os.remove(temp_file1) os.remove(temp_file2) except: pass # Return exist status return return_value ## Export a point count for a specific height range and set of classes for all 10 m cells in a tile def odm_export_point_count(tile_id, name = 'vegetation_point_count', lower_limit = -1, upper_limit = 50.0, point_classes = None): """ Exports point counts for each 10 m x 10 m cell in an ODM and a given height interval specified by the lower and upper limit parameters, as well as a given set of point classes specified by the pint_classes parameter. Height parameters are given as normalised height above ground. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param name: identifier name for the point count - used in file and folder naming during export :param lower_limit: lower limit for the height interval (normalised height in m) :param upper_limit: upper limit for the height interval (normalised height in m) :param point_classes: classes to subset from :return: execution status """ # Initiate return value and log_output return_value = '' log_file = open('log.txt', 'a+') # Get temporary working directory wd = os.getcwd() # Initiate point_classes default value if no value is provided: if point_classes is None: point_classes = [3,4,5] # Veg class points # Generate paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' out_folder = settings.output_folder + '/point_count' if lower_limit < 10 and lower_limit >= 0: lower_limit_str = '0' + str(lower_limit) elif lower_limit == -1: lower_limit_str = '-01' else: lower_limit_str = str(lower_limit) if upper_limit < 10 and upper_limit >= 0: upper_limit_str = '0' + str(upper_limit) else: upper_limit_str = str(upper_limit) prefix = name + '_' + lower_limit_str + 'm-' + upper_limit_str + 'm' temp_file = wd + '/temp_' + tile_id + '.tif' out_file = out_folder + '/' + prefix + '/' + prefix + '_' + tile_id + '.tif' # Create folders if they don't exist if not os.path.exists(out_folder): os.mkdir(out_folder) if not os.path.exists(out_folder + '/' + prefix): os.mkdir(out_folder + '/' + prefix) # Export point count try: # Initialise exporter export_point_count = opals.Cell.Cell() # Specificy filter strings: height_filter = 'generic[NormalizedZ >= ' + str(lower_limit) + ' and NormalizedZ < ' + str(upper_limit) + ']' class_filter = 'Generic[Classification == ' + \ ' OR Classification == '.join([str(point_class) for point_class in point_classes]) + ']' # Export point count export_point_count.inFile = odm_file export_point_count.outFile = temp_file export_point_count.filter = height_filter + ' AND ' + class_filter export_point_count.feature = 'pcount' export_point_count.cellSize = settings.out_cell_size export_point_count.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_point_count.noData = 0 export_point_count.commons.screenLogLevel = opals.Types.LogLevel.none export_point_count.commons.nbThreads = settings.nbThreads export_point_count.run() return_value = 'success' except: return_value = 'opalsError' # Convert to 16 bit integer and set no data value to -9999 try: # Construct gdal command cmd = settings.gdal_translate_bin + \ '-ot Int16 -a_nodata -9999 ' + \ temp_file + ' ' + \ out_file # Execute and log command log_file.write('\n' + tile_id + ' converting to Int16 success. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file) return_value = 'success' except: if return_value == 'opalsError': pass else: log_file.write('\n' + tile_id + ' converting to Int16 for ' + prefix + ' failed. \n') return_value = 'gdalError' # Tidy up try: os.remove(temp_file) except: pass # Close log file log_file.close() # Return exist status return return_value ## Export point counts for a pre-defined set of height ranges and classes def odm_export_point_counts(tile_id): """ Exports point counts for multiple pre-defined classes and height intervals by calling the odm_export_point_count() function. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate empty list for return values return_values = [] ## Ground point count return_values.append(odm_export_point_count(tile_id, 'ground_point_count', -1, 1, [2])) ## Water point count return_values.append(odm_export_point_count(tile_id, 'water_point_count', -1, 1, [9])) ## Ground and water point count return_values.append(odm_export_point_count(tile_id, 'ground_and_water_point_count', -1, 1, [2,9])) ## Vegetation point count return_values.append(odm_export_point_count(tile_id, 'vegetation_point_count', 0, 50, [3,4,5])) ## Building point counts return_values.append(odm_export_point_count(tile_id, 'building_point_count', -1, 50, [6])) ## All classes return_values.append(odm_export_point_count(tile_id, 'total_point_count', -1, 50, [2,3,4,5,6,9])) ## Vegetation point counts for continous height bins # 0-2 m at 0.5 m intervals for lower in numpy.arange(0, 2.0, 0.5): return_values.append(odm_export_point_count(tile_id, 'vegetation_point_count', lower, lower + 0.5, [3,4,5])) # 2-20 m at 1 m intervals for lower in range(2, 20, 1): return_values.append(odm_export_point_count(tile_id, 'vegetation_point_count', lower, lower + 1, [3,4,5])) # 20-25 m at 5 m interval return_values.append(odm_export_point_count(tile_id, 'vegetation_point_count', 20, 25, [3,4,5])) # 25 m to 50 m return_values.append(odm_export_point_count(tile_id, 'vegetation_point_count', 25, 50, [3,4,5])) # Set return value status # There are only two return value states so if there is more than one return value in the list # one of them has to be an opalsError return_values = set(return_values) if len(return_values) > 1: return_value = "opalsError" else: return_value = list(return_values)[0] return return_value ## Calculate proportions based on two point counts def odm_calc_proportions(tile_id, prop_name, point_count_id1, point_count_id2): """ Function to calculate point count proportions for two point counts. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :param prop_name: name to be assinged to the proportions output :param point_count_id1: name of point count to be rationed (numerator) :param point_count_id2: name of point count to be rationed to (denominator) :return: execution status """ return_value = '' log_file = open('log.txt', 'a+') # Generate paths for numerator and denominator num_file = settings.output_folder + '/point_count/' + point_count_id1 + '/' + point_count_id1 + '_' + tile_id + '.tif' den_file = settings.output_folder + '/point_count/' + point_count_id2 + '/' + point_count_id2 + '_' + tile_id + '.tif' out_folder = settings.output_folder + '/proportions' out_file = out_folder + '/' + prop_name + '/' + prop_name + '_' + tile_id + '.tif' # Create folders if they do not exist if not os.path.exists(out_folder): os.mkdir(out_folder) if not os.path.exists(out_folder + '/' + prop_name): os.mkdir(out_folder + '/' + prop_name) # get wd temp_wd = os.getcwd() # specify temp file path temp_file = temp_wd + '/temp.tif' # Attempt calculating the proportions using gdal_calc try: # Construct gdal command nb. needed to use true_divide here. A cast into int16 will have to follow separately cmd = settings.gdal_calc_bin + \ '-A ' + num_file + ' ' +\ '-B ' + den_file + ' ' +\ '--outfile=' + temp_file + ' ' + \ '--type=Int16 ' +\ '--calc=rint(10000*true_divide(A,B)) ' + \ '--NoDataValue=-9999' log_file.write(cmd) # Execute gdal command log_file.write('\n' + tile_id + ' calculated proportions ' + prop_name + '... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Round and convert to int16 cmd = settings.gdal_translate_bin + \ '-ot Int16 ' + \ '-a_nodata -9999 ' +\ temp_file + ' ' +\ out_file + ' ' log_file.write(cmd) # Execute gdal command log_file.write('\n' + tile_id + ' calculated proportions ' + prop_name + '... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) os.remove(temp_file) # Apply mask(s) common.apply_mask(out_file) return_value = 'success' except: log_file.write('\n' + tile_id + ' calculation of proportions ' + prop_name + ' failed. gdalError \n') return_value = 'gdalError' # Close log file log_file.close() return return_value ## Export a pre-defined list of proportions for a tile def odm_export_proportions(tile_id): """ Exports proportions for: canopy openness, canopy height profile, buildings point counts :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: exit status """ # Initiate return values return_values = [] ## Export canopy openness return_values.append(odm_calc_proportions(tile_id, 'canopy_openness', 'ground_and_water_point_count_-01m-01m', 'total_point_count_-01m-50m')) ## Export vegeation density return_values.append(odm_calc_proportions(tile_id, 'vegetation_density', 'vegetation_point_count_00m-50m', 'total_point_count_-01m-50m')) ## Export canopy height profile # 0-2 m at 0.5 m intervals for lower in numpy.arange(0, 2, 0.5): veg_height_bin = 'vegetation_point_count_0' + str(lower) + 'm-0' + str(lower + 0.5) + 'm' prop_variable_bin = 'vegetation_proportion_0' + str(lower) + 'm-0' + str(lower + 0.5) + 'm' return_values.append(odm_calc_proportions(tile_id, prop_variable_bin, veg_height_bin, 'vegetation_point_count_00m-50m')) # 2-9 m at 1 m intervals for lower in range(2, 9, 1): veg_height_bin = 'vegetation_point_count_0' + str(lower) + 'm-0' + str(lower + 1) + 'm' prop_variable_bin = 'vegetation_proportion_0' + str(lower) + 'm-0' + str(lower + 1) + 'm' return_values.append(odm_calc_proportions(tile_id, prop_variable_bin, veg_height_bin, 'vegetation_point_count_00m-50m')) # 9-10 m return_values.append(odm_calc_proportions(tile_id, 'vegetation_proportion_09m-10m', 'vegetation_point_count_09m-10m', 'vegetation_point_count_00m-50m')) # 10-20 m at 1 m intervals for lower in range(10, 20, 1): veg_height_bin = 'vegetation_point_count_' + str(lower) + 'm-' + str(lower + 1) + 'm' prop_variable_bin = 'vegetation_proportion_' + str(lower) + 'm-' + str(lower + 1) + 'm' return_values.append(odm_calc_proportions(tile_id, prop_variable_bin, veg_height_bin, 'vegetation_point_count_00m-50m')) # 20-25 m return_values.append(odm_calc_proportions(tile_id, 'vegetation_proportion_20m-25m', 'vegetation_point_count_20m-25m', 'vegetation_point_count_00m-50m')) # 25-50 m return_values.append(odm_calc_proportions(tile_id, 'vegetation_proportion_25m-50m', 'vegetation_point_count_25m-50m', 'vegetation_point_count_00m-50m')) # Export building proportion return_values.append(odm_calc_proportions(tile_id, 'building_proportion', 'building_point_count_-01m-50m', 'total_point_count_-01m-50m')) # Set return value status # There are only two return value states so if there is more than one return value in the list # one of them has to be an opalsError return_values = set(return_values) if len(return_values) > 1: return_value = "gdalError" else: return_value = list(return_values)[0] return return_value ## Export mean and sd in the amplitude variable for all 10 m cells in a tile def odm_export_amplitude(tile_id): """ Exports mean and sd for the lidar amplitude for all 10 m x 10 m cells in a tile. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' # Generate file paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' out_folder = settings.output_folder + '/amplitude' out_file_mean = out_folder + '/amplitude_mean/amplitude_mean_' + tile_id + '.tif' out_file_sd = out_folder + '/amplitude_sd/amplitude_sd_' + tile_id + '.tif' # Create folders if they do not exist if not os.path.exists(out_folder): os.mkdir(out_folder) if not os.path.exists(out_folder + '/amplitude_mean'): os.mkdir(out_folder + '/amplitude_mean') if not os.path.exists(out_folder + '/amplitude_sd'): os.mkdir(out_folder + '/amplitude_sd') # Export amplitude mean and sd using OPALS Cell try: # Initialise exporter export_amplitude = opals.Cell.Cell() # Export mean export_amplitude.inFile = odm_file export_amplitude.outFile = out_file_mean export_amplitude.attribute = 'amplitude' export_amplitude.feature = 'mean' export_amplitude.cellSize = settings.out_cell_size export_amplitude.filter = settings.all_classes # all classes (2,3,4,5,6,9) export_amplitude.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_amplitude.commons.screenLogLevel = opals.Types.LogLevel.none export_amplitude.commons.nbThreads = settings.nbThreads export_amplitude.run() # Reset exporter export_amplitude.reset() # Export sd export_amplitude = opals.Cell.Cell() export_amplitude.inFile = odm_file export_amplitude.outFile = out_file_sd export_amplitude.attribute = 'amplitude' export_amplitude.feature = 'stdDev' export_amplitude.cellSize = settings.out_cell_size export_amplitude.filter = settings.all_classes # all classes (2,3,4,5,6,9) export_amplitude.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_amplitude.commons.screenLogLevel = opals.Types.LogLevel.none export_amplitude.commons.nbThreads = settings.nbThreads export_amplitude.run() # Apply mask(s) common.apply_mask(out_file_mean) common.apply_mask(out_file_sd) return_value = 'success' except: return_value = 'opalsError' # Return exist status return return_value ## Export flight strip information for all 10 m cells in a tile def odm_export_point_source_info(tile_id): """ Extracts point source statistics for the 10 m x 10 m cells of the point cloud. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initate return value return_value = '' # Initiate log output log_file = open('log.txt', 'a+') # get current work dir string temp_wd = os.getcwd() # Set file paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' out_folder = settings.output_folder + '/point_source_info' out_folder_ids = out_folder + '/point_source_ids' out_folder_nids = out_folder + '/point_source_nids' out_folder_counts = out_folder + '/point_source_counts' out_folder_prop = out_folder + '/point_source_proportion' # Create folders if they do not exist if not os.path.exists(out_folder): os.mkdir(out_folder) if not os.path.exists(out_folder_ids): os.mkdir(out_folder_ids) if not os.path.exists(out_folder_nids): os.mkdir(out_folder_nids) if not os.path.exists(out_folder_counts): os.mkdir(out_folder_counts) if not os.path.exists(out_folder_prop): os.mkdir(out_folder_prop) ## Look up unique point source ids found in odm file try: # Open odm in python DM dm = opals.pyDM.Datamanager.load(odm_file) # Create layout and add 'PointSourceID' column lf = opals.pyDM.AddInfoLayoutFactory() lf.addColumn(dm, 'PointSourceId', True) layout = lf.getLayout() # Get set of histograms for layout # (this will only have one item, the histogram for the point source id column) histograms_set = dm.getHistogramSet(layout) # Initate empty list of point source ids point_source_ids = [] # Load unique point source ids from histogram # Note: The histogram set does not allow subsetting, so we loop through it. # As it will only have one object, this does not matter for histo in histograms_set.histograms(): for value, count in histo.values(): point_source_ids.append(value) # Remove dm object and close connection to odm file for later use del dm ## Use opals cell to extract point counts for each point source id for point_source_id in point_source_ids: # Initate opals cell module export_point_count = opals.Cell.Cell() # Initate filter string point_classes = [2, 3, 4, 5, 6, 9] class_filter = 'Generic[Classification == ' + \ ' OR Classification == '.join([str(point_class) for point_class in point_classes]) + ']' # Export point count export_point_count.inFile = odm_file export_point_count.outFile = 'temp_count_' + str(point_source_id) + '.tif' export_point_count.filter = class_filter + \ ' AND Generic[PointSourceId == ' + str(point_source_id) + ']' export_point_count.feature = 'pcount' export_point_count.cellSize = settings.out_cell_size export_point_count.limit = 'corner' # This switch is really important when working with tiles! # It sets the ROI to the extent to the bounding box of points in the ODM export_point_count.noData = 0 export_point_count.commons.screenLogLevel = opals.Types.LogLevel.none export_point_count.commons.nbThreads = settings.nbThreads export_point_count.run() export_point_count.reset() ## Convert to int16 and set no data to -9999 and apply mask for point_source_id in point_source_ids: cmd = settings.gdal_translate_bin + '-ot Int16 -a_nodata -9999 ' + \ temp_wd + '/temp_count_' + str(point_source_id) + '.tif ' + \ out_folder_counts + '/point_source_counts_' + tile_id + '_' + str(point_source_id) + '.tif ' log_file.write('\n' + tile_id + ' ' + str(point_source_id) + ' converted point source file to int16. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) common.apply_mask(out_folder_counts + '/point_source_counts_' + tile_id + '_' + str(point_source_id) + '.tif ') ## Determine the number of uniuqe point source ids per cell using gdal_calc. # Prepare in file string and equation string alphabet = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z'] files_string = '' equation = [] for i in range(len(point_source_ids)): files_string = files_string + ' -' + alphabet[i] + ' ' + out_folder_counts + \ '/point_source_counts_' + tile_id + '_' + str(point_source_ids[i]) + '.tif ' equation.append(alphabet[i]) equation = '1*greater(' + ',0)+1*greater('.join(equation) + ',0)' # Construct gdal command cmd = settings.gdal_calc_bin + files_string + \ '--outfile=' + out_folder_nids + '/point_source_nids_' + tile_id + '.tif' + \ ' --calc=' + equation + \ ' --type=Int16 --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' extracted number of unique point source ids. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask('--outfile=' + out_folder_nids + '/point_source_nids_' + tile_id + '.tif') ## Calculate proportion of hits pre cell per point source using gdal_calc # Calculate total sum of points per cell, prepare gdal command alphabet = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z'] files_string = '' equation = [] for i in range(len(point_source_ids)): files_string = files_string + ' -' + alphabet[i] + ' ' + out_folder_counts + \ '/point_source_counts_' + tile_id + '_' + str(point_source_ids[i]) + '.tif ' equation.append(alphabet[i]) equation = '+'.join(equation) # Construct gdal command cmd = settings.gdal_calc_bin + files_string + '--outfile=' + temp_wd + '/temp_total_points.tif ' + \ '--calc=' + equation + ' --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' created temporary total point count file. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) ## Calculate proportions using gdal_calc, round, stretch by 10000 for point_source_id in point_source_ids: # Construct gdal command cmd = settings.gdal_calc_bin + \ '-A ' + out_folder_counts + '/point_source_counts_' + tile_id + '_' + str(point_source_id) + '.tif ' + \ '-B ' + temp_wd + '/temp_total_points.tif ' + \ '--outfile=' + temp_wd + '/point_source_prop_' + str(point_source_id) + '.tif ' + \ '--calc=rint(true_divide(A,B)*10000) ' + '--type=Float32 --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' calculated proportions for ' + str(point_source_id) + '. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) ## Convert proportions to int16 for point_source_id in point_source_ids: cmd = settings.gdal_translate_bin + '-a_nodata -9999 -ot Int16 ' + \ temp_wd + '/point_source_prop_' + str(point_source_id) + '.tif ' + \ out_folder_prop + '/point_source_prop_' + tile_id + '_' + str(point_source_id) + '.tif ' log_file.write('\n' + tile_id + ' ' + str(point_source_id) + ' converted proportion to int16. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_folder_prop + '/point_source_prop_' + tile_id + '_' + str(point_source_id) + '.tif') ## Create a layer with presence / absence of point source id indicated by the point source id itself for point_source_id in point_source_ids: # Construct gdal command cmd = settings.gdal_calc_bin + \ '-A ' + out_folder_counts + '/point_source_counts_' + tile_id + '_' + str(point_source_id) + '.tif ' + \ '--outfile=' + temp_wd + '/temp_presence_' + str(point_source_id) + '.tif ' + \ '--calc=' + str(point_source_id) + '*greater(A,0)' + \ ' --type=Int32 --NoDataValue=-9999' # Execute gdal command log_file.write('\n' + tile_id + ' created temporary presence layer for ' + \ str(point_source_id) + '. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) ## Merge files into one using gdal_merge # Prepare gdal command in_files_string = '.tif ' + temp_wd + '/temp_presence_' in_files_string = temp_wd + '/temp_presence_' + in_files_string.join( [str(i) for i in point_source_ids]) + '.tif' # Construct gdal command cmd = settings.gdal_merge_bin + '-a_nodata -9999 -separate ' + \ '-o ' + out_folder_ids + '/point_source_ids_' + tile_id + '.tif ' + \ in_files_string # Execute gdal command log_file.write('\n' + tile_id + ' merged temporary layers in point source ids file. \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_folder_ids + '/point_source_ids_' + tile_id + '.tif') # The 'majority' stat produced by opals is not reliable... I'm leaving the below code for leagcy reasons. # 'majority statistics will have to be calculate from the above generated rasters by hand. # ## Extract mode of point count ids # # Initate opals cell module # export_point_mode = opals.Cell.Cell() # # # Export point count # export_point_mode.inFile = odm_file # export_point_mode.outFile = out_folder_mode + '/point_source_mode_' + tile_id + '.tif' # export_point_mode.filter = settings.ground_and_veg_classes_filter # export_point_mode.attribute = 'PointSourceId' # export_point_mode.feature = 'majority' # export_point_mode.cellSize = settings.out_cell_size # export_point_mode.limit = 'corner' # This switch is really important when working with tiles! # # It sets the ROI to the extent to the bounding box of points in the ODM # export_point_mode.commons.screenLogLevel = opals.Types.LogLevel.none # export_point_mode.commons.nbThreads = settings.nbThreads # export_point_mode.run() return_value = 'success' except: return_value = 'opalsError' # Close log file log_file.close() # remove temporary files for temp_file in glob.glob(temp_wd + '/*.tif'): try: os.remove(temp_file) except: pass return return_value ## Export mean and sd of height above ground for all 10 m cells in a tile def odm_export_date_stamp(tile_id): """ Export estimated most common date stamp for each 10 m x 10 m cells. This is done by converting GPS time to a Int32 with the format YYYYMMDD. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # Initiate return value return_value = '' log_file = open('log.txt', 'a+') # Set file and folder paths odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' temp_file_maj = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_maj.tif' temp_file_min = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_min.tif' temp_file_max = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_max.tif' temp_file_maj_2 = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_maj_2.tif' temp_file_min_2 = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_min_2.tif' temp_file_max_2 = re.sub('\\\\', '/', os.getcwd()) + '/temp_' + tile_id + '_max_2.tif' out_folder_all = settings.output_folder + '/date_stamp' out_folder_maj = out_folder_all + '/date_stamp_mode' out_folder_min = out_folder_all + '/date_stamp_min' out_folder_max = out_folder_all + '/date_stamp_max' out_file_maj = out_folder_maj + '/date_stamp_mode_' + tile_id + '.tif' out_file_min = out_folder_min + '/date_stamp_min_' + tile_id + '.tif' out_file_max = out_folder_max + '/date_stamp_max_' + tile_id + '.tif' # Create folders if they do not already exists for folder in [out_folder_all, out_folder_maj, out_folder_min, out_folder_max]: if not os.path.exists(folder): os.mkdir(folder) # Add GPSDate attribute to point cloud (veg points only) try: log_file.write('\n' + tile_id + ' adding GPSdate. \n') add_GPSDate = opals.AddInfo.AddInfo() add_GPSDate.inFile = odm_file add_GPSDate.attribute = '_GPSDay=floor(GPSTime/(60*60*24))' add_GPSDate.commons.screenLogLevel = opals.Types.LogLevel.none add_GPSDate.commons.nbThreads = settings.nbThreads add_GPSDate.filter = settings.veg_classes_filter add_GPSDate.run() add_GPSDate.reset() add_GPSDate = None log_file.write(tile_id + ' success. \n') except: return_value = 'opalsError' # Export time stamp mode, min and max try: # Initialise exporter export_time_stamp = opals.Cell.Cell() # Export most common date stamp (mode) log_file.write(tile_id + ' exporting mode GPSDate raster... \n') export_time_stamp.inFile = odm_file export_time_stamp.outFile = temp_file_maj export_time_stamp.attribute = '_GPSDay' export_time_stamp.feature = 'majority' export_time_stamp.cellSize = settings.out_cell_size export_time_stamp.limit = 'corner' # This switch is really important when working with tiles! export_time_stamp.filter = settings.veg_classes_filter export_time_stamp.noData = 0.0 export_time_stamp.commons.screenLogLevel = opals.Types.LogLevel.none export_time_stamp.commons.nbThreads = settings.nbThreads export_time_stamp.run() export_time_stamp.reset() log_file.write(tile_id + ' success. \n') # Export earliest date stamp (min) log_file.write(tile_id + ' exporting min GPSDate raster... \n') export_time_stamp.inFile = odm_file export_time_stamp.outFile = temp_file_min export_time_stamp.attribute = '_GPSDay' export_time_stamp.feature = 'min' export_time_stamp.cellSize = settings.out_cell_size export_time_stamp.limit = 'corner' # This switch is really important when working with tiles! export_time_stamp.filter = settings.veg_classes_filter export_time_stamp.noData = 0.0 export_time_stamp.commons.screenLogLevel = opals.Types.LogLevel.none export_time_stamp.commons.nbThreads = settings.nbThreads export_time_stamp.run() export_time_stamp.reset() log_file.write(tile_id + ' success. \n') # Export most recent date stamp (max) log_file.write(tile_id + ' exporting max GPSDate raster... \n') export_time_stamp.inFile = odm_file export_time_stamp.outFile = temp_file_max export_time_stamp.attribute = '_GPSDay' export_time_stamp.feature = 'max' export_time_stamp.cellSize = settings.out_cell_size export_time_stamp.limit = 'corner' # This switch is really important when working with tiles! export_time_stamp.filter = settings.veg_classes_filter export_time_stamp.noData = 0.0 export_time_stamp.commons.screenLogLevel = opals.Types.LogLevel.none export_time_stamp.commons.nbThreads = settings.nbThreads export_time_stamp.run() export_time_stamp.reset() log_file.write(tile_id + ' success. \n') except: return_value = 'opalsError' # Convert GPSDate to YYYYMMDD date and save as 32 bit integer try: ## GPSDate mode ---- log_file.write(tile_id + ' converting mode raster... \n') # Load raster as numpy array temp_raster_maj = gdal_array.LoadFile(temp_file_maj) # Generate mask temp_raster_mask = temp_raster_maj == 0 # Convert to seconds and add 10^9 to all values (as this has been previously substracted) and copy to new object temp_raster = temp_raster_maj * (60*60*24) + 10**9 # close original file connection temp_raster_maj = None # Convert time stamp to CET date as integer # Note we assume that the difference in leap seconds is constant despite a shift # on 1 July 2015 (it is 36 afterwards) -> we drop the hours anyways # https://hpiers.obspm.fr/eop-pc/index.php?index=TAI-UTC_tab&lang=en # Conversion based on https://stackoverflow.com/questions/33415475/how-to-get-current-date-and-time-from-gps-unsegment-time-in-python # Thanks to user jfs convert_to_utc = lambda t: int((datetime(1980, 1, 6) + timedelta(seconds=t - (35 - 19) - 3600)).strftime("%Y%m%d")) convert_to_utc_vec = numpy.vectorize(convert_to_utc) temp_raster = convert_to_utc_vec(temp_raster) # Re-assing NA values temp_raster[temp_raster_mask] = -9999 # Set any values in 2011 to NA (these result from GPS time stamps that # were not converted from GPS seconds per week to GPS time => they end up in Sept. 2011). # We know that there were no flights in 2011 which allows us to create this mask. temp_raster[numpy.logical_and(temp_raster >= 20110101, temp_raster <= 20111231)] = -9999 # Write array as raster (Int32 as default - needed for 8 digit date format) temp_raster = gdal_array.SaveArray(temp_raster, temp_file_maj_2, format = "GTiff", prototype = temp_file_maj) log_file.write(tile_id + ' done. \n') # Close file connection temp_raster = None # Reset no data value cmd = settings.gdal_translate_bin + ' -of GTiff -a_nodata -9999 ' + \ temp_file_maj_2 + ' ' + out_file_maj # Execute gdal command log_file.write('\n' + tile_id + ' re-setting no data value... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file_maj) # Log file output log_file.write('\n' + tile_id + ' time_stamp export successful. \n') ## GPSDate min ----- log_file.write(tile_id + ' converting min raster... \n') # Load raster as numpy array temp_raster_min = gdal_array.LoadFile(temp_file_min) # Generate mask temp_raster_mask = temp_raster_min == 0 # Convert to seconds and add 10^9 to all values (as this has been previously substracted) and copy to new object temp_raster = temp_raster_min * (60*60*24) + 10**9 # close original file connection temp_raster_min = None # Convert time stamp to CET date as integer # Note we assume that the difference in leap seconds is constant despite a shift # on 1 July 2015 (it is 36 afterwards) -> we drop the hours anyways # https://hpiers.obspm.fr/eop-pc/index.php?index=TAI-UTC_tab&lang=en # Conversion based on https://stackoverflow.com/questions/33415475/how-to-get-current-date-and-time-from-gps-unsegment-time-in-python # Thanks to user jfs convert_to_utc = lambda t: int((datetime(1980, 1, 6) + timedelta(seconds=t - (35 - 19) - 3600)).strftime("%Y%m%d")) convert_to_utc_vec = numpy.vectorize(convert_to_utc) temp_raster = convert_to_utc_vec(temp_raster) # Re-assing NA values temp_raster[temp_raster_mask] = -9999 # Set any values in 2011 to NA (these result from GPS time stamps that # were not converted from GPS seconds per week to GPS time => they end up in Sept. 2011). # We know that there were no flights in 2011 which allows us to create this mask. temp_raster[numpy.logical_and(temp_raster >= 20110101, temp_raster <= 20111231)] = -9999 # Write array as raster (Int32 as default - needed for 8 digit date format) temp_raster = gdal_array.SaveArray(temp_raster, temp_file_min_2, format = "GTiff", prototype = temp_file_min) log_file.write(tile_id + ' done. \n') # Close file connection temp_raster = None # Reset no data value cmd = settings.gdal_translate_bin + ' -of GTiff -a_nodata -9999 ' + \ temp_file_min_2 + ' ' + out_file_min # Execute gdal command log_file.write('\n' + tile_id + ' re-setting no data value... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file_min) ## GPSDate max ----- log_file.write(tile_id + ' converting max raster... \n') # Load raster as numpy array temp_raster_max = gdal_array.LoadFile(temp_file_max) # Generate mask temp_raster_mask = temp_raster_max == 0 # Convert to seconds and add 10^9 to all values (as this has been previously substracted) and copy to new object temp_raster = temp_raster_max * (60*60*24) + 10**9 # close original file connection temp_raster_max = None # Convert time stamp to CET date as integer # Note we assume that the difference in leap seconds is constant despite a shift # on 1 July 2015 (it is 36 afterwards) -> we drop the hours anyways # https://hpiers.obspm.fr/eop-pc/index.php?index=TAI-UTC_tab&lang=en # Conversion based on https://stackoverflow.com/questions/33415475/how-to-get-current-date-and-time-from-gps-unsegment-time-in-python # Thanks to user jfs convert_to_utc = lambda t: int((datetime(1980, 1, 6) + timedelta(seconds=t - (35 - 19) - 3600)).strftime("%Y%m%d")) convert_to_utc_vec = numpy.vectorize(convert_to_utc) temp_raster = convert_to_utc_vec(temp_raster) # Re-assing NA values temp_raster[temp_raster_mask] = -9999 # Set any values in 2011 to NA (these result from GPS time stamps that # were not converted from GPS seconds per week to GPS time => they end up in Sept. 2011). # We know that there were no flights in 2011 which allows us to create this mask. temp_raster[numpy.logical_and(temp_raster >= 20110101, temp_raster <= 20111231)] = -9999 # Write array as raster (Int32 as default - needed for 8 digit date format) temp_raster = gdal_array.SaveArray(temp_raster, temp_file_max_2, format = "GTiff", prototype = temp_file_max) log_file.write(tile_id + ' done. \n') # Close file connection temp_raster = None # Reset no data value cmd = settings.gdal_translate_bin + ' -of GTiff -a_nodata -9999 ' + \ temp_file_max_2 + ' ' + out_file_max # Execute gdal command log_file.write('\n' + tile_id + ' re-setting no data value... \n' + \ subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT)) # Apply mask(s) common.apply_mask(out_file_max) # Log file output log_file.write('\n' + tile_id + ' time_stamp export successful. \n') return_value = 'success' except: if return_value == 'opalsError': pass else: return_value = 'gdalError' log_file.write('\n' + tile_id + ' date_stamp export failed. \n') # Tidy up try: os.remove(temp_file_maj) os.remove(temp_file_min) os.remove(temp_file_max) os.remove(temp_file_maj_2) os.remove(temp_file_min_2) os.remove(temp_file_max_2) except: pass # Close log file log_file.close() # Return exist status return return_value def odm_remove_temp_files(tile_id): """ Removes footprint and odm files to clear up space for subsequent processing. :param tile_id: tile id in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: execution status """ # initiate return value return_value = '' odm_file = settings.odm_folder + '/odm_' + tile_id + '.odm' odm_footprint_files = glob.glob(settings.odm_footprint_folder + '/footprint_' + tile_id + '.*') try: os.remove(odm_file) return_value = 'success' except: return_value = 'unable to delete odm file' try: for file in odm_footprint_files: os.remove(file) return_value = 'success' except: return_value = return_value + 'unable to delete odm footprint file' # return execution status return return_value
jakobjassmann/ecodes-dk-lidar
dklidar/common.py
### Common module for the dklidar reporcessing - general functions likely used by all scripts ### <NAME> <EMAIL> 29 January 2019 # Imports import os import glob import pandas import re import shutil import datetime import subprocess from dklidar import settings ## Function definitons ## Logging function to initalise logging process key to progress managment. def init_log_folder(script_name, tile_ids): """ Initiates a log folder for storing the processing output and progress management :param script_name: name of the processing script for which to initialise the logging database / folder :param tile_ids: tile ids in the format "rrrr_ccc" where rrrr is the row number and ccc is the column number :return: pandas DataFrame with progress data """ # Check and create root folder for script log_folder = settings.log_folder + '/' + script_name if not os.path.exists(log_folder): os.mkdir(log_folder) # Check whether processing status file exists progress_file = log_folder + '/' + 'overall_progress.csv' if not os.path.exists(progress_file): print(datetime.datetime.now().strftime('%X') + ' No progress file found, creating log folder and progress file...'), ## Initiate pandas data frame with tile_ids as rows and processing_steps as columns # Create empty list to hold columns cols = [] # Add processing status column cols.append(['pending'] * len(tile_ids)) # prepare colnames colnames = ['processing'] # Zip into pandas data frame progress_df = pandas.DataFrame(zip(*cols), index = tile_ids, columns = colnames) progress_df.index.name = 'tile_id' # Status update print(' done.') else: print(datetime.datetime.now().strftime('%X') + ' Progress file found, loading previous processing status...'), ## Load progress status_file try: progress_df = pandas.read_csv(progress_file, index_col='tile_id') print(' done.') # update progress dataframe progress_df = update_progress_df(script_name, progress_df) except: print('\n' + datetime.datetime.now().strftime('%X') + 'Can\'t load progress file. Exiting script!') quit() # Compare tile_id column with tile_ids list if not progress_df.index.values.tolist() == tile_ids: print('\n' +datetime.datetime.now().strftime('%X') + 'Warning: lists of tile_ids in laz folder( ' + settings.laz_folder + ') and progress file (' + progress_file + ') do not match.' + '\nPlease remove manually to reset.\nExiting script!') quit() # Export progress_df as CSV progress_df.to_csv(progress_file, index=True, header=True) # return progress_df return(progress_df) ## Function to gather progress update, key to progress management and logging. def update_progress_df(script_name, progress_df): """ Searches a script's log folder for subfolders matching the tile_id pattern (rrrr_ccc), then crawls these folders for status.csv files, compiling them into a single pandas dataframe and returning it. :param script_name: name of the script (for folder matching) :param progress_df: progress dataframe to be updated :return: returns an updated progress_df """ # Status update print(datetime.datetime.now().strftime('%X') + ' Updating progress management...'), # Check log root folder for script if not existing... quit! log_folder = settings.log_folder + '/' + script_name if not os.path.exists(log_folder): print('\n' + datetime.datetime.now().strftime('%X') + 'Warning: script log folder does not exist. Exiting script') quit() # Gather list of tile folders tile_folders = glob.glob(log_folder + '/*/') # Filter only those folders matching a tile/id tile_id_pattern = re.compile('.*\d+_\d+.*') tile_folders = filter(tile_id_pattern.match, tile_folders) # Loop through each tile_id_folder updating the log for tile_folder in tile_folders: # set path to status csv file status_csv = tile_folder + '/status.csv' # Check whether status.csv file exists. if os.path.exists(status_csv): # load status csv status_df = pandas.read_csv(status_csv, index_col='tile_id') # get tile id from first value in index tile_id = status_df.index.values.tolist()[0] # get list of colnames col_names = status_df.columns.values.tolist() # for each col name copy cell value into progress_df for col_name in col_names: # set cell value status_value = status_df.at[tile_id, col_name] # check whether col_name exists in progress data frame if not add empty col to data frame if col_name in progress_df.columns.values.tolist(): # set cell value for col and tile id progress_df.at[tile_id, col_name] = status_value else: # create empty 'pending' colum for colname that does not exist. progress_df[col_name] = ['pending'] * len(progress_df.index) progress_df.at[tile_id, col_name] = status_value else: # if the status_csv does not exist, something must have gone wrong. # do not update the progress for this tile, re-process tile. pass # Status update print(' done.') # Return progress_df return(progress_df) ## Define function to gather logs def gather_logs(script_name, step_name, tile_id): """ Gather logs from temporary work dir. This wee function is to be run out of a temporary working directory by a pool process from a multiprocessing workflow. It then copies all log files in the temporary working directory to the main log folder for the processing script, where it stores them in a sub-subfolder according to the step_name and tile_id parameters. :param script_name: name of the script that is calling the function :param step_name: name of the step that should be logged for :param tile_id: tile id in the usual format (rrrr_ccc) :return: nothing """ # Generate string for log folder for the tile and create the directory if it does not exist log_folder_tile = settings.log_folder + '/' + script_name + '/' + tile_id if not os.path.exists(log_folder_tile): os.mkdir(log_folder_tile) # Generate string for log folder and create the directory if it does not exist log_folder_step = log_folder_tile + '/' + step_name if not os.path.exists(log_folder_step): os.mkdir(log_folder_step) # Confirm function is executed from temporary work dir using regex wd = os.getcwd() temp_re = re.compile('.*temp_.*') if not temp_re.match(wd): print(datetime.datetime.now().strftime('%X') + ' Error: gather_logs function called from outside temporary work dir.') print(wd) return('Error: gather_logs.') # Check whether dklidar logfile exists if yes copy: if os.path.exists(wd + '/log.txt'): # copy file to tile log directory shutil.copy(wd + '/log.txt', log_folder_step) # remove log file from temp directory os.remove(wd + '/log.txt') # Check whether opalslog logfile exists if yes copy: if os.path.exists(wd + '/opalsLog.xml'): # copy file to tile log directory shutil.copy(wd + '/opalsLog.xml', log_folder_step) # remove log file from temp directory os.remove(wd + '/opalsLog.xml') # Check whether opalsError logfile exists if yes copy: if os.path.exists(wd + '/opalsErrors.txt'): # copy file to tile log directory shutil.copy(wd + '/opalsErrors.txt', log_folder_step) # remove log file from temp directory os.remove(wd + '/opalsErrors.txt') ## Function to generate sea and inland water masks for a tile def generate_water_masks(tile_id): """ Generates both sea and inland water mask rasters with 10 m grain size, based on the dtm as a template and the the nationwide sea and inland water masks vector files specified in the settings file. :param tile_id: id of the tile :return: execution status """ # Initiate return valule and log output return_value = '' log_file = open('log.txt', 'a+') # get temporary work directory wd = os.getcwd() # Prepare output folder sea_out_folder = settings.output_folder + '/masks/sea_mask' inland_water_out_folder = settings.output_folder + '/masks/inland_water_mask' if not os.path.exists(settings.output_folder + '/masks'): os.mkdir(settings.output_folder + '/masks') if not os.path.exists(sea_out_folder): os.mkdir(sea_out_folder) if not os.path.exists(inland_water_out_folder): os.mkdir(inland_water_out_folder) sea_mask_file = sea_out_folder + '/sea_mask_' + tile_id + '.tif ' inland_mask_file = inland_water_out_folder + '/inland_water_mask_' + tile_id + '.tif ' temp_file = os.getcwd() + '/temp.tif' try: ## Aggregate dtm cmd = settings.gdalwarp_bin + \ '-tr 10 10 -r min -ot Int16 -dstnodata -9999 -overwrite ' + \ settings.dtm_folder + '/DTM_1km_' + tile_id + '.tif ' + \ temp_file # Execute gdal command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' aggregating dtm to 10 m for mask successful.\n\n') # Set all cells with data in raster to 0 and set it as no data value cmd = settings.gdal_calc_bin + \ '--calc=1 --NoDataValue=-9999 --overwrite --type Int16 ' + \ '-A ' + temp_file + ' ' + \ '--outfile=' + sea_mask_file # Execute gdal command log_file.write(subprocess.check_output(cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + tile_id + ' set all cells with data to 1.\n\n') # Dublicate file shutil.copyfile(sea_mask_file, inland_mask_file) # Make a local copies of the nationwide masks to speed up simulatneaous access dk_sea_mask_path = re.sub('(.*/)(.*)\.shp$', '\g<1>', settings.dk_coastline_poly) dk_inland_mask_path = re.sub('(.*/)(.*)\.shp$', '\g<1>', settings.dk_lakes_poly) dk_sea_mask_file_base = re.sub('(.*/)(.*)\.shp$', '\g<2>', settings.dk_coastline_poly) dk_inland_mask_file_base = re.sub('(.*/)(.*)\.shp$', '\g<2>', settings.dk_lakes_poly) for file in glob.glob(dk_sea_mask_path + dk_sea_mask_file_base + '.*'): temp_file = os.getcwd() + '/temp_' + re.sub('(.*\\\)(.*)\.*$', '\g<2>', file) shutil.copy(file, temp_file) for file in glob.glob(dk_inland_mask_path + dk_inland_mask_file_base + '.*'): temp_file = os.getcwd() + '/temp_' + re.sub('(.*\\\)(.*)\.*$', '\g<2>', file) shutil.copy(file, temp_file) dk_sea_mask_temp_file = os.getcwd() + '/temp_' + dk_sea_mask_file_base + '.shp' dk_inland_mask_temp_file = os.getcwd() + '/temp_' + dk_inland_mask_file_base + '.shp' # Generate sea mask cmd = settings.gdal_rasterize_bin + \ '-b 1 ' + '-burn -9999 ' + '-i ' + '-at ' + \ dk_sea_mask_temp_file + ' ' + \ sea_mask_file log_file.write('\n' + \ subprocess.check_output( cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + sea_mask_file + ' sea mask created. \n\n ') # Generate inland water mask cmd = settings.gdal_rasterize_bin + \ '-b 1 ' + '-burn -9999 ' + '-at ' + \ dk_inland_mask_temp_file + ' ' + \ inland_mask_file log_file.write('\n' + \ subprocess.check_output( cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + inland_mask_file + ' inland water mask created. \n\n ') # Remove temporary files temp_file_list = glob.glob(os.getcwd() + '/temp*.*') for file in temp_file_list: os.remove(file) return_value = 'success' except: log_file.write('\n' + tile_id + ' creating mask rasters failed.\n\n') return_value = 'gdalError' # Close log file log_file.close() return return_value ## Function to apply water masks, sea or inland water. def apply_mask(target_raster = '', sea_mask = False, inland_water_mask = False): """ For a given target raster, this function masks all sea off the coastline of Denmark (sea_mask = True), or all inland water bodies such as lakes or ponds (inland_water_mask = True) or both. Requires raster masks to be generated using generate_water_masks(). :param sea_mask: boolean switch for applying sea mask :param inland_water_mask: boolean switch for applying the inland water mask :param target_raster: target raster file path :return: execution status """ # initiate return value and log ouptut return_value = '' log_file = open('log.txt', 'a+') # Check whether input raster was provided if (target_raster == ''): raise Exception('No input raster provided.') # Get current wd temp_wd = os.getcwd() # Get tile_id from path tile_id = re.sub('.*?_(\d*_)(\d*)(_\d*)?\.tif *', '\g<1>\g<2>', target_raster) # set mask paths sea_out_folder = settings.output_folder + '/masks/sea_mask' inland_water_out_folder = settings.output_folder + '/masks/inland_water_mask' sea_mask_file = sea_out_folder + '/sea_mask_' + tile_id + '.tif' inland_mask_file = inland_water_out_folder + '/inland_water_mask_' + tile_id + '.tif' temp_file = temp_wd + '/temp_raster.tif' # Apply sea mask if (sea_mask == True): try: # Construct gdal command cmd = settings.gdal_calc_bin + \ '-A ' + sea_mask_file + ' ' + \ '-B ' + target_raster + ' ' + \ '--outfile=' + temp_file + ' ' + \ '--calc=B --NoDataValue=-9999 --overwrite --type Int16 ' log_file.write('\n' + \ subprocess.check_output( cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + target_raster + ' sea mask applied. \n\n ') shutil.copyfile(temp_file, target_raster) os.remove(temp_file) return_value = 'success' except: log_file.write('\n' + target_raster + ' applying sea mask failed. \n\n ') return_value = 'gdalError' # Apply lake mask if (inland_water_mask == True): try: # Construct gdal command cmd = settings.gdal_calc_bin + \ '-A ' + inland_mask_file + ' ' + \ '-B ' + target_raster + ' ' + \ '--outfile=' + temp_file + ' ' + \ '--calc=B --NoDataValue=-9999 --overwrite --type Int16 ' log_file.write('\n' + \ subprocess.check_output( cmd, shell=False, stderr=subprocess.STDOUT) + \ '\n' + target_raster + ' inland water mask applied. \n\n ') shutil.copyfile(temp_file, target_raster) os.remove(temp_file) except: log_file.write('\n' + target_raster + ' applying inland water mask failed. \n\n ') return_value = 'gdalError' if (sea_mask == False & inland_water_mask == False): log_file.write('\n' + target_raster + ' no masks to be applied. \n\n ') return_value = 'success' # Close log file log_file.close() return return_value
jakobjassmann/ecodes-dk-lidar
scripts/remove_missing_tiles.py
# Short script to remove tiles that have missing point cloud or dtm data # This script can be run after executing checksum_qa.py and directly uses the output files # generated by the script. # <NAME> <EMAIL> 16 January 2020 import os from dklidar import settings # Prompt user input confirmation_string = raw_input('Are you sure you would like to remove the files for all incompleted tiles?\n[yes/no]: ') if(confirmation_string == 'yes' or confirmation_string == 'Yes'): print('Deleting LAZ files with missing DTM tiles...') # Remove LAZ files with missing DTMs with open(settings.laz_folder + '../laz_files_with_missing_dtm.txt') as f: files = f.read().splitlines() for file in files: os.remove(file) print('Done.') print('Deleting DTM files with missing LAZ tiles...') # Remove DTM files with missing LAZs with open(settings.laz_folder + '../dtm_files_with_missing_laz.txt') as f: files = f.read().splitlines() for file in files: os.remove(file) print('Done.') else: print("No or invalid input, aborting script...")
jakobjassmann/ecodes-dk-lidar
scripts/checksum_qa.py
<gh_stars>0 # Short script to confirm transferred file integrity using md5 checksums. # Prerequisite is that checsum were generated using create_checksums.bat # A comparsions of the sets of laz tiles and dtm tiles is also carried out. # <NAME> <EMAIL> 16 January 2020 # Imports import os import glob import numpy as np import pandas import re from dklidar import settings ### Read in filenames using unix filename extensions from glob ### 1) Pointcloud files (laz) ##orig_md5_files = glob.glob(settings.laz_folder + '*.md5') ##local_md5_files = glob.glob(settings.laz_folder + '*.local_md5') ### 2) DTM files (tif) ##orig_md5_files.extend(glob.glob(settings.dtm_folder + "*.md5")) ##local_md5_files.extend(glob.glob(settings.dtm_folder + "*.local_md5")) ## ### Initiate empty lists ##md5_orig = list() ##md5_local = list() ## ### Fill lists with md5 sums from files ##for file_name in orig_md5_files: ## file = open(file_name) ## md5_orig.append(file.read(32)) ##for file_name in local_md5_files: ## file = open(file_name) ## md5_local.append(file.read(32)) ## ### Zip all lists into one data frame ##df = pandas.DataFrame(zip(orig_md5_files, md5_orig, local_md5_files, md5_local), ## columns=['orig_file', 'orig_md5', 'local_file', 'local_md5']) ## ### Add md5_vheck comparison column to df ##md5_check = df['orig_md5'] == df['local_md5'] ##df['md5_check'] = md5_check ## ### Print dataframe overiview to console ##print() ##print('df overview:') ##print(df.head()) ##print() ## ### Filter rows where the check returned false ##print('Non matching md5 checksums:' + str(np.sum([not i for i in df['md5_check']]))) ##damaged_files = df[df['md5_check'] == False] ##print(damaged_files) ## ### Export csv ##damaged_files.to_csv(settings.laz_folder + '../damaged_files.csv', index=False) ## # --------------------------------------------------- # Check for completeness of datasets # Load file names dtm_files = glob.glob(settings.dtm_folder + '/*.tif') laz_files = glob.glob(settings.laz_folder + '/*.laz') # initiate empty lists for tile_ids dtm_tile_ids = [] laz_tile_ids = [] # fill dictionaries with tile_id, as well as row number and column number for each file name: for file_name in dtm_files: tile_id = re.sub('.*DTM_1km_(\d*_\d*).tif', '\g<1>', file_name) dtm_tile_ids.append(tile_id) for file_name in laz_files: tile_id = re.sub('.*PUNKTSKY_1km_(\d*_\d*).laz', '\g<1>', file_name) laz_tile_ids.append(tile_id) # Determine differences between sets of tiles missing_laz_tiles = set(dtm_tile_ids) - set(laz_tile_ids) missing_dtm_tiles = set(laz_tile_ids) - set(dtm_tile_ids) df_missing_dtm = pandas.DataFrame(zip(missing_dtm_tiles), columns=['tile_id']) df_missing_dtm.to_csv(settings.dtm_folder + '../missing_dtm_tile_ids.csv', index=False) # Print out a quick overview of data frame for control print(df_missing_dtm.head()) df_missing_laz = pandas.DataFrame(zip(missing_laz_tiles), columns=['tile_id']) df_missing_laz.to_csv(settings.laz_folder + '../missing_laz_tile_ids.csv', index=False) # Print out a quick overview of data frame for control print(df_missing_laz.head()) # Export lists to files # DTMs with missing LAZs out_file = open(settings.dtm_folder + '../dtm_files_with_missing_laz.txt', 'a+') for tile_id in missing_laz_tiles: out_file.write(settings.dtm_folder + 'DTM_1km_' + tile_id + '.tif\n') out_file.close() # LAZs with missing DTMs out_file = open(settings.laz_folder + '../laz_files_with_missing_dtm.txt', 'a+') for tile_id in missing_dtm_tiles: out_file.write(settings.laz_folder +'PUNKTSKY_1km_' + tile_id + '.laz\n') out_file.close()
jakobjassmann/ecodes-dk-lidar
scripts/progress_monitor.py
### Script to mointor progress of the process_tiles.py script ### Execute in separate terminal in when running process_tile.py ### <NAME> <EMAIL> 20 April 2020 ## Imports import glob import os import datetime import time from dklidar import settings # Set working directory os.chdir(settings.wd) # Set number of parallel processes: n_processes = 62 # 54 # set update interval update_interval = 60 # 60 s # Load tile file names and derive total number of files to process laz_files = glob.glob(settings.laz_folder + '/*.laz') n_total = len(laz_files) # Check nubmer of files already processed n_processed_at_start = len(glob.glob(settings.log_folder + '/process_tiles/*_*')) # set start date and time: start_time = datetime.datetime.fromtimestamp(os.path.getmtime(settings.log_folder + '/process_tiles/overall_progress.csv')) # Initate progress variables progress = 0 # Update progress till complete while progress < 1: # Obtain n sum of n tiles in process and completed n_sum = len(glob.glob(settings.log_folder + '/process_tiles/*_*')) # Calculate n of fully processed tiles n_processed = n_sum - n_processes progress = float(n_processed) / float(n_total) # Calculate time differences time_passed = datetime.datetime.now() - start_time if (n_processed - n_processed_at_start) <= 54: time_estimated = 'estimating' else: time_estimated = (time_passed / (n_processed - n_processed_at_start)) * (n_total - n_processed) # Check whether the overall_progress.csv has been updated, if so # assume the processing of the last 54 files is done overal_progress_timestamp = datetime.datetime.fromtimestamp( os.path.getmtime(settings.log_folder + '/process_tiles/overall_progress.csv')) if not overal_progress_timestamp == start_time: progress = 1 n_processed = n_total # Print stats on screen os.system('cls') print('\n') print('-' * 80), print(' ' * (79 - len('<NAME> <EMAIL> 2020 ')) + '<NAME> <EMAIL> 2020 ') print(' \'process_tiles.py\' progress:') print(' - processing with ' + str(n_processes) + ' parallel threads') print(' - start time: ' + str(start_time.ctime())) print(' - execute \'stop.bat\' to pause \ interrupt processing') print('\n') print(' update interval: ' + str(update_interval) + ' s' + ' ' * (79 - len(' update interval: ' + str(update_interval) + ' s' + 'last update: ' + str(datetime.datetime.now().ctime()))) + 'last update: ' + str(datetime.datetime.now().ctime())) print('-' * 80) print('\n' * 2) print(' ' + str(n_processed) + ' / ' + str(n_total) + ' tiles' + ' ' * (80 - len(' ' + str(n_processed) + ' / ' + str(n_total) + ' tiles' + str(int(round(progress * 100))) + '% ')) + str(int(round(progress * 100))) + '%'), print('-' * 80), print('#' * int(round(78*progress))) print('-' * 80), print('passed: ' + str(time_passed).split('.')[0] + ' ' * (79 - len('passed: ' + str(time_passed).split('.')[0] + 'remaining (estimate): ' + str(time_estimated).split('.')[0] + ' ')) + 'remaining (estimate): ' + str(time_estimated).split('.')[0] + ' ') print(' ' * (79 - len('press CRTL+C to exit progress monitor ')) + 'press CRTL+C to exit progress monitor ') # Wait for next update time.sleep(update_interval) # End of while loop
jakobjassmann/ecodes-dk-lidar
documentation/source_data/merger_scripts/copy_files.py
<reponame>jakobjassmann/ecodes-dk-lidar ## DHM2018+ and DHM2015 merger ## Script to copy the files and create a physical version of the merger ## using the outputs from dhm201415_merger.R ## <NAME> <EMAIL> ## Dependencies import os import shutil import pandas import glob import re ## Status print('\n' + '#' * 80) print('Merging DHM2018+, DHM2015 and GST2014\n') print('Setting up work environment...') ## Prepare environment # Set wd os.chdir('D:/Jakob/dhm201415_merger/') # Set file paths gst2014_laz = 'O:/Nat_Ecoinformatics/B_Read/LegacyData/Denmark/Elevation/GST_2014/Punktsky/laz/' dhm2015_laz = 'D:/Jakob/datafordeler_downloads/DHM2015_punktsky' dhm2018_laz = 'D:/Jakob/dk_nationwide_lidar/data/laz' gst2014_dtm = 'O:/Nat_Ecoinformatics/B_Read/LegacyData/Denmark/Elevation/GST_2014/DTM_tif' dhm2015_dtm = 'D:/Jakob/datafordeler_downloads/DHM2015_terraen' dhm2018_dtm = 'D:/Jakob/dk_nationwide_lidar/data/dtm' laz_out = 'laz/' dtm_out = 'dtm/' os.mkdir(laz_out) os.mkdir(dtm_out) # Load tile_ids to source from DHM2015 gst2014_tile_ids = pandas.read_csv('tiles_from_GST2014.csv') # Load tile_ids to source from DHM2015 dhm2015_tile_ids = pandas.read_csv('tiles_from_DHM2015.csv') # Load tile_ids to source from DHM2018+ dhm2018_tile_ids = pandas.read_csv('tiles_from_DHM2018.csv') # Load file names gst2014_laz_files = glob.glob(gst2014_laz + '/*.laz') dhm2015_laz_files = glob.glob(dhm2015_laz + '/*.laz') dhm2018_laz_files = glob.glob(dhm2018_laz + '/*.laz') gst2014_dtm_files = glob.glob(gst2014_dtm + '/*.tif') dhm2015_dtm_files = glob.glob(dhm2015_dtm + '/*.tif') dhm2018_dtm_files = glob.glob(dhm2018_dtm + '/*.tif') # Helper function to retrieve tile_ids from file names def get_tile_ids(file_names): # initiate empty list for tile_ids tile_ids = [] # clean up files names for i in range(0, len(file_names)): file_names[i] = re.sub('\\\\', '/', file_names[i]) # fill dictionaries with tile_id, as well as row number and column number for each file name: for file_name in file_names: tile_id = re.sub('.*(\d{4}_\d{3}).*', '\g<1>', file_name) tile_ids.append(tile_id) # combine to data frame files_df = pandas.DataFrame(zip(*[tile_ids, file_names]), columns = ['tile_id', 'file_name']) # return files_df return(files_df) # Appy helper function to all file names gst2014_laz_files_df = get_tile_ids(gst2014_laz_files) dhm2015_laz_files_df = get_tile_ids(dhm2015_laz_files) dhm2018_laz_files_df = get_tile_ids(dhm2018_laz_files) gst2014_dtm_files_df = get_tile_ids(gst2014_dtm_files) dhm2015_dtm_files_df = get_tile_ids(dhm2015_dtm_files) dhm2018_dtm_files_df = get_tile_ids(dhm2018_dtm_files) ## Copy files # Helper function to select and copy files with progress bar def copy_tiles(tiles_to_copy, files_df, out_folder): # Subset files to copy files_to_copy = files_df[files_df['tile_id']. isin(tiles_to_copy)]['file_name'].tolist() # Set counter to 0: progress = 0 # Copy files for i in range(0, len(files_to_copy)): # Check whether file exists, if not copy if not os.path.isfile(os.getcwd() + '/' + out_folder + re.sub('.*/(.*)', '\g<1>', files_to_copy[i])): # Copy file shutil.copy(files_to_copy[i], os.getcwd() + '/' + out_folder + re.sub('.*/(.*)', '\g<1>', files_to_copy[i])) # Update progress progress = float(i + 1) / float(len(files_to_copy)) # Update progress bar print('\r|' + '#' * int(round(progress * 54)) + '-' * int(round((1 - progress) * 54)) + '| ' + str(int(round(progress * 100))) + '%'), # Copy laz files print('\nCopying GST2014 laz files...') copy_tiles(gst2014_tile_ids['tile_id'], gst2014_laz_files_df, laz_out) print('\n\nCopying DHM2015 laz files...') copy_tiles(dhm2015_tile_ids['tile_id'], dhm2015_laz_files_df, laz_out) print('\n\nCopying DHM2018 laz files...') copy_tiles(dhm2018_tile_ids['tile_id'], dhm2018_laz_files_df, laz_out) # Copy dtm files print('\n\nCopying GST2014 dtm files...') copy_tiles(gst2014_tile_ids['tile_id'], gst2014_dtm_files_df, dtm_out) print('\n\nCopying DHM2015 dtm files...') copy_tiles(dhm2015_tile_ids['tile_id'], dhm2015_dtm_files_df, dtm_out) print('\n\nCopying DHM2018 dtm files...') copy_tiles(dhm2018_tile_ids['tile_id'], dhm2018_dtm_files_df, dtm_out) ## Remove incomplete laz / dtm pairs from dataset # Status print('\n\nDetermining completness of laz / dtm pairs in merged data set...\n') # Retrieve files names and tile ids in merged data set laz_merged_files = glob.glob(laz_out + '/*.laz') dtm_merged_files = glob.glob(dtm_out + '/*.tif') laz_merged_files_df = get_tile_ids(laz_merged_files) dtm_merged_files_df = get_tile_ids(dtm_merged_files) # Work out incomplete pairs laz_missing_dtm = list(set(laz_merged_files_df['tile_id']) - set(dtm_merged_files_df['tile_id'])) dtm_missing_laz = list(set(dtm_merged_files_df['tile_id']) - set(laz_merged_files_df['tile_id'])) # Save incomplte pairs to file incomplete_pairs = pandas.concat( [laz_merged_files_df[laz_merged_files_df['tile_id'].isin(laz_missing_dtm)], dtm_merged_files_df[dtm_merged_files_df['tile_id'].isin(dtm_missing_laz)]]) incomplete_pairs.to_csv('incomplete_tile_pairs.csv', index = False) # Status print(str(len(laz_missing_dtm)) + ' laz files have no corresponding dtm file.' + '\n') print(str(len(dtm_missing_laz)) + ' dtm files have no corresponding laz file.' + '\n') print('Incomplete tile_ids and file names saved to' + ' "incomplete_tile_pairs.csv".\n') # Prompt for choice to delete del_choice = raw_input('Would you like to remove the incomplete tile pairs?' + '[y/n]') # Evaluate choice if(del_choice == 'n'): print('\nNo files will be deleted.') elif(del_choice == 'y'): print('\nDeleting files...') # Delete files progress = 0 files_to_delete = incomplete_pairs['file_name'].tolist() for i in range(0, len(files_to_delete)): os.remove(files_to_delete[i]) progress = float(i + 1) / float(len(files_to_delete)) print('\r|' + '#' * int(round(progress * 54)) + '-' * int(round((1 - progress) * 54)) + '| ' + str(int(round(progress * 100))) + '%'), progress = None print('\n') else: print('\nInvalid input.') # Status print('Merger complete.\n\n' + '#' * 80) # EOF
murfrosoft/pygame-demos
pygame02_Animation.py
<filename>pygame02_Animation.py import pygame # initalize pygame modules pygame.init() clock = pygame.time.Clock() # create a clock to control update speed # create a screen on which to draw width, height = 500, 400 screensize = [width, height] # [width, height] in pixels screen = pygame.display.set_mode(screensize) pygame.display.set_caption('demo02: Basic Annimation') # title of screen window # initialize font font = pygame.font.SysFont(None, 16) # font size = 16 # create a textbox: red text with white background text = font.render('Score: %d' %(0), True, (255,0,0), (255,255,255)) textRect = text.get_rect() textRect.topleft = screen.get_rect().topleft # import an image to move image = pygame.image.load(r'Images\fireball.png') topx = width//2 topy = height//2 xspeed = 1 yspeed = 1 score = 0 # ready to run game loop run = True while run: for event in pygame.event.get(): if event.type == pygame.QUIT: run = False if event.type == pygame.KEYDOWN and event.key == pygame.K_ESCAPE: run = False screen.fill((255,255,255)) # fill screen with white text = font.render('Score: %d' %(score), True, (0,0,0), (255,255,255)) screen.blit(text, textRect) # draw text to screen screen.blit(image, (topx, topy)) pygame.display.update() # update screen drawing # update positions topx += xspeed topy += yspeed if topx <= 0 or topx >= width - 32: xspeed *= -1 score += 1 if topy <= 0 or topy >= height - 32: yspeed *= -1 score += 1 clock.tick(60) # advance at 60 fps
murfrosoft/pygame-demos
pygame03_AnimationTranslation.py
<filename>pygame03_AnimationTranslation.py import pygame # initalize pygame modules pygame.init() clock = pygame.time.Clock() # create a clock to control update speed # create a screen on which to draw width, height = 500, 400 screensize = [width, height] # [width, height] in pixels screen = pygame.display.set_mode(screensize) pygame.display.set_caption('demo03: More Annimation') # title of screen window # initialize font font = pygame.font.SysFont(None, 16) # font size = 16 # create a textbox: red text with white background text = font.render('Score: %d' %(0), True, (255,0,0), (255,255,255)) textRect = text.get_rect() textRect.topleft = screen.get_rect().topleft # import an image to move image = pygame.image.load(r'Images\fireball.png') img_x = width//2 img_y = height//2 xspeed = 1 yspeed = 1 score = 0 angle = 0 # ready to run game loop run = True while run: for event in pygame.event.get(): if event.type == pygame.QUIT: run = False if event.type == pygame.KEYDOWN and event.key == pygame.K_ESCAPE: run = False screen.fill((255,255,255)) # fill screen with white text = font.render('Score: %d' %(score), True, (0,0,0), (255,255,255)) screen.blit(text, textRect) # draw text to screen fireball = pygame.transform.rotate(image,angle); img_rect = fireball.get_rect() img_rect.centerx = img_x img_rect.centery = img_y screen.blit(fireball, img_rect)#screen.blit(fireball, (img_x, img_y)) pygame.display.update() # update screen drawing # update positions img_x += xspeed img_y += yspeed angle -= 9 if angle <= 0: angle = 360 if img_x <= 16 or img_x >= width - 16: xspeed *= -1 score += 1 if img_y <= 16 or img_y >= height - 16: yspeed *= -1 score += 1 clock.tick(60) # advance at 60 fps
murfrosoft/pygame-demos
pygame04_ScrollingBackground.py
<reponame>murfrosoft/pygame-demos import pygame # initalize pygame modules pygame.init() clock = pygame.time.Clock() # create a clock to control update speed # create a screen on which to draw width, height = 500, 300 screensize = [width, height] # [width, height] in pixels screen = pygame.display.set_mode(screensize) pygame.display.set_caption('demo04: Scrolling background') # title of screen window # initialize font font = pygame.font.SysFont(None, 16) # font size = 16 # create a textbox: red text with white background text = font.render('Score: %d' %(0), True, (255,255,255), (0,0,0)) textRect = text.get_rect() textRect.topleft = screen.get_rect().topleft # import an image to move image_bg = pygame.image.load(r'Images\starbg.png') # ready to run game loop bg_x = 0 bg_speed = 1 score = 0 run = True while run: for event in pygame.event.get(): if event.type == pygame.QUIT: run = False if event.type == pygame.KEYDOWN and event.key == pygame.K_ESCAPE: run = False if event.type == pygame.KEYDOWN and event.key == pygame.K_UP: if bg_speed < 30: bg_speed += 1 if event.type == pygame.KEYDOWN and event.key == pygame.K_DOWN: if bg_speed > 1: bg_speed -= 1 screen.fill((255,255,255)) # fill screen with white text = font.render('Score: %d' %(score), True, (255,255,255), (0,0,0)) screen.blit(image_bg, (bg_x,0)) screen.blit(image_bg, (bg_x+600,0)) #screen.blit(fireball, (img_x, img_y)) screen.blit(text, textRect) # draw text to screen pygame.display.update() # update screen drawing bg_x = bg_x - bg_speed if bg_x <= -600: bg_x = 0 score += 1 clock.tick(60) # advance at 60 fps
murfrosoft/pygame-demos
pygame01_HelloWorld.py
<filename>pygame01_HelloWorld.py import pygame # initalize pygame modules pygame.init() clock = pygame.time.Clock() # create a clock to control update speed # create a screen on which to draw screensize = [300, 200] # [width, height] in pixels screen = pygame.display.set_mode(screensize) pygame.display.set_caption('demo01: Hello World') # title of screen window # initialize font font = pygame.font.SysFont(None, 40) # font size = 40 # create a textbox: red text with white background text = font.render(' Hello world!', True, (255,0,0), (255,255,255)) textRect = text.get_rect() textRect.centerx = screen.get_rect().centerx textRect.centery = screen.get_rect().centery # ready to run game loop red = 255 run = True while run: for event in pygame.event.get(): if event.type == pygame.QUIT: run = False if event.type == pygame.KEYDOWN and event.key == pygame.K_ESCAPE: run = False screen.fill((255,255,255)) # fill screen with white text = font.render('%003d Hello world!' %(red), True, (red,0,0), (255,255,255)) screen.blit(text, textRect) # draw text to screen pygame.display.update() # update screen drawing red -= 1 if red <= 0: red = 255 clock.tick(60) # advance at 60 fps
weynelucas/drf-nested-views
drf_nested_views/mixins.py
""" Basic building blocks for generic class based views. Use theese mixins with GenericAPIView from drf_nested_views. With exception of CreateModelMixin and UpdateModelMixin, other mixins does not perform or override any action of his parents (rest_framework.mixins). Use them for track future changes. """ from rest_framework import mixins class CreateModelMixin(mixins.CreateModelMixin): """ Create a nested model instance. """ def perform_create(self, serializer): serializer.save(**self.get_parent_lookup()) class ListModelMixin(mixins.ListModelMixin): """ List a nested queryset. """ pass class RetrieveModelMixin(mixins.RetrieveModelMixin): """ Retrieve a nested model instance. """ pass class UpdateModelMixin(mixins.UpdateModelMixin): """ Update a nested model instance. """ def perform_update(self, serializer): serializer.save(**self.get_parent_lookup()) class DestroyModelMixin(mixins.DestroyModelMixin): """ Destroy a nested model instance. """ pass
weynelucas/drf-nested-views
drf_nested_views/generics.py
<reponame>weynelucas/drf-nested-views from rest_framework.serializers import ModelSerializer from rest_framework import generics from rest_framework_nested.serializers import NestedHyperlinkedModelSerializer class GenericAPIView(generics.GenericAPIView): """ Base class for all other generic views that are nested. """ parent_lookup_kwargs = {} def get_queryset(self): """ Get the list of items for this view. This must be an iterable, and may be a queryset. If `serializer_class` attribute is a ModelSerializer subclass, the queryset will be retrieved through `model` attribute. """ if issubclass(self.serializer_class, ModelSerializer): return self.serializer_class.Meta.model.objects.all() return super(GenericAPIView, self).get_queryset() def get_parent_lookup_kwargs(self): """ Returns the parent lookup kwargs, a dictionary that maps URL parameters to object properties. Defaults to using `self.parent_lookup_kwargs`. If `serializer_class` attribute is a NestedHyperlinkedModelSerializer subclass, the kwargs will be retrieved from it through `parent_lookup_kwargs` attribute. """ if issubclass(self.serializer_class, NestedHyperlinkedModelSerializer): return self.serializer_class.parent_lookup_kwargs assert self.parent_lookup_kwargs, ( "'%s' should either include a `parent_lookup_kwargs` attribute, " "or override the `get_parent_lookup_kwargs()` method." % self.__class__.__name__ ) return self.parent_lookup_kwargs def get_parent_lookup(self): """ Returns the entire lookup for parent based on `parent_lookup_kwargs` and URL keyword arguments. """ parent_lookup_kwargs = self.get_parent_lookup_kwargs() try: return { v: self.kwargs[k] for k, v in parent_lookup_kwargs.items() } except KeyError as error: raise AttributeError( "Keyword argument %s from `parent_lookup_kwargs` does not " "have any match on URL keyword arguments." % str(error) ) def filter_queryset(self, queryset): """ Given a queryset, filter it with whichever filter backend is in use. After filter backend process, filter it with `parent_lookup_kwargs` attribute. """ queryset = super(GenericAPIView, self).filter_queryset(queryset) return queryset.filter(**self.get_parent_lookup())
dave-s477/SoMeNLP
somenlp/utils/time_marker.py
<gh_stars>0 import datetime def get_time_marker(format='full'): """Get current time as a formatted string. Args: format (string): full (with time) vs. short (only date) Returns: string: current time """ if format=='full': out_s = datetime.datetime.now().strftime("%d-%m-%Y_%H-%M-%S") else: out_s = datetime.datetime.now().strftime("%d-%m-%Y") return out_s
dave-s477/SoMeNLP
somenlp/entity_disambiguation/feature_calculator.py
<filename>somenlp/entity_disambiguation/feature_calculator.py import numpy as np import pandas as pd import time import re import torch import json import nltk nltk.download('stopwords') from itertools import combinations from nltk.corpus import stopwords from multiprocessing import Pool from os.path import join, exists from Levenshtein import distance as levenshtein_distance from Levenshtein import jaro from articlenizer import articlenizer MENTION_SCALING_FACTOR = 60 URL_SCALING_FACTOR = 120 DEVEL_SCALING_FACTOR = 30 VERSION_SCALING_FACTOR = 10 class EntityDisambiguationFeatureGenerator: """Calculates features for entity disambiguation """ def __init__(self, dbpedia): """Init Args: dbpedia (str): location of a pandas dataframe containing DBpedia information for software entities """ self.stops = stopwords.words('english') # self.dbpedia_names = pd.read_csv(dbpedia, compression='gzip') # self.dbpedia_names.drop(self.dbpedia_names.columns.difference(['unique','label']), 1, inplace=True) # self.dbpedia_names.dropna(inplace=True) # self.dbpedia_unique_mapping = {} # self.dbpedia_label_mapping = {} # for index, row in self.dbpedia_names.iterrows(): # if row['unique'] not in self.dbpedia_unique_mapping: # self.dbpedia_unique_mapping[row['unique']] = set() # self.dbpedia_unique_mapping[row['unique']].update([row['label']]) # if row['label'] not in self.dbpedia_label_mapping: # self.dbpedia_label_mapping[row['label']] = set() # self.dbpedia_label_mapping[row['label']].update([row['unique']]) with open(dbpedia, 'r') as j_in: self.dbpedia_data = json.load(j_in) #self.dbpedia_unique_mapping = dbpedia_data['unique_mapping'] #self.dbpedia_label_mapping = dbpedia_data['label_mapping'] self.string_features_to_extract = [ { # String based features 'name': 'LenFirst', 'function': self._len_first, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'LenSecond', 'function': self._len_second, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Jaro', 'function': self._jaro, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Levenshtein', 'function': self._levenshtein, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Substring', 'function': self._substring, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Norm_string_Jaro', 'function': self._norm_string_jaro, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Norm_string_Levenshtein', 'function': self._norm_string_levenshtein, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'KnownAbbr', 'function': self._known_abbreviation, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'MenGenAbbr', 'function': self._mention_generated_abbreviation, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'NormGenAbbr', 'function': self._norm_generated_abbreviation, 'data_type': np.float32, 'init': 1.0 } ] self.context_features_to_extract = [ { # Abbreviations and Alternative Names 'name': 'GivenAbbr', 'function': self._given_abbreviation, 'data_type': np.float32, 'init': 1.0 }, { # URLs 'name': 'URL_LenFirst', 'function': self._url_len_first, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'URL_LenSecond', 'function': self._url_len_second, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'URL_Jaro', 'function': self._url_jaro, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'URL_Substring', 'function': self._url_substring, 'data_type': np.float32, 'init': 1.0 }, { # Developers 'name': 'Devel_LenFirst', 'function': self._devel_len_first, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Devel_LenSecond', 'function': self._devel_len_second, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Devel_Jaro', 'function': self._devel_jaro, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Devel_Substring', 'function': self._devel_substring, 'data_type': np.float32, 'init': 1.0 }, { # Versions 'name': 'Version_LenFirst', 'function': self._version_len_first, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Version_LenSecond', 'function': self._version_len_second, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Version_Jaro', 'function': self._version_jaro, 'data_type': np.float32, 'init': 1.0 }, { 'name': 'Version_Substring', 'function': self._version_substring, 'data_type': np.float32, 'init': 1.0 } ] self.feature_length = len(self.string_features_to_extract) + len(self.context_features_to_extract) def _len_fct(self, s, factor): return len(s) / factor def _jaro_fct(self, s0, s1): return 1 - jaro(s0, s1) def _levenshtein_fct(self, s0, s1): return levenshtein_distance(s0, s1) / max(len(s0), len(s1)) def _substring_fct(self, s0, s1): return 1 - (s0 in s1 or s1 in s0) def _len_first(self, pair): return self._len_fct(pair[0]['string'], MENTION_SCALING_FACTOR) def _len_second(self, pair): return self._len_fct(pair[1]['string'], MENTION_SCALING_FACTOR) def _jaro(self, pair): """Jaro distance between entity mentions Args: pair (list): entity pair """ return self._jaro_fct(pair[0]['string'], pair[1]['string']) def _levenshtein(self, pair): """Levenshtein distance between entity mentions Args: pair (list): entity pair Returns: int: Levenshtein edit distance """ return self._levenshtein_fct(pair[0]['string'], pair[1]['string']) def _substring(self, pair): """Test substring relation between entity mentions Args: pair (list): entity pair Returns: bool: result """ return self._substring_fct(pair[0]['string'], pair[1]['string']) def _normalize(self, s): norm_s = re.sub('[^0-9a-zA-Z]+', ' ', s.casefold()).rstrip('0123456789 ,.').lstrip(' ') norm_s = ' '.join([w for w in norm_s.split() if w not in self.stops]) if not norm_s: norm_s = s return norm_s def _remove_spaces(self, s): replace_regex = re.compile(r'\s(?P<to_keep>[\+\-#™_/\d]+)\s?') matches = replace_regex.findall(s) return replace_regex.sub(r'\g<to_keep>', s) def _norm_string_jaro(self, pair): return self._jaro_fct(pair[0]['norm'], pair[1]['norm']) def _norm_string_levenshtein(self, pair): return self._levenshtein_fct(pair[0]['norm'], pair[1]['norm']) def _add_info_len(self, pair, idx, name, factor): num = 0 length = 0 for rel in pair[idx]['relations']: if rel['type'] == name: num += 1 length += len(rel['string']) length = length / num if num > 0 else 0 return length / factor def _add_info_jaro(self, pair, name): x_infos = [] for rel in pair[0]['relations']: if rel['type'] == name: x_infos.append(rel['string']) if not x_infos: return 1.0 y_infos = [] for rel in pair[1]['relations']: if rel['type'] == name: y_infos.append(rel['string']) if not y_infos: return 1.0 min_dist = 1.0 for i_x in x_infos: for i_y in y_infos: dist = self._jaro_fct(i_x, i_y) if dist < min_dist: min_dist = dist return min_dist def _add_info_substring(self, pair, name): x_infos = [] for rel in pair[0]['relations']: if rel['type'] == name: x_infos.append(rel['string']) if not x_infos: return 1 y_infos = [] for rel in pair[1]['relations']: if rel['type'] == name: y_infos.append(rel['string']) if not y_infos: return 1 for i_x in x_infos: for i_y in y_infos: if i_x in i_y or i_y in i_x: return 0 return 1 def _url_len_first(self, pair): return self._add_info_len(pair, 0, 'URL_of', URL_SCALING_FACTOR) def _url_len_second(self, pair): return self._add_info_len(pair, 1, 'URL_of', URL_SCALING_FACTOR) def _url_jaro(self, pair): return self._add_info_jaro(pair, 'URL_of') def _url_substring(self, pair): return self._add_info_substring(pair, 'URL_of') def _devel_len_first(self, pair): return self._add_info_len(pair, 0, 'Developer_of', DEVEL_SCALING_FACTOR) def _devel_len_second(self, pair): return self._add_info_len(pair, 1, 'Developer_of', DEVEL_SCALING_FACTOR) def _devel_jaro(self, pair): return self._add_info_jaro(pair, 'Developer_of') def _devel_substring(self, pair): return self._add_info_substring(pair, 'Developer_of') def _version_len_first(self, pair): return self._add_info_len(pair, 0, 'Version_of', VERSION_SCALING_FACTOR) def _version_len_second(self, pair): return self._add_info_len(pair, 1, 'Version_of', VERSION_SCALING_FACTOR) def _version_jaro(self, pair): return self._add_info_jaro(pair, 'Version_of') def _version_substring(self, pair): return self._add_info_substring(pair, 'Version_of') def _mention_generated_abbreviation(self, pair): men_str_x = pair[0]['string'].replace('-', ' ') men_abbr_x = ''.join([s[0] for s in men_str_x.split()]) if len(men_str_x.split()) > 2 else '' men_str_y = pair[1]['string'].replace('-', ' ') men_abbr_y = ''.join([s[0] for s in men_str_y.split()]) if len(men_str_y.split()) > 2 else '' if not men_abbr_x and not men_abbr_y: return 1.0 min_dist = min(self._jaro_fct(men_abbr_x, pair[1]['string']), self._jaro_fct(men_abbr_y, pair[0]['string']), self._jaro_fct(men_abbr_x, men_abbr_y)) return min_dist def _norm_generated_abbreviation(self, pair): men_str_x = pair[0]['norm'].replace('-', ' ') men_abbr_x = ''.join([s[0] for s in men_str_x.split()]) if len(men_str_x.split()) > 2 else '' men_str_y = pair[1]['norm'].replace('-', ' ') men_abbr_y = ''.join([s[0] for s in men_str_y.split()]) if len(men_str_y.split()) > 2 else '' if not men_abbr_x and not men_abbr_y: return 1.0 min_dist = min(self._jaro_fct(men_abbr_x, pair[1]['norm']), self._jaro_fct(men_abbr_y, pair[0]['norm']), self._jaro_fct(men_abbr_x, men_abbr_y)) return min_dist def _known_abbreviation(self, pair): x_altnames = set([pair[0]['string']]) if pair[0]['string'] in self.dbpedia_data['unique_mapping']: x_altnames.update(self.dbpedia_data['unique_mapping'][pair[0]['string']]) elif pair[0]['string'] in self.dbpedia_data['label_mapping']: for key in self.dbpedia_data['label_mapping'][pair[0]['string']]: x_altnames.update([key]) x_altnames.update(self.dbpedia_data['unique_mapping'][key]) y_altnames = set([pair[1]['string']]) if pair[1]['string'] in self.dbpedia_data['unique_mapping']: y_altnames.update(self.dbpedia_data['unique_mapping'][pair[1]['string']]) elif pair[1]['string'] in self.dbpedia_data['label_mapping']: for key in self.dbpedia_data['label_mapping'][pair[1]['string']]: y_altnames.update([key]) y_altnames.update(self.dbpedia_data['unique_mapping'][key]) min_dist = 1.0 for i_x in x_altnames: dist = self._jaro_fct(i_x, pair[1]['string']) if dist < min_dist: min_dist = dist for i_y in y_altnames: dist = self._jaro_fct(i_y, pair[0]['string']) if dist < min_dist: min_dist = dist for i_x in x_altnames: for i_y in y_altnames: dist = self._jaro_fct(i_x, i_y) if dist < min_dist: min_dist = dist return min_dist def _given_abbreviation(self, pair): x_infos = [] for rel in pair[0]['relations']: if rel['type'] in ['Abbreviation_of', 'AlternativeName_of']: x_infos.append(rel['string']) y_infos = [] for rel in pair[1]['relations']: if rel['type'] in ['Abbreviation_of', 'AlternativeName_of']: y_infos.append(rel['string']) min_dist = 1.0 for i_x in x_infos: dist = self._jaro_fct(i_x, pair[1]['string']) if dist < min_dist: min_dist = dist for i_y in y_infos: dist = self._jaro_fct(i_y, pair[0]['string']) if dist < min_dist: min_dist = dist for i_x in x_infos: for i_y in y_infos: dist = self._jaro_fct(i_x, i_y) if dist < min_dist: min_dist = dist return min_dist def features_for_pair(self, pair): results = [] for fct in self.string_features_to_extract: result = fct['function'](pair) results.append(result) for fct in self.context_features_to_extract: result = fct['function'](pair) results.append(result) #print(results) #print(torch.tensor(result).data) return results def features_for_pre_clustered(self, pair): string_results = [] #string_results = np.empty((len(self.string_features_to_extract))) for fct in self.string_features_to_extract: result = fct['function'](pair) string_results.append(result) #string_results[idx] = result context_results = [] #context_results = np.empty((len(pair[0]['contexts'])*len(pair[1]['contexts']), len(self.context_features_to_extract))) #count = 0 for context_first in pair[0]['contexts']: context_first['string'] = pair[0]['string'] for context_second in pair[1]['contexts']: context_second['string'] = pair[1]['string'] context_results.append(string_results.copy()) for fct in self.context_features_to_extract: result = fct['function']([context_first, context_second]) context_results[-1].append(result) #context_results[count, idx] += result #count += 1 return context_results#string_results, context_results def apply_features(self, entities, ncores=5): """ Applies an arbitrary number of features onto a list of entity names. The feature is calculated once for each possible pair of entity names. Arguments: entities (list): list of entity names """ matrices = {} for entity in entities: entity['string'] = self._remove_spaces(entity['mention']) entity['norm'] = self._normalize(entity['mention']) pairs_to_compare = list(combinations(entities, 2)) # for fct in self.features_to_extract: # start_time = time.time() # if ncores > 1: # with Pool(ncores) as p: # result = p.map(fct['function'], pairs_to_compare) # matrices[fct['name']] = result # else: # result = np.full((sum(range(len(entities)))), fct['init'], dtype=fct['data_type']) # for idx, pair in enumerate(pairs_to_compare): # distance = fct['function'](pair) # result[idx] = distance # matrices[fct['name']] = result # end_time = time.time() # print("It took {} seconds to calculate feature {} for {} inputs.".format(round(end_time-start_time, 3), fct['name'], len(entities))) for fct in self.string_features_to_extract: start_time = time.time() if ncores > 1: with Pool(ncores) as p: result = p.map(fct['function'], pairs_to_compare) matrices[fct['name']] = result else: result = np.full((sum(range(len(entities)))), fct['init'], dtype=fct['data_type']) for idx, pair in enumerate(pairs_to_compare): distance = fct['function'](pair) result[idx] = distance matrices[fct['name']] = result end_time = time.time() print("It took {} seconds to calculate feature {} for {} inputs.".format(round(end_time-start_time, 3), fct['name'], len(entities))) for fct in self.context_features_to_extract: start_time = time.time() if ncores > 1: with Pool(ncores) as p: result = p.map(fct['function'], pairs_to_compare) matrices[fct['name']] = result else: result = np.full((sum(range(len(entities)))), fct['init'], dtype=fct['data_type']) for idx, pair in enumerate(pairs_to_compare): distance = fct['function'](pair) result[idx] = distance matrices[fct['name']] = result end_time = time.time() print("It took {} seconds to calculate feature {} for {} inputs.".format(round(end_time-start_time, 3), fct['name'], len(entities))) return matrices def get_labels(self, entities, gold): for g in gold: g_sentences = articlenizer.get_tokenized_sentences(g['sentence']) g['prosentence'] = [' '.join(s) for s in g_sentences] g['proconcat'] = ' '.join(g['prosentence']) for ent in entities: for idx, g in enumerate(gold): if ent['paper_id'] == g['paper_id'] and any([ent['sentence'] == s for s in g['prosentence']]) or ent['sentence'] == g['proconcat']: if ent['mention'] == ' '.join(articlenizer.get_tokenized_sentences(g['mention'])[0]): ent['gold_id'] = idx break pairs_to_compare = list(combinations(entities, 2)) start_time = time.time() result = np.full((sum(range(len(entities)))), 0, dtype=np.int16) for idx, pair in enumerate(pairs_to_compare): if 'gold_id' in pair[0] and 'gold_id' in pair[1] and gold[pair[0]['gold_id']]['link'] == gold[pair[1]['gold_id']]['link']: result[idx] = 1 end_time = time.time() print("It took {} seconds to get labels for {} inputs.".format(round(end_time-start_time, 3), len(entities))) return result
dave-s477/SoMeNLP
somenlp/feature_engineering/candidate_rules.py
<reponame>dave-s477/SoMeNLP first_char_lower = lambda s: s[:1].lower() + s[1:] if s else '' def pan_top_1(x): '''use <> software''' left_context = ['use'] right_context = ['software'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=1, style='lemma') right_lemmas = x.sentence.get_right_tokens(x.end_idx, size=1, style='lemma') if not left_lemmas or not right_lemmas or len(left_context) != len(left_lemmas) or len(right_context) != len(right_lemmas): return False for cont, feat in zip(left_context, left_lemmas): if cont != feat: return False for cont,feat in zip(right_context, right_lemmas): if cont != feat: return False return True def pan_top_2(x): '''perform use <>''' left_context = ['perform', 'use'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=2, style='lemma') if len(left_context) != len(left_lemmas): return False for c,l in zip(left_context, left_lemmas): if c != l: return False return True def pan_top_3(x): '''be perform use <>''' left_context = ['be', 'perform', 'use'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=3, style='lemma') if len(left_context) != len(left_lemmas): return False for c,l in zip(left_context, left_lemmas): if c != l: return False return True def pan_top_4(x): '''analysis be perform use <>''' left_context = ['analysis', 'be', 'perform', 'use'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=4, style='lemma') if len(left_context) != len(left_lemmas): return False for c,l in zip(left_context, left_lemmas): if c != l: return False return True def pan_top_5(x): '''analyze use <>''' left_context_ae = ['analyze', 'use'] left_context_be = ['analyse', 'use'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=2, style='lemma') if len(left_context_ae) != len(left_lemmas): return False for c_ae, c_be, l in zip(left_context_ae, left_context_be, left_lemmas): if c_ae != l and c_be != l: return False return True def pan_top_6(x): '''analysis be perform with <>''' left_context = ['analysis', 'be', 'perform', 'with'] left_lemmas = x.sentence.get_left_tokens(x.start_idx, size=4, style='lemma') if len(left_context) != len(left_lemmas): return False for c,l in zip(left_context, left_lemmas): if c != l: return False return True def pan_top_7(x): '''<> statistical software''' right_context = ['statistical', 'software'] right_lemmas = x.sentence.get_right_tokens(x.end_idx, size=2, style='lemma') if not right_lemmas or len(right_context) != len(right_lemmas): return False for c,l in zip(right_context, right_lemmas): if c != l: return False return True def pan_top_8(x): '''<> software be use''' right_context = ['software', 'be', 'use'] right_lemmas = x.sentence.get_right_tokens(x.end_idx, size=3, style='lemma') if not right_lemmas or len(right_context) != len(right_lemmas): return False for c,l in zip(right_context, right_lemmas): if c != l: return False return True def text_is_in_brackets(x): left_context = x.sentence.get_left_tokens(x.start_idx, 1, style='plain') right_context = x.sentence.get_right_tokens(x.end_idx, 1, style='plain') if len(x.base_span) != 1 and left_context and right_context and left_context[0] in ['(', '[', '{'] and right_context[0] in [')', ']', '}']: return True else: return False def developer(x): '''software developer match''' left_context = x.sentence.get_left_tokens(x.start_idx, 1, style='plain') right_context = x.sentence.get_right_tokens(x.end_idx, 1, style='plain') if len(x.base_span) != 1 and left_context and right_context and left_context[0] == '(' and right_context[0] == ')': for tok in x.tokens: if tok.lower() in ['inc.', 'ltd.', 'corp.', 'apply', 'inc', 'ltd', 'corp']: return True return False
dave-s477/SoMeNLP
somenlp/feature_engineering/word_rules.py
def upper_cased(x): if x.isalpha() and x.isupper(): return True else: return False def first_char_upper(x): if x.isalpha() and x[0].isupper(): if all(c.islower() for c in x[1:]): return True return False def mixed_case(x): if x.isalpha(): if any([c.isupper() for c in x[1:]]) and any([c.islower() for c in x]): return True return False def lower_case(x): if x.isalpha() and x.islower(): return True else: return False
dave-s477/SoMeNLP
somenlp/utils/gpu_setup.py
<gh_stars>0 import torch import subprocess, re # Nvidia-smi GPU memory parsing. # Tested on nvidia-smi 370.23 def run_command(cmd): """Run command, return output as string.""" output = subprocess.Popen(cmd, stdout=subprocess.PIPE, shell=True).communicate()[0] return output.decode("ascii") def list_available_gpus(): """Returns list of available GPU ids.""" output = run_command("nvidia-smi -L") # lines of the form GPU 0: TITAN X gpu_regex = re.compile(r"GPU (?P<gpu_id>\d+):") result = [] for line in output.strip().split("\n"): m = gpu_regex.match(line) assert m, "Couldnt parse "+line result.append(int(m.group("gpu_id"))) return result def gpu_memory_map(): """Returns map of GPU id to memory allocated on that GPU.""" output = run_command("nvidia-smi") gpu_output = output[output.find("GPU Memory"):] # lines of the form # | 0 8734 C python 11705MiB | memory_regex = re.compile(r"[|]\s+?(?P<gpu_id>\d+)\D+?(?P<pid>\d+).+[ ](?P<gpu_memory>\d+)MiB") rows = gpu_output.split("\n") result = {gpu_id: 0 for gpu_id in list_available_gpus()} for row in gpu_output.split("\n"): m = memory_regex.search(row) if not m: continue gpu_id = int(m.group("gpu_id")) gpu_memory = int(m.group("gpu_memory")) result[gpu_id] += gpu_memory return result def pick_gpu_lowest_memory(): """Returns GPU with the least allocated memory""" memory_gpu_map = [(memory, gpu_id) for (gpu_id, memory) in gpu_memory_map().items()] best_memory, best_gpu = sorted(memory_gpu_map)[0] return best_gpu def setup_cuda(gpu): if gpu and torch.cuda.device_count() > 0: GPU = str(pick_gpu_lowest_memory()) print("Working on GPU: {}".format(GPU)) device = torch.device("cuda:{}".format(GPU)) else: print("Working on CPU") device = torch.device("cpu") return device
dave-s477/SoMeNLP
somenlp/NER/models/combined_lstm.py
<gh_stars>0 import torch import torch.nn as nn from .crf import CRF class CombinedLSTM(nn.Module): def __init__(self, device, char_size, word_size, tag2idx, emb_vecs, char_config, emb_config, drop_config, gen_config, **kwargs): super(CombinedLSTM, self).__init__() self.device = device self.char_size = char_size self.word_size = word_size self.tag2idx = tag2idx self.tag_size = len(tag2idx) self.char_config = char_config self.emb_config = emb_config self.drop_config = drop_config self.gen_config = gen_config if not 'feature_dim' in kwargs: raise(RuntimeError("FeatureLSTM requires feature_dim to be given.")) self.feature_dim = kwargs['feature_dim'] self.feature_projection = self.gen_config['feature_emb_dim'] > 0 self.char_embeds = nn.Embedding(self.char_size, self.char_config['emb_dim']) self.char_lstm = nn.LSTM(self.char_config['emb_dim'], self.char_config['hidden_dim'], num_layers=self.char_config['layers'], bidirectional=True) self.word_embeds = nn.Embedding.from_pretrained(emb_vecs) if not self.emb_config['train']: self.word_embeds.requires_grad = False if self.feature_projection: self.feature_remap = nn.Linear(self.feature_dim, self.gen_config['feature_emb_dim']) feature_out_dim = self.gen_config['feature_emb_dim'] else: feature_out_dim = self.feature_dim self.lstm = nn.ModuleList() for i in range(self.gen_config['layers']): input_size = ( self.emb_config['dim'] + (self.char_config['hidden_dim'] * 2) + feature_out_dim ) if i == 0 else self.gen_config['hidden_dim'] * 2 self.lstm.append(nn.LSTM(input_size, self.gen_config['hidden_dim'], num_layers=1, bidirectional=True)) self.hidden2tag = nn.Linear(self.gen_config['hidden_dim'] * 2 * self.gen_config['layers'], self.tag_size) self.crf = CRF(self.tag_size, device, init_parameters=None) self.drop_char_emb = nn.Dropout(self.drop_config['char_emb']) self.drop_char_feat = nn.Dropout(self.drop_config['char_feat']) self.drop_word_emb = nn.Dropout(self.drop_config['word_emb']) self.drop_lstm_feat = nn.Dropout(self.drop_config['lstm_feat']) self.drop_dense_feat = nn.Dropout(self.drop_config['dense_feat']) self.drop_features_in = nn.Dropout(self.drop_config['in_feat']) self.drop_features_proj = nn.Dropout(self.drop_config['proj_feat']) if self.gen_config['zero_init']: self.hidden = self._init_hidden_zero() self.char_hidden = self._init_hidden_zero() else: self.hidden = self._init_hidden_xavier_norm() self.char_hidden = self._init_hidden_xavier_norm() # init hidden idea: # create param -> copy over batch -> concat with input length -> input to dense network # -> use as initial lstm state... def _init_hidden_zero(self, batch_size=1, hidden_dim=100): """Randomly initializes the hidden state of the bi-lstm. Args: tag_indices: A [batch_size, max_seq_len] matrix of tag indices. Returns: (float tensor, float tensor): Random initial states for forward and backward cells """ return (torch.zeros(2, batch_size, hidden_dim).to(self.device), torch.zeros(2, batch_size, hidden_dim).to(self.device)) def _init_hidden_xavier_norm(self, batch_size=1, hidden_dim=100): """Randomly initializes the hidden state of the bi-lstm. Args: tag_indices: A [batch_size, max_seq_len] matrix of tag indices. Returns: (float tensor, float tensor): Random initial states for forward and backward cells """ return (nn.init.xavier_normal_(torch.empty(2, batch_size, hidden_dim).to(self.device)), nn.init.xavier_normal_(torch.empty(2, batch_size, hidden_dim).to(self.device))) def _get_character_features(self, char_sentence): """Create character based features by applying a bi-lstm on a char embedding Args: char_sentence: A [batch_size, max_seq_len, max_word_len] tensor of char indices. Returns: char_features: A [batch_size, max_seq_len, feat_len] tensor of char features for each word """ original_size = char_sentence.size() if self.gen_config['zero_init']: self.char_hidden = self._init_hidden_zero(original_size[0] * original_size[1], self.char_config['hidden_dim']) else: self.char_hidden = self._init_hidden_xavier_norm(original_size[0] * original_size[1], self.char_config['hidden_dim']) # Get character embedding char_embeds = self.char_embeds(char_sentence) char_embeds = self.drop_char_emb(char_embeds) char_embeds = char_embeds.view(original_size[0] * original_size[1], original_size[2], char_embeds.size()[-1]).permute(1, 0, 2) # Get char bi-lstm features char_lstm_out, self.char_hidden = self.char_lstm(char_embeds, self.char_hidden) concat_lstm_hidden = torch.cat((self.char_hidden[0][0], self.char_hidden[0][1]), dim=-1) concat_lstm_hidden = concat_lstm_hidden.view(original_size[0], original_size[1], -1) concat_lstm_hidden = self.drop_char_feat(concat_lstm_hidden) return concat_lstm_hidden def _get_lstm_features(self, sentence, char_feats, features, sequence_lengths): """Create word based features by applying a bi-lstm on a sentence Args: sentence: A [batch_size, max_seq_len] tensor of word indices for a word emb lookup char_feats: A [batch_size, max_seq_len, char_feat_len] tensor of char features. Returns: lstm_feats: A [batch_size, max_seq_len, lstm_feat_len] tensor of lstm features. """ batch_size = sentence.size()[0] # Get word embeddings embeds = self.word_embeds(sentence) embeds = self.drop_word_emb(embeds) given_features = self.drop_features_in(features) if self.feature_projection: given_features = self.feature_remap(given_features) given_features = self.drop_features_proj(given_features) # Combine word and character features and mask inputs concat_feats = torch.cat((embeds, char_feats, given_features), dim=-1) masks = torch.arange(sentence.shape[1]).to(self.device)[None, :] < sequence_lengths[:, None] masks = masks.squeeze(1).long().unsqueeze(2) concat_feats = torch.mul(concat_feats, masks).permute(1, 0, 2) # Get bi-lstm features lstm_outputs = [] for i in range(self.gen_config['layers']): if self.gen_config['zero_init']: self.hidden = self._init_hidden_zero(batch_size, self.gen_config['hidden_dim']) else: self.hidden = self._init_hidden_xavier_norm(batch_size, self.gen_config['hidden_dim']) concat_feats, self.hidden = self.lstm[i](concat_feats, self.hidden) lstm_outputs.append(concat_feats.permute(1, 0, 2)) lstm_out = torch.cat(lstm_outputs, -1) lstm_out = self.drop_lstm_feat(lstm_out) return lstm_out def get_features(self, char_sentence, sentence, features, lengths): char_feats = self._get_character_features(char_sentence) word_feats = self._get_lstm_features(sentence, char_feats, features, lengths) # Dense transform to num of tags feats = self.hidden2tag(word_feats) feats = self.drop_dense_feat(feats) return feats def neg_log_likelihood(self, tags, **features): """Calculates log_likelihood for training a bi-lstm-crf Args: char_sentence: A [batch_size, max_seq_len, max_word_len] tensor of char indices. sentence: A [batch_size, max_seq_len] tensor of word indices tags: A [batch_size, max_seq_len] tensor of ground-truth tags. lengths: A [batch_size, 1] tensor of actual sequence lengths for the padded input. Returns: log_likelihood: A scalar value for the log_likelihood over the entire batch """ # Getting the features feats = self.get_features(features['char_sentence'], features['sentence'], features['feature_sentence'], features['lengths']) # Getting the CRF score log_likelihood = self.crf.crf_log_likelihood(feats, tags, features['lengths'].squeeze(1)) if self.gen_config['crf_norm']: log_likelihood = log_likelihood / features['lengths'].squeeze(-1) tag_weights = tags != self.tag2idx['O'] tag_weights = torch.where(tag_weights.any(1), torch.tensor(self.gen_config['sample_weight']).to(self.device), torch.tensor(1.0).to(self.device)) log_likelihood = log_likelihood * tag_weights neg_log_likelihood = torch.mean(-log_likelihood) return neg_log_likelihood def forward(self, **features): """Calculates tag sequence and its score based on the bi-lstm-crf Args: char_sentence: A [batch_size, max_seq_len, max_word_len] tensor of char indices. sentence: A [batch_size, max_seq_len] tensor of word indices lengths: A [batch_size, 1] tensor of actual sequence lengths for the padded input. Returns: log_likelihood: A scalar value for the log_likelihood over the entire batch """ # Getting the features feats = self.get_features(features['char_sentence'], features['sentence'], features['feature_sentence'], features['lengths']) # Find the best path, given the features. tag_seq, seq_score, seq_mask = self.crf.viterbi_decode_batch(feats, features['lengths'].squeeze(1)) return tag_seq, seq_score, seq_mask
dave-s477/SoMeNLP
somenlp/distant_supervision/perform_wiki_data_queries.py
import re import json import requests import xml.etree.ElementTree as ET from bs4 import BeautifulSoup from urllib.parse import quote_plus WIKIDATA_ID = re.compile(r'Q\d{5,}') def parse_query(plain_query, lang, default_address='https://query.wikidata.org/bigdata/namespace/wdq/sparql?query='): """Encode Wikidata query to pass it as an URL Args: plain_query (str): plain text query lang (str): wikidata language identifier for the query default_address (str, optional): url for wikidata sparql interface. Defaults to 'https://query.wikidata.org/bigdata/namespace/wdq/sparql?query='. Returns: str: encoded query """ plain_query = re.sub("'en'", '"' + lang + '"', plain_query) encoded_query = quote_plus(plain_query, safe=r"()\{\}") encoded_query = re.sub(r"\+", "%20", encoded_query) complete_url = '{}{}'.format(default_address, encoded_query) return complete_url def execute_query(query_url): """Perform wikidata query. Might not be handled if no user agent is provided. Args: query_url (str): url query to execute Returns: str: json formatted response text """ #user_agent = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_3) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/35.0.1916.47 Safari/537.36' user_agent = '' headers = { 'Accept': 'json', 'User-Agent': user_agent } response = requests.get(query_url, headers=headers) if response.status_code != 200: print("Received HTTP-Statuscode " + str(response.status_code)) if response.status_code == 400: print("Error is probably in the query syntax.") sys.exit(1) elif response.status_code == 429: print("Too many requests were sent: ") print("Sleeping: " + int(response.headers["Retry-After"])) time.sleep(int(response.headers["Retry-After"])) print("Continuing with query") response = requests.get(query_url, headers=headers) if response.status_code != 200: raise(RuntimeError("Got another error while querying wikidata {}\nShutting down..".format(response.status_code))) else: return response.text else: return response.text def query_wikidata(query_config): """Perform a series of wikidata queries based on a given configuration Args: query_config (dictionary): configuration of queries saved as json Returns: dictionary: merged responses for given queries """ results = {} with open(query_config) as qc: queries = json.load(qc) for query_name, query_data in queries.items(): print("Performing wikidata queries for {}".format(query_name)) target_main_names = '{}_main_name'.format(query_data['target']) target_alt_names = '{}_alt_name'.format(query_data['target']) if target_main_names not in results.keys(): results[target_main_names] = set() results[target_alt_names] = set() for lang in query_data['languages']: count = 0 query_string = parse_query(query_data["query_string"], lang) query_response = execute_query(query_string) soup = BeautifulSoup(query_response, 'lxml') for res in soup.findAll('result'): for bind in res.findAll('binding'): count += 1 if bind['name'] == 'itemLabel': results[target_main_names].update([bind.text.rstrip().lstrip()]) elif bind['name'] == 'abbreviation': results[target_alt_names].update([bind.text.rstrip().lstrip()]) print("Processed {} entries for lang {}".format(count, lang)) return results
dave-s477/SoMeNLP
somenlp/NER/data_handler.py
import re import torch import torch.nn as nn import torch.nn.functional as F import numpy as np import json import math import random import copy from torch.nn.utils.rnn import pad_sequence from torch.utils.data import DataLoader, RandomSampler, SequentialSampler from pathlib import Path from gensim.models import KeyedVectors from itertools import zip_longest from articlenizer import articlenizer as art from transformers import BertTokenizer from .LSTM_dataset import LSTMDataset from .BERT_dataset import BERTDataset, BERTMultiDataset BERT_MAX_LENGTH = 256 class DataHandler(): def __init__(self, data_config=None, data_file_extension='.data.txt', label_file_extension='.labels.txt', feature_file_extension='', relation_file_extension='', output_handler=None, checkpoint=None, padding='<PAD>', unknown='<UNK>', batch_size=32, max_word_length=-1, max_sent_length=-1, data_files=None, prepro=False, tokenizer=None, multi_task=False): self.data_config = copy.deepcopy(data_config) self.data_file_extension = data_file_extension self.label_file_extension = label_file_extension self.feature_file_extension = feature_file_extension self.relation_file_extension = relation_file_extension self.checkpoint = checkpoint self.padding = padding self.unknown = unknown self.output_handler = output_handler self.batch_size = batch_size self.max_word_length = max_word_length self.max_sent_length = max_sent_length self.data_files = data_files self.prepro = prepro self.feature_dim = None self.multi_task_mapping = multi_task self.data = [] self.features = [] self.labels = [] self._load_tag_remapping() self._load_relation_remapping() self.tokenizer = tokenizer if tokenizer is not None: self._setup_tokenizer(tokenizer) def _setup_tokenizer(self, location): self.tokenizer = BertTokenizer.from_pretrained(location, do_lower_case=False) self.padding = "[PAD]" self.special_toks = { 'pad_tok': self.tokenizer.vocab["[PAD]"], 'sep_tok': self.tokenizer.vocab["[SEP]"], 'cls_tok': self.tokenizer.vocab["[CLS]"] } def _load_tag_remapping(self): if self.data_config is not None and 'transform' in self.data_config and 'mapping' in self.data_config['transform'] and self.data_config['transform']['mapping']: print("Loading tag remapping: {}".format(self.data_config['transform']['mapping'])) mapping = Path(self.data_config['transform']['mapping']) with mapping.open(mode='r') as mapping_j: self.tag_remapping = json.load(mapping_j) elif self.data_config is not None and 'transform' in self.data_config and len(self.data_config['transform']) > 0: print("Loading tag mappings for a multi-label tagging problem") self.tag_remapping = {} self.multi_task_mapping = True for k, v in self.data_config['transform'].items(): with Path(v).open(mode='r') as mapping_j: self.tag_remapping[k] = json.load(mapping_j) else: self.tag_remapping = None def _load_relation_remapping(self): if self.data_config is not None and 'transform' in self.data_config and 'relation_mapping' in self.data_config['transform'] and self.data_config['transform']['relation_mapping']: print("Loading relation remapping: {}".format(self.data_config['transform']['relation_mapping'])) mapping = Path(self.data_config['transform']['relation_mapping']) with mapping.open(mode='r') as mapping_j: self.relation_remapping = json.load(mapping_j) else: self.relation_remapping = None def _adjust_tag(self, tag, key=''): if self.tag_remapping is None or tag == 'O': return tag else: tag_prefix, tag_name = tag.split('-') if not key: remapped_tag = self.tag_remapping[tag_name] else: remapped_tag = self.tag_remapping[key][tag_name] if remapped_tag == 'O': return 'O' else: return '{}-{}'.format(tag_prefix, remapped_tag) def stream_files(self): for f_conf in self.data_files: if self.prepro: text = self._prepro_text_file(f_conf['in']) else: plain_text = self._read_text_file(f_conf['in']) text = [] for line in plain_text: text.append(line.split()) if 'feat' in f_conf: features = self._read_feature_file(f_conf['feat']) else: features = [] if self.tokenizer is None: characters, ids, tags, features = self._prepare_prediction_data(text, features) input_data = LSTMDataset(characters, ids, tags, features, self.encoding['char2idx'], self.padding, self.max_word_length, self.max_sent_length) sampler = SequentialSampler(input_data) data_loader = DataLoader(input_data, sampler=sampler, batch_size=self.batch_size, collate_fn=self._collate_fn) else: ids, tags, masks, lengths = self._prepare_bert_prediction_data(text) ids = self._pad_to_length(ids, length=BERT_MAX_LENGTH, fill_value=self.special_toks['pad_tok'], dtype=torch.long) masks = self._pad_to_length(masks, length=BERT_MAX_LENGTH, fill_value=0, dtype=torch.long) if not self.multi_task_mapping: tags = self._pad_to_length(tags, length=BERT_MAX_LENGTH, fill_value=self.encoding['tag2idx']['O'], dtype=torch.long) input_data = BERTDataset(ids, tags, masks) else: for k in tags.keys(): tags[k] = self._pad_to_length(tags[k], length=BERT_MAX_LENGTH, fill_value=self.encoding['tag2idx'][k]['O'], dtype=torch.long) input_data = BERTMultiDataset(ids, tags, masks, lengths) sampler = SequentialSampler(input_data) data_loader = DataLoader(input_data, sampler=sampler, batch_size=self.batch_size) yield [{'out': f_conf['out'], 'out-text': f_conf['out-text']}, data_loader, text] def load_data_from_config(self): for dataset, dataset_setup in self.data_config['sets'].items(): for sub_dataset in dataset_setup: sub_dataset['all_files'] = [] for folder in sub_dataset['folder']: files = list(Path(folder).rglob('*{}'.format(self.data_file_extension))) for f in files: base_file_name = f.name.split(self.data_file_extension)[0] sample_files = {} if self.data_file_extension: f_entry = Path('{}/{}{}'.format(str(f.parents[0]), base_file_name, self.data_file_extension)) sample_files[self.data_file_extension] = f_entry if self.label_file_extension: f_entry = Path('{}/{}{}'.format(str(f.parents[0]), base_file_name, self.label_file_extension)) if not f_entry.is_file(): raise(RuntimeError("Label file {} not found".format(f_entry))) sample_files[self.label_file_extension] = f_entry if self.feature_file_extension: f_entry = Path('{}/{}{}'.format(str(f.parents[0]), base_file_name, self.feature_file_extension)) if not f_entry.is_file(): raise(RuntimeError("Feature file {} not found".format(f_entry))) sample_files[self.feature_file_extension] = f_entry if self.relation_file_extension: f_entry = Path('{}/{}{}'.format(str(f.parents[0]), base_file_name, self.relation_file_extension)) if not f_entry.is_file(): raise(RuntimeError("Relation file {} not found".format(f_entry))) sample_files[self.relation_file_extension] = f_entry sub_dataset['all_files'].append(sample_files) def encoding(self, tags_only=False): if self.checkpoint is not None and self.checkpoint['model']: print("Loading given encodings") self.encoding = self.output_handler.load_encoding() for k, v in self.encoding.items(): if self.multi_task_mapping and k.endswith('tag2name'): for sk, vk in v.items(): vk_new = {int(key): value for key, value in vk.items()} self.encoding[k][sk] = vk_new elif k.endswith('name'): v_new = {int(key): value for key, value in v.items()} self.encoding[k] = v_new else: print("Generating new encodings") tag2idx, tag2name, word2idx, word2name, char2idx, char2name = {}, {}, {}, {}, {}, {} if not tags_only: for dataset, dataset_setup in self.data_config['sets'].items(): for sub_dataset in dataset_setup: for f in sub_dataset['all_files']: with f[self.data_file_extension].open(mode='r') as in_f: for line in in_f: for word in line.rstrip().split(): if word not in word2idx: word2name[len(word2idx)] = word word2idx[word] = len(word2idx) for char in word: if char not in char2idx: char2name[len(char2idx)] = char char2idx[char] = len(char2idx) if self.multi_task_mapping: print("Considering a multi-task problem") for dataset, dataset_setup in self.data_config['sets'].items(): for sub_dataset in dataset_setup: for f in sub_dataset['all_files']: with f[self.label_file_extension].open(mode='r') as in_f: for line in in_f: for tag in line.rstrip().split(): for mapping_key in self.tag_remapping.keys(): if mapping_key not in tag2idx: tag2idx[mapping_key] = {} tag2name[mapping_key] = {} t = self._adjust_tag(tag, mapping_key) if tags_only and '-' in t: t = t.split('-')[-1] if t not in tag2idx[mapping_key]: tag2name[mapping_key][len(tag2idx[mapping_key])] = t tag2idx[mapping_key][t] = len(tag2idx[mapping_key]) for k, v in tag2idx.items(): to_add = set() for tk, vk in v.items(): if tk.startswith('B-') and 'I-' + tk.split('B-')[-1] not in v: to_add.update(['I-' + tk.split('B-')[-1]]) for i in to_add: tag2name[k][len(tag2idx[k])] = i tag2idx[k][i] = len(tag2idx[k]) else: print("considering a single-task problem") for dataset, dataset_setup in self.data_config['sets'].items(): for sub_dataset in dataset_setup: for f in sub_dataset['all_files']: with f[self.label_file_extension].open(mode='r') as in_f: for line in in_f: for tag in line.rstrip().split(): t = self._adjust_tag(tag) if tags_only and '-' in t: t = t.split('-')[-1] if t not in tag2idx: tag2name[len(tag2idx)] = t tag2idx[t] = len(tag2idx) if not tags_only: word2name[len(word2idx)] = self.padding word2idx[self.padding] = len(word2idx) char2name[len(char2idx)] = self.padding char2idx[self.padding] = len(char2idx) word2name[len(word2idx)] = self.unknown word2idx[self.unknown] = len(word2idx) char2name[len(char2idx)] = self.unknown char2idx[self.unknown] = len(char2idx) self.encoding = { 'tag2idx': tag2idx, 'tag2name': tag2name, 'word2idx': word2idx, 'word2name': word2name, 'char2idx': char2idx, 'char2name': char2name } self.output_handler.save_json(self.encoding, name='encoding') def _prepro_text_file(self, path): with path.open(mode='r') as in_f: text_in = in_f.read() text_prepro = art.get_tokenized_sentences(text_in) return text_prepro def _read_text_file(self, path, read_empty=False): text = [] with path.open(mode='r') as in_f: for line in in_f: clean_line = line.rstrip() if clean_line: text.append(clean_line) elif read_empty: text.append([]) return text def _read_feature_file(self, path): features = np.load(str(path), allow_pickle=True) if self.feature_dim is None: self.feature_dim = features['features'][0].shape[-1] return features['features'] def _read_relation_file(self, path): relations = [] with path.open(mode='r') as in_f: for line in in_f: if not line.rstrip(): relations.append([]) else: sentence_rels = [] rel_strings = line.split(';;') for rel in rel_strings: if rel.rstrip(): rel_type, ent_1_str, ent_1_beg, ent_1_num, ent_2_str, ent_2_beg, ent_2_num = rel.split('\t') if self.relation_remapping is not None: rel_type = self.relation_remapping[rel_type] sentence_rels.append({ 'type': rel_type, 'ent1_s': ent_1_str, 'ent1_b': ent_1_beg, 'ent1_n': ent_1_num, 'ent2_s': ent_2_str, 'ent2_b': ent_2_beg, 'ent2_n': ent_2_num }) relations.append(sentence_rels) return relations def _collate_fn(self, batch): characters = [x['characters'] for x in batch] ids = [x['ids'] for x in batch] tags = [x['tags'] for x in batch] features = [x['features'] for x in batch] max_word_length_batch = 0 if self.max_word_length > 0: max_word_length_batch = self.max_word_length else: max_word_length_batch = 0 for sent in characters: if sent.size()[-1] > max_word_length_batch: max_word_length_batch = sent.size()[-1] length = torch.tensor([x.size() for x in ids]) padded_ids = pad_sequence(ids, batch_first=True, padding_value=self.encoding['word2idx'][self.padding]) padded_tags = pad_sequence(tags, batch_first=True, padding_value=self.encoding['tag2idx']['O']) max_sent_length_batch = 0 if self.max_sent_length > 0: padded_ids = F.pad(padded_ids, (0, self.max_sent_length-padded_ids.shape[-1]), "constant", self.encoding['word2idx'][self.padding]) padded_tags = F.pad(padded_tags, (0, self.max_sent_length-padded_tags.shape[-1]), "constant", self.encoding['tag2idx']['O']) characters = [F.pad(x, (0,(max_word_length_batch-x.size()[-1]),0,(self.max_sent_length-x.size()[0])), "constant", self.encoding['char2idx'][self.padding]).unsqueeze(0) for x in characters] else: characters = [F.pad(x, (0,(max_word_length_batch-x.size()[-1]),0,(padded_ids.size()[-1]-x.size()[0])), "constant", self.encoding['char2idx'][self.padding]).unsqueeze(0) for x in characters] if features[0] is not None: padded_features = pad_sequence(features, batch_first=True) if self.max_sent_length > 0: padded_features = F.pad(padded_features, (0, 0, 0, self.max_sent_length-padded_features.shape[-2]), "constant", 0) else: padded_features = None characters = torch.cat(characters, dim=0) return { 'chars': characters, 'ids': padded_ids, 'tags': padded_tags, 'lengths': length, 'features': padded_features } def _prepare_prediction_data(self, sentences, features, tags=[]): character_ids = [] input_ids = [] tag_ids = [] feat_ids = [] for sentence, feat, tag in zip_longest(sentences, features, tags, fillvalue=[]): # TODO how to handle unknowns? character_sentence = [[self.encoding['char2idx'][i] if i in self.encoding['char2idx'] else self.encoding['char2idx'][self.unknown] for i in w] for w in sentence] tokenized_sentence = [self.encoding['word2idx'][w] if w in self.encoding['word2idx'] else self.encoding['word2idx'][self.unknown] for w in sentence] character_ids.append(character_sentence) input_ids.append(tokenized_sentence) tag_ids.append(tag) feat_ids.append(feat) return character_ids, input_ids, tag_ids, feat_ids def _prepare_training_data(self, sentences, tags, features, keep_prob=1.0): character_ids = [] input_ids = [] tags_ids = [] feat_ids = [] for sentence, tag, feat in zip_longest(sentences, tags, features, fillvalue=[]): labels = tag.split() if any([not t.startswith('O') for t in labels]) or math.isclose(keep_prob, 1.0, rel_tol=1e-5) or random.random() < keep_prob: words = sentence.split() labels = [self.encoding['tag2idx'][self._adjust_tag(t)] for t in labels] character_sentence = [[self.encoding['char2idx'][i] for i in j] for j in words] tokenized_sentence = [self.encoding['word2idx'][w] for w in words] character_ids.append(character_sentence) input_ids.append(tokenized_sentence) tags_ids.append(labels) feat_ids.append(feat) return character_ids, input_ids, tags_ids, feat_ids def _prepare_bert_prediction_data(self, sentences, tags=[]): input_ids = [] if not self.multi_task_mapping: tags_ids = [] else: tags_ids = {} for k in self.encoding['tag2idx'].keys(): tags_ids[k] = [] attention_masks = [] length = [] for sentence, tag in zip_longest(sentences, tags, fillvalue=[]): tokenized_sentence = [] for word in sentence: tokenized_word = self.tokenizer.tokenize(word) tokenized_sentence.extend(tokenized_word) inputs = self.tokenizer.encode_plus(tokenized_sentence, add_special_tokens=True, return_attention_mask=True) input_ids.append(inputs["input_ids"]) attention_masks.append(inputs['attention_mask']) length.append(min(len(inputs['input_ids']), BERT_MAX_LENGTH)) if not self.multi_task_mapping: tags_ids.append(tag) else: for mapping_key in self.encoding['tag2idx'].keys(): tags_ids[mapping_key].append(tag) return input_ids, tags_ids, attention_masks, length def _prepare_bert_training_data(self, sentences, tags, keep_prob): input_ids = [] if not self.multi_task_mapping: tags_ids = [] else: tags_ids = {} for k in self.tag_remapping.keys(): tags_ids[k] = [] attention_masks = [] length = [] for sentence, tag in zip(sentences, tags): labels = tag.split() if any([not t.startswith('O') for t in labels]) or math.isclose(keep_prob, 1.0, rel_tol=1e-5) or random.random() < keep_prob: tokenized_sentence = [] tokenized_labels = [] words = sentence.split() for word, label in zip(words, labels): tokenized_word = self.tokenizer.tokenize(word) n_subwords = len(tokenized_word) tokenized_sentence.extend(tokenized_word) if not label.startswith('B-'): tokenized_labels.extend([label] * n_subwords) else: tokenized_labels.extend([label]) tokenized_labels.extend([label.replace('B-', 'I-')] * (n_subwords-1)) inputs = self.tokenizer.encode_plus(tokenized_sentence, add_special_tokens=True, return_attention_mask=True) tokenized_labels = ["O"] + tokenized_labels + ["O"] input_ids.append(inputs["input_ids"]) attention_masks.append(inputs['attention_mask']) length.append(min(len(inputs["input_ids"]), BERT_MAX_LENGTH)) if not self.multi_task_mapping: label_ids = [self.encoding['tag2idx'][self._adjust_tag(t)] for t in tokenized_labels] tags_ids.append(label_ids) else: for mapping_key in self.tag_remapping.keys(): label_ids = [self.encoding['tag2idx'][mapping_key][self._adjust_tag(t, mapping_key)] for t in tokenized_labels] tags_ids[mapping_key].append(label_ids) return input_ids, tags_ids, attention_masks, length def load_input(self): for dataset, dataset_setup in self.data_config['sets'].items(): for sub_dataset in dataset_setup: sub_dataset['sentences'] = [] sub_dataset['tags'] = [] sub_dataset['features'] = [] sub_dataset['relations'] = [] for file_config in sub_dataset['all_files']: if self.data_file_extension: sub_dataset['sentences'].extend(self._read_text_file(file_config[self.data_file_extension])) if self.label_file_extension: sub_dataset['tags'].extend(self._read_text_file(file_config[self.label_file_extension])) if self.feature_file_extension: sub_dataset['features'].extend(self._read_feature_file(file_config[self.feature_file_extension])) if self.relation_file_extension: sub_dataset['relations'].extend(self._read_relation_file(file_config[self.relation_file_extension])) def _pad_to_length(self, sequences, length=512, fill_value=0, dtype=torch.long): out_seqs = torch.full((len(sequences), length), fill_value, dtype=dtype) for idx, ids in enumerate(sequences): ids_to_insert = ids[:length] out_seqs[idx, :len(ids_to_insert), ...] = torch.tensor(ids_to_insert) return out_seqs def data_loaders(self): for dataset, dataset_setup in self.data_config['sets'].items(): for idx, sub_dataset in enumerate(dataset_setup): if self.tokenizer is None: characters, ids, tags, features = self._prepare_training_data(sub_dataset['sentences'], sub_dataset['tags'], sub_dataset['features'], sub_dataset['keep_neg_sample_prob']) input_data = LSTMDataset(characters, ids, tags, features, self.encoding['char2idx'], self.padding, self.max_word_length, self.max_sent_length) sampler = RandomSampler(input_data) sub_dataset['dataloader'] = DataLoader(input_data, sampler=sampler, batch_size=self.batch_size, collate_fn=self._collate_fn) else: input_ids, tag_ids, attention_masks, lengths = self._prepare_bert_training_data(sub_dataset['sentences'], sub_dataset['tags'], sub_dataset['keep_neg_sample_prob']) input_ids = self._pad_to_length(input_ids, length=BERT_MAX_LENGTH, fill_value=self.special_toks['pad_tok'], dtype=torch.long) attention_masks = self._pad_to_length(attention_masks, length=BERT_MAX_LENGTH, fill_value=0, dtype=torch.long) if not self.multi_task_mapping: tag_ids = self._pad_to_length(tag_ids, length=BERT_MAX_LENGTH, fill_value=self.encoding['tag2idx']['O'], dtype=torch.long) input_data = BERTDataset(input_ids, tag_ids, attention_masks) else: for k in tag_ids.keys(): tag_ids[k] = self._pad_to_length(tag_ids[k], length=BERT_MAX_LENGTH, fill_value=self.encoding['tag2idx'][k]['O'], dtype=torch.long) input_data = BERTMultiDataset(input_ids, tag_ids, attention_masks, lengths) sampler = RandomSampler(input_data) sub_dataset['dataloader'] = DataLoader(input_data, sampler=sampler, batch_size=self.batch_size) def word_embedding(self, emb_conf): if self.checkpoint is not None and self.checkpoint['model']: print("Using a pre-trained model --- word embedding is loaded as part of the model") embedding_weights = torch.zeros(len(self.encoding['word2idx']), emb_conf['dim']) else: print("Loading word embedding: {}".format(emb_conf['file'])) pretrained_word_vectors = KeyedVectors.load_word2vec_format(emb_conf['file'], binary=True) if emb_conf['zero_init']: embedding_weights = torch.zeros(len(self.encoding['word2idx']), emb_conf['dim']) else: embedding_weights = torch.randn(len(self.encoding['word2idx']), emb_conf['dim']) unknown_word_count = 0 for word, idx in self.encoding['word2idx'].items(): try: embedding_weights[idx] = torch.tensor(pretrained_word_vectors[word].copy()) except KeyError: unknown_word_count += 1 print("{}/{} word from the dataset do no exist in the word embedding".format(unknown_word_count, len(self.encoding['word2idx']))) return embedding_weights
dave-s477/SoMeNLP
setup.py
<reponame>dave-s477/SoMeNLP import sys if sys.version_info <= (3,7): sys.exit('Python >= 3.7 is required') from setuptools import setup def readme(): with open('Readme.md') as f: return f.read() setup(name='SoMeNLP', version='0.1', description='NLP procedures for scientific articles and software extraction.', long_description=readme(), classifiers=[ 'License :: OSI Approved :: GPLv3', 'Programming Language :: Python :: 3.7', 'Topic :: NER :: NLP', ], keywords='scientific entity relation software paper article', url='https://github.com/dave-s477/SoMeNLP', author='<NAME>', author_email='<EMAIL>', license='GPLv3', packages=['somenlp'], scripts=[ 'bin/train_word_emb', 'bin/distant_supervision', 'bin/custom_feature_gen', 'bin/train_model', 'bin/tune_model', 'bin/tune_relext', 'bin/predict', 'bin/split_data', 'bin/train_relext', 'bin/load_dbpedia_info', 'bin/entity_disambiguation', 'bin/somesci_disambiguation_input', 'bin/map_unique_names_to_files', 'bin/predict_relext', 'bin/combine_annotations', 'bin/generate_file_list' ], install_requires=[ 'pytest', 'gensim>=4.0.1', 'torch', 'tensorboard', 'pandas', 'numpy', 'beautifulsoup4', 'wiktextract', 'wget', 'NLTK', 'scikit-learn', 'transformers==4.6.1', 'SPARQLWrapper', 'python-levenshtein', 'articlenizer @ https://github.com/dave-s477/articlenizer/tarball/master#egg=package' ], include_package_data=True, zip_safe=False)
dave-s477/SoMeNLP
somenlp/RE/RE_model.py
<filename>somenlp/RE/RE_model.py import pandas as pd import numpy as np import pickle from torch import cross from sklearn.preprocessing import StandardScaler from sklearn.ensemble import RandomForestClassifier from sklearn.neural_network import MLPClassifier from sklearn.metrics import confusion_matrix, classification_report from sklearn.model_selection import cross_validate from itertools import combinations class REmodel(): """Relation extraction model """ def __init__(self, gen_config, model_config, data_handler, output_handler, output_config, feature_generator=None): """Init Args: model_config (dict): configuration of the RE model data_handler (somenlp.NER.DataHandler) output_handler (somenlp.NER.OutputHandler) """ self.gen_config = gen_config self.model_config = model_config self.data_handler = data_handler self.output_handler = output_handler self.feature_generator = feature_generator self.output_config = output_config if gen_config['type'] == 'RF': self.model = RandomForestClassifier( n_estimators=model_config["n_estimators"], criterion=model_config["criterion"], max_depth=model_config["max_depth"], min_samples_split=model_config["min_samples_split"], min_samples_leaf=model_config["min_samples_leaf"], min_weight_fraction_leaf=model_config["min_weight_fraction_leaf"], max_features=model_config["max_features"], max_leaf_nodes=model_config["max_leaf_nodes"], max_samples=model_config["max_samples"] ) elif gen_config['type'] == 'NN': if model_config['scaler']: self.scaler = StandardScaler() self.model = MLPClassifier( solver = model_config['solver'], batch_size = gen_config['batch_size'], max_iter = model_config['epochs'], learning_rate_init = model_config['lr'], hidden_layer_sizes=tuple(model_config['layers']) ) else: raise(RuntimeError("Got unsupported model type {}".format(model_config['type']))) def _train_model(self, data): """Train model based on provided data Args: data (list): features """ input_data = pd.DataFrame(data) y_train = input_data['label'].values X_train = input_data.loc[:, input_data.columns != 'label'] if self.gen_config['type'] == 'NN' and self.model_config['scaler']: self.scaler.fit(X_train) X_train = self.scaler.transform(X_train) #self.X_train = X_train self.model.fit(X_train, y_train) def train(self): """Train RE model """ all_train_samples = [] for idx, dataset in enumerate(self.data_handler.data_config['sets']['train']): print("Getting samples from {} dataset from train set".format(idx)) all_train_samples.extend(dataset['relext_feature_list']) self._train_model(all_train_samples) def cross_val(self): all_train_samples = [] for idx, dataset in enumerate(self.data_handler.data_config['sets']['train']): print("Getting samples from {} dataset from train set".format(idx)) all_train_samples.extend(dataset['relext_feature_list']) input_data = pd.DataFrame(all_train_samples) y = input_data['label'].values X = input_data.loc[:, input_data.columns != 'label'] if self.gen_config['type'] == 'NN' and self.model_config['scaler']: self.scaler.fit(X) X = self.scaler.transform(X) cv_results = cross_validate(self.model, X, y, cv=5, scoring=('precision_micro', 'recall_micro', 'f1_micro', 'precision_macro', 'recall_macro', 'f1_macro')) for k, v in cv_results.items(): if k.startswith('test'): print('{}: {}'.format(k, v)) def test(self): """Test RE model """ out_s = '' cls_for_latex = {} for idx, dataset in enumerate(self.data_handler.data_config['sets']['test']): print("Start testing on corpus {}".format(idx)) out_s += 'Test Corpus {}: {}\n\n'.format(idx, dataset['name']) input_data = pd.DataFrame(dataset['relext_feature_list']) y_test = input_data['label'].values X_test = input_data.loc[:, input_data.columns != 'label'] if self.gen_config['type'] == 'NN' and self.model_config['scaler']: X_test = self.scaler.transform(X_test) predictions = self.model.predict(X_test) if "confusion" in self.output_config and self.output_config['confusion']: cm = confusion_matrix(y_test, predictions, labels=self.model.classes_) out_s += "Confusion Matrix:\n{}\n{}\n\n".format(self.model.classes_, cm) if "no_neg" in self.output_config and self.output_config['no_neg']: cl_no_negatives = classification_report(y_test, predictions, labels=self.model.classes_[self.model.classes_!='none']) out_s += "Classification without negative samples:\n{}\n\n".format(cl_no_negatives) if "with_neg" in self.output_config and self.output_config['with_neg']: cl_with_negatives = classification_report(y_test, predictions, labels=self.model.classes_) out_s += "Classification with negative samples:\n{}\n\n".format(cl_with_negatives) if "latex" in self.output_config and self.output_config['latex']: cl_no_negatives_dict = classification_report(y_test, predictions, labels=self.model.classes_[self.model.classes_!='none'], output_dict=True) cls_for_latex[dataset['name']] = cl_no_negatives_dict if "latex" in self.output_config and self.output_config['latex']: out_s += "Formatted for latex: {}".format(self.output_handler.cl_for_latex(cls_for_latex)) print(out_s) if "save_log" in self.output_config and self.output_config['save_log']: with open('{}/rel_extraction_result.log'.format(self.output_handler.log_dir), 'w') as out_file: out_file.write(out_s) def predict(self, output): print("Starting prediction") iterator = self.feature_generator.stream_files() for input in iterator: prediction_outputs = [] for idx, (sentence, features, entities) in enumerate(zip(input['sentences'], input['relext_feature_list'], input['entity_list'])): prediction_outputs.append([]) if features is not None: # print(sentence) # print(input['tags']) # print() #print(features) X = pd.DataFrame(features) if self.gen_config['type'] == 'NN' and self.model_config['scaler']: X = self.scaler.transform(X) predictions = self.model.predict(X) #print(X) for ent_pair, assigned_label in zip(entities, predictions): if assigned_label != 'none': prediction_outputs[-1].append([ent_pair, assigned_label]) #print(predictions) #print() with input['out_name'].open(mode='w') as out_f: for line_pred in prediction_outputs: line_string = '' for pred in line_pred: line_string += '{}\t{}\t{}\t{}\t{}\t{}\t{};;'.format(pred[1], pred[0][0]['string'], pred[0][0]['beg'], pred[0][0]['idx'], pred[0][1]['string'], pred[0][1]['beg'], pred[0][1]['idx']) out_f.write(line_string + '\n') #prediction_outputs def show_features_importance(self): """Print a summary of the feature importance provided by sklearn """ feat_importances = pd.Series(self.model.feature_importances_, index=self.X_train.columns) print("Feature importance for classifier:") print(feat_importances.sort_values()) def save(self): output_name = '{}/model.sav'.format(self.output_handler.save_dir) pickle.dump(self.model, open(output_name, 'wb')) print("Saved model at {}".format(output_name)) def load(self, checkpoint): self.model = pickle.load(open(checkpoint['model'], 'rb')) print("Loaded model from {}".format(checkpoint['model']))
dave-s477/SoMeNLP
somenlp/word_embedding/calculate_embedding.py
import os import gensim import random import logging from pathlib import Path logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO) class SentenizedInput(object): """Simple linebased iterator for data. Assumes sentences are split by newline character and tokens are split by whitespaces. """ def __init__(self, file_list, seed): random.seed(seed) random.shuffle(file_list) self.file_list = file_list def __iter__(self): for f in self.file_list: with f.open() as text: for line in text: if line.strip(): yield line.split() def resume_training_embedding(checkpoint, in_files, out_path, epochs, replace=False, format=2, seed=42, ncores=8): """Re-train a Word2Vec embedding from a gensim model checkpoint Args: checkpoint (string): path to pre-trained model in_files (list of PosixPaths): files of training corpus out_path (string): output path epochs (int): number of training epochs replace (bool, optional): removes the original checkpoint. Defaults to False. format (int, optional): determines the output format between bin/gensim-model/both. Defaults to 2. seed (int, optional): data shuffling seed. Defaults to 42. ncores (int, optional): number of gensim workers. Defaults to 8. """ iterator = SentenizedInput(in_files, seed) model = gensim.models.Word2Vec.load(checkpoint) model.build_vocab(iterator, update=True) model.train(iterator, total_examples=model.corpus_count, epochs=epochs) out_loc = checkpoint.split('.model')[0] + "_re-{}".format(epochs) if format == 0: print("Saving model as bin") model.wv.save_word2vec_format(out_loc + '.bin', binary=True) elif format == 1: print("Saving trainable model") model.save(out_loc + '.model') else: print("Saving model as bin and trainable") model.wv.save_word2vec_format(out_loc + '.bin', binary=True) model.save(out_loc + '.model') if replace: os.remove(checkpoint) def train_embedding(name, in_files, out_path, emb_dim=200, win_size=5, min_count=5, epochs=1, format=2, seed=42, ncores=8): """Train a gensim Word2Vec embedding on a corpus Args: name (string): name for writing the model in_files (list of PosixPaths): files of training corpus out_path (string): output path emb_dim (int, optional): size of the embedding. Defaults to 200. win_size (int, optional): w2v window size. Defaults to 200. min_count (int, optional): word mincount to be in embedding. Defaults to 200. epochs (int, optional): number of training epochs format (int, optional): determines the output format between bin/gensim-model/both. Defaults to 2. seed (int, optional): data shuffling seed. Defaults to 42. ncores (int, optional): number of gensim workers. Defaults to 8. """ iterator = SentenizedInput(in_files, seed) model = gensim.models.Word2Vec(iterator, vector_size=emb_dim, window=win_size, min_count=min_count, workers=ncores, sg=1, epochs=epochs) out_loc = out_path + '/' + name if format == 0: print("Saving model as bin") model.wv.save_word2vec_format(out_loc + '.bin', binary=True) elif format == 1: print("Saving trainable model") model.save(out_loc + '.model') else: print("Saving model as bin and trainable") model.wv.save_word2vec_format(out_loc + '.bin', binary=True) model.save(out_loc + '.model')
dave-s477/SoMeNLP
read_data.py
<gh_stars>0 #!/usr/bin/env python import tensorflow as tf import argparse import sys import numpy as np from os import write from pathlib import Path from tensorflow.python.summary.summary_iterator import summary_iterator tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR) def format_latex_string(values, separator='&'): s = '' for task_k, task_v in values.items(): s += "Task: {}\n".format(task_k) for label_k, label_v in task_v.items(): s += '{} {} {} ({}) {} {} ({}) {} {} ({}) {}\n'.format( label_k, separator, label_v['Precision_test_0'], label_v['Precision_devel_0'], separator, label_v['Recall_test_0'], label_v['Recall_devel_0'], separator, label_v['FScore_test_0'], label_v['FScore_devel_0'], separator ) s += '\n' return s def format_latex_average_string(averages, separator='&'): s = '' for task_k, task_v in averages.items(): s += "Task: {}\n".format(task_k) for label_k, label_v in task_v.items(): s += '{} {} {}$\pm${} {} {}$\pm${} {} {}$\pm${} {} {}$\pm${} {} {}$\pm${} {} {}$\pm${}\\\\ {}\n'.format( label_k, separator, round(label_v['Precision_test_0']['mean'], 3), round(label_v['Precision_test_0']['std'], 3), separator, round(label_v['Precision_devel_0']['mean'], 3), round(label_v['Precision_devel_0']['std'], 3), separator, round(label_v['Recall_test_0']['mean'], 3), round(label_v['Recall_test_0']['std'], 3), separator, round(label_v['Recall_devel_0']['mean'], 3), round(label_v['Recall_devel_0']['std'], 3), separator, round(label_v['FScore_test_0']['mean'], 3), round(label_v['FScore_test_0']['std'], 3), separator, round(label_v['FScore_devel_0']['mean'], 3), round(label_v['FScore_devel_0']['std'], 3), len(label_v['Precision_test_0']['values']) ) s += '\n' return s def get_epochs(file, args): epoch_summary = {} for e in summary_iterator(file): if e.step >= args.epochs[0] and e.step < args.epochs[1]: if e.step not in epoch_summary: epoch_summary[e.step] = {} for v in e.summary.value: if v.tag.count('/') == 3: task, entity, value, dataset = v.tag.split('/') else: task = 'default' entity, value, dataset = v.tag.split('/') if task not in epoch_summary[e.step]: epoch_summary[e.step][task] = {} if entity not in epoch_summary[e.step][task]: epoch_summary[e.step][task][entity] = {} epoch_summary[e.step][task][entity][value + '_' + dataset] = v.simple_value return epoch_summary def get_max(file, args): max_fscore = 0 max_epoch = 0 #print(file) for e in summary_iterator(file): for v in e.summary.value: if v.tag.count('/') == 3 and v.tag == '{}/{}/FScore/devel_0'.format(args.get_max_task, args.get_max_label): if v.simple_value > max_fscore: max_fscore = v.simple_value max_epoch = e.step elif v.tag.count('/') == 2 and v.tag == '{}/FScore/devel_0'.format(args.get_max_label): if v.simple_value > max_fscore: max_fscore = v.simple_value max_epoch = e.step #if e.step < 100: # print("HEappfaefa") # return None epoch_summary = { max_epoch: {} } for e in summary_iterator(file): if e.step == max_epoch: for v in e.summary.value: if v.tag.count('/') == 3: task, entity, value, dataset = v.tag.split('/') else: task = 'default' entity, value, dataset = v.tag.split('/') if task not in epoch_summary[e.step]: epoch_summary[e.step][task] = {} if entity not in epoch_summary[e.step][task]: epoch_summary[e.step][task][entity] = {} epoch_summary[e.step][task][entity][value + '_' + dataset] = v.simple_value #print(epoch_summary) return epoch_summary def get_average_performance(files, args): averages = {} for file in files: epoch_summary = get_max(str(file), args) if epoch_summary is None: continue epoch_result = list(epoch_summary.values())[0] for task, task_res in epoch_result.items(): if task not in averages: averages[task] = {} for label, label_res in task_res.items(): if label not in averages[task]: averages[task][label] = {} for score, score_res in label_res.items(): if score not in averages[task][label]: averages[task][label][score] = { 'values': [] } averages[task][label][score]['values'].append(score_res) for _, task_res in averages.items(): for _, label_res in task_res.items(): for score, score_res in label_res.items(): mean_val = np.mean(score_res['values']) score_res['mean'] = mean_val std_val = np.std(score_res['values']) score_res['std'] = std_val return averages if __name__ == "__main__": parser = argparse.ArgumentParser(description = "Get values from tensorboard.") parser.add_argument("--path", required=True, help="Full path to tensorboard log or path to folder containing multiple logs") parser.add_argument("--get-max", action="store_true", help="Get the max score based on devel F-Score.") parser.add_argument("--get-max-task", default="software") parser.add_argument("--get-max-label", default="Application") parser.add_argument("--epochs", default=[10,12], nargs=2, help="Epochs to search") parser.add_argument("--round", default='2', help="How to round results") parser.add_argument("--merge-all", action="store_true") args = parser.parse_args() in_path = Path(args.path) if in_path.is_file(): file_list = {'single': [in_path]} elif in_path.is_dir(): file_list = {} events = list(in_path.rglob('events.out.tfevents*')) for event in events: if args.merge_all: config_string = 'all' else: config_string = str(event).rsplit('/', maxsplit=2)[-2].split('_', maxsplit=2)[-1] if config_string not in file_list: file_list[config_string] = [] file_list[config_string].append(event) else: raise(RuntimeError("Invalid input path {}".format(args.path))) for k,v in file_list.items(): print(k) if not args.get_max: for file in v: args.epochs = [int(x) for x in args.epochs] epoch_summary = get_epochs(str(file), args) for k,v in epoch_summary.items(): print("Epoch {}".format(k)) print(format_latex_string(v)) else: averages = get_average_performance(v, args) print(format_latex_average_string(averages)) print()
dave-s477/SoMeNLP
somenlp/word_embedding/__init__.py
from .calculate_embedding import train_embedding, resume_training_embedding
dave-s477/SoMeNLP
somenlp/entity_disambiguation/linking_data.py
<gh_stars>0 from hashlib import new import json import sys import random import re import numpy as np from articlenizer import articlenizer from nltk.corpus import stopwords random.seed(42) STOPS = stopwords.words('english') def normalize(s): norm_s = re.sub('[^0-9a-zA-Z]+', ' ', s.casefold()).rstrip('0123456789 ,.').lstrip(' ') norm_s = ' '.join([w for w in norm_s.split() if w not in STOPS]) if not norm_s: norm_s = s return norm_s def remove_spaces(s): replace_regex = re.compile(r'\s(?P<to_keep>[\+\-#™_/\d]+)\s?') return replace_regex.sub(r'\g<to_keep>', s) class LinkingData: def __init__(self, config): self.config = config self.read_data_files(config) def read_data_files(self, config): self.train = {} for entry in config['train']: with open(entry['file'], 'r') as data_f: self.train[entry['name']] = json.load(data_f) for idx, entry in enumerate(self.train[entry['name']]): entry['init_id'] = idx self.test = {} for entry in config['test']: with open(entry['file'], 'r') as data_f: self.test[entry['name']] = json.load(data_f) for idx, entry in enumerate(self.test[entry['name']]): entry['init_id'] = idx if config['gold']: with open(config['gold'], 'r') as gold_f: self.gold = json.load(gold_f) if self.config['data_augmentation']: self.all = [] self.all_names = [] self.all_augmented_names = [] self.short_samples = [] self.long_samples = [] with open(config['all'], 'r') as data_f: all = json.load(data_f) for sample in all: self.all_names.append(sample['mention'].lower()) tokens = sample['mention'].split() sample['tokens'] = tokens self.all.append(sample) if len(tokens) <= 1: self.short_samples.append(sample) else: self.long_samples.append(sample) self._get_augmented_samples() for s in self.all: for idx, g in enumerate(self.gold): if s['paper_id'] == g['paper_id'] and s['mention'] == ' '.join(articlenizer.get_tokenized_sentences(g['mention'])[0]): s['gold_id'] = idx break for s in self.augmented_samples: self.all.append(s) random.shuffle(self.all) self.link_lookup_table = np.zeros((len(self.all), len(self.all)), dtype=bool) for idx, s in enumerate(self.all): s['origin_id'] = idx s['string'] = remove_spaces(s['mention']) s['norm'] = normalize(s['mention']) for idx_2, s_2 in enumerate(self.all[idx+1:]): if 'gold_id' in s and 'gold_id' in s_2 and self.gold[s['gold_id']]['link'] == self.gold[s_2['gold_id']]['link']: self.link_lookup_table[idx][idx+1+idx_2] = True self.link_lookup_table[idx+1+idx_2][idx] = True self.all_names = None self.all_augmented_names = None self.short_samples = None self.long_samples = None def _get_augmented_samples(self): num_augmentation_samples = 0 for _, v in self.train.items(): num_augmentation_samples += len(v) print(num_augmentation_samples) num_augmentation_samples *= self.config['num_augmentation_samples'] self.augmented_samples = [] for _ in range(int(num_augmentation_samples/2)): # Augment short sample s_1 = self.short_samples[random.randint(0, len(self.short_samples)-1)] s_2 = self.short_samples[random.randint(0, len(self.short_samples)-1)] self._recombine(s_1, s_2) # Augment long sample s_1 = self.long_samples[random.randint(0, len(self.long_samples)-1)] s_2 = self.long_samples[random.randint(0, len(self.long_samples)-1)] self._recombine(s_1, s_2) def _recombine_words(self, s1, s2): split_idx_1 = random.randint(1, len(s1)-1) if len(s1) > 1 else random.randint(0,1) split_idx_2 = random.randint(1, len(s2)-1) if len(s2) > 1 else random.randint(0,1) if isinstance(s1, str) and isinstance(s2, str): s1_new = s1[:split_idx_1] + s2[split_idx_2:] s2_new = s2[:split_idx_2] + s1[split_idx_1:] elif isinstance(s1, list) and isinstance(s2, list): s1_new = ' '.join(s1[:split_idx_1] + s2[split_idx_2:]) s2_new = ' '.join(s2[:split_idx_2] + s1[split_idx_1:]) elif isinstance(s1, str) and isinstance(s2, list): s1_new = s1[:split_idx_1] + ' ' + ' '.join(s2[split_idx_2:]) s2_new = ' '.join(s2[:split_idx_2]) + ' ' + s1[split_idx_1:] elif isinstance(s1, list) and isinstance(s2, str): s1_new = ' '.join(s1[:split_idx_1]) + ' ' + s2[split_idx_2:] s2_new = s2[:split_idx_2] + ' ' + ' '.join(s1[split_idx_1:]) return s1_new, s2_new def _get_add_info(self, r): i = {} for v in r: tokens = v['string'].split() i[v['type']] = tokens if len(tokens) > 1 else v['string'] return i def _recombine(self, s_1, s_2): if s_1['mention'].lower() == s_2['mention'].lower(): return [] new_strings = self._recombine_words(s_1['mention'], s_2['mention']) add_info_1 = self._get_add_info(s_1['relations']) add_info_2 = self._get_add_info(s_2['relations']) for k, v in add_info_1.items(): if k in add_info_2 and v == add_info_2[k]: return [] new_developers = self._recombine_words(add_info_1['Developer_of'], add_info_2['Developer_of']) if 'Developer_of' in add_info_1 and 'Developer_of' in add_info_2 else None new_url = self._recombine_words(add_info_1['URL_of'], add_info_2['URL_of']) if 'URL_of' in add_info_1 and 'URL_of' in add_info_2 else None new_version = self._recombine_words(add_info_1['Version_of'], add_info_2['Version_of']) if 'Version_of' in add_info_1 and 'Version_of' in add_info_2 else None for idx, n_s in enumerate(new_strings): if not n_s.lower() in self.all_names and not n_s.lower() in self.all_augmented_names: new_sample = { 'mention': n_s, 'relations': [] } if new_developers is not None: new_sample['relations'].append({ 'type': 'Developer_of', 'string': new_developers[idx] }) if new_url is not None: new_sample['relations'].append({ 'type': 'URL_of', 'string': new_url[idx] }) if new_version is not None: new_sample['relations'].append({ 'type': 'Version_of', 'string': new_version[idx] }) self.augmented_samples.append(new_sample) self.all_augmented_names.append(n_s)
dave-s477/SoMeNLP
somenlp/distant_supervision/packages.py
import json import urllib.request import os from pathlib import Path from bs4 import BeautifulSoup def get_pypi_package_names(default_address='https://pypi.org/simple/'): """Download PyPi package names Args: default_address (str, optional): url for pypi. Defaults to 'https://pypi.org/simple/'. Returns: list: pypi package names """ print("Loading pypi names") pypi_package_names = [] try: content = urllib.request.urlopen(default_address) except urllib.error.URLError as e: print("Parsing pypi went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, "lxml") for a in soup.findAll('a', href=True): pypi_package_names.append(a.text) return pypi_package_names def get_R_forge_package_names(default_address='https://r-forge.r-project.org/softwaremap/trove_list.php?cat=c&form_cat=307&page='): """Download R-forge package names Args: default_address (str, optional): url for R-forge. Defaults to 'https://r-forge.r-project.org/softwaremap/trove_list.php?cat=c&form_cat=307&page='. Returns: list: list of R-forge packages """ print("Loading R packages") rforge_packages = [] counter = 1 prev_rforge_packages = [] current_rforge_packages = [] try: content = urllib.request.urlopen('{}{}'.format(default_address, counter)) except urllib.error.URLError as e: print("Parsing R packages went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, "lxml") main_div = soup.find("div", {"id": "maindiv"}) for idx, tab in enumerate(soup.findAll("table",{"class":"fullwidth"})): if idx != 0: for ref in tab.findAll("a", href=True): if ref.text != '[Filter]' and ref.text.strip(): current_rforge_packages.append(ref.text.rstrip()) while current_rforge_packages != prev_rforge_packages: if counter % 10 == 0: print("Currently at R forge page {}".format(counter)) prev_rforge_packages = current_rforge_packages rforge_packages.extend(prev_rforge_packages) counter += 1 current_rforge_packages = [] try: content = urllib.request.urlopen('{}{}'.format(default_address, counter)) except urllib.error.URLError as e: print("Parsing R packages went wrong on page {}: {}".format(counter, e)) return rforge_packages soup = BeautifulSoup(content, "lxml") main_div = soup.find("div", {"id": "maindiv"}) for idx, tab in enumerate(soup.findAll("table",{"class":"fullwidth"})): if idx == 0: pass else: for ref in tab.findAll("a", href=True): if ref.text != '[Filter]' and ref.text.strip(): current_rforge_packages.append(ref.text.rstrip()) return rforge_packages def get_swMATH_software_names(default_address='https://swmath.org/?which_search=browse&sel=all&sortby=-rank&&page='): """Download list of SwMATH software names Args: default_address (str, optional): swMATH url. Defaults to 'https://swmath.org/?which_search=browse&sel=all&sortby=-rank&&page='. Returns: list: swMATH software names """ print("Loading SwMATH names") swMATH_packages = [] counter = 1 prev_swMATH = [] current_swMATH = [] try: content = urllib.request.urlopen('{}{}'.format(default_address, counter)) except urllib.error.URLError as e: print("Parsing swMATH names went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, "lxml") for h1 in soup.findAll("h1"): current_swMATH.append(h1.text) current_swMATH.remove('swMATH') while current_swMATH != prev_swMATH: if counter % 10 == 0: print("Currently at swMATH page {}".format(counter)) prev_swMATH = current_swMATH swMATH_packages.extend(prev_swMATH) counter += 1 current_swMATH = [] try: content = urllib.request.urlopen('{}{}'.format(default_address, counter), timeout=20) soup = BeautifulSoup(content, "lxml") except urllib.error.URLError as e: print("Parsing swMATH names went wrong on page {}: {}".format(counter, e)) return swMATH_packages except: print("Probably a timeout in loading {}... trying again".format(counter)) counter -= 1 else: for h1 in soup.findAll("h1"): current_swMATH.append(h1.text) current_swMATH.remove('swMATH') return swMATH_packages def get_CRAN_package_names(default_address='https://cran.r-project.org/web/packages/available_packages_by_name.html'): """Download CRAN package names Args: default_address (str, optional): url for CRAN. Defaults to 'https://cran.r-project.org/web/packages/available_packages_by_name.html'. Returns: list: CRAN package names """ print("Loading CRAN packages") cran_packages = [] try: content = urllib.request.urlopen(default_address) except urllib.error.URLError as e: print("Parsing pypi went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, 'lxml') for a in soup.findAll('a', href=True): cran_packages.append(a.text) return cran_packages def get_Bioconductor_package_names(default_address='https://www.bioconductor.org/packages/release/bioc/'): """Bioconductor package names Args: default_address (str, optional): url for Bioconductor. Defaults to 'https://www.bioconductor.org/packages/release/bioc/'. Returns: list: Bioconductor package names """ print("Loading Bioconductor packages") bioconductor_packages = [] try: content = urllib.request.urlopen(default_address) except urllib.error.URLError as e: print("Parsing pypi went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, 'lxml') div = soup.find("div", {"id": "PageContent"}) for row in div.find_all('table')[0].find_all('tr'): for a in row.findAll('a', href=True): bioconductor_packages.append(a.text) return bioconductor_packages def get_Anaconda_package_names(repo='anaconda', default_address='https://anaconda.org/{}/repo?page='): """Get Anaconda package names Args: repo (str, optional): repo location for anaconda. Defaults to 'anaconda'. default_address (str, optional): url for anaconda. Defaults to 'https://anaconda.org/{}/repo?page='. Returns: list: Anaconda package names """ conda_packages = [] counter = 1 prev_conda_packages = [] current_conda_packages = [] try: req = urllib.request.Request('{}{}'.format(default_address.format(repo), counter), headers={'User-Agent': 'Mozilla/5.0'}) content = urllib.request.urlopen(req) except urllib.error.URLError as e: print("Parsing Anaconda went wrong due to: {}".format(e)) return None soup = BeautifulSoup(content, 'lxml') for span in soup.findAll("span", {"class": "packageName"}): current_conda_packages.append(span.text) while current_conda_packages != prev_conda_packages: if (counter+1) % 1 == 0: print("Currently at Anaconda ({}) page {}".format(repo, counter+1)) prev_conda_packages = current_conda_packages conda_packages.extend(prev_conda_packages) counter += 1 current_conda_packages = [] try: req = urllib.request.Request('{}{}'.format(default_address.format(repo), counter), headers={'User-Agent': 'Mozilla/5.0'}) content = urllib.request.urlopen(req, timeout=20) soup = BeautifulSoup(content, 'lxml') except urllib.error.URLError as e: print("Parsing Anaconda names went wrong on page {}: {}".format(counter, e)) return conda_packages except: print("Probably a timeout in loading {}... trying again".format(counter)) counter -= 1 else: for span in soup.findAll("span", {"class": "packageName"}): current_conda_packages.append(span.text) return conda_packages def get_if_not_exists(date, function, name, location='/tmp', *args): """Load file if it exists, else call a function to create it. Args: date (str): string identifier of file function (fct): function to call is file does not exist name (str): base name of file location (str, optional): path to file. Defaults to '/tmp'. Returns: json-serializeable object: result from generating or loading a json file """ loc = Path(location) / '{}_{}'.format(date, name) output = None if loc.is_file(): print("Loading {} from {}".format(name, str(loc))) with loc.open(mode='r') as j_in: output = json.load(j_in) else: output = function(*args) with loc.open(mode='w') as j_out: json.dump(output, j_out, indent=4) print("Saved {} to {}".format(name, str(loc))) return output def load_package_names(date, location='/tmp'): """Load package names from a number of repositories Args: date (str): date-based identifier location (str, optional): output/lookup path. Defaults to '/tmp'. Returns: dictionary: packages repositories with their respective software lists """ data_dictionary = { 'pypi': get_if_not_exists(date, get_pypi_package_names, 'pypi.json', location), 'rforge': get_if_not_exists(date, get_R_forge_package_names, 'rforge.json', location), 'swMATH': get_if_not_exists(date, get_swMATH_software_names, 'swMATH.json', location), 'CRAN': get_if_not_exists(date, get_CRAN_package_names, 'CRAN.json', location), 'Bioconductor': get_if_not_exists(date, get_Bioconductor_package_names, 'Bioconductor.json', location), 'Anaconda': get_if_not_exists(date, get_Anaconda_package_names, 'Anaconda.json', location, 'anaconda'), 'Conda-forge': get_if_not_exists(date, get_Anaconda_package_names, 'Conda-forge.json', location, 'conda-forge') } return data_dictionary
dave-s477/SoMeNLP
somenlp/NER/tuner.py
<reponame>dave-s477/SoMeNLP import json import copy from pathlib import Path from somenlp.utils import find_type_in_dict, getFromDict, setInDict, get_abbr class Tuner(): def __init__(self, config, time): self.config = config self.values_to_vary = find_type_in_dict(self.config, list) self._gen_all_parameter_combinations() self._gen_all_configs() def _entry_to_combinations(self, entry): res = [] for val in entry['values']: res.append([{ 'path': entry['path'], 'values': val }]) return res def _gen_all_parameter_combinations(self): old_combinations = self._entry_to_combinations(self.values_to_vary[0]) for val in self.values_to_vary[1:]: #print(self._entry_to_combinations(val)) new_combinations = [] for combination in old_combinations: for next_param in self._entry_to_combinations(val): comb = combination.copy() comb.extend(next_param) new_combinations.append(comb) old_combinations = new_combinations self.combinations = old_combinations def _gen_config_name(self, parameter_config): name = '' for parameter in parameter_config: for sub_path in parameter['path']: name += '{}-'.format(get_abbr(sub_path)) name += '{}_'.format(parameter['values']) for char in ['{', '}', ' ', "'", '/', '\\', ':', ',']: name = name.replace(char, '') return name.rstrip('_')[:200] def _gen_all_configs(self): self.configs_to_execute = {} for combination in self.combinations: config = copy.deepcopy(self.config) config_name = self._gen_config_name(combination) for entry in combination: setInDict(config, entry['path'], entry['values']) data_c_file = config.pop('data', None) if data_c_file is None: raise(RuntimeError("No data config was provided in tuning config.")) data_c_path = Path(data_c_file) with data_c_path.open(mode='r') as data_c_json: data_conf = json.load(data_c_json) self.configs_to_execute[config_name] = { 'data': data_conf, 'model': config } def yield_configs(self): for c_name, v in self.configs_to_execute.items(): yield c_name, v['data'], v['model']
dave-s477/SoMeNLP
somenlp/distant_supervision/combine_info.py
<gh_stars>0 import json from articlenizer import articlenizer as art def load_dicts(locations): """Load and merge information from json dictionaries Args: locations (list of Posix paths): locations to json files Returns: dictionary: dictionary merged from json files """ results = {} for f in locations: name = f.name.split('.dic')[0] results[name] = [] with f.open(mode='r', errors='ignore') as in_f: for line in in_f: if line.rstrip(): results[name].append(line.rstrip()) return results def normalize_entries(entries, max_length): """Shorten entries in list based on number of tokens after tokenization Args: entries (list): list of candidates max_length (int): max allowed token length Returns: list: list with filtered long entries """ entry_list = [] for entry in entries: tokens = art.tokenize_text(entry) if len(tokens) <= max_length: entry_list.append(' '.join(tokens)) return entry_list def merge_results(out_file, max_token_length, inputs): """Merge and write a distant supervision ditionary from filtered dictionaries Args: out_file (Posix path): output location max_token_length (int): max allowed token length inputs (list of dictionaries): individual dicts to be merged """ distant_supervision_dictionary = {} for i in inputs: if i is not None: for k, v in i.items(): distant_supervision_dictionary[k] = normalize_entries(v, max_token_length) with out_file.open(mode='w') as json_out: json.dump(distant_supervision_dictionary, json_out, indent=4)
dave-s477/SoMeNLP
somenlp/NER/models/crf.py
import torch import torch.nn as nn import torch.nn.functional as F import os os.environ['CUDA_LAUNCH_BLOCKING'] = "1" class CRF(nn.Module): def __init__(self, tagset_size, device, init_parameters=None): super(CRF, self).__init__() self.device = device if init_parameters is None: #self.transition_params = nn.Parameter(torch.randn(tagset_size, tagset_size)) self.transition_params = nn.Parameter(nn.init.xavier_normal_(torch.empty(tagset_size, tagset_size))) else: self.transition_params = nn.Parameter(init_parameters.to(self.device)) def crf_sequence_score(self, inputs, tag_indices, sequence_lengths): """Computes the unnormalized score for a tag sequence. Args: inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer. tag_indices: A [batch_size, max_seq_len] matrix of tag indices for which we compute the unnormalized score. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: sequence_scores: A [batch_size] vector of unnormalized sequence scores. """ # If max_seq_len is 1, we skip the score calculation and simply gather the # unary potentials of the single tag. def _single_seq_fn(): batch_size = inputs.shape[0] example_inds = torch.arange(batch_size, dtype=tag_indices.dtype).to(self.device).unsqueeze(1) sequence_scores = inputs.squeeze(1)[example_inds, tag_indices].squeeze(1) sequence_scores = torch.where(sequence_lengths <= 0, torch.zeros(sequence_scores.shape).to(self.device), sequence_scores) return sequence_scores # Compute the scores of the given tag sequence. def _multi_seq_fn(): unary_scores = self.crf_unary_score(tag_indices, sequence_lengths, inputs) binary_scores = self.crf_binary_score(tag_indices, sequence_lengths) sequence_scores = unary_scores + binary_scores return sequence_scores if inputs.shape[1] == 1: return _single_seq_fn() else: return _multi_seq_fn() def crf_forward(self, inputs, state, sequence_lengths): """Computes the alpha values in a linear-chain CRF. See http://www.cs.columbia.edu/~mcollins/fb.pdf for reference. Args: inputs: A [batch_size, num_tags] matrix of unary potentials. state: A [batch_size, num_tags] matrix containing the previous alpha values. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: new_alphas: A [batch_size, num_tags] matrix containing the new alpha values. """ batch_size = inputs.shape[0] sequence_lengths, _ = torch.max( torch.stack( (torch.zeros(sequence_lengths.shape, dtype=sequence_lengths.dtype).to(self.device), sequence_lengths - 2), dim=1), dim=1) inputs = inputs.permute(1, 0, 2) transition_params_unsq = self.transition_params.unsqueeze(0) all_alphas = [] for idx in range(inputs.shape[0]): state = state.unsqueeze(2) transition_scores = state + transition_params_unsq new_alphas = inputs[idx] + torch.logsumexp(transition_scores, dim=1) state = new_alphas all_alphas.append(new_alphas) all_alphas = torch.stack(all_alphas, dim=1) return all_alphas[torch.arange(sequence_lengths.shape[0]).to(self.device), sequence_lengths] def crf_log_norm(self, inputs, sequence_lengths): """Computes the normalization for a CRF. Args: inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: log_norm: A [batch_size] vector of normalizers for a CRF. """ # Split up the first and rest of the inputs in preparation for the forward # algorithm. first_input = inputs.narrow(1, 0, 1).squeeze(1) # If max_seq_len is 1, we skip the algorithm and simply reduce_logsumexp over # the "initial state" (the unary potentials). def _single_seq_fn(): log_norm = torch.logsumexp(first_input, dim=1) log_norm = torch.where(sequence_lengths <= 0, torch.zeros(log_norm.shape).to(self.device), log_norm) return log_norm def _multi_seq_fn(): rest_of_input = inputs.narrow(1, 1, inputs.shape[1]-1) alphas = self.crf_forward(rest_of_input, first_input, sequence_lengths) log_norm = torch.logsumexp(alphas, dim=1) log_norm = torch.where(sequence_lengths <= 0, torch.zeros(log_norm.shape).to(self.device), log_norm) return log_norm if inputs.shape[1] == 1: return _single_seq_fn() else: return _multi_seq_fn() def crf_log_likelihood(self, inputs, tag_indices, sequence_lengths): """Computes the log-likelihood of tag sequences in a CRF. Args: inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer. tag_indices: A [batch_size, max_seq_len] matrix of tag indices for which we compute the log-likelihood. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: log_likelihood: A [batch_size] `Tensor` containing the log-likelihood of each example, given the sequence of tag indices. """ # Get shape information. num_tags = inputs.shape[2] sequence_scores = self.crf_sequence_score(inputs, tag_indices, sequence_lengths) log_norm = self.crf_log_norm(inputs, sequence_lengths) log_likelihood = sequence_scores - log_norm return log_likelihood def crf_unary_score(self, tag_indices, sequence_lengths, inputs): """Computes the unary scores of tag sequences. Args: tag_indices: A [batch_size, max_seq_len] matrix of tag indices. sequence_lengths: A [batch_size] vector of true sequence lengths. inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials. Returns: unary_scores: A [batch_size] vector of unary scores. """ batch_size = inputs.shape[0] max_seq_len = inputs.shape[1] num_tags = inputs.shape[2] flattened_inputs = torch.flatten(inputs) offsets = torch.unsqueeze(torch.arange(batch_size).to(self.device) * max_seq_len * num_tags, 1) offsets = torch.add(offsets, torch.unsqueeze(torch.arange(max_seq_len).to(self.device) * num_tags, 0)) flattened_tag_indices = offsets + tag_indices flattened_tag_indices = torch.flatten(flattened_tag_indices) unary_scores = torch.gather(flattened_inputs, 0, flattened_tag_indices).view(batch_size, max_seq_len) masks = torch.arange(unary_scores.shape[1]).to(self.device)[None, :] < sequence_lengths[:, None] masks = masks.squeeze(1).long() unary_scores = torch.sum(unary_scores * masks, dim=1) return unary_scores def crf_binary_score(self, tag_indices, sequence_lengths): """Computes the binary scores of tag sequences. Args: tag_indices: A [batch_size, max_seq_len] matrix of tag indices. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: binary_scores: A [batch_size] vector of binary scores. """ # Get shape information. batch_size = tag_indices.shape[0] num_tags = self.transition_params.shape[0] num_transitions = tag_indices.shape[1] - 1 start_tag_indices = tag_indices.narrow(1, 0, num_transitions) end_tag_indices = tag_indices.narrow(1, 1, num_transitions) # Encode the indices in a flattened representation. flattened_transition_indices = start_tag_indices * num_tags + end_tag_indices flattened_transition_indices = flattened_transition_indices.flatten() flattened_transition_params = self.transition_params.flatten() # Get the binary scores based on the flattened representation. binary_scores = torch.gather(flattened_transition_params, 0, flattened_transition_indices).view(batch_size, -1) masks = torch.arange(tag_indices.shape[1]).to(self.device)[None, :] < sequence_lengths[:, None] masks = masks.squeeze(1).long() truncated_masks = masks.narrow(1, 1, masks.shape[1]-1) binary_scores = torch.sum(binary_scores * truncated_masks, dim=1) return binary_scores def viterbi_decode_batch(self, feats, sequence_lengths): """Computes the most likely sequence with the viterbi algorithm for batch inputs. Args: inputs: A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer. sequence_lengths: A [batch_size] vector of true sequence lengths. Returns: viterbi_sequence: A [batch_size, max_seq_len] vector of most likely sequences. viterbi_score: The corresponding score of the viterbi_sequence bool_mask: Input mask corresponding to the given lengths """ bool_masks = torch.arange(feats.shape[1]).to(self.device)[None, :] < sequence_lengths[:, None] bool_masks = bool_masks.squeeze(1) trellis = torch.zeros(feats.shape).to(self.device) backpointers = torch.zeros(feats.shape, dtype=torch.int64).to(self.device) trellis[:, 0] = feats[:, 0] for t in range(1, feats.shape[1]): v = torch.unsqueeze(trellis[:, t - 1], 2) + self.transition_params val_max, arg_max = torch.max(v, dim=1) bool_mask = bool_masks[:, t] trellis[:, t] = torch.where(bool_mask.unsqueeze(1), feats[:, t] + val_max, trellis[:, t-1]) backpointers[:, t]= torch.where( bool_mask.unsqueeze(1), arg_max, torch.argmax( trellis[:, t-1], dim=1).unsqueeze(1).expand( arg_max.shape)) viterbi_cand_val, viterbi_cand_idx = torch.max(trellis[:, -1], dim=1) viterbi = [viterbi_cand_idx] reverse_bps = torch.flip(backpointers[:, 1:], dims=[1]) for idx in range(reverse_bps.shape[1]): next_viterbi = torch.gather( reverse_bps[:, idx], 1, viterbi[-1].unsqueeze(1) ).squeeze(1) viterbi.append(next_viterbi) viterbi.reverse() viterbi_score, _ = torch.max(trellis[:, -1], dim=1) return torch.stack(viterbi, dim=1), viterbi_score, bool_masks
dave-s477/SoMeNLP
somenlp/utils/__init__.py
import torch import argparse import operator from functools import reduce from .time_marker import get_time_marker def str2bool(v): """Tranform a string value into bool Args: v (str): string indicating a bool value Raises: argparse.ArgumentTypeError: str is unsuited for parsing to boolean Returns: bool: result """ if isinstance(v, bool): return v if v.lower() in ('yes', 'True', 'true', 't', 'y', '1'): return True elif v.lower() in ('no', 'False', 'false', 'f', 'n', '0'): return False else: raise argparse.ArgumentTypeError('Boolean value expected.') def set_dropout(model, drop_rate=0.3): """Recursively set dropout in pytorch model Args: model (pytorch model): model drop_rate (float, optional): value for dropouts. Defaults to 0.3. """ for name, child in model.named_children(): if isinstance(child, torch.nn.Dropout): child.p = drop_rate set_dropout(child, drop_rate=drop_rate) # TODO recursive would be nicer.. def find_type_in_dict(dictionary, data_type=list): """Get all entries in a dictionary that are of a certain type up to depth 3 Args: dictionary (dict): input dictionary data_type (python data type identifier, optional): data type to match. Defaults to list. Returns: list: paths of matched elements within the dictionary """ paths_to_examine = [] for main_key, main_values in dictionary.items(): current_path = [main_key] if isinstance(main_values, data_type): paths_to_examine.append({ 'path': current_path, 'values': main_values }) elif isinstance(main_values, dict): for sub_key, sub_values in main_values.items(): current_path = [main_key, sub_key] if isinstance(sub_values, data_type): paths_to_examine.append({ 'path': current_path, 'values': sub_values }) elif isinstance(sub_values, dict): for sub_sub_key, sub_sub_values in sub_values.items(): current_path = [main_key, sub_key, sub_sub_key] if isinstance(sub_sub_values, data_type): paths_to_examine.append({ 'path': current_path, 'values': sub_sub_values }) return paths_to_examine def getFromDict(dataDict, mapList): return reduce(operator.getitem, mapList, dataDict) def setInDict(dataDict, mapList, value): getFromDict(dataDict, mapList[:-1])[mapList[-1]] = value def get_abbr(s): abbr = s[0] prev_char = s[0] for c in s[1:]: if prev_char in ['_', ' ']: abbr += c prev_char = c return abbr
dave-s477/SoMeNLP
somenlp/NER/__init__.py
from .output_handler import OutputHandler from .tuner import Tuner from .LSTM_dataset import LSTMDataset from .BERT_dataset import BERTDataset, BERTMultiDataset from .data_handler import DataHandler from .model_wrapper import ModelWrapper from .models import BiLSTM_CRF, FeatureLSTM, CombinedLSTM from .trainer import Trainer from .run_model import main, predict, tune
dave-s477/SoMeNLP
somenlp/NER/tuning_wrapper_lstm.py
<filename>somenlp/NER/tuning_wrapper_lstm.py import pickle import random import argparse import json from os.path import exists, join from shutil import copytree from itertools import product import main_lstm as lstm import main_lstm_with_features as feat_lstm import main_lstm_combined as comb_lstm def generate_all_configs(tuning_config): model_confs = [dict(zip(tuning_config["model"], v)) for v in product(*tuning_config["model"].values())] all_configs = [] for mc in model_confs: mc['test_name'] = get_model_name(mc) all_configs.append({ "mode": tuning_config['mode'], "data": tuning_config['data'], "model": mc, "analyses": tuning_config['analyses'] }) return all_configs def get_abbr(s): abbr = s[0] prev_char = s[0] for c in s[1:]: if prev_char == '_': abbr += c prev_char = c return abbr def get_model_name(model_config): base_name = 'O' for k, v in model_config.items(): corr_v = v if isinstance(corr_v, str): if not corr_v: corr_v = 'none' else: corr_v = corr_v.replace(r'/', r'-') corr_v = corr_v.replace(r'.', r'-') corr_v = corr_v.replace(r'>', r'') corr_v = corr_v.replace(r'<', r'') if isinstance(corr_v, dict): corr_s = '' for k2, v2 in corr_v.items(): corr_v2 = v2 if isinstance(corr_v2, str): corr_v2 = corr_v2.replace(r'/', r'-') corr_s = '{}_{}_{}'.format(corr_s, get_abbr(k2), corr_v2) corr_v = corr_s if isinstance(corr_v, bool): corr_v = str(corr_v)[0] base_name = '{}_{}_{}'.format(base_name, get_abbr(k), corr_v) if len(base_name) > 255: print(RuntimeWarning("Maximum naming lengths is reached, consider chaning the setup.")) return base_name if __name__ == "__main__": parser = argparse.ArgumentParser( description = "Run hyper-parameter tuning for Bi-LSTM-CRF.") parser.add_argument("--log-loc", default='', help="Name of log-dir.") parser.add_argument("--reset", action='store_true', help="Use new configs.") parser.add_argument("--config", required=True, help="name of config") args = parser.parse_args() with open(args.config, 'r') as t_file: config = json.load(t_file) for key, value in config['data'].items(): if key in ['pretrain', 'train', 'devel', 'test']: if value['text'] and not value['text'].startswith('/'): value['text'] = config['data']['base_dir'] + value['text'] if value['text'] else '' value['features'] = config['data']['base_dir'] + value['features'] if value['features'] else '' value['labels'] = config['data']['base_dir'] + value['labels'] if value['labels'] else '' if 'prepro' in value.keys() and 'arg' in value['prepro'].keys(): value['prepro']['arg'] = config['data']['base_dir'] + value['prepro']['arg'] value['relations']['tag_names'] = config['data']['base_dir'] + value['relations']['tag_names'] value['relations']['relations'] = config['data']['base_dir'] + value['relations']['relations'] training_confs = generate_all_configs(config) for c in training_confs: print("Training a model with config:") print(json.dumps(c, indent=4)) if 'combined' in args.config: print("Running a combined LSTM") comb_lstm.train(c) elif 'with_features' in args.config: print("Running a custom feature LSTM") feat_lstm.train(c) else: print("Running a plain LSTM") lstm.train(c)
dave-s477/SoMeNLP
somenlp/NER/seqeval_custom/__init__.py
from .metrics.sequence_labeling import classification_report, precision_recall_fscore_support
dave-s477/SoMeNLP
somenlp/NER/output_handler.py
<reponame>dave-s477/SoMeNLP import json from torch.utils.tensorboard import SummaryWriter from pathlib import Path from sklearn.metrics import confusion_matrix from articlenizer.formatting import bio_to_brat class OutputHandler(): def __init__(self, name, time='0_0_0', checkpoint={}, log_dir='logs', save_dir='save'): if 'model' in checkpoint and checkpoint['model']: self.model_loc = checkpoint['model'] else: self.model_loc = '' if self.model_loc and 'log_dir' in checkpoint: self.log_dir = Path(checkpoint['log_dir']) else: self.log_dir = Path('{}/{}/{}'.format(log_dir, name, time)) self.writer = SummaryWriter(self.log_dir) if self.model_loc and 'save_dir' in checkpoint: self.save_dir = Path(checkpoint['save_dir']) else: self.save_dir = Path('{}/{}/{}'.format(save_dir, name, time)) self.save_dir.mkdir(parents=True, exist_ok=True) def save_json(self, data, name='encoding'): with open('{}/{}.json'.format(self.save_dir, name), 'w') as json_file: json.dump(data, json_file, indent=4) def load_encoding(self): with open('{}/encoding.json'.format(self.save_dir), 'r') as json_file: encoding_dict = json.load(json_file) return encoding_dict def print_scalars(self, scalars, epoch, name, meta_name=''): out_s = 'Classification result on {}{} ep {}:\n'.format(meta_name, name, epoch) for idx, (k, v) in enumerate(scalars.items()): if idx % 3 == 0: out_s += '\n' out_s += '{}:\t{}\n'.format(k, round(v, 3)) out_s += 'Done\n\n' print(out_s) def write_scalars(self, scalars, epoch): for scalar_key, scalar_value in scalars.items(): self.writer.add_scalar(scalar_key, scalar_value, epoch) def print_errors(self, labels, predictions, sentences, max_output_length, data_set_name, word2name): out_s = '' for sent_true, sent_pred, sent_coded_words in zip(labels, predictions, sentences): if sent_true != sent_pred: out_s += "Wrong sentence:\n" output_string_true, output_string_pred, output_string_words = '', '', '' if isinstance(word2name, dict): sentence_words = [word2name[w] for w in sent_coded_words] else: sentence_words = word2name.convert_ids_to_tokens(sent_coded_words) for true_label, predicted_label, word in zip(sent_true, sent_pred, sentence_words): next_length = max(len(word), len(true_label), len(predicted_label)) + 1 if (len(output_string_words) + next_length) > max_output_length: out_s += 'Sent:\t{}\n'.format(output_string_words) out_s += 'True:\t{}\n'.format(output_string_true) out_s += 'Pred:\t{}\n'.format(output_string_pred) output_string_true, output_string_pred, output_string_words = '', '', '' output_string_words += '{:{}s}'.format(word, next_length) output_string_true += '{:{}s}'.format(true_label, next_length) output_string_pred += '{:{}s}'.format(predicted_label, next_length) if output_string_words: out_s += 'Sent:\t{}\n'.format(output_string_words) out_s += 'True:\t{}\n'.format(output_string_true) out_s += 'Pred:\t{}\n'.format(output_string_pred) out_s += '\n\n' out_loc = self.save_dir / '{}_tagging_errors.txt'.format(data_set_name) with out_loc.open(mode='w') as out_f: out_f.write(out_s) def c_matrix(self, names, labels, predictions, tag_mode): tags = [] for n in names: if n != 'O': tags.append('B-{}'.format(n)) tags.append('I-{}'.format(n)) if tag_mode == 'bioes': tags.append('S-{}'.format(n)) tags.append('E-{}'.format(n)) tags.append('O') t = [item for sublist in labels for item in sublist] p = [item for sublist in predictions for item in sublist] unique_tags = set(tags) cm = confusion_matrix(t, p, labels=tags) out_s = """ Confusion Matrix for: {} {} """.format(tags, cm) print(out_s) def save_predictions_fct(self, path, predictions, text): with path['out'].open(mode='w') as out_f: for preds in predictions: out_f.write('{}\n'.format(' '.join(preds).rstrip())) with path['out-text'].open(mode='w') as out_t: for line in text: out_t.write('{}\n'.format(' '.join(line).rstrip())) def save_predictions(self, path, predictions, text): if not isinstance(predictions, dict): self.save_predictions_fct(path, predictions, text) else: path_out = str(path['out']) for k, v in predictions.items(): path['out'] = Path(path_out + '.' + k) self.save_predictions_fct(path, v, text) def summarize_predictions_fct(self, path, predictions, text): print("Predicted entities for {}".format(path.name)) out_path = Path(str(path) + '.sum') entities, _, _ = bio_to_brat(text, predictions, split_sent=False, split_words=False) out_s = '' for e in entities: out_s += '{}\t{} {} {}\t{}\n'.format(e['id'], e['type'], e['beg'], e['end'], e['string']) with out_path.open(mode='w') as out_f: out_f.write(out_s) print(out_s) def summarize_predictions(self, path, predictions, text): if not isinstance(predictions, dict): self.summarize_predictions_fct(path['out'], predictions, text) else: for k, v in predictions.items(): self.summarize_predictions_fct(Path(str(path['out']) + k), v, text) def cl_for_latex(self, dictionary, round_n=2): new_mapping = {} for k,v in dictionary.items(): for label, values in v.items(): if label not in new_mapping: new_mapping[label] = {} for metric, result in values.items(): new_mapping[label][metric + '_' + k] = round(result, round_n) separator='&' s = '' for k,v in new_mapping.items(): s += '{} {} {} ({}) {} {} ({}) {} {} ({}) {} {} ({})\n'.format( k, separator, v['precision_test_0'], v['precision_devel_0'], separator, v['recall_test_0'], v['recall_devel_0'], separator, v['f1-score_test_0'], v['f1-score_devel_0'], separator, v['support_test_0'], v['support_devel_0'] ) return s
dave-s477/SoMeNLP
somenlp/feature_engineering/__init__.py
<filename>somenlp/feature_engineering/__init__.py from .gen_custom_features import calculate_features_parallel
dave-s477/SoMeNLP
somenlp/distant_supervision/gen_sequences.py
from itertools import product GEN_SYMBOLS = ['T', 'C', 'G', 'A'] def generate_triplets(): """Generate all possible triples Returns: dictionary: dictionary with list of possible GEN triplets """ gen_triplets = [''.join(seq) for seq in product(GEN_SYMBOLS, repeat = 3)] return {'gen_triplets': gen_triplets}
dave-s477/SoMeNLP
somenlp/entity_disambiguation/efficient_prediction.py
<gh_stars>0 import torch import torch.nn as nn import os import json import pickle import random import copy import time import re import math import numpy as np import nltk import itertools nltk.download('stopwords') from functools import partial from nltk.corpus import stopwords from torch.utils.data import DataLoader, IterableDataset, Dataset #torch.multiprocessing.set_sharing_strategy('file_system') from heapq import merge from pathlib import Path from . import EntityDisambiguationFeatureGenerator from .model import DisambiguationModel from sklearn.metrics import classification_report BLOCK_SIZE = 5000000 STOPS = stopwords.words('english') def normalize(s): norm_s = re.sub('[^0-9a-zA-Z]+', ' ', s.casefold()).rstrip('0123456789 ,.').lstrip(' ') norm_s = ' '.join([w for w in norm_s.split() if w not in STOPS]) if not norm_s: norm_s = s return norm_s def remove_spaces(s): replace_regex = re.compile(r'\s(?P<to_keep>[\+\-#™_/\d]+)\s?') matches = replace_regex.findall(s) return replace_regex.sub(r'\g<to_keep>', s) class ReducedSampleSet(): def __init__(self, in_path, overview_file, save_path='/tmp'): self.output_path = '{}/reduced_features.json'.format(save_path) if os.path.isfile(self.output_path): self._load() else: self._sample_overview(overview_file) self._generate(in_path) def _sample_overview(self, overview_file): self.sample_overview = {} with open(overview_file, 'r') as f_in: for line in f_in: num, key = line.rstrip().split(maxsplit=1) self.sample_overview[key] = int(num) def _load(self): with open(self.output_path, 'r') as j_in: self.sample_set = json.load(j_in) def _save(self): with open(self.output_path, 'w') as j_out: json.dump(self.sample_set, j_out, indent=4) def _generate(self, in_path): file_count = 0 self.sample_set = {} print("Gathering files..") file_list = Path(in_path).rglob('*.linking') for f in file_list: file_count += 1 if file_count % 10000 == 0: print("At file {}".format(file_count)) with f.open() as f_in: entities = json.load(f_in) for entity in entities: name = entity.pop("mention") if name not in self.sample_set: self.sample_set[name] = { "mention": name, "string": remove_spaces(name), "norm": normalize(name), "contexts": [] } if len(self.sample_set[name]['contexts']) < 5: self.sample_set[name]['contexts'].append(copy.deepcopy(entity)) else: keep_val = random.random() occurrence_num = self.sample_overview[name] if name in self.sample_overview else 1000 if keep_val <= 1 / occurrence_num: #print("Replace a value of {} with {} and num {}".format(name, keep_val, self.sample_overview[name])) keep_pos = random.randint(0, 4) self.sample_set[name]['contexts'][keep_pos] = copy.deepcopy(entity) self._save() # class IterDataset(IterableDataset): # def __init__(self, data, sample_set, feature_calc, compare_set, labels=None): # self.data = data # self.sample_set = sample_set # self.feature_calc = feature_calc # self.compare_set = compare_set # self.labels = labels # def __iter__(self): # for x in self.data: # if self.compare_set is None: # sample = [self.sample_set[x[0]], self.sample_set[x[1]]] # else: # sample = [self.sample_set[x[0]], self.compare_set[x[1]]] # features = self.feature_calc.features_for_pair(sample) # origin = [sample[0]['origin_id'], sample[1]['origin_id']] # if self.labels is None: # label = 0 # else: # label = self.labels[sample[0]['origin_id']][sample[1]['origin_id']] # yield torch.tensor(features), label, x, origin class IterDataset(Dataset): def __init__(self, data, sample_set, feature_calc, compare_set, labels=None): self.data = data self.sample_set = sample_set self.feature_calc = feature_calc self.compare_set = compare_set self.labels = labels def __len__(self): return len(self.data) def __getitem__(self, idx): x = self.data[idx] if self.compare_set is None: sample = [self.sample_set[x[0]], self.sample_set[x[1]]] else: sample = [self.sample_set[x[0]], self.compare_set[x[1]]] features = self.feature_calc.features_for_pair(sample) origin = [sample[0]['origin_id'], sample[1]['origin_id']] if self.labels is None: label = 0 else: label = self.labels[sample[0]['origin_id']][sample[1]['origin_id']] return torch.tensor(features), label, x, origin def worker_init_fn(_): worker_info = torch.utils.data.get_worker_info() dataset = worker_info.dataset worker_id = worker_info.id split_size = math.ceil(len(dataset.data) / worker_info.num_workers) dataset.data = dataset.data[worker_id * split_size:(worker_id + 1) * split_size] class DistanceMap(): def __init__(self, sample_set, dbpedia, model_config, start_row, end_row, threshold=0.01, compare_set=None, n_cores=8, save_path='/tmp', batch_size=500, device='cpu'): self.device = torch.device(device) self.distance_map = [] self.sample_set = sample_set self.compare_set = compare_set self.sample_num = len(self.sample_set) self.max_idx = int(self.sample_num * ( self.sample_num - 1 ) / 2) self.threshold = threshold self.cores = n_cores self.batch_size = batch_size self.start_row = start_row self.end_row = end_row self.dbpedia = dbpedia self.model_config = model_config self.output_path = '{}/distance_map_{}_{}.p'.format(save_path, start_row, end_row) def calculate_distance_map(self, compare=False): if os.path.isfile(self.output_path): self._load() else: self.feature_generator = EntityDisambiguationFeatureGenerator(self.dbpedia) self._setup_model(self.model_config) if not compare: self._generate() else: self._compare() def _setup_model(self, config): self.model = DisambiguationModel(len(self.feature_generator.string_features_to_extract) + len(self.feature_generator.context_features_to_extract), self.model_config['layer_sizes'], self.model_config['drop_outs']) if not 'checkpoint' in config or not config['checkpoint']: raise(RuntimeError("A pretrained model is required for prediction..")) checkpoint_data = torch.load(config['checkpoint'], map_location=self.device) self.model.load_state_dict(checkpoint_data['model_state_dict']) self.model.eval() def _load(self): with open(self.output_path, 'rb') as p_in: self.distance_map = pickle.load(p_in) def _save(self): with open(self.output_path, 'wb') as p_out: pickle.dump(self.distance_map, p_out) def _generator(self): for idx_i in range(self.start_row, self.end_row): for idx_j in range(idx_i+1, len(self.sample_set)): yield idx_i, idx_j def _compare_generator(self): for idx_i in range(len(self.sample_set)): for idx_j in range(len(self.compare_set)): yield idx_i, idx_j def _prune_predictions(self, predictions): pruned = [] last_source = [] for prediction in predictions: if prediction[2] != last_source: pruned.append(prediction) else: if prediction[0] < pruned[-1][0]: pruned[-1][0] = prediction[0] last_source = prediction[2] return pruned def _generate(self): generator = self._generator() self._run_generation(generator) def _compare(self): print("Doing compare") generator = self._compare_generator() self._run_generation(generator) def _run_generation(self, generator): count = 0 while True: print("At block {}".format(count)) count += 1 start = time.time() block = list(itertools.islice(generator, BLOCK_SIZE)) if not block: break iterable_dataset = IterDataset(list(block), self.sample_set, self.feature_generator, self.compare_set) loader = DataLoader(iterable_dataset, batch_size=self.batch_size, num_workers=self.cores, pin_memory=True)#self.batch_size, self.cores) predictions = [] with torch.no_grad(): for sample in loader: #print(type(sample)) pred = self.model(sample[0].to(self.device)) pred_class = 1 - torch.squeeze(torch.sigmoid(pred)).cpu().numpy() results = [[x, [int(idx_x), int(idx_y)], [int(org_x), int(org_y)]] for x, idx_x, idx_y, org_x, org_y in zip(pred_class, sample[2][0], sample[2][1], sample[3][0], sample[3][1]) if x <= self.threshold] predictions.extend(results) #print(results) print("preds") print(len(predictions)) # pruned_predictions = self._prune_predictions(predictions) # print("pruned") # print(len(pruned_predictions)) predictions.sort() self.distance_map = list(merge(self.distance_map, predictions)) end = time.time() print("Took {}".format(round(end-start, 5))) self._save() class EfficientClustering(): def __init__(self, sorted_list, threshold, sample_set, save_path='/tmp', ncores=8): self.dim = len(sample_set) self.entities = sample_set self.threshold = threshold self.output_path = '{}/clusters.json'.format(save_path) self.clusters = list(range(len(sample_set))) self.features = sorted_list self.ncores = ncores self.reverse_index = {} def cluster(self): if os.path.isfile(self.output_path): self._load() else: self._cluster() def _load(self): with open(self.output_path, 'r') as j_in: self.clusters = json.load(j_in) def _save(self): with open(self.output_path, 'w') as j_out: json.dump(self.clusters, j_out, indent=4) def _get_cluster_idx(self, idx): cluster_value = self.clusters[idx] while idx != cluster_value: idx = cluster_value cluster_value = self.clusters[idx] return idx def _match_clusters(self, matches): clusters_to_match = [set(x) for x in matches] global_index = len(clusters_to_match) - 1 while global_index > 0: next_match = clusters_to_match[global_index] for idx, match in enumerate(clusters_to_match[:global_index]): if next_match & match: new_matches = next_match.union(match) clusters_to_match[idx] = new_matches clusters_to_match.pop(global_index) break global_index -= 1 return clusters_to_match def _generator(self, start, end, clusters): for idx_i in range(start, end): for idx_j in range(idx_i+1, len(clusters)): yield idx_i, idx_j def _get_postprocessing_merges(self, rows, clusters): print("Len clusters {}".format(len(clusters))) gen = self._generator(rows[0], rows[1], clusters) merges = [] for i in gen: if clusters[i[0]] & clusters[i[1]]: merges.append(i) return merges def _iterate_distances(self, threshold, cut_off_idx=0): start = time.time() if cut_off_idx > 0: features = self.features[cut_off_idx:] else: features = self.features for idx, i in enumerate(features): if (idx+1) % 10000 == 0: end = time.time() print("At {}: {}, last step took {}".format(idx, i, round(end-start, 4))) start = end dist_val, pair_indices, _ = i if dist_val > threshold: break x_idx, y_idx = pair_indices x_pointer = self.clusters[x_idx] y_pointer = self.clusters[y_idx] if x_pointer == y_pointer: # already a correct link -> do nothing pass elif x_pointer == x_idx and y_pointer == y_idx: # two "original" samples -> new cluster self.clusters[x_idx] = len(self.clusters) self.clusters[y_idx] = len(self.clusters) self.reverse_index[len(self.clusters)] = [x_idx, y_idx] self.clusters.append(len(self.clusters)) elif x_pointer != x_idx and y_pointer == y_idx: # first point does already belong to a cluster, second is "original" -> add second point to cluster self.clusters[y_idx] = x_pointer self.reverse_index[x_pointer].append(y_idx) elif y_pointer != y_idx and x_pointer == x_idx: # second point does already belong to a cluster, first is "original" -> add first point to cluster self.clusters[x_idx] = y_pointer self.reverse_index[y_pointer].append(x_idx) elif y_pointer != y_idx and x_pointer != x_idx: # both points already belong to clusters -> merge clusters (we will add one "empty" cluster in the result) reverse_y_pointers = self.reverse_index.pop(y_pointer) for p in reverse_y_pointers: self.clusters[p] = x_pointer self.reverse_index[x_pointer].append(p) return idx + cut_off_idx def _values_to_clusters(self): cluster_buckets = {} for i in range(len(self.clusters)-1, -1, -1): if i == self.clusters[i]: cluster_buckets[i] = { 'name': i, 'indices': [i], 'entities': [] } else: bucket_idx = self._get_cluster_idx(i) cluster_buckets[bucket_idx]['indices'].append(i) return cluster_buckets def _print_values(self, n=100): for i in [x for x in self.features if x[0] < self.threshold][-100:]: dist, pair_indices, _ = i x_idx, y_idx = pair_indices print(dist) print('x = {}'.format(self.entities[x_idx])) print('y = {}'.format(self.entities[y_idx])) print() def _cluster(self): self._pre_cluster() print("Iterating distances..") self._iterate_distances(self.threshold) print("Writing values to clusters..") cluster_buckets = self._values_to_clusters() for bucket in cluster_buckets: cluster_buckets[bucket]['indices'] = [x for x in cluster_buckets[bucket]['indices'] if x < self.dim] for cluster in cluster_buckets: for idx in cluster_buckets[cluster]['indices']: cluster_buckets[cluster]['entities'].append(self.entities[idx]) self.clusters = cluster_buckets print("Removing empty clusters..") self.clusters = {x:y for x,y in self.clusters.items() if y['entities']} for cluster, values in self.clusters.items(): values['origin_ids'] = [x['origin_id'] for x in values['entities']] self._save() def _pre_cluster(self): print("Performing pre-clustering on exact string matches.") name_index = {} for idx, ent in enumerate(self.entities): if idx % 10000 == 0: print("At entity {}".format(idx)) if ent['mention'] not in name_index: name_index[ent['mention']] = [] name_index[ent['mention']].append(idx) for idx, (k, v) in enumerate(name_index.items()): if idx % 10000 == 0: print("At pre-cluster {}".format(idx)) if len(v) > 1: cluster_id = len(self.clusters) self.reverse_index[cluster_id] = [] self.clusters.append(cluster_id) for ent_id in v: self.clusters[ent_id] = cluster_id self.reverse_index[cluster_id].append(ent_id) def evaluate(self, gold_table, step_size, start_eval_threshold): print("Evaluating clustering..") self._pre_cluster() current_threshold = 0 iteration_index = 0 prev_num_clusters = len(self.clusters) while current_threshold < self.threshold: current_threshold = min(self.threshold, current_threshold + step_size) iteration_index = self._iterate_distances(current_threshold, cut_off_idx=iteration_index) if current_threshold >= start_eval_threshold and len(self.clusters) > prev_num_clusters: print("Added clusters - start evaluation") prev_num_clusters = len(self.clusters) eval_entities_with_clusters = [[x, y] for x, y in zip(self.entities, self.clusters) if 'eval' in x] print(len(eval_entities_with_clusters)) # for x in eval_entities_with_clusters: # print(x) # print() # break preds = [] trues = [] for idx, entity in enumerate(eval_entities_with_clusters): for comp_entity in eval_entities_with_clusters[idx+1:]: pred_link = entity[1] == comp_entity[1] preds.append(pred_link) true_link = gold_table[entity[0]['origin_id']][comp_entity[0]['origin_id']] trues.append(true_link) print("Threshold {}".format(current_threshold)) # print("Iteration idx {}".format(iteration_index)) print("Num clusters: {} (not correct - just for debugging)".format(len(self.clusters))) # print(classification_report(trues, preds)) output_dict = classification_report(trues, preds, output_dict=True) logging_string = 'Graph_out:\t{},{},{},{}'.format(current_threshold, output_dict['True']['precision'], output_dict['True']['recall'], output_dict['True']['f1-score']) print(logging_string) print() # def _post_pro_merge(self): # total_size = len(self.clusters) * (len(self.clusters) - 1) / 2 # max_samples = math.ceil(total_size / self.ncores) # row_overview = [0] # curr_size = 0 # for row in range(len(self.clusters)-1): # curr_size += len(self.clusters)-1-row # if curr_size > max_samples: # row_overview.append(row) # curr_size = 0 # row_overview.append(row) # row_pairs = [[row_overview[x-1], row_overview[x]] for x in range(1, len(row_overview))] # print(row_overview) # print("pairs") # print(row_pairs) # with Pool(len(row_overview)) as p: # origin_ids = [x['origin_ids'] for x in self.clusters.values()] # fct = partial(self._get_postprocessing_merges, clusters=origin_ids) # post_processing_merges = p.map(fct, row_pairs) # TODO -> flatten list # post_processing_merges = [item for sublist in post_processing_merges for item in sublist] # post_processing_merges = [] # for idx_x, (cluster_x, values_x) in enumerate(self.clusters.items()): # if idx_x % 5000 == 0: # print("At cluster {}".format(idx_x)) # for cluster_y, values_y in list(self.clusters.items())[idx_x+1:]: # for ent_x in values_x['entities']: # matched = False # for ent_y in values_y['entities']: # if ent_x['mention'] == ent_y['mention']: # matched = True # #if values_x['origin_ids'] & values_y['origin_ids']: # post_processing_merges.append([cluster_x, cluster_y]) # break # if matched: # break # p_merges = self._match_clusters(post_processing_merges) # for ppm in p_merges: # ppm = list(ppm) # for i in ppm[1:]: # values = self.clusters.pop(i) # self.clusters[ppm[0]]['indices'].extend(values['indices']) # self.clusters[ppm[0]]['entities'].extend(values['entities']) # self.clusters[ppm[0]]['origin_ids'].update(values['origin_ids'])
dave-s477/SoMeNLP
somenlp/entity_disambiguation/clustering.py
import numpy as np import math import time import bisect import sys from statistics import mean from multiprocessing import Pool class SimpleCluster: """Cluster a set of entites """ def __init__(self, config, entities, features): """Init Args: entities (list): list of entity mentions with additional information features (dictionary): dictionary of triangular numpy matricies containing features for clustering """ self.config = config self.features = [] for idx, i in enumerate(features): if i <= config['threshold']: self.features.append([i[0], idx]) self.features.sort() self.dim = len(entities) self.entities = entities self.reverse_index = {} #self.clusters = {idx: {'entities': [x]} for idx, x in enumerate(entities)} self.clusters = list(range(len(entities))) #self.clusters = {x: x for x in range(len(entities))} def _get_eval_indices(self): eval_indices = [] for _, cluster in self.clusters.items(): indices = [x['init_id'] for x in cluster['entities']] eval_indices.append(indices) return eval_indices def _idx_to_matrix_pos(self, idx): i = self.dim - 2 - int(math.sqrt(-8*idx + 4*self.dim*(self.dim-1)-7)/2.0 - 0.5) j = idx + i + 1 - self.dim*(self.dim-1)/2 + (self.dim-i)*((self.dim-i)-1)/2 return i, int(j) def _get_cluster_idx(self, idx): #print("idx..") #print(idx) cluster_value = self.clusters[idx] #print(cluster_value) while idx != cluster_value: idx = cluster_value #print(idx) cluster_value = self.clusters[idx] #print() return idx def cluster(self): for i in self.features: _, pair_index = i x_idx, y_idx = self._idx_to_matrix_pos(pair_index) x_pointer = self.clusters[x_idx] y_pointer = self.clusters[y_idx] if x_pointer == y_pointer: # already a correct link -> do nothing pass elif x_pointer == x_idx and y_pointer == y_idx: # two "original" samples -> new cluster self.clusters[x_idx] = len(self.clusters) self.clusters[y_idx] = len(self.clusters) self.reverse_index[len(self.clusters)] = [x_idx, y_idx] self.clusters.append(len(self.clusters)) elif x_pointer != x_idx and y_pointer == y_idx: # first point does already belong to a cluster, second is "original" -> add second point to cluster self.clusters[y_idx] = x_pointer self.reverse_index[x_pointer].append(y_idx) elif y_pointer != y_idx and x_pointer == x_idx: # second point does already belong to a cluster, first is "original" -> add first point to cluster self.clusters[x_idx] = y_pointer self.reverse_index[y_pointer].append(x_idx) elif y_pointer != y_idx and x_pointer != x_idx: # both points already belong to clusters -> merge clusters (we will add one "empty" cluster in the result) #self.clusters = [i if i == idx or i != y_pointer else x_pointer for idx, i in enumerate(self.clusters)] #self.clusters[:self.dim] = [i if i == idx or i != y_pointer else x_pointer for idx, i in enumerate(self.clusters[:self.dim])] reverse_y_pointers = self.reverse_index.pop(y_pointer) for p in reverse_y_pointers: self.clusters[p] = x_pointer self.reverse_index[x_pointer].append(p) #print(len(self.clusters)) cluster_buckets = {} for i in range(len(self.clusters)-1, -1, -1): #print(i) #print(self.clusters[i]) if i == self.clusters[i]: cluster_buckets[i] = { 'name': i, 'indices': [i], 'entities': [] } else: bucket_idx = self._get_cluster_idx(i) cluster_buckets[bucket_idx]['indices'].append(i) #print(len(cluster_buckets)) for bucket in cluster_buckets: #print(bucket) #print(cluster_buckets[bucket]) cluster_buckets[bucket]['indices'] = [x for x in cluster_buckets[bucket]['indices'] if x < self.dim] #print(cluster_buckets[bucket]) #print() for cluster in cluster_buckets: for idx in cluster_buckets[cluster]['indices']: cluster_buckets[cluster]['entities'].append(self.entities[idx]) self.clusters = cluster_buckets class Clustering: """Cluster a set of entites """ def __init__(self, config, entities, features): """Init Args: entities (list): list of entity mentions with additional information features (dictionary): dictionary of triangular numpy matricies containing features for clustering """ self.config = config self.features = features # self._calculate_distances() self.threshold = config['threshold'] self.dim = len(entities) self.clusters = {idx: {'entities': [x]} for idx, x in enumerate(entities)} self.new_clusters = list(self.clusters.keys()) self.cluster_count = len(self.clusters) # def _calculate_distances(self): # if self.config['feat_to_dist'] == 'average': # self.distances = np.mean(self.features, axis=-1) # print(self.distances) # elif self.config['feat_to_dist'] == 'sum': # self.distances = np.sum(self.features, axis=-1) # else: # raise(RuntimeError("Got unknown feat to dist config: {}".format(self.config['feat_to_dist']))) # self.min = self.distances.min() # print(self.min) # self.max = self.distances.max() # print(self.max) # self.threshold = self.min + (self.max - self.min) * self.config['threshold'] # print(self.threshold) def _get_eval_indices(self): eval_indices = [] for _, cluster in self.clusters.items(): indices = [x['init_id'] for x in cluster['entities']] eval_indices.append(indices) return eval_indices def _get_matrix_index(self, x, y): if x > y: x1 = y x2 = x else: x1 = x x2 = y return int((x1 * self.dim) - ((x1+1) * ((x1+1) + 1) / 2 ) + x2) def _calculate_distance(self, cluster_id): distances = {} for c_k, cluster in self.clusters.items(): if c_k == cluster_id: distances[cluster_id] = 1.0 else: cluster_distances = [] for c_ent in cluster['entities']: for id_ent in self.clusters[cluster_id]['entities']: cluster_distances.append(self.features[self._get_matrix_index(c_ent['init_id'], id_ent['init_id'])]) if self.config['linkage'] == 'single': distances[c_k] = float(min(cluster_distances)) elif self.config['linkage'] == 'complete': distances[c_k] = float(max(cluster_distances)) elif self.config['linkage'] == 'average': distances[c_k] = mean([float(x) for x in cluster_distances]) else: raise(RuntimeError("Received unkown type of linkage {}".format(self.config['linkage']))) return distances def _calculate_distances_cluster_parallel(self): pass def _match_clusters(self, matches): clusters_to_match = [set(x) for x in matches] global_index = len(clusters_to_match) - 1 while global_index > 0: next_match = clusters_to_match[global_index] for idx, match in enumerate(clusters_to_match[:global_index]): if next_match & match: new_matches = next_match.union(match) clusters_to_match[idx] = new_matches clusters_to_match.pop(global_index) break global_index -= 1 return clusters_to_match def cluster(self): # max_count = 0 while len(self.new_clusters) > 0: # and max_count < 1000: # max_count += 1 if len(self.new_clusters) > 1: # calculate distance to all other clusters parallel with defined linkage type for cluster in self.new_clusters: dist = self._calculate_distance(cluster) self.clusters[cluster]['distances'] = dist else: # calculate distance to all other clusters parallel with defined linkage type self.clusters[self.new_clusters[0]]['distances'] = self._calculate_distance(self.new_clusters[0]) # find smallest distance between all existing clusters.. test if clusters do still exist! smallest_dist = 1.0 matches = [] for c_k, cluster in self.clusters.items(): keys_to_pop = [] for dist_k, dist_value in cluster['distances'].items(): # TODO: test if clusters do still exist if dist_k in self.clusters: if math.isclose(smallest_dist, dist_value): to_append = sorted([c_k, dist_k]) if to_append not in matches: matches.append(to_append) elif dist_value < smallest_dist: smallest_dist = dist_value to_append = sorted([c_k, dist_k]) matches = [to_append] else: keys_to_pop.append(dist_k) for k_pop in keys_to_pop: cluster['distances'].pop(k_pop) if smallest_dist > self.threshold: print("Stopped clustering due to threshold criterion: T {}, D {}".format(round(self.threshold, 4), round(smallest_dist, 4))) break matched_clusters = self._match_clusters(sorted(matches)) # pop clusters to be merged from cluster list self.new_clusters = [] for cluster_group in matched_clusters: new_entities = [] for cluster_idx in cluster_group: clust_entities = self.clusters.pop(cluster_idx) new_entities.extend(clust_entities['entities']) self.clusters[self.cluster_count] = { 'entities': new_entities } self.new_clusters.append(self.cluster_count) self.cluster_count += 1 class IntervalClustering: """Cluster a set of entites """ def __init__(self, config, entities, features): """Init Args: entities (list): list of entity mentions with additional information features (dictionary): dictionary of triangular numpy matricies containing features for clustering """ self.config = config self.features = features self.threshold = config['threshold'] self.n_intervals = config['intervals'] self.n_cores = config['n_cores'] self.drop_below = config['drop_below_threshold'] self.dim = len(entities) self.clusters = {idx: {'entities': [x]} for idx, x in enumerate(entities)} self.new_clusters = list(self.clusters.keys()) self.cluster_count = len(self.clusters) self._calculate_intervals() self.cluster_distances = [] self.cluster_distances_keys = [] def _calculate_intervals(self): self.intervals = np.linspace(0, self.threshold, num=self.n_intervals, endpoint=True) def _get_eval_indices(self): eval_indices = [] for _, cluster in self.clusters.items(): indices = [x['init_id'] for x in cluster['entities']] eval_indices.append(indices) return eval_indices def _get_matrix_index(self, x, y): if x > y: x1 = y x2 = x else: x1 = x x2 = y return int((x1 * self.dim) - ((x1+1) * ((x1+1) + 1) / 2 ) + x2) def _calculate_distance(self, cluster_id): distances = {} for c_k, cluster in self.clusters.items(): if c_k == cluster_id: distances[cluster_id] = 1.0 else: cluster_distances = [] for c_ent in cluster['entities']: for id_ent in self.clusters[cluster_id]['entities']: cluster_distances.append(self.features[self._get_matrix_index(c_ent['init_id'], id_ent['init_id'])]) if self.config['linkage'] == 'single': distances[c_k] = float(min(cluster_distances)) elif self.config['linkage'] == 'complete': distances[c_k] = float(max(cluster_distances)) elif self.config['linkage'] == 'average': distances[c_k] = mean([float(x) for x in cluster_distances]) else: raise(RuntimeError("Received unkown type of linkage {}".format(self.config['linkage']))) return distances def _match_clusters(self, matches): clusters_to_match = [set(x) for x in matches] global_index = len(clusters_to_match) - 1 while global_index > 0: next_match = clusters_to_match[global_index] for idx, match in enumerate(clusters_to_match[:global_index]): if next_match & match: new_matches = next_match.union(match) clusters_to_match[idx] = new_matches clusters_to_match.pop(global_index) break global_index -= 1 return clusters_to_match def _calculate_distance(self, distance_tuple): single_distances = [] for first_entity in self.clusters[distance_tuple[0]]['entities']: for second_entity in self.clusters[distance_tuple[1]]['entities']: single_distances.append(self.features[self._get_matrix_index(first_entity['init_id'], second_entity['init_id'])]) if self.config['linkage'] == 'single': distance = float(min(single_distances)) elif self.config['linkage'] == 'complete': distance = float(max(single_distances)) elif self.config['linkage'] == 'average': distance = mean([float(x) for x in single_distances]) else: raise(RuntimeError("Received unkown type of linkage {}".format(self.config['linkage']))) if self.drop_below and distance > self.threshold: return None return distance def _calculate_distances_cluster_parallel(self, distance_tuples): with Pool(self.n_cores) as p: distances = p.map(self._calculate_distance, distance_tuples) return distances def _update_distance_list(self): distances_to_calculate = set() print(len(self.new_clusters)) while self.new_clusters: cluster = self.new_clusters.pop() distances_to_calculate.update([tuple([x, cluster]) if x < cluster else tuple([cluster, x]) for x in self.clusters.keys() if x != cluster]) print(len(distances_to_calculate)) distances = self._calculate_distances_cluster_parallel(distances_to_calculate) distances_with_keys = [[x, y] for x, y in zip(distances, distances_to_calculate) if x is not None] print("Benchmark") start = time.time() sorted(distances_with_keys) end = time.time() print("First sorting option: {}".format(round(end-start, 4))) start = time.time() qsort(distances_with_keys) end = time.time() print("First sorting option: {}".format(round(end-start, 4))) if not self.cluster_distances: distance_tuples = sorted(distances_with_keys) self.cluster_distances = [x[0] for x in distance_tuples] self.cluster_distances_keys = [x[1] for x in distance_tuples] else: for distance, keys in distances_with_keys: insertion_index = bisect.bisect(self.cluster_distances, distance) self.cluster_distances.insert(insertion_index, distance) self.cluster_distances_keys.insert(insertion_index, keys) # TODO: use bisect to get right insertion point for each element def _remove_from_distance_list(self, elements): indices = [] for idx, cluster_keys in enumerate(self.cluster_distances_keys): if cluster_keys[0] in elements or cluster_keys[1] in elements: indices.append(idx) for idx in reversed(indices): self.cluster_distances_keys.pop(idx) self.cluster_distances.pop(idx) def cluster(self): for interval in self.intervals[1:]: start = time.time() self._update_distance_list() # for cluster in self.new_clusters: # dist = self._calculate_distance(cluster) # self.clusters[cluster]['distances'] = dist end = time.time() print("Calculating distances took {}".format(round(end-start, 4))) start = end # # find smallest distance between all existing clusters.. test if clusters do still exist! # matches = [] bisection_point = bisect.bisect_right(self.cluster_distances, interval) print(bisection_point) clusters_to_merge = self.cluster_distances_keys[:bisection_point] matched_clusters = self._match_clusters(clusters_to_merge) end = time.time() print("Matching clusters took {}".format(round(end-start, 4))) start = end self.cluster_distances_keys = self.cluster_distances_keys[bisection_point:] self.cluster_distances = self.cluster_distances[bisection_point:] clusters_to_remove = set(sorted([item for sublist in matched_clusters for item in sublist])) self._remove_from_distance_list(clusters_to_remove) # for c_k, cluster in self.clusters.items(): # keys_to_pop = [] # for dist_k, dist_value in cluster['distances'].items(): # # TODO: test if clusters do still exist # if dist_k in self.clusters: # if dist_value <= interval: # to_append = sorted([c_k, dist_k]) # if to_append not in matches: # matches.append(to_append) # else: # keys_to_pop.append(dist_k) # for k_pop in keys_to_pop: # cluster['distances'].pop(k_pop) # matched_clusters = self._match_clusters(sorted(matches)) end = time.time() print("Removing entries took {}".format(round(end-start, 4))) start = end # # pop clusters to be merged from cluster list self.new_clusters = [] for cluster_group in matched_clusters: new_entities = [] for cluster_idx in cluster_group: clust_entities = self.clusters.pop(cluster_idx) new_entities.extend(clust_entities['entities']) self.clusters[self.cluster_count] = { 'entities': new_entities } self.new_clusters.append(self.cluster_count) self.cluster_count += 1 end = time.time() print("New clustering took {}\n".format(round(end-start, 4))) start = end
dave-s477/SoMeNLP
somenlp/entity_disambiguation/feature_writer.py
import json import os import numpy as np class FeatureWriter: def __init__(self, save_dir): self.output_path = save_dir def read_features(self, path): with open(path, 'r') as j_in: feature_locations = json.load(j_in) feature_matrices = {} for k,v in feature_locations.items(): if k not in feature_matrices: feature_matrices[k] = {} for k2, v2 in v.items(): if k2 not in feature_matrices[k]: feature_matrices[k][k2] = {} for k3, v3 in v2.items(): feature_matrices[k][k2][k3] = np.load(v3) return feature_matrices def write_features(self, matrices): os.makedirs(self.output_path, exist_ok=True) save_locations = {} for set_k, set_v in matrices.items(): if set_k not in save_locations: save_locations[set_k] = {} for set_k2, set_v2 in set_v.items(): if set_k2 not in save_locations[set_k]: save_locations[set_k][set_k2] = {} for feature_k, feature_m in set_v2.items(): output_name = '{}/{}_{}_{}'.format(self.output_path, set_k, set_k2, feature_k) save_locations[set_k][set_k2][feature_k] = output_name + '.npy' np.save(output_name, feature_m) with open(self.output_path + '/save_locations.json', 'w') as j_out: json.dump(save_locations, j_out, indent=4) def save_triangular_matrix(self, tri_matrix, save_location): """ This function saves a triangular matrix in a flat representation so unnecessary zeros can be excluded. Only necessary indicies from the matrix are extracted line by line. Arguments: tri_matrix (np.matrix with an arbitrary d_type): the matrix to save save_location (string): path to output file """ indices = np.triu_indices(tri_matrix.shape[0], 1) flat_representation = tri_matrix[indices] np.save(save_location, flat_representation)
dave-s477/SoMeNLP
somenlp/NER/BERT_dataset.py
<reponame>dave-s477/SoMeNLP import torch import numpy as np import pandas as pd from torch.utils.data import Dataset class BERTDataset(Dataset): """PyTorch Dataset for BERT data """ def __init__(self, ids, tags, masks, transforms=None): self.ids = ids self.tags = tags self.masks = masks self.transforms = transforms def __len__(self): return len(self.ids) def __getitem__(self, idx): if torch.is_tensor(idx): idx = idx.tolist() sample = { 'ids': self.ids[idx], 'masks': self.masks[idx], 'tags': self.tags[idx] } if self.transforms: sample = self.transforms(sample) return sample class BERTMultiDataset(Dataset): """PyTorch Dataset for BERT data """ def __init__(self, ids, tags, masks, lengths, transforms=None): self.ids = ids self.tags = tags self.masks = masks self.lengths = lengths self.transforms = transforms def __len__(self): return len(self.ids) def __getitem__(self, idx): if torch.is_tensor(idx): idx = idx.tolist() sample = { 'ids': self.ids[idx], 'masks': self.masks[idx], 'lengths': self.lengths[idx] } for k in self.tags.keys(): sample[k] = self.tags[k][idx] if self.transforms: sample = self.transforms(sample) return sample
dave-s477/SoMeNLP
somenlp/entity_disambiguation/model.py
import torch import torch.nn as nn import torch.optim as optim import numpy as np from torch.utils.data import Dataset from sklearn.metrics import classification_report from torch.utils.tensorboard import SummaryWriter class DisambiguationModel(nn.Module): def __init__(self, dim, layer_sizes, drop_outs): super(DisambiguationModel, self).__init__() self.lin_layers = nn.ModuleList([nn.Linear(dim, layer_sizes[0])]) for i in range(len(layer_sizes)-1): self.lin_layers.append(nn.ReLU()) self.lin_layers.append(nn.Dropout(drop_outs[i])) self.lin_layers.append(nn.Linear(layer_sizes[i], layer_sizes[i+1])) def forward(self, x): for l in self.lin_layers: x = l(x) return x class ModelWrapper(): def __init__(self, config, input_dim, save_path=None, device='cpu'): self.config = config self.device = device self.writer = SummaryWriter('{}/log'.format(save_path)) self.model = DisambiguationModel(input_dim, config['layer_sizes'], config['drop_outs']).to(device) self.loss_fn = nn.BCEWithLogitsLoss().to(device) self.optim = optim.Adam(self.model.parameters()) self.epoch = 0 self.save_path = save_path def train(self, train_set): self.model.train() cum_loss = 0 for idx, sample in enumerate(train_set): self.optim.zero_grad() pred = self.model(sample[0].to(self.device)) loss = self.loss_fn(torch.squeeze(pred), sample[1].float().to(self.device)) loss.backward() self.optim.step() cum_loss += loss.detach() if idx != 0 and idx % 4000 == 0: print("Batch {} Avg. Loss {}".format(idx, cum_loss / 100)) cum_loss = 0 def test(self, test_set): self.model.eval() predictions = [] true = [] for idx, sample in enumerate(test_set): with torch.no_grad(): pred = self.model(sample[0].to(self.device)) pred_class = torch.squeeze(torch.sigmoid(pred)) predictions.extend(pred_class.cpu().numpy()) true.extend(sample[1]) if idx != 0 and idx % 4000 == 0: print("At batch {}".format(idx)) return true, predictions def predict(self, test_set='train', test_set_ext='train'): self.model.eval() predictions = [] for idx, sample in enumerate(self.inputs[test_set][test_set_ext]['loader']): with torch.no_grad(): pred = self.model(sample[0]) pred_class = torch.squeeze(torch.sigmoid(pred)) predictions.extend(pred_class.cpu().numpy()) return predictions def eval(self, true, predictions, threshold=.5, epoch=0, write=True): predictions = [0 if pred <= threshold else 1 for pred in predictions] print(classification_report(true, predictions)) classification_dict = classification_report(true, predictions, output_dict=True) if write and 'True' in classification_dict: for k,v in classification_dict['True'].items(): self.writer.add_scalar(k, v, epoch) def save(self): print("Saving model") torch.save({ 'epoch': self.epoch, 'model_state_dict': self.model.state_dict(), 'optimizer_state_dict': self.optim.state_dict(), }, '{}/model.pth'.format(self.save_path)) def load(self): if 'checkpoint' in self.config and self.config['checkpoint']: print("Loading model from checkpoint") checkpoint_data = torch.load(self.config['checkpoint'], map_location=self.device) self.model.load_state_dict(checkpoint_data['model_state_dict']) self.optim.load_state_dict(checkpoint_data['optimizer_state_dict']) self.epoch = checkpoint_data['epoch']
dave-s477/SoMeNLP
somenlp/RE/__init__.py
<filename>somenlp/RE/__init__.py from .features import FeatureGenerator from .RE_model import REmodel
dave-s477/SoMeNLP
somenlp/feature_engineering/distant_supervision_rules.py
<reponame>dave-s477/SoMeNLP import numpy as np def distant_supervision_by_dict(candidate, dictionary, mapping): result = np.zeros(len(mapping), dtype=np.bool) span_to_examine = candidate.base_span values = [mapping[x] for x in dictionary[span_to_examine]] if span_to_examine in dictionary.keys() else [] if values: for val in values: result[val] = True return result
dave-s477/SoMeNLP
somenlp/feature_engineering/sentence_rep.py
<gh_stars>0 import string import nltk import re import unicodedata import numpy as np from nltk.corpus import wordnet from . import word_rules lemmatizer = nltk.stem.WordNetLemmatizer() URL = re.compile(r"^(https?\:\/\/[a-zA-Z0-9\-\.]+[\w\/\._\-\:~\?=#%]*[\w\/_\-\:~\?=#%]|ftp\:\/\/[a-zA-Z0-9\-\.]+[\w\/\._\-\:~\?=#%]*[\w\/_\-\:~\?=#%]|www\.[a-zA-Z0-9\-\.]+[\w\/\._\-\:~\?=#%]*|[a-zA-Z0-9\-\.]+\.(org|edu)/[\w\/_\-\:~\?=#%]*)$") CITATION = re.compile(r'^\[[0-9\-,\?]+\]$') FLOAT_NUM = re.compile(r'^\d+\.\d*$') FLOAT_NON_LEADING = re.compile(r'^\.\d+$') VERSION_LIKE_NUM = re.compile(r'^(\d+\.){2}\w*$') LONG_VERSION_LIKE = re.compile(r'^(\d+\.){3,8}\w*$') LONG_NUM = re.compile(r'^(\d{1,3}\,){1,8}\d{3}$') HEADWORDS = ['software', 'package', 'program', 'tool', 'toolbox', 'web', 'service', 'spreadsheet', 'database', 'registry', 'data', 'model', 'algorithm', 'kit', 'standard', 'method', 'procedure'] def get_wordnet_pos(tag): """Map POS tag to first character lemmatize() accepts""" s_tag = tag[0].upper() tag_dict = {"J": wordnet.ADJ, "N": wordnet.NOUN, "V": wordnet.VERB, "R": wordnet.ADV} return tag_dict.get(s_tag, wordnet.NOUN) class SentenceRepresentation: def __init__(self, input_string): self.features = {} # Basic words self.tokens = input_string.split() self.tokens_lower = [x.lower() for x in self.tokens] pos_tags = nltk.pos_tag(self.tokens) self.lemmas = [] for w, t in pos_tags: wntag = get_wordnet_pos(t) self.lemmas.append(lemmatizer.lemmatize(w, wntag)) self.length = len(self.tokens) self.pos_tags = [x[1] for x in pos_tags] # additional information self.features['token_length'] = [len(x) for x in self.tokens] self.features['punct'] = [x in string.punctuation for x in self.tokens] self.features['math_chars'] = [unicodedata.category(x) == 'Sm' if len(x) == 1 else False for x in self.tokens] self.features['hypen'] = [x == '-' for x in self.tokens] self.features['slash'] = [x == '/' for x in self.tokens] self.features['bracket_open'] = [x == '(' for x in self.tokens] self.features['bracket_close'] = [x == ')' for x in self.tokens] # digit information self.features['digit'] = [x.isdigit() for x in self.tokens] self.features['float_num'] = [bool(FLOAT_NUM.match(x)) for x in self.tokens] self.features['float_non_leading'] = [bool(FLOAT_NON_LEADING.match(x)) for x in self.tokens] self.features['version_like_num'] = [bool(VERSION_LIKE_NUM.match(x)) for x in self.tokens] self.features['long_version_like'] = [bool(LONG_VERSION_LIKE.match(x)) for x in self.tokens] self.features['long_num'] = [bool(LONG_NUM.match(x)) for x in self.tokens] # further token based information self.features['citation'] = [bool(CITATION.match(x)) for x in self.tokens] self.features['url'] = [bool(URL.match(x)) for x in self.lemmas] # casing info self.features['upper'] = [word_rules.upper_cased(x) for x in self.tokens] self.features['first_char_upper'] = [word_rules.first_char_upper(x) for x in self.tokens] self.features['mixed_case'] = [word_rules.mixed_case(x) for x in self.tokens] self.features['lower_case'] = [word_rules.lower_case(x) for x in self.tokens] # individual head words for headword in HEADWORDS: self.features[headword] = [x == headword for x in self.lemmas] def get_features(self): matrix = [] for _, v in self.features.items(): matrix.append(v) return np.array(matrix) def get_left_tokens(self, idx, size, style='plain'): if style == 'plain': return self.tokens[max(idx-size, 0):idx] if style == 'lower': return self.tokens_lower[max(idx-size, 0):idx] elif style == 'lemma': return self.lemmas[max(idx-size, 0):idx] else: raise(RuntimeError("Style '{}' is not defined".format(style))) def get_right_tokens(self, idx, size, style='plain'): if style == 'plain': return self.tokens[idx:idx+size] if style == 'lower': return self.tokens_lower[idx:idx+size] elif style == 'lemma': return self.lemmas[idx:idx+size] else: raise(RuntimeError("Style '{}' is not defined".format(style))) def get_candidate(self, start_idx, end_idx): return self.tokens[start_idx:end_idx], self.lemmas[start_idx:end_idx] class Candidate: def __init__(self, sentence, start_idx, end_idx): self.tokens, self.lemmas = sentence.get_candidate(start_idx, end_idx) self.sentence = sentence self.start_idx = start_idx self.end_idx = end_idx self.base_span = ' '.join(self.tokens)
dave-s477/SoMeNLP
somenlp/entity_disambiguation/__init__.py
from .linking_data import LinkingData from .feature_writer import FeatureWriter from .feature_calculator import EntityDisambiguationFeatureGenerator from .model import ModelWrapper from .clustering import Clustering, IntervalClustering, SimpleCluster from .ed_main import main from .efficient_prediction import ReducedSampleSet, DistanceMap, EfficientClustering, IterDataset, worker_init_fn
dave-s477/SoMeNLP
somenlp/entity_disambiguation/ed_main.py
from somenlp.entity_disambiguation.efficient_prediction import BLOCK_SIZE import torch import torch.nn as nn import torch.optim as optim torch.multiprocessing.set_sharing_strategy('file_system') import json import math import pickle import os from somenlp.entity_disambiguation.model import DisambiguationModel import sys import random import itertools import time as timeit import numpy as np from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, IterableDataset, Dataset from itertools import combinations from articlenizer.util import chunk_list import nltk nltk.download('stopwords') from nltk.corpus import stopwords from sklearn.metrics import classification_report from somenlp.entity_disambiguation import LinkingData, EntityDisambiguationFeatureGenerator, FeatureWriter, ModelWrapper, Clustering, IntervalClustering, SimpleCluster from somenlp.entity_disambiguation.efficient_prediction import IterDataset, worker_init_fn def concat_features(vectors, labels=False): dim = -1 if not vectors: raise(RuntimeError("No features were given")) feature_matrix = {} for k, v in vectors.items(): if k not in feature_matrix: feature_matrix[k] = {} for k2, v2 in v.items(): if k2 not in feature_matrix[k]: feature_matrix[k][k2] = {} for idx, (k3, v3) in enumerate(v2.items()): if idx == 0: matrix = np.expand_dims(v3, -1) elif labels and k3 == 'labels': feature_matrix[k][k2]['labels'] = v3 else: matrix = np.append(matrix, np.expand_dims(v3, -1), axis=1) if dim < 0: dim = matrix.shape[1] feature_matrix[k][k2]['features'] = matrix return feature_matrix, dim def evaluate(entities, indicies, labels): pairs_to_compare = list(combinations(entities, 2)) tp_same_name, fp_same_name, fn_same_name, tn_same_name = 0, 0, 0, 0 tp_diff_name, fp_diff_name, fn_diff_name, tn_diff_name = 0, 0, 0, 0 for pair, label in zip(pairs_to_compare, labels): exact_match = False if pair[0]['mention'] == pair[1]['mention']: exact_match = True pos = False for ind in indicies: if pair[0]['init_id'] in ind and pair[1]['init_id'] in ind: pos = True if pos and label: if exact_match: tp_same_name += 1 else: tp_diff_name += 1 elif pos and not label: if exact_match: fp_same_name += 1 else: fp_diff_name += 1 elif not pos and label: if exact_match: fn_same_name += 1 else: fn_diff_name += 1 elif not pos and not label: if exact_match: tn_same_name += 1 else: tn_diff_name += 1 tp = tp_same_name + tp_diff_name fp = fp_same_name + fp_diff_name fn = fn_same_name + fn_diff_name tn = tn_same_name + tn_diff_name print('Overall: tp ', tp, ', fp ', fp, ', fn ', fn, ', tn ', tn) precision = tp / (tp + fp) recall = tp / (tp + fn) fscore = (2 * precision * recall) / (precision + recall) print('Precision: {}, Recall {}, FScore: {}\n'.format( round(precision, 2), round(recall, 2), round(fscore, 2) )) print('Same name: tp ', tp_same_name, ', fp ', fp_same_name, ', fn ', fn_same_name, ', tn ', tn_same_name) precision = tp_same_name / (tp_same_name + fp_same_name) recall = tp_same_name / (tp_same_name + fn_same_name) fscore = (2 * precision * recall) / (precision + recall) print('Precision: {}, Recall {}, FScore: {}\n'.format( round(precision, 2), round(recall, 2), round(fscore, 2) )) print('Diff name: tp ', tp_diff_name, ', fp ', fp_diff_name, ', fn ', fn_diff_name, ', tn ', tn_diff_name) precision = tp_diff_name / (tp_diff_name + fp_diff_name) recall = tp_diff_name / (tp_diff_name + fn_diff_name) fscore = (2 * precision * recall) / (precision + recall) print('Precision: {}, Recall {}, FScore: {}\n'.format( round(precision, 2), round(recall, 2), round(fscore, 2) )) def generator_fct(sample_num, start_row, end_row): for idx_i in range(start_row, end_row): for idx_j in range(idx_i+1, sample_num): yield idx_i, idx_j def main(config, curr_time, ncores, gpu): if gpu < 0: device = torch.device("cpu") else: device = torch.device("cuda:{}".format(gpu)) if config['gold']: read_labels = True else: read_labels = False save_path = '{}/{}/{}'.format( config['logdir'].rstrip('/'), config['name'].rstrip('/'), curr_time ) if not os.path.isdir(save_path): os.makedirs(save_path) with open('{}/config.json'.format(save_path), 'w') as j_out: json.dump(config, j_out, indent=4) linking_data = LinkingData(config) feature_generator = EntityDisambiguationFeatureGenerator(config['dbpedia']) input_dim = len(feature_generator.string_features_to_extract) + len(feature_generator.context_features_to_extract) print("Augmentated Dataset has {} samples".format(len(linking_data.all))) test_set_size = int(len(linking_data.all) * 0.18) model_w = ModelWrapper(config['model'], input_dim, save_path, device) print("MODEL") print(model_w.model) print() if config['model']['checkpoint']: print("Loading model for disambiguation.. (or re-training)") model_w.load() for e in range(config['model']['epochs']): model_w.epoch += 1 print("At Epoch {} ({})".format(e, model_w.epoch)) print(test_set_size) train_generator = generator_fct(len(linking_data.all), test_set_size, len(linking_data.all)) count = 0 while True: print("At block {}".format(count)) count += 1 block = list(itertools.islice(train_generator, BLOCK_SIZE)) if not block: break train_dataset = IterDataset(list(block), linking_data.all, feature_generator, None, labels=linking_data.link_lookup_table) train_loader = DataLoader(train_dataset, batch_size=config['model']['batch_size'], num_workers=ncores, pin_memory=True) model_w.train(train_loader) print("Starting evaluation..") test_generator = generator_fct(len(linking_data.all), 0, test_set_size) count = 0 true = [] predictions = [] while True: print("At block {}".format(count)) count += 1 block = list(itertools.islice(test_generator, BLOCK_SIZE)) if not block: break test_dataset = IterDataset(list(block), linking_data.all, feature_generator, None, labels=linking_data.link_lookup_table) test_loader = DataLoader(test_dataset, batch_size=config['model']['batch_size'], num_workers=ncores, pin_memory=True) t, p = model_w.test(test_loader) true.extend(t) predictions.extend(p) model_w.eval(true, predictions, epoch=e) print("Starting evaluation..") test_generator = generator_fct(len(linking_data.all), 0, test_set_size) count = 0 true = [] predictions = [] while True: print("At block {}".format(count)) count += 1 block = list(itertools.islice(test_generator, BLOCK_SIZE)) if not block: break test_dataset = IterDataset(list(block), linking_data.all, feature_generator, None, labels=linking_data.link_lookup_table) test_loader = DataLoader(test_dataset, batch_size=config['model']['batch_size']) t, p = model_w.test(test_loader) true.extend(t) predictions.extend(p) model_w.eval(true, predictions, epoch=0, write=False) model_w.save()
dave-s477/SoMeNLP
somenlp/RE/features.py
import re import numpy as np from gensim.models import KeyedVectors from itertools import combinations from articlenizer import CUSTOM_STOPWORDS from articlenizer.formatting import bio_to_brat class FeatureGenerator(): def __init__(self, data_handler, embedding_location, emb_dim=200): self.main_entities = ['Application_Creation', 'Application_Deposition', 'Application_Usage', 'Application_Mention', 'PlugIn_Creation', 'PlugIn_Deposition', 'PlugIn_Usage', 'PlugIn_Mention', 'ProgrammingEnvironment_Usage', 'ProgrammingEnvironment_Mention', 'OperatingSystem_Usage', 'OperatingSystem_Mention', 'SoftwareCoreference_Deposition'] self.data_handler = data_handler self.type2idx, self.type2name, self.mention2idx, self.mention2name = {}, {}, {}, {} for k in sorted(data_handler.encoding['tag2name']): if '_' in data_handler.encoding['tag2name'][k]: ent_type, mention_type = data_handler.encoding['tag2name'][k].split('_') else: ent_type = data_handler.encoding['tag2name'][k] mention_type = 'None' if ent_type not in self.type2idx: self.type2name[len(self.type2idx)] = ent_type self.type2idx[ent_type] = len(self.type2idx) if mention_type not in self.mention2idx: self.mention2name[len(self.mention2idx)] = mention_type self.mention2idx[mention_type] = len(self.mention2idx) self.default_ent_type = 'Application' self.default_men_type = 'Usage' # self.word_emb = KeyedVectors.load_word2vec_format(embedding_location, binary=True) # self.emb_dim = emb_dim def correct_types(self, ent_0, ent_1, men_0, men_1): ent_0 = self.default_ent_type if ent_0 == 'Unknown' else ent_0 ent_1 = self.default_ent_type if ent_1 == 'Unknown' else ent_1 men_0 = self.default_men_type if men_0 == 'Unknown' else men_0 men_1 = self.default_men_type if men_1 == 'Unknown' else men_1 return ent_0, ent_1, men_0, men_1 def acronym(self, tokens_in): """Get acronym of tokens Args: tokens_in (list): list of tokens Returns: str: acronym """ tokens = [x.casefold() for x in tokens_in] tokens = [re.sub('[^0-9a-zA-Z]+', ' ', x) for x in tokens] tokens = [x.rstrip('0123456789 ,.').lstrip(' ') for x in tokens] tokens = [x for x in tokens if x not in CUSTOM_STOPWORDS] tokens = [x for x in tokens if x] if len(tokens) <= 2: return None acronym = ''.join([x[0] for x in tokens]) return acronym def is_substring(self, e1, e2): """Substring relation between entities Args: e1 (str): entity string e2 (str): entity string Returns: bool: result """ if e1 is None or e2 is None: return False return ''.join(e1).casefold() in ''.join(e2).casefold() def get_left_context_word(self, sentence, position): """Get word left of position in sentence Args: sentence (list): list of string tokens position (int): position to extract Returns: str: string of token left of position """ sub_sent = sentence[:position] sub_tokens = sub_sent.split() if len(sub_tokens) > 0: return sub_tokens[-1] else: return None def get_right_context_word(self, sentence, position): """Get word right of position in sentence Args: sentence (list): list of string tokens position (int): position to extract Returns: str: string of token left of position """ sub_sent = sentence[position:] sub_tokens = sub_sent.split() if len(sub_tokens) > 0: return sub_tokens[0] else: return None def one_hot_encoding(self, dictionary, name, ent_type, encoding): """Generate a one-hot encoding Args: dictionary (dict): dictionary with information, is written and changed name (str): addition to output feature name ent_type ([type]): restriction for entries to consider encoding ([type]): mapping for names """ found_key = False for k, v in encoding.items(): feature_name = '{}_{}'.format(name, k) if ent_type == k: dictionary[feature_name] = 1 found_key = True else: dictionary[feature_name] = 0 if not found_key: raise(RuntimeError("Unknown entity type: {}".format(ent_type))) # def get_entities_inbetween(self, tags, ind1, ind2): # """Get number of tagged elements inbetween two indicies # Args: # tags (list): sequence of tags # ind1 (int): index # ind2 (int): index # Returns: # int: number of tagged entities # """ # sub_tags = tags[ind1:ind2].split() # return sum([1 if x.startswith('B-') else 0 for x in sub_tags]) def get_features(self, pair, sentence, tags, main_entity_count): """Get feature dictionary for pair of entities Args: pair (tuple): pair of entities sentence (list): sentence containing entities as tokens tags (list): IOB2 labels assigned to sentence main_entity_count (int): number of main entities in sentence entity_encoding (dictionary): numeric encoding of entities Returns: dictionary: features for pair """ if pair[0]['beg'] > pair[1]['beg']: larger = pair[0] smaller = pair[1] else: larger = pair[1] smaller = pair[0] distance_string = sentence[smaller['end']:larger['beg']] tokens_distance_string = distance_string.split() # ent_0_left_context = self.get_left_context_word(sentence, pair[0]['beg']) # ent_1_left_context = self.get_left_context_word(sentence, pair[1]['beg']) # ent_0_right_context = self.get_right_context_word(sentence, pair[0]['end']) # ent_1_right_context = self.get_right_context_word(sentence, pair[1]['end']) ent_0_tokens = pair[0]['string'].split() ent_1_tokens = pair[1]['string'].split() ent_0_acronym = self.acronym(ent_0_tokens) ent_1_acronym = self.acronym(ent_1_tokens) #entities_in_between = self.get_entities_inbetween(tags, smaller['end'], larger['beg']) # if ent_0_left_context in self.word_emb: # ent_0_left_context_emb = self.word_emb[ent_0_left_context] # else: # ent_0_left_context_emb = np.zeros(self.emb_dim) # if ent_1_left_context in self.word_emb: # ent_1_left_context_emb = self.word_emb[ent_1_left_context] # else: # ent_1_left_context_emb = np.zeros(self.emb_dim) # if ent_0_right_context in self.word_emb: # ent_0_right_context_emb = self.word_emb[ent_0_right_context] # else: # ent_0_right_context_emb = np.zeros(self.emb_dim) # if ent_1_right_context in self.word_emb: # ent_1_right_context_emb = self.word_emb[ent_1_right_context] # else: # ent_1_right_context_emb = np.zeros(self.emb_dim) features = { 'entity_distance_abs': len(distance_string), ###'entity_distance_rel': len(distance_string) / len(sentence), 'entity_distance_tok': len(tokens_distance_string), ###'entity_distance_tok_rel': len(tokens_distance_string) / len(sentence.split()), 'entity_order': pair[1]['beg'] > pair[0]['beg'], 'entity_0_char_length': pair[0]['end'] - pair[0]['beg'], 'entity_0_token_length': len(ent_0_tokens), 'entity_1_char_length': pair[1]['end'] - pair[1]['beg'], 'entity_1_token_length': len(ent_1_tokens), ###'entity_0_left_context_for': ent_0_left_context == 'for', ###'entity_1_left_context_for': ent_1_left_context == 'for', #'ent_0_left_context_emb': ent_0_left_context_emb, #'ent_1_left_context_emb': ent_1_left_context_emb, #'ent_0_right_context_emb': ent_0_right_context_emb, #'ent_1_right_context_emb': ent_1_right_context_emb, 'num_main_entities': main_entity_count, 'entity_0_substring_of_entity_1': self.is_substring(pair[0]['string'], pair[1]['string']), 'acronym_entity_0_substring_of_entity_1': self.is_substring(ent_0_acronym, pair[1]['string']), 'acronym_entity_0_substring_of_acronym_entity_1': self.is_substring(ent_0_acronym, ent_1_acronym), 'entity_1_substring_of_entity_0': self.is_substring(pair[1]['string'], pair[0]['string']), 'acronym_entity_1_substring_of_entity_0': self.is_substring(ent_1_acronym, pair[0]['string']), 'acronym_entity_1_substring_of_acronym_entity_0': self.is_substring(ent_1_acronym, ent_0_acronym) } ent_0_type = pair[0]['type'].split('_')[0] mention_0_type = pair[0]['type'].split('_')[1] if '_' in pair[0]['type'] else 'None' ent_1_type = pair[1]['type'].split('_')[0] mention_1_type = pair[1]['type'].split('_')[1] if '_' in pair[1]['type'] else 'None' ent_0_type, ent_1_type, mention_0_type, mention_1_type = self.correct_types(ent_0_type, ent_1_type, mention_0_type, mention_1_type) self.one_hot_encoding(features, 'entity_0_type', ent_0_type, self.type2idx) self.one_hot_encoding(features, 'mention_0_type', mention_0_type, self.mention2idx) self.one_hot_encoding(features, 'entity_1_type', ent_1_type, self.type2idx) self.one_hot_encoding(features, 'mention_1_type', mention_1_type, self.mention2idx) # for idx, (x, y, z, a) in enumerate(zip(ent_0_left_context_emb, ent_1_left_context_emb, ent_0_right_context_emb, ent_1_right_context_emb)): # features['ent_0_left_context_emb_{}'.format(idx)] = x # features['ent_1_left_context_emb_{}'.format(idx)] = y # features['ent_0_right_context_emb_{}'.format(idx)] = z # features['ent_1_right_context_emb_{}'.format(idx)] = a return features def get_sentence_relations_and_features(self, sentence, tags, entities, relations=None): """Span all potential relations in sentence and calculate its respective features Args: sentence (list): sentence as token list tags (list): IOB2 labels assigned to sentence entities (dict): entities in sentence relations (dict): true relations in sentence Returns: list: relations in sentence including features and labels """ sentence_feature_list = [] entities = sorted(entities, key=lambda item: item['beg']) for idx, x in enumerate(entities): x['idx'] = idx # TODO add this index as training data? main_entity_count = 0 for ent in entities: if ent['type'] in self.main_entities: main_entity_count += 1 entity_combinations = combinations(entities, 2) entity_pairs = [] for comb in entity_combinations: entity_pairs.extend([list(comb), list(reversed(comb))]) assigned_count = 0 for idx, pair in enumerate(entity_pairs): features = self.get_features(pair, sentence, tags, main_entity_count) if relations is not None: features['label'] = 'none' for relation in relations: if int(relation['ent1_b']) == int(pair[0]['beg']) and int(relation['ent2_b']) == int(pair[1]['beg']): features['label'] = relation['type'] assigned_count += 1 break sentence_feature_list.append(features) return sentence_feature_list, entity_pairs def generate_relation_extraction_features(self): for dataset, dataset_setup in self.data_handler.data_config['sets'].items(): for sub_dataset in dataset_setup: sub_dataset['relext_feature_list'] = [] for sentence, tags, relations in zip(sub_dataset['sentences'], sub_dataset['tags'], sub_dataset['relations']): if sum([1 if t.startswith('B-') else 0 for t in tags.split()]) > 1: entities, _, _ = bio_to_brat(sentence, tags) if self.data_handler.tag_remapping is not None: for ent in entities: ent['type'] = self.data_handler.tag_remapping[ent['type']] pairs, _ = self.get_sentence_relations_and_features(sentence, tags, entities, relations) sub_dataset['relext_feature_list'].extend(pairs) def stream_files(self): for file_names in self.data_handler.data_files: article = { 'out_name': file_names['out'], 'sentences': self.data_handler._read_text_file(file_names['in']), 'tags': self.data_handler._read_text_file(file_names['entities']), 'entity_list': [], 'relext_feature_list': [] } for idx, (sentence, tags) in enumerate(zip(article['sentences'], article['tags'])): if sum([1 if t.startswith('B-') else 0 for t in tags.split()]) > 1: entities, _, _ = bio_to_brat(sentence, tags) if self.data_handler.tag_remapping is not None: for ent in entities: ent['type'] = self.data_handler.tag_remapping[ent['type']] pairs, entity_pairs = self.get_sentence_relations_and_features(sentence, tags, entities) article['entity_list'].append(list(entity_pairs)) article['relext_feature_list'].append(pairs) else: article['entity_list'].append(None) article['relext_feature_list'].append(None) yield article
dave-s477/SoMeNLP
somenlp/NER/models/__init__.py
<gh_stars>0 from .crf import CRF from .lstm import BiLSTM_CRF from .feature_lstm import FeatureLSTM from .combined_lstm import CombinedLSTM from .multi_bert import BERTMultiTask, BERTMultiTaskOpt2, BERTMultiTaskCRF, BERTMultiTaskOpt2CRF
dave-s477/SoMeNLP
somenlp/RE/run_relation_extraction.py
<filename>somenlp/RE/run_relation_extraction.py<gh_stars>0 from somenlp.NER import OutputHandler, DataHandler, Tuner from . import REmodel, FeatureGenerator def main(model_config, data_config, time, data_file_ext, ent_file_ext, rel_file_ext): """Main function for performing relation extraction training Args: model_config (dict): configuration for RE model data_config (dict): configuration of input data time (str): time marker for saving data_file_ext (str): identifying extension for automatic file loading ent_file_ext (str): identifying extension for automatic file loading rel_file_ext (str): identifying extension for automatic file loading """ print("\nSetting up output handler") output_handler = OutputHandler(model_config['general']['name'], time, model_config['general']['checkpoint']) output_handler.save_json(model_config, name='model_conf') output_handler.save_json(data_config, name='data_conf') print("\nSetting up data handler") data_handler = DataHandler(data_config=data_config, data_file_extension=data_file_ext, label_file_extension=ent_file_ext, relation_file_extension=rel_file_ext, output_handler=output_handler, checkpoint=model_config['general']['checkpoint'], max_word_length=model_config['general']['max_word_length'], max_sent_length=model_config['general']['max_sentence_length']) data_handler.load_data_from_config() data_handler.encoding(tags_only=True) data_handler.load_input() feature_generator = FeatureGenerator(data_handler, model_config['model']['word_embedding']) feature_generator.generate_relation_extraction_features() print("\nSetting up model") model_w = REmodel(model_config['general'], model_config['model'], data_handler, output_handler, model_config['output']) if model_config['general']['train']: if 'cross_val' in model_config['general'] and model_config['general']['cross_val']: model_w.cross_val() else: model_w.train() model_w.test() model_w.save() #model_w.show_features_importance() else: model_w.load(model_config['general']['checkpoint']) model_w.test() def tune(config, time, data_file_extension, ent_file_ext, rel_file_ext): """Hyper-parameter tuning for RE model Args: config (dict): configuration including model and data time (str): time marker for saving data_file_ext (str): identifying extension for automatic file loading ent_file_ext (str): identifying extension for automatic file loading rel_file_ext (str): identifying extension for automatic file loading """ tuner = Tuner(config, time) iterator = tuner.yield_configs() for name, data_conf, model_conf in iterator: print(model_conf) time_name = '{}_{}'.format(time, name) print("Training model {}".format(time_name)) main(model_conf, data_conf, time_name, data_file_extension, ent_file_ext, rel_file_ext) def predict(model_config, files, prepro, output): print("\nSetting up output handler") output_handler = OutputHandler(model_config['general']['name'], checkpoint=model_config['general']['checkpoint']) print("\nSetting up data handler") data_handler = DataHandler(data_files=files, prepro=prepro, output_handler=output_handler, checkpoint=model_config['general']['checkpoint'], max_word_length=model_config['general']['max_word_length'], max_sent_length=model_config['general']['max_sentence_length']) data_handler.encoding(tags_only=True) feature_generator = FeatureGenerator(data_handler, model_config['model']['word_embedding']) print("\nSetting up model") model_w = REmodel(model_config['general'], model_config['model'], data_handler, output_handler, model_config['output'], feature_generator=feature_generator) model_w.load(model_config['general']['checkpoint']) model_w.predict(output)
dave-s477/SoMeNLP
somenlp/feature_engineering/gen_custom_features.py
<gh_stars>0 import json import numpy as np from multiprocessing import Pool from functools import partial from . import candidate_rules from . import distant_supervision_rules from .sentence_rep import SentenceRepresentation, Candidate FUNCTION_NAMES = [ 'pan_top_1', 'pan_top_2', 'pan_top_3', 'pan_top_4', 'pan_top_5', 'pan_top_6', 'pan_top_7', 'pan_top_8', 'text_is_in_brackets', 'developer'] fcts = [getattr(candidate_rules, n) for n in FUNCTION_NAMES] def encode_sentence(text, max_length, distant_supervision_dict, key_mapping, pos_tag_encoding): sentence = SentenceRepresentation(text) lengths = sentence.length pos_tags = np.expand_dims(np.array([pos_tag_encoding[x] for x in sentence.pos_tags]), 0) features = sentence.get_features() sentence_counts = np.zeros((len(FUNCTION_NAMES), sentence.length), np.int32) distant_supervision_counts = np.zeros((len(key_mapping), sentence.length), np.int32) for cand_length in range(1, max_length+1): for cand_beg in range((sentence.length - cand_length) + 1): cand_end = cand_beg + cand_length candidate = Candidate(sentence, cand_beg, cand_end) for fct_idx, fct in enumerate(fcts): match_res = fct(candidate) if match_res == 1: for pos in range(candidate.start_idx, candidate.end_idx): sentence_counts[fct_idx][pos] += 1 distant_supervision_result = distant_supervision_rules.distant_supervision_by_dict(candidate, distant_supervision_dict, key_mapping) for pos in range(candidate.start_idx, candidate.end_idx): distant_supervision_counts[:,pos] += distant_supervision_result all_feats = np.concatenate((pos_tags, features, sentence_counts, distant_supervision_counts)) return np.transpose(all_feats) def calculate_features(input_files, max_length, distant_supervision_dict, key_mapping, pos_tag_encoding): sentence_feature_encodings = [] with input_files[0].open(mode='r') as in_text: for line in in_text: sentence_feature_encodings.append(encode_sentence(line, max_length, distant_supervision_dict, key_mapping, pos_tag_encoding)) array_to_write = np.array(sentence_feature_encodings) np.savez_compressed('{}.npz'.format(str(input_files[1])), features=array_to_write) def update_words(all_words, in_list, name): for word in in_list: if word not in all_words.keys(): all_words[word] = [] if name not in all_words[word]: all_words[word].append(name) def load_distant_supervision(loc): with loc.open(mode='r') as in_f: distant = json.load(in_f) all_distsup_words = {} key_mapping = {} for k, v in distant.items(): update_words(all_distsup_words, v, k) key_mapping[k] = len(key_mapping) return all_distsup_words, key_mapping def load_pos_tag_encoding(loc): with loc.open(mode='r') as in_f: encoding = json.load(in_f) return encoding def calculate_features_parallel(files, max_length, distant_supervision_dictionary_location, pos_tag_encoding_location, n_cores=8): distant_supervision_dictionary, distant_supervision_key_mapping = load_distant_supervision(distant_supervision_dictionary_location) pos_tag_encoding = load_pos_tag_encoding(pos_tag_encoding_location) fct_to_execute = partial(calculate_features, max_length=max_length, distant_supervision_dict=distant_supervision_dictionary, key_mapping=distant_supervision_key_mapping, pos_tag_encoding=pos_tag_encoding) with Pool(n_cores) as p: p.map(fct_to_execute, files)
dave-s477/SoMeNLP
somenlp/distant_supervision/__init__.py
<reponame>dave-s477/SoMeNLP from .packages import load_package_names from .perform_wiki_data_queries import query_wikidata from .gen_sequences import generate_triplets from .gen_wiktionary_dict import load_wiktionary from .combine_info import merge_results, load_dicts
dave-s477/SoMeNLP
somenlp/NER/LSTM_dataset.py
<gh_stars>0 import torch from torch.nn.utils.rnn import pad_sequence from torch.utils.data import Dataset class LSTMDataset(Dataset): """PyTorch dataset for LSTM input data """ def __init__(self, characters, ids, tags, features, character2idx, padding, max_word_length, max_sent_length, transforms=None): self.characters = characters self.ids = ids self.tags = tags self.transforms = transforms self.character2idx = character2idx self.max_sent_length = max_sent_length self.max_word_length = max_word_length self.padding = padding self.features = features def __len__(self): return len(self.ids) def __getitem__(self, idx): if torch.is_tensor(idx): idx = idx.tolist() if self.max_sent_length > 0: sample = { 'ids': torch.tensor(self.ids[idx][:self.max_sent_length]), 'features': torch.tensor(self.features[idx][:self.max_sent_length]) if self.features is not None else None, 'tags': torch.tensor(self.tags[idx][:self.max_sent_length])} else: sample = { 'ids': torch.tensor(self.ids[idx]), 'features': torch.tensor(self.features[idx]) if self.features is not None else None, 'tags': torch.tensor(self.tags[idx])} if self.max_word_length < 0: sample['characters'] = pad_sequence([torch.tensor(x) for x in self.characters[idx]], batch_first=True, padding_value=self.character2idx[self.padding]) else: sample['characters'] = pad_sequence([torch.tensor(x[:self.max_word_length]) for x in self.characters[idx]], batch_first=True, padding_value=self.character2idx[self.padding]) if self.transforms: sample = self.transforms(sample) return sample
dave-s477/SoMeNLP
somenlp/distant_supervision/gen_wiktionary_dict.py
import json import wget from pathlib import Path def load_wiktionary(download_location='/tmp/', default_address='https://kaikki.org/dictionary/English/kaikki.org-dictionary-English.json'): """Download a pre-processed wiktionary dump to generate an english dictionary for distant supervision. Args: download_location (str, optional): where to write the download. Defaults to '/tmp/'. default_address (str, optional): url from where to download the dump. Defaults to 'https://kaikki.org/dictionary/English/kaikki.org-dictionary-English.json'. Returns: dictionary: list of english words from Wiktionary """ print("Loading English dictionary") dict_location = '{}/{}'.format(download_location.rstrip('/'), 'wiktionary_dict.json') wget.download(default_address, out=dict_location) results = {} with open(dict_location, 'r') as w_file: for line in w_file: entry = json.loads(line) if entry['pos'] not in results.keys(): results[entry['pos']] = set() results[entry['pos']].update([entry['word']]) return results
cjavaphp/keybinder3
examples/example.py
#!/usr/bin/env python """ example.py Created in 2010 by <NAME> <<EMAIL>> This work is placed in the public domain. """ import gi gi.require_version('Gtk', '3.0') gi.require_version('Keybinder', '3.0') from gi.repository import Gtk from gi.repository import Keybinder def callback(keystr, user_data): print "Handling", keystr, user_data print "Event time:", Keybinder.get_current_event_time() Gtk.main_quit() if __name__ == '__main__': keystr = "<Ctrl>A" Keybinder.init() Keybinder.bind(keystr, callback, "Keystring %s (user data)" % keystr) print "Press", keystr, "to handle keybinding and quit" Gtk.main()
lvrbanec/100DaysOfCode_Python
Project06, Beginner, Hangman/main.py
<reponame>lvrbanec/100DaysOfCode_Python # 02.01.21, Frollo # level: Beginner # Project: Create a Hangman game import random from hangman_words import word_list #import wordlist from hangman_art import logo, stages # imports pictures of a stages of hanging man # pick a word randomly chosen_word = random.choice(word_list) # print(f"the secret word is {chosen_word}") # show logo of the game print(logo) # create a drawing of the chosen_word drawing = [] for i in chosen_word: drawing.append('_') print(drawing) #create a drawing of a hangman print(stages[6]) # guess the word until you win or die end_of_game = False lives = 6 previous_guesses = [] while not end_of_game: # guess a letter guess = input('Please choose a letter\n').lower() # check if you guessed right index = 0 #index od a letter for letter in chosen_word: # or to get a position can write for i in len(chosen_word) index += 1 if guess == letter: drawing[index-1] = letter print(drawing) # if the letter isnt guessed, remove one life if (guess not in chosen_word) and (guess not in previous_guesses): lives -= 1 print(f'Letter {guess} is not in the chosen word.') print(stages[lives]) # if guess in previous_guesses: print(f'You already tried letter {guess}.') previous_guesses.append(guess) # win condition if "_" not in drawing: end_of_game = True print("You win.") # lose condition if lives == 0: end_of_game = True print("You lost.")
lvrbanec/100DaysOfCode_Python
Project07, Beginner, caesar ciphone/main.py
# 02.02.2021, Frollo # project: create Casear Cipher # level Beginner from art import logo print(logo) alphabet = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z'] # define a function for encryption def caesar(word, shift_ammount, direction): encrypted_word = '' if direction == 'decode': shift_ammount *= -1 # substract the position if decoding for char in word: if char in alphabet: position = alphabet.index(char) # gives only the first index encrypted_letter = alphabet[position + shift_ammount] encrypted_word += encrypted_letter else: encrypted_word += char print(f"The {direction}d text is '{encrypted_word}'.") should_continue = True # run again? while should_continue: # give input to encript desired_direction = input("Type 'encode' to encrypt, type 'decode' to decrypt:\n") text = input("Type your message:\n").lower() shift = int(input("Type the shift number:\n")) # if the user imputs huge number find the one that corresponds to alphabet, 26 letters our alphabet shift = shift % 26 # call function caesar(word = text, shift_ammount = shift, direction = desired_direction ) # run again? run_again = input("Type 'yes' if you want to run again. Otherwise type 'no'\n.") if run_again == "no": should_continue = False print("Goodbye")
lvrbanec/100DaysOfCode_Python
Project15, Intermediate, Plagiate D.Hirst Painting/main.py
# 08.02.2021, Frollo # Level: Intermediate # Project: Create a Damien Hirst painting using module turtle import turtle import random turtle.colormode(255) # change the module to use the rgb color code tim = turtle.Turtle() tim.speed("fastest") tim.penup() # remove the lines connecting the dots tim.hideturtle() # remove the turtle ## extract the colors of the painting "painting.jpg" # import colorgram as cg # rgb_colors = [] # list of tuples # colors = cg.extract("painting.jpg", 30) # for color in colors: # r = color.rgb.r # g = color.rgb.g # b = color.rgb.b # new_color = (r, g, b) # rgb_colors.append(new_color) # # print(rgb_colors) # copy the output rgb_colors and remove first # two colors as they are the background of the painting color_list = [(238, 246, 244), (249, 243, 247), (1, 12, 31), (53, 25, 17), (218, 127, 106), (10, 104, 159), (242, 213, 68), (149, 83, 39), (215, 87, 63), (155, 6, 24), (165, 162, 31), (157, 62, 102), (10, 64, 33), (206, 74, 104), (11, 96, 57), (95, 6, 20), (174, 135, 163), (1, 61, 145), (7, 172, 216), (3, 213, 207), (159, 33, 24), (8, 140, 85), (145, 227, 217), (122, 193, 147), (220, 177, 216), (100, 218, 229), (117, 171, 192), (79, 135, 178)] # draw 10 x 10 dots in a random pattern using the used_colors # put the cursor at the left, down part of the screen tim.setheading(255) tim.forward(325) tim.setheading(0) number_of_dots = 100 for dot_count in range(1, number_of_dots + 1): # in total moved 500 tim.dot(20, random.choice(color_list)) tim.forward(50) if dot_count % 10 == 0: tim.setheading(90) tim.forward(50) tim.setheading(180) tim.forward(500) tim.setheading(0) #create a screen screen = turtle.Screen() screen.exitonclick()
lvrbanec/100DaysOfCode_Python
Project12, Beginner, HigherLowerGame/main.py
<reponame>lvrbanec/100DaysOfCode_Python<gh_stars>0 # 05/02/2021, Frollo # Level: Beginner # Project: Higher Lower game import art from game_data import data import random from replit import clear # select a random datapoint def select_datapoint(): """"Selects a random datapoint from a list, returns it as a dictionary""" index = random.randint(0, len(data)-1) # !can also use random.choice(list) datapoint_dict = data[index] return datapoint_dict # check if the player made a right decision and count the score def compare_datapoints(dic_A, dic_B, player_decision, prev_score): ''' Looks if the player made a right decision and if yes increases the final score by 1. Output is guessed (True) or not guessed (False). ''' # if the player is right print it if (dic_A['follower_count'] > dic_B['follower_count'] and player_decision == "a") or (dic_A['follower_count'] < dic_B['follower_count'] and player_decision == "b"): prev_score += 1 print(f"You're right! Current score: {prev_score}") return [True, prev_score] # if the player is wrong print it else: print(f"Sorry, that's wrong. Current score: {prev_score}") return [False, prev_score] print(art.logo) A = select_datapoint() game_over = False score = 0 while not game_over: print(f"Compare A: {A['name']}, {A['description']}, from {A['country']}.") print(art.vs) B = select_datapoint() while A == B: # if two datapoints are selected, pick again B = select_datapoint() print(f"Against B: {B['name']}, {B['description']}, from {B['country']}.") answer = input("Who has more followers? Type 'A' or 'B': ").lower() clear() print(art.logo) guessed, score = compare_datapoints(dic_A = A, dic_B = B, player_decision = answer, prev_score = score ) # returns true if player guessed A = B if not guessed: game_over = True
lvrbanec/100DaysOfCode_Python
Project11, Beginner, NumberGuess/art.py
logo = """" _______ (_______) _ ___ _ _ _____ ___ ___ | | (_ | | | | ___ |/___)/___) | |___) | |_| | ____|___ |___ | \_____/|____/|_____(___/(___/ _ | | _| |_| |__ _____ (_ _| _ \| ___ | | |_| | | | ____|_ \__|_| |_|_____) | ____ _ _ ____ | |__ _____ ____ | _ \| | | | \| _ \| ___ |/ ___) | | | | |_| | | | | |_) | ____| | |_| |_|____/|_|_|_|____/|_____|_| """
lvrbanec/100DaysOfCode_Python
Project04, Beginner, FizzBuzz/main.py
# 01.02.2021 # level: Beginner # computer solves the Fizz Buzz game from number 1 to 100 (if the number is devisible by 3, print Fizz. If the number is devisible by 5 print Buzz. If its devisible by both, print FizzBuzz. Otherwise print the number.) for number in range(1, 101): if number % 3 == 0 and number % 5 == 0: print('FizzBuzz') elif number % 3 == 0: print ('Fizz') elif number % 5 == 0: print('Buzz') else: print(number)
lvrbanec/100DaysOfCode_Python
Project11, Beginner, NumberGuess/main.py
# 05.02.2021, Frollo # Level Beginner # Project: Number Guessing Game # computer randlomly picks an integer between 1 and 100, player needs to guess the number import random from art import logo EASY_LEVEL_TURNS = 10 # constants are accessible in functions HARD_LEVEL_TURNS = 5 # compare two numbers def compare(number, num_guess): if num_guess > number: print("Too high.") case = 0 elif num_guess < number: print("Too low.") case = 1 elif num_guess == number: print("You guessed! The number was {number}") case = 2 return case #chose a difficulty def set_difficulty(): def_difficulty = input("Choose a difficulty. Type 'easy' or 'hard': ") if def_difficulty == "hard": return HARD_LEVEL_TURNS elif def_difficulty == "easy": return EASY_LEVEL_TURNS def play_game(): # select a random number between 1 and 100 all_numbers = list(range(1,101)) # can be written as ? randint(1,100) rand_number = random.choice(all_numbers) #print(number) print(logo) print("Welcome to the Number Guessing Game!") attempts = set_difficulty() game_over = False while not game_over: print(f"You have {attempts} attempts remaining to guess the number.") num_guess = int(input("Make a guess: ")) case = compare(rand_number, num_guess) # reduce the number of attempts attempts -= 1 # end the game if you run out of attempts, game over if attempts == 0: print("Game lost") game_over = True # if not guessed, try again elif case == 0 or case == 1: print("Guess again.") # if guesssed, stop the game elif case == 2: game_over = True play_game()
lvrbanec/100DaysOfCode_Python
Project16, Intermediate, TurtleRace using turtle module/main.py
# 08.02.2021, Frollo # Level: Intermediate # Project: Make a turtle race betting game from turtle import Turtle, Screen import random is_race_on = False screen = Screen() screen.setup(width=500, height=400) user_bet = screen.textinput(title="Make your bet", prompt="Which turtle will win the race? Enter a color: ") colors = ["red", "green", "purple", "blue", "yellow", "black"] all_turtles = [] y_position = -100 for turtle_index in range(6): new_turtle = Turtle("turtle") new_turtle.color(colors[turtle_index]) new_turtle.penup() new_turtle.goto(x=-230, y=y_position) y_position += 40 all_turtles.append(new_turtle) if user_bet: # of variable exists (wait for the input of the user) is_race_on = True while is_race_on: for turtle in all_turtles: if turtle.xcor() > 230: is_race_on = False # stop the race winning_color = turtle.pencolor() if winning_color == user_bet: print(f"You've won! The {winning_color} turtle is the winner!") else: print(f"You've lost! The {winning_color} turtle is the winner!") rand_distance = random.randint(0, 20) turtle.forward(rand_distance) screen.exitonclick()
lvrbanec/100DaysOfCode_Python
Project09, Beginner, Calculator/main.py
# 03.02.21, Frollo # Level: beginner # Project: Easy calculator from art import logo print(logo) # Operations # Add def add(n1, n2): return n1 + n2 # Substract def substract(n1, n2): return n1 - n2 # Multipy def multipy(n1, n2): return n1 * n2 # Divide def divide(n1, n2): return n1 / n2 operations = { "+" : add, "-" : substract, "*" : multipy, "/" : divide } def call_again(): # to restart the code upon conditional switch = True while switch: # pick a first number num1 = float(input("What's the first number?: ")) # pick a second number and an operation, and if wanted claculate with a previous result while switch: print("Pick one of the following operation:") for key in operations: print(key) operation_symbol= input() chosen_function = operations[operation_symbol] num2 = float(input("What's the next number?: ")) result = chosen_function(num1, num2) print(f"{num1} {operation_symbol} {num2} = {result}") should_continue = input(f"Type 'y' to continue calculating with {result}, type 'n' to start a new calculation, or type 'esc' to exit.\n") if should_continue == "esc": switch = False print("Thank you for using our calculator.") elif should_continue == 'y': num1 = result elif should_continue == 'n': call_again() call_again()
lvrbanec/100DaysOfCode_Python
Project14, Intermediate, KnowledgeQuiy, OOP/quiz_brain.py
<filename>Project14, Intermediate, KnowledgeQuiy, OOP/quiz_brain.py class QuizBrain: # load in the question list of objects, set the question number as 0 def __init__(self, question_list): self.question_number = 0 self.score = 0 self.question_list = question_list # check if there are unused objects left in the list of objects def still_has_questions(self): return len(self.question_list) > self.question_number # increase the question number and show the next question def next_question(self): current_question = self.question_list[self.question_number] # enters the # specific object in a list, no matter of its name # current_question.text fetches the object's attribute text self.question_number += 1 user_answer = input(f"Q.{self.question_number}: {current_question.text} (True/False): ") self.check_answer(user_answer, current_question.answer) # checks if the answer was correct and tracks the number of right answers def check_answer(self, user_answer, correct_answer): if user_answer.lower() == correct_answer.lower(): self.score += 1 print("You got it right!") else: print("That's wrong.") print(f"The correct answer was: {correct_answer}") print(f"Your current score is: {self.score}/{self.question_number}.\n")
lvrbanec/100DaysOfCode_Python
Project14, Intermediate, KnowledgeQuiy, OOP/main.py
from question_model import Question from data import question_data from quiz_brain import QuizBrain # Reformat the data: create a list of objects containing .question and .answer # from a list of dictionaries with "question" and "answer" question_bank = [] for dictionary in question_data: question = dictionary["text"] answer = dictionary["answer"] new_question = Question(question, answer) question_bank.append(new_question) # create an object quiz, on which the program is ran quiz = QuizBrain(question_bank) while quiz.still_has_questions(): quiz.next_question() print("You've completed the quiz") print(f"Your final score was: {quiz.score}/{quiz.question_number}.")
lvrbanec/100DaysOfCode_Python
Project02, Beginner, Camel Game/main.py
<reponame>lvrbanec/100DaysOfCode_Python # 30.01.2021 # level Beginner # The point of the game is to find the water in the deser to escape Mr. Death print(''' ********************************************************************** ### #### # ### #### ### ######## #### ######### ## #### # ### # ###### # ### # ### # ### # ##### # ### # ##### # # #### ######### ## ### # ################# ###### ##### # ################## ######## ##### ## ############### ########## ####### ############## ## ##### ##### ##### ##### # #### ###### ### ## # # #### ######### ## ## # #### ######### ## ## # #### # ####### ## ## # ##### # ##### ## ## # #### # ###### ## ## # ###### ######### ## ## # #################### # ##################### ####################### ##################### ################### ###### ####### #### ###### ### ##### ### ### # ### ### # ### ## # ### #### ## ### ************************************************************************ ''') # the art is taken from https://ascii.co.uk/art/camels print("Welcome to The Desert.") print("You were riding cammels, when your cammel (with you on the back) ran away and you got lost in the desert.") # \' escapes the ' print("You\'r mission is to find the water, or otherwise you will meet Mr. Death. You have 3 days until you dehydrate.") dome = input("You see one huge sand dome on the right. Other than that there is not much to see. The sand is everywhere. Where should you go? Write \"right\" to climb the dome, or \"left\" to walk next to the dome.\n").lower() if dome == "left": night = input("You tried to walk next to the dome but a huge wind came over and the dome fell apart. Good think you didn't climb that dome, or you would have died. Lucky you. The night has finally fallen. You see the sky. Should you follow the brighest star and move towards the north, or would you rather rest and sleep? Write \"walk\" or \"rest\".\n").lower() if night == "walk": elephants = input("Good choice! You walked towards the north until you met a group of elephants. Should you continue walking north, should you try to shoe the elephants away, or should you follow the elephants? Write \"walk\", \"shoo\" or \"follow\".\n").lower() if elephants == "walk": print("You continued walking norht, but you never reached the water. Sorry man.") elif elephants == "shoo": print("The elephants felt attached and smashed you with their huge legs. Be smarter next time.") elif elephants == "follow": print ("It is well known that the elephants remember the paths to the water in desert. Seems like you watched some documentaries on netflix. You found water. You are the survivor.") else: print("You decided to rest, but the temparatures fell under zero. You froze. Try again.") else: print("You climbed the dome and from the top saw your city, but the strong winds came and the dome collapsed. You got burried under the sand. Mr. Death came. Bad luck.")
lvrbanec/100DaysOfCode_Python
Project10, Beginner, Balckjack game/main.py
# 04.02.2021, Frollo # Level: Beginner # project: simulation of a blackjack game ############### Our Blackjack House Rules ##################### ## The deck is unlimited in size. ## There are no jokers. ## The Jack/Queen/King all count as 10. ## The the Ace can count as 11 if sum < 21 or 1 if sum > 21. ## The cards in the list have equal probability of being drawn. ## Cards are not removed from the deck as they are drawn. from art import logo from replit import clear import random # fuction for dealing 1 card def deal_one_card(): """ returns a random card from the deck """ cards = [11, 2, 3, 4, 5, 6, 7, 8, 9, 10, 10, 10, 10] return random.choice(cards) # fuction for dealing 2 cards def deal_two_cards(): """ returns two random cards from the deck as a tuple""" cards = [11, 2, 3, 4, 5, 6, 7, 8, 9, 10, 10, 10, 10] first_card = random.choice(cards) second_card = random.choice(cards) return first_card, second_card # returns tuple # fuction for calculating sum of cards and declaring a blackjack def calculate_score(cards): """ calculates a score of a deck of cards (from a tuple) """ if sum(cards) == 21 and len(cards) == 2: return 0 # blackjack! if 11 in cards and sum(cards) > 21: cards = tuple(list(cards).remove(11).append(1)) # check if works return sum(cards) # function to compare player_cards, and pc scores def compare(player_sum, pc_sum): if pc_sum > 21 and player_sum > 21: return "You went over. You lose." elif player_sum == pc_sum: return "Draw" elif pc_sum == 0: return "Lose, opponen has Blackjack" elif player_sum == 0: return "Win with a Blackjack" elif player_sum > 21: return "You went over, you lose." elif pc_sum > 21: return "Computer went over, you win." elif player_sum > pc_sum: return "Your win." else: return "You lose." # blackjack game def play_game(): print (logo) player_cards = deal_two_cards() pc_cards = deal_two_cards() # decisions of a player, ie does she/he wants another card? is_game_over = False while not is_game_over: player_sum = calculate_score(player_cards) pc_sum = calculate_score(pc_cards) print(f"Your cards are: {player_cards}. Current score : {player_sum}.") print(f"Computer's first card: {pc_cards[0]}") if player_sum == 0 or pc_sum == 0 or player_sum > 21 : is_game_over = True else: additional_card = input("Type 'y' to get another card, type 'n' to pass: ") if additional_card == 'y': player_cards += (deal_one_card(),) # adding an int to tuple. int has to be in a (int,) form else: is_game_over = True # decisions of a computer, ie if computer has less than 17, needs to draw one more card while pc_sum !=0 and pc_sum < 17: # 0 is blackjack pc_cards += (deal_one_card(),) pc_sum = calculate_score(pc_cards) print(f"Your cards are: {player_cards}. Final score : {player_sum}.") print(f"Computer's cards are: {pc_cards}. Final score : {pc_sum}.") print(compare(player_sum, pc_sum)) # Do you want to restart a game? while input("Do you want to play the game of blackjack? Type 'y' or 'n': ") == 'y': clear() play_game()
lvrbanec/100DaysOfCode_Python
Project01, Beginner, Love Calculator/main.py
<gh_stars>0 # 30.01.2021 # level: Beginner # The aim of this code is to create a love calculator, which "predicts" how well two people match based on their names. The logic is based on the game we played as a kids, and is explained at https://www.buzzfeed.com/ariannarebolini/what-are-the-chances-your-crush-is-actually-your-true-love print("Welcome to the Love Calculator!") name1 = input("What is your name? \n") name2 = input("What is their name? \n") name_together = name1 + name2 name_together_low = name_together.lower() #print(name_together_low) score_true = name_together_low.count("t")+ name_together_low.count("r") + name_together_low.count("u") + name_together_low.count("e") #print(score_true) score_love= name_together_low.count('l') + name_together_low.count('o') + name_together_low.count('v') + name_together_low.count('e') #print(score_love) final_score = int(f"{score_true}{score_love}") #print(final_score) if final_score < 10 or final_score>90: print(f"Your score is {final_score}, you go together like coke and mentos.") elif final_score <= 50 and final_score >= 40: print(f"Your score is {final_score}, you are alright together.") else: print(f"Your score is {final_score}.")
lvrbanec/100DaysOfCode_Python
Project05, Beginner, Password Generator/main.py
<filename>Project05, Beginner, Password Generator/main.py # 01.02.2021 # Lever: Beginner # Password Generator Project # generate a password with the number of letters, numbers and symbols specified by a subject import random letters = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z'] numbers = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'] symbols = ['!', '#', '$', '%', '&', '(', ')', '*', '+'] print("Welcome to the PyPassword Generator!") nr_letters= int(input("How many letters would you like in your password?\n")) nr_symbols = int(input(f"How many symbols would you like?\n")) nr_numbers = int(input(f"How many numbers would you like?\n")) # select a random letter selected_letters=[] # for a string, we would write selected_letters = "" for i in range(nr_letters): rand_letter = letters[random.randint(0, len(letters)-1)] # or rand_letter = random.choice(letters) selected_letters.append(rand_letter) # or use sellected_letters += rand_letter print(selected_letters) # select a random number selected_numbers=[] for i in range(nr_numbers): rand_number = numbers[random.randint(0, len(numbers)-1)] selected_numbers.append(rand_number) print(selected_numbers) # select a random symbol selected_symbols=[] for i in range(nr_symbols): rand_symbol = symbols[random.randint(0, len(symbols)-1)] selected_symbols.append(rand_symbol) print(selected_symbols) password_list = selected_letters + selected_numbers + selected_symbols random.shuffle(password_list) # automatically overwrites variable. if new variable needed use random.sample(x, len(x)) password_str = "".join(password_list) print(password_str)
lvrbanec/100DaysOfCode_Python
Project08. Beginner, Blind bid decision/main.py
# 03.02.21, Frollo # level: Beginner # project: Create a blind bid program from replit import clear from art import logo print(logo) # add a new bidder function bidder_info = [] # could be also saved as {name1: amount1, name2: amount2} def add_bidder_info(val_name, val_bid): new_bidder = {"name": val_name, "bid": val_bid} bidder_info.append(new_bidder) # find the highest bidder function def find_highest_bidder(list_of_dictionary): max_bid = 0 for dictionary in list_of_dictionary: bid = dictionary["bid"] if bid > max_bid: max_bid = bid max_name = dictionary["name"] print(f'Highest bid is {max_bid}$ bidded by {max_name}.') # save info about the bidders will_continue = True while will_continue: input_name = input("What is your name?\n") input_bid = float(input("What's your bid?\n$")) input_continue = input("Are there any other bidders? Type 'yes' or 'no'.\n") add_bidder_info(val_name = input_name, val_bid = input_bid) clear() if input_continue == "no": will_continue = False print(bidder_info) find_highest_bidder(bidder_info)
SubZero12556/Cats2dogs_ONNX
networks.py
<filename>networks.py import torch import torch.nn as nn from torch.nn.parameter import Parameter class ResnetGenerator(nn.Module): def __init__(self, input_nc, output_nc, ngf=64, n_blocks=6, img_size=256, light=False): assert(n_blocks >= 0) super(ResnetGenerator, self).__init__() self.input_nc = input_nc self.output_nc = output_nc self.ngf = ngf self.n_blocks = n_blocks self.img_size = img_size self.light = light DownBlock = [] DownBlock += [nn.ReflectionPad2d(3), nn.Conv2d(input_nc, ngf, kernel_size=7, stride=1, padding=0, bias=False), nn.InstanceNorm2d(ngf), nn.ReLU(True)] # Down-Sampling n_downsampling = 2 for i in range(n_downsampling): mult = 2**i DownBlock += [nn.ReflectionPad2d(1), nn.Conv2d(ngf * mult, ngf * mult * 2, kernel_size=3, stride=2, padding=0, bias=False), nn.InstanceNorm2d(ngf * mult * 2), nn.ReLU(True)] # Down-Sampling Bottleneck mult = 2**n_downsampling for i in range(n_blocks): DownBlock += [ResnetBlock(ngf * mult, use_bias=False)] # Class Activation Map self.gap_fc = nn.Linear(ngf * mult, 1, bias=False) self.gmp_fc = nn.Linear(ngf * mult, 1, bias=False) self.conv1x1 = nn.Conv2d(ngf * mult * 2, ngf * mult, kernel_size=1, stride=1, bias=True) self.relu = nn.ReLU(True) # Gamma, Beta block if self.light: FC = [nn.Linear(ngf * mult, ngf * mult, bias=False), nn.ReLU(True), nn.Linear(ngf * mult, ngf * mult, bias=False), nn.ReLU(True)] else: FC = [nn.Linear(img_size // mult * img_size // mult * ngf * mult, ngf * mult, bias=False), nn.ReLU(True), nn.Linear(ngf * mult, ngf * mult, bias=False), nn.ReLU(True)] self.gamma = nn.Linear(ngf * mult, ngf * mult, bias=False) self.beta = nn.Linear(ngf * mult, ngf * mult, bias=False) # Up-Sampling Bottleneck for i in range(n_blocks): setattr(self, 'UpBlock1_' + str(i+1), ResnetAdaILNBlock(ngf * mult, use_bias=False)) # Up-Sampling UpBlock2 = [] for i in range(n_downsampling): mult = 2**(n_downsampling - i) UpBlock2 += [nn.Upsample(scale_factor=2, mode='nearest'), nn.ReflectionPad2d(1), nn.Conv2d(ngf * mult, int(ngf * mult / 2), kernel_size=3, stride=1, padding=0, bias=False), ILN(int(ngf * mult / 2)), nn.ReLU(True)] UpBlock2 += [nn.ReflectionPad2d(3), nn.Conv2d(ngf, output_nc, kernel_size=7, stride=1, padding=0, bias=False), nn.Tanh()] self.DownBlock = nn.Sequential(*DownBlock) self.FC = nn.Sequential(*FC) self.UpBlock2 = nn.Sequential(*UpBlock2) def forward(self, input): x = self.DownBlock(input) gap = torch.nn.functional.adaptive_avg_pool2d(x, 1) gap_logit = self.gap_fc(gap.view(x.shape[0], -1)) gap_weight = list(self.gap_fc.parameters())[0] gap = x * gap_weight.unsqueeze(2).unsqueeze(3) gmp = torch.nn.functional.adaptive_max_pool2d(x, 1) gmp_logit = self.gmp_fc(gmp.view(x.shape[0], -1)) gmp_weight = list(self.gmp_fc.parameters())[0] gmp = x * gmp_weight.unsqueeze(2).unsqueeze(3) cam_logit = torch.cat([gap_logit, gmp_logit], 1) x = torch.cat([gap, gmp], 1) x = self.relu(self.conv1x1(x)) heatmap = torch.sum(x, dim=1, keepdim=True) if self.light: x_ = torch.nn.functional.adaptive_avg_pool2d(x, 1) x_ = self.FC(x_.view(x_.shape[0], -1)) else: x_ = self.FC(x.view(x.shape[0], -1)) gamma, beta = self.gamma(x_), self.beta(x_) for i in range(self.n_blocks): x = getattr(self, 'UpBlock1_' + str(i+1))(x, gamma, beta) out = self.UpBlock2(x) return out, cam_logit, heatmap class ResnetBlock(nn.Module): def __init__(self, dim, use_bias): super(ResnetBlock, self).__init__() conv_block = [] conv_block += [nn.ReflectionPad2d(1), nn.Conv2d(dim, dim, kernel_size=3, stride=1, padding=0, bias=use_bias), nn.InstanceNorm2d(dim), nn.ReLU(True)] conv_block += [nn.ReflectionPad2d(1), nn.Conv2d(dim, dim, kernel_size=3, stride=1, padding=0, bias=use_bias), nn.InstanceNorm2d(dim)] self.conv_block = nn.Sequential(*conv_block) def forward(self, x): out = x + self.conv_block(x) return out class ResnetAdaILNBlock(nn.Module): def __init__(self, dim, use_bias): super(ResnetAdaILNBlock, self).__init__() self.pad1 = nn.ReflectionPad2d(1) self.conv1 = nn.Conv2d(dim, dim, kernel_size=3, stride=1, padding=0, bias=use_bias) self.norm1 = adaILN(dim) self.relu1 = nn.ReLU(True) self.pad2 = nn.ReflectionPad2d(1) self.conv2 = nn.Conv2d(dim, dim, kernel_size=3, stride=1, padding=0, bias=use_bias) self.norm2 = adaILN(dim) def forward(self, x, gamma, beta): out = self.pad1(x) out = self.conv1(out) out = self.norm1(out, gamma, beta) out = self.relu1(out) out = self.pad2(out) out = self.conv2(out) out = self.norm2(out, gamma, beta) return out + x class adaILN(nn.Module): def __init__(self, num_features, eps=1e-5): super(adaILN, self).__init__() self.eps = eps self.rho = Parameter(torch.Tensor(1, num_features, 1, 1)) self.rho.data.fill_(0.9) def forward(self, input, gamma, beta): in_mean, in_var = torch.mean(input, dim=[2, 3], keepdim=True), torch.var(input, dim=[2, 3], keepdim=True) out_in = (input - in_mean) / torch.sqrt(in_var + self.eps) ln_mean, ln_var = torch.mean(input, dim=[1, 2, 3], keepdim=True), torch.var(input, dim=[1, 2, 3], keepdim=True) out_ln = (input - ln_mean) / torch.sqrt(ln_var + self.eps) out = self.rho.expand(input.shape[0], -1, -1, -1) * out_in + (1-self.rho.expand(input.shape[0], -1, -1, -1)) * out_ln out = out * gamma.unsqueeze(2).unsqueeze(3) + beta.unsqueeze(2).unsqueeze(3) return out class ILN(nn.Module): def __init__(self, num_features, eps=1e-5): super(ILN, self).__init__() self.eps = eps self.rho = Parameter(torch.Tensor(1, num_features, 1, 1)) self.gamma = Parameter(torch.Tensor(1, num_features, 1, 1)) self.beta = Parameter(torch.Tensor(1, num_features, 1, 1)) self.rho.data.fill_(0.0) self.gamma.data.fill_(1.0) self.beta.data.fill_(0.0) def forward(self, input): in_mean, in_var = torch.mean(input, dim=[2, 3], keepdim=True), torch.var(input, dim=[2, 3], keepdim=True) out_in = (input - in_mean) / torch.sqrt(in_var + self.eps) ln_mean, ln_var = torch.mean(input, dim=[1, 2, 3], keepdim=True), torch.var(input, dim=[1, 2, 3], keepdim=True) out_ln = (input - ln_mean) / torch.sqrt(ln_var + self.eps) out = self.rho.expand(input.shape[0], -1, -1, -1) * out_in + (1-self.rho.expand(input.shape[0], -1, -1, -1)) * out_ln out = out * self.gamma.expand(input.shape[0], -1, -1, -1) + self.beta.expand(input.shape[0], -1, -1, -1) return out class Discriminator(nn.Module): def __init__(self, input_nc, ndf=64, n_layers=5): super(Discriminator, self).__init__() model = [nn.ReflectionPad2d(1), nn.utils.spectral_norm( nn.Conv2d(input_nc, ndf, kernel_size=4, stride=2, padding=0, bias=True)), nn.LeakyReLU(0.2, True)] for i in range(1, n_layers - 2): mult = 2 ** (i - 1) model += [nn.ReflectionPad2d(1), nn.utils.spectral_norm( nn.Conv2d(ndf * mult, ndf * mult * 2, kernel_size=4, stride=2, padding=0, bias=True)), nn.LeakyReLU(0.2, True)] mult = 2 ** (n_layers - 2 - 1) model += [nn.ReflectionPad2d(1), nn.utils.spectral_norm( nn.Conv2d(ndf * mult, ndf * mult * 2, kernel_size=4, stride=1, padding=0, bias=True)), nn.LeakyReLU(0.2, True)] # Class Activation Map mult = 2 ** (n_layers - 2) self.gap_fc = nn.utils.spectral_norm(nn.Linear(ndf * mult, 1, bias=False)) self.gmp_fc = nn.utils.spectral_norm(nn.Linear(ndf * mult, 1, bias=False)) self.conv1x1 = nn.Conv2d(ndf * mult * 2, ndf * mult, kernel_size=1, stride=1, bias=True) self.leaky_relu = nn.LeakyReLU(0.2, True) self.pad = nn.ReflectionPad2d(1) self.conv = nn.utils.spectral_norm( nn.Conv2d(ndf * mult, 1, kernel_size=4, stride=1, padding=0, bias=False)) self.model = nn.Sequential(*model) def forward(self, input): x = self.model(input) gap = torch.nn.functional.adaptive_avg_pool2d(x, 1) gap_logit = self.gap_fc(gap.view(x.shape[0], -1)) gap_weight = list(self.gap_fc.parameters())[0] gap = x * gap_weight.unsqueeze(2).unsqueeze(3) gmp = torch.nn.functional.adaptive_max_pool2d(x, 1) gmp_logit = self.gmp_fc(gmp.view(x.shape[0], -1)) gmp_weight = list(self.gmp_fc.parameters())[0] gmp = x * gmp_weight.unsqueeze(2).unsqueeze(3) cam_logit = torch.cat([gap_logit, gmp_logit], 1) x = torch.cat([gap, gmp], 1) x = self.leaky_relu(self.conv1x1(x)) heatmap = torch.sum(x, dim=1, keepdim=True) x = self.pad(x) out = self.conv(x) return out, cam_logit, heatmap class RhoClipper(object): def __init__(self, min, max): self.clip_min = min self.clip_max = max assert min < max def __call__(self, module): if hasattr(module, 'rho'): w = module.rho.data w = w.clamp(self.clip_min, self.clip_max) module.rho.data = w
SubZero12556/Cats2dogs_ONNX
utils.py
from scipy import misc import os, cv2, torch import numpy as np def load_test_data(image_path, size=256): img = misc.imread(image_path, mode='RGB') img = misc.imresize(img, [size, size]) img = np.expand_dims(img, axis=0) img = preprocessing(img) return img def preprocessing(x): x = x/127.5 - 1 # -1 ~ 1 return x def save_images(images, size, image_path): return imsave(inverse_transform(images), size, image_path) def inverse_transform(images): return (images+1.) / 2 def imsave(images, size, path): return misc.imsave(path, merge(images, size)) def merge(images, size): h, w = images.shape[1], images.shape[2] img = np.zeros((h * size[0], w * size[1], 3)) for idx, image in enumerate(images): i = idx % size[1] j = idx // size[1] img[h*j:h*(j+1), w*i:w*(i+1), :] = image return img def check_folder(log_dir): if not os.path.exists(log_dir): os.makedirs(log_dir) return log_dir def str2bool(x): return x.lower() in ('true') def cam(x, size = 256): x = x - np.min(x) cam_img = x / np.max(x) cam_img = np.uint8(255 * cam_img) cam_img = cv2.resize(cam_img, (size, size)) cam_img = cv2.applyColorMap(cam_img, cv2.COLORMAP_JET) return cam_img / 255.0 def imagenet_norm(x): mean = [0.485, 0.456, 0.406] std = [0.299, 0.224, 0.225] mean = torch.FloatTensor(mean).unsqueeze(0).unsqueeze(2).unsqueeze(3).to(x.device) std = torch.FloatTensor(std).unsqueeze(0).unsqueeze(2).unsqueeze(3).to(x.device) return (x - mean) / std def denorm(x): return x * 0.5 + 0.5 def tensor2numpy(x): return x.detach().cpu().numpy().transpose(1,2,0) def RGB2BGR(x): return cv2.cvtColor(x, cv2.COLOR_RGB2BGR)
jamesstidard/pipenv-sublime
vendor/pipenvlib.py
<gh_stars>10-100 import os import json from . import toml from . import delegator class Dependency(object): """A Dependency.""" def __init__(self, name, constraint, locked=False): self.name = name self.constraint = constraint self.locked = locked def __repr__(self): return "<Dependency '{0}' constraint='{1}'>".format( self.name, self.constraint ) class LockedDependency(object): """A Locked Dependency.""" def __init__(self, name, constraint, hashes): self.name = name self.constraint = constraint self.hashes = hashes def __repr__(self): return "<LockedDependency '{0}{1}'>".format( self.name, self.constraint ) class Requirement(object): """A Requirement.""" def __init__(self, name, constraint): self.name = name self.constraint = constraint def __repr__(self): return "<Requirement '{0}' constraint='{1}'>".format( self.name, self.constraint ) class PipenvProject(object): """A Pipenv project.""" def __init__(self, home, pipfile='Pipfile', create=False): self.home = home self.pipfile = pipfile if not create: # Assert that the Pipfile exists. self.assert_has_pipfile() else: # Cheat a project-creation by installing, then uninstalling # the Requests library. :) self.install('requests') self.uninstall('requests') @property def _pipfile_path(self): return os.path.sep.join([self.home, self.pipfile]) @property def _lockfile_path(self): return os.path.sep.join([self.home, '{0}.lock'.format(self.pipfile)]) def assert_has_pipfile(self): """Asserts that the Pipfile exists.""" assert os.path.isfile(self._pipfile_path) def assert_has_lockfile(self): """Asserts that the Pipfile.lock exists.""" assert os.path.isfile(self._lockfile_path) def _get_section_of_pipfile(self, section, target): def gen(): pipfile = toml.load(self._pipfile_path) for package in pipfile[section]: name = package constraint = pipfile[section][package] yield target(name, constraint) return [p for p in gen()] @property def packages(self): """Returns a list of Dependency objects (for [packages]) for the Pipenv project""" return self._get_section_of_pipfile('packages', Dependency) @property def dev_packages(self): """Returns a list of Dependency objects (for [dev-packages]) for the Pipenv project. """ return self._get_section_of_pipfile('dev-packages', Dependency) @property def requires(self): """Returns a list of Requirement objects for the Pipenv project. """ return self._get_section_of_pipfile('requires', Requirement) @property def locked_packages(self): """Returns a list of LockedDependency objects for the Pipenv project. """ self.assert_has_lockfile() def gen(): with open(self._lockfile_path) as f: lockfile = json.load(f) for package in lockfile['default']: name = package constraint = lockfile['default'][package]['version'] hashes = lockfile['default'][package]['hashes'] yield LockedDependency(name, constraint, hashes) return [l for l in gen()] @property def locked_dev_packages(self): """Returns a list of LockedDependency objects for the Pipenv project. """ self.assert_has_lockfile() def gen(): with open(self._lockfile_path) as f: lockfile = json.load(f) for package in lockfile['develop']: name = package constraint = lockfile['develop'][package]['version'] hashes = lockfile['develop'][package]['hashes'] yield LockedDependency(name, constraint, hashes) return [l for l in gen()] @property def locked_requirements(self): """Returns a list of Requirement objects for the Pipenv project, from the Pipfile.lock. """ self.assert_has_lockfile() def gen(): with open(self._lockfile_path) as f: lockfile = json.load(f) for require in lockfile['_meta']['requires']: name = require constraint = lockfile['_meta']['requires'][require] yield Requirement(name, constraint) return [l for l in gen()] def run(self, cmd, block=True): """Run a Pipenv command for the Pipenv project.""" return delegator.run('pipenv {0}'.format(cmd), cwd=self.home, block=block) def install(self, package_name, constraint=None, dev=False): """Installs a given package to the Pipenv project.""" # If no constraint was if constraint is not None: # Append the constraint to the package name. package_name = 'package_name{0}'.format(constraint) dev = '' if not dev else '--dev' return self._run('install {0} {1}'.format(package_name, dev)).return_code == 0 def uninstall(self, package_name): """Uninstalls a given package from the Pipenv project.""" return self.run('uninstall {0}'.format(package_name)).return_code == 0 def check(self): """Runs Pipenv check on the Pipenv project.""" return self.run('check').return_code == 0 @property def virtualenv_location(self): return self.run('--venv').out.strip()
jamesstidard/pipenv-sublime
subl_pipenv.py
import os import sublime import sublime_plugin from .vendor import requests from .vendor import parse from .vendor import pipenvlib TEMPLATE = "<a href='{0}'>{0}</a><br/>" ALL_PACKAGES = list() def plugin_loaded(): pass class PipenvIsEnabledMixin: def is_enabled(self): open_files = [view.file_name() for view in sublime.active_window().views()] for o_f in open_files: o_f = os.path.abspath(o_f) dirname = os.path.dirname(o_f) dirname = os.path.sep.join([dirname, '..', '..']) for root, dirs, files in os.walk(dirname, followlinks=True): if 'Pipfile' in files: return True return False class InstallHandler(sublime_plugin.ListInputHandler): def __init__(self): super(InstallHandler, self).__init__() def _yield_packages(self): # Set the status message. sublime.status_message("Fetching all available packages from PyPi (just a sec!)…") # Make the HTTP Request. r = requests.get('https://pypi.python.org/simple/') sublime.status_message("") # Yield results. for result in parse.findall(TEMPLATE, r.text): yield result[0] @property def _all_packages(self): return ALL_PACKAGES @_all_packages.setter def function(self, value): global ALL_PACKAGES ALL_PACKAGES = value @property def all_packages(self): if self._all_packages: return self._all_packages for package in self._yield_packages(): self._all_packages.append(package) # List special packages first, because I can. kr_packages = ( 'requests', 'requests-html', 'maya', 'records', 'httpbin', 'crayons', 'delegator.py', 'tablib', 'background', 'clint', 'xerox' ) # Special shout-outs. kr_favorites = ('django', 'flask', 'docopt', 'parse', 'apistar') kr_favorites = list(kr_packages + kr_favorites) # Reverse order. kr_favorites.reverse() for kr_package in kr_favorites: package = self._all_packages.pop(self._all_packages.index(kr_package)) self._all_packages.insert(0, package) return self._all_packages def list_items(self): return self.all_packages def initial_text(self, *args): return "" class pipenv_install(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_install, self).__init__(text) # def is_enabled(self): # return super(pipenv_install, self).is_enabled() def input(self, *args): return InstallHandler() def run(self, install_handler): # The package to install. package = install_handler # The home directory for the current file name. home = os.path.dirname(sublime.active_window().active_view().file_name()) p = pipenvlib.PipenvProject(home) # Update package status. sublime.status_message("Installing {!r} with Pipenv…".format(package)) # Show the console. sublime.active_window().active_view().window().run_command('show_panel', {'panel': 'console'}) # Run the install command. c = p.run('install {}'.format(package), block=False) # Update the status bar. sublime.status_message("Waiting for {!r} to install…".format(package)) # Block on subprocess… c.block() # Print results to console. print(c.out) # Assure that the intallation was successful. try: # Ensure installation was successful. assert c.return_code == 0 # Update the status bar. sublime.status_message("Success installing {!r}!".format(package)) # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile') # Hide the console. sublime.active_window().active_view().window().run_command('hide_panel', {'panel': 'console'}) except AssertionError: # Update the status bar. sublime.status_message("Error installing {!r}!".format(package)) # Report the error. print(c.err) class pipenv_install_dev(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_install_dev, self).__init__(text) # def is_enabled(self): # return super(pipenv_install, self).is_enabled() def input(self, *args): return InstallHandler() def run(self, install_handler): # The package to install. package = install_handler # The home directory for the current file name. home = os.path.dirname(sublime.active_window().active_view().file_name()) p = pipenvlib.PipenvProject(home) # Update package status. sublime.status_message("Installing {!r} with Pipenv…".format(package)) # Show the console. sublime.active_window().active_view().window().run_command('show_panel', {'panel': 'console'}) # Run the install command. c = p.run('install --dev {}'.format(package), block=False) # Update the status bar. sublime.status_message("Waiting for {!r} to install…".format(package)) # Block on subprocess… c.block() # Print results to console. print(c.out) # Assure that the intallation was successful. try: # Ensure installation was successful. assert c.return_code == 0 # Update the status bar. sublime.status_message("Success installing {!r}!".format(package)) # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile') # Hide the console. sublime.active_window().active_view().window().run_command('hide_panel', {'panel': 'console'}) except AssertionError: # Update the status bar. sublime.status_message("Error installing {!r}!".format(package)) # Report the error. print(c.err) class UninstallHandler(sublime_plugin.ListInputHandler): def __init__(self): super(UninstallHandler, self).__init__() def list_items(self): home = os.path.dirname(sublime.active_window().active_view().file_name()) p = pipenvlib.PipenvProject(home) return list(set([p.name for p in p.packages + p.dev_packages])) def initial_text(self, *args): return "" class pipenv_uninstall(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_uninstall, self).__init__(text) # def is_enabled(self): # return super(pipenv_uninstall, self).is_enabled() def input(self, *args): return UninstallHandler() def run(self, uninstall_handler): # The package to install. package = uninstall_handler # The home directory for the current file name. home = os.path.dirname(sublime.active_window().active_view().file_name()) p = pipenvlib.PipenvProject(home) # Update package status. sublime.status_message("Un–installing {!r} with Pipenv…".format(package)) # Show the console. sublime.active_window().active_view().window().run_command('show_panel', {'panel': 'console'}) # Run the uninstall command. c = p.run('uninstall {}'.format(package), block=False) # Update the status bar. sublime.status_message("Waiting for {!r} to un–install…".format(package)) # Block on subprocess… c.block() # Print results to console. print(c.out) # Assure that the intallation was successful. try: # Ensure installation was successful. assert c.return_code == 0 # Update the status bar. sublime.status_message("Success un–installing {!r}!".format(package)) # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile') # Hide the console. sublime.active_window().active_view().window().run_command('hide_panel', {'panel': 'console'}) except AssertionError: # Update the status bar. sublime.status_message("Error un–installing {!r}!".format(package)) # Report the error. print(c.err) class pipenv_open_pipfile(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_open_pipfile, self).__init__(text) # def is_enabled(self): # return super(pipenv_open_pipfile, self).is_enabled() def run(self): # Update package status. sublime.status_message("Opening {!r} with Pipenv…".format('Pipfile')) # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile') class pipenv_open_pipfile_lock(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_open_pipfile_lock, self).__init__(text) def is_enabled(self): return super(pipenv_open_pipfile_lock, self).is_enabled() def run(self): # Update package status. sublime.status_message("Opening {!r} with Pipenv…".format('Pipfile.lock')) # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile.lock') class pipenv_lock(PipenvIsEnabledMixin, sublime_plugin.WindowCommand): def __init__(self, text): super(pipenv_lock, self).__init__(text) # def is_enabled(self): # return super(pipenv_lock, self).is_enabled() def run(self): # Update package status. sublime.status_message("Locking {!r} with Pipenv…".format('Pipfile')) home = os.path.dirname(sublime.active_window().active_view().file_name()) p = pipenvlib.PipenvProject(home) c = p.run('lock', block=False) c.block() try: assert c.return_code == 0 # Update locking status. sublime.status_message("Success!") # Open the Pipfile. sublime.active_window().active_view().window().open_file('Pipfile.lock') sublime.status_message("") except AssertionError: # Show the console. sublime.active_window().active_view().window().run_command('show_panel', {'panel': 'console'}) # Update locking status. sublime.status_message("Error while locking!") print(c.err) if __name__ == '__main__': if sublime.version() < '3000': plugin_loaded()
philskillz-coder/discord-py-paginator
examples/normal_list.py
from abc import ABC from discord import app_commands, Embed, Color, Interaction from discord.ext.paginator import paginator class ColorPaginator(paginator.Paginator, ABC): pass @app_commands.command( name="colors", description="Show some colors" ) async def show_colors(interaction: Interaction): await interaction.response.send_message( content="Nice colors", view=await ColorPaginator.from_list( interaction.client, interaction.user, data=[ { "content": "The color red", "embed": Embed(title="Red", color=Color.red()), }, { "content": "The color green", "embed": Embed(title="Green", color=Color.green()), }, { "content": "The color blue", "embed": Embed(title="Blue", color=Color.blue()), } ] ).run() )
philskillz-coder/discord-py-paginator
discord/ext/paginator/button.py
from __future__ import annotations from typing import Any, Dict, Callable, List, Coroutine, Type, Optional, Union from discord import ButtonStyle, Emoji, PartialEmoji, Interaction from discord.ui import Button from . import errors EH = Callable[[Any, Interaction, errors.ButtonException], Coroutine[Any, Any, Any]] CH = Callable[[Any, Interaction], Coroutine[Any, Any, bool]] class ButtonErrorHandler: def __init__(self, callback: EH, exception_type: Type[errors.ButtonException]): self.callback = callback self.exception_type: Type[errors.ButtonException] = exception_type async def invoke(self, instance, interaction: Interaction, error: errors.ButtonException): return await self.callback(instance, interaction, error) class ButtonCheck: def __init__(self, callback: CH, priority: int): self.callback = callback self.priority = priority async def invoke(self, instance, interaction: Interaction): return await self.callback(instance, interaction) class ButtonMeta(type): __error_handlers__: Dict[Type[errors.ButtonException], ButtonErrorHandler] = {} __checks__: Dict[int, ButtonCheck] = {} def __new__(mcs, *args: Any, **kwargs: Any): name, bases, attrs = args error_handlers = {} checks = {} new_cls = super().__new__(mcs, *args, **kwargs) for base in reversed(new_cls.__mro__): for elem, value in base.__dict__.items(): if isinstance(value, ButtonErrorHandler): error_handlers[value.exception_type] = value if isinstance(value, ButtonCheck): checks[value.priority] = value new_cls.__error_handlers__ = error_handlers new_cls.__checks__ = checks return new_cls def __init__(cls, *args: Any, **kwargs: Any) -> None: super().__init__(*args) def button_on_error(exception_type: Type[errors.ButtonException]): def deco(method: EH) -> ButtonErrorHandler: return ButtonErrorHandler(method, exception_type) return deco def button_check(priority: int): def deco(method: CH) -> ButtonCheck: return ButtonCheck(method, priority) return deco class BetterButton(Button, metaclass=ButtonMeta): def __init__( self, style: ButtonStyle = ButtonStyle.secondary, label: Optional[str] = None, disabled: bool = False, custom_id: Optional[str] = None, url: Optional[str] = None, emoji: Optional[Union[str, Emoji, PartialEmoji]] = None, row: Optional[int] = None ): super().__init__( style=style, label=label, disabled=disabled, custom_id=custom_id, url=url, emoji=emoji, row=row ) async def on_error(self, interaction: Interaction, error: errors.ButtonException): raise error async def interaction_check(self, interaction: Interaction) -> bool: return True async def on_click(self, interaction: Interaction): raise NotImplementedError("On click method not implemented!") async def callback(self, interaction: Interaction) -> Optional[Any]: try: value = await self.interaction_check(interaction) if value is not True: raise errors.ButtonFailed(f"Interaction check {self.interaction_check.__name__!r} with priority 0 failed!") _checks: List[ButtonCheck] = sorted(self.__checks__.values(), key=lambda c: c.priority) for check in _checks: value = await check.invoke(self, interaction) if value is not True: raise errors.ButtonFailed("Button check failed!") except errors.ButtonException as button_exception: for exception_type, callback in self.__error_handlers__.items(): if isinstance(button_exception, exception_type): await callback.invoke(self, interaction, button_exception) return None await self.on_error(interaction, button_exception) return None return await self.on_click(interaction) def __init__button__(self, parent): self.parent = parent
philskillz-coder/discord-py-paginator
discord/ext/paginator/modals.py
from __future__ import annotations from typing import TYPE_CHECKING, Optional from discord import User, Interaction, ui from discord.utils import MISSING if TYPE_CHECKING: from .paginator import Paginator class QuickNav(ui.Modal, title='Quick Navigation'): def __init__(self, *, parent: Paginator, user: User, title: str = MISSING, timeout: Optional[float] = None, custom_id: str = MISSING) -> None: super().__init__(title=title, timeout=timeout, custom_id=custom_id) self.parent = parent self.user = user page = ui.TextInput(label='Page') async def interaction_check(self, interaction: Interaction) -> bool: if interaction.user != self.user: await interaction.response.send_message( content="You are not allowed to do this", ephemeral=True ) raise ValueError("You are not allowed to do this!") # add better error message return True async def on_submit(self, interaction: Interaction): if not str(self.page).isdigit(): await interaction.response.send_message(f"`{self.page}` is not a number!") raise ValueError("Not a number") # add better error message await self.parent.update_page_number(interaction, int(str(self.page))) await self.parent.update_page_content(interaction)
philskillz-coder/discord-py-paginator
discord/ext/paginator/view_buttons.py
<gh_stars>1-10 from __future__ import annotations from abc import ABC from discord import ButtonStyle, User, Interaction from discord.ext.commands import Bot from . import button, errors, modals from typing import TYPE_CHECKING if TYPE_CHECKING: from .paginator import Paginator class FirstElement(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.update_page_number(interaction, 0) await self.parent.update_page_content(interaction) class PreviousElement(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.update_page_number(interaction, self.parent.page - 1) await self.parent.update_page_content(interaction) class NextElement(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.update_page_number(interaction, self.parent.page + 1) await self.parent.update_page_content(interaction) class LastElement(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.update_page_number(interaction, await self.parent.acquire_page_count(interaction)-1) await self.parent.update_page_content(interaction) class Stop(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.paginator_stop(interaction) await interaction.response.send_message( content="Stopped", ephemeral=True ) class Start(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await self.parent.paginator_start(interaction) await interaction.response.defer() values = await self.parent.acquire_page_content(interaction) values["view"] = self.parent ws = interaction.followup await ws.edit_message( (await interaction.original_message()).id, **values ) class QuickNav(button.BetterButton): def __init__( self, style: ButtonStyle, label: str, emoji: str, /, client: Bot, parent: Paginator, user: User, using: bool, disabled: bool = True ): super().__init__( style=style, label=label, emoji=emoji, disabled=disabled ) self.client = client self.parent = parent self.user = user self.using = using @button.button_check(1) async def check_author(self, interaction: Interaction) -> bool: if interaction.user != self.user: raise errors.NotAuthor("You are not allowed to do this!") return True async def on_error(self, interaction: Interaction, error: errors.ButtonException): await interaction.response.send_message( content=error.message, ephemeral=True ) async def on_click(self, interaction: Interaction): await interaction.response.send_modal(modals.QuickNav(parent=self.parent, user=self.user)) class Placeholder(button.BetterButton, ABC): def __init__( self, style: ButtonStyle, label: str, emoji: str, / ): super().__init__( style=style, label=label, emoji=emoji, disabled=True )
philskillz-coder/discord-py-paginator
discord/ext/paginator/paginator.py
<filename>discord/ext/paginator/paginator.py from discord import ui, User, Interaction, Webhook, Client, ButtonStyle from discord.ext.commands import Bot from . import view_buttons from typing import Dict, Any, List, Union, Optional, Callable, Coroutine GCP_TYPE = Callable[[Interaction, int], Coroutine[Any, Any, Dict[str, Any]]] class Paginator(ui.View): __CONFIG__: Dict = { # DO NOT CHANGE THIS! You can add custom config attrs in Paginator.CONFIG "paginator_view_timeout": 180, "paginator_ephemeral": None, # this setting overwrites ephemeral= from get_page_content. if None, not overwritten "paginator_delete_when_finished": True, # only works when paginator is not ephemeral "paginator_delete_delay": 10, "start_button_enabled": True, # option not changable! "start_button_style": ButtonStyle.success, "start_button_label": "Start", "start_button_emoji": None, "stop_button_enabled": True, # option not changable! "stop_button_style": ButtonStyle.danger, "stop_button_label": "Quit", "stop_button_emoji": None, "quick_navigation_button_enabled": True, "quick_navigation_button_style": ButtonStyle.blurple, "quick_navigation_button_label": "Nav", "quick_navigation_button_emoji": None, "quick_navigation_error_message": "%s is not a number!", # False means no message "quick_navigation_error_ephemeral": True, "first_element_button_enabled": True, "first_element_button_style": ButtonStyle.secondary, "first_element_button_label": "\U000025c0 \U000025c0", # None means no label "first_element_button_emoji": None, "prev_element_button_enabled": True, "prev_element_button_style": ButtonStyle.secondary, "prev_element_button_label": "\U000025c0", "prev_element_button_emoji": None, "next_element_button_enabled": True, "next_element_button_style": ButtonStyle.secondary, "next_element_button_label": "\U000025b6", "next_element_button_emoji": None, "last_element_button_enabled": True, "last_element_button_style": ButtonStyle.secondary, "last_element_button_label": "\U000025b6 \U000025b6", "last_element_button_emoji": None, "placeholder_button_enabled": True, "placeholder_button_style": ButtonStyle.secondary, "placeholder_button_label": "\U0001f6ab", "placeholder_button_emoji": None, } CONFIG = __CONFIG__.copy() # will count for all instances of your paginator @staticmethod def parse_config(config: Dict) -> Dict: if config is None: return Paginator.CONFIG _config = Paginator.CONFIG.copy() _config.update(config) return _config def __init__( self, client: Union[Bot, Client], user: User, config: Optional[Dict] = None, *args, **kwargs ): config = self.parse_config(config) self.config = config super().__init__(timeout=config["view_timeout"]) self.client = client self.user = user self.page = 0 self.first_elem_btn = view_buttons.FirstElement( config["first_element_button_style"], config["first_element_button_label"], config["first_element_button_emoji"], client=client, parent=self, user=user, using=config["first_element_button_enabled"] ) self.prev_elem_btn = view_buttons.PreviousElement( config["prev_element_button_style"], config["prev_element_button_label"], config["prev_element_button_emoji"], client=client, parent=self, user=user, using=config["prev_element_button_enabled"] ) self.next_elem_btn = view_buttons.NextElement( config["next_element_button_style"], config["next_element_button_label"], config["next_element_button_emoji"], client=client, parent=self, user=user, using=config["next_element_button_enabled"] ) self.last_elem_btn = view_buttons.LastElement( config["last_element_button_style"], config["last_element_button_label"], config["last_element_button_emoji"], client=client, parent=self, user=user, using=config["last_element_button_enabled"] ) self.start_btn = view_buttons.Start( config["start_button_style"], config["start_button_label"], config["start_button_emoji"], client=client, parent=self, user=user, using=True ) self.stop_btn = view_buttons.Stop( config["stop_button_style"], config["stop_button_label"], config["stop_button_emoji"], client=self.client, parent=self, user=user, using=True ) self.quick_nav_btn = view_buttons.QuickNav( config["quick_navigation_button_style"], config["quick_navigation_button_label"], config["quick_navigation_button_emoji"], client=self.client, parent=self, user=user, using=config["quick_navigation_button_enabled"] ) self.static_data: Optional[List[Dict[str, Any]]] = kwargs.get("static_data") self.static_data_page_count = len(self.static_data or []) or kwargs.get("static_page_count") or None @classmethod def from_list( cls, client: Union[Bot, Client], user: User, config: Optional[Dict] = None, data: List[Dict[str, Any]] = None ): return cls( client=client, user=user, config=config, static_data=data ) async def setup(self, *args, **kwargs): pass # noinspection PyArgumentList async def add_buttons(self): placeholder_config = ( self.config["placeholder_button_style"], self.config["placeholder_button_label"], self.config["placeholder_button_emoji"] ) self.add_item(self.first_elem_btn) self.add_item(self.prev_elem_btn) self.add_item(self.start_btn) self.add_item(self.next_elem_btn) self.add_item(self.last_elem_btn) if self.config["placeholder_button_enabled"]: self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(self.stop_btn) if self.config["placeholder_button_enabled"]: self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(view_buttons.Placeholder(*placeholder_config)) async def run(self, *args, **kwargs): await self.add_buttons() await self.setup(*args, **kwargs) return self async def paginator_start(self, interaction: Interaction): await self._paginator_start() await self.on_start(interaction) # noinspection PyArgumentList async def _paginator_start(self): self.clear_items() if self.config["first_element_button_enabled"]: self.first_elem_btn.disabled = False if self.config["prev_element_button_enabled"]: self.prev_elem_btn.disabled = False if self.config["next_element_button_enabled"]: self.next_elem_btn.disabled = False if self.config["last_element_button_enabled"]: self.last_elem_btn.disabled = False self.quick_nav_btn.disabled = self.config["quick_navigation_button_enabled"] placeholder_config = ( self.config["placeholder_button_style"], self.config["placeholder_button_label"], self.config["placeholder_button_emoji"] ) self.add_item(self.first_elem_btn) self.add_item(self.prev_elem_btn) self.add_item(self.quick_nav_btn) self.add_item(self.next_elem_btn) self.add_item(self.last_elem_btn) if self.config["placeholder_button_enabled"]: self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(self.stop_btn) if self.config["placeholder_button_enabled"]: self.add_item(view_buttons.Placeholder(*placeholder_config)) self.add_item(view_buttons.Placeholder(*placeholder_config)) async def paginator_stop(self, interaction: Interaction): if self.config["paginator_delete_when_finished"]: if interaction.message.flags.ephemeral: await interaction.message.delete( delay=self.config["paginator_delete_delay"] ) await self.on_stop(interaction) super().stop() async def on_start(self, interaction: Interaction): pass async def on_stop(self, interaction: Interaction): pass async def acquire_page_count(self, interaction: Interaction) -> int: if self.static_data_page_count is None: return await self.get_page_count(interaction) return self.static_data_page_count async def get_page_count(self, interaction: Interaction) -> int: raise NotImplementedError("get_page_count must be implemented!") async def update_page_number(self, interaction: Interaction, page: int): count = await self.acquire_page_count(interaction) self.page = (page % count) async def update_page_content(self, interaction: Interaction): await interaction.response.defer() contents = await self.acquire_page_content(interaction) ephemeral = self.config["paginator_ephemeral"] if ephemeral is not None: contents["ephemeral"] = ephemeral # noinspection PyTypeChecker ws: Webhook = interaction.followup await ws.edit_message( (await interaction.original_message()).id, **contents ) async def acquire_page_content(self, interaction: Interaction): if self.static_data is None: return await self.get_page_content(interaction, self.page) return self.static_data[self.page] async def get_page_content(self, interaction: Interaction, page: int) -> Dict[str, Any]: raise NotImplementedError("get_page_content must be implemented!")
philskillz-coder/discord-py-paginator
examples/default.py
<reponame>philskillz-coder/discord-py-paginator from discord import app_commands, Embed, Color, Interaction from discord.ext.paginator import paginator from typing import Dict, Any class GuildPaginator(paginator.Paginator): async def get_page_count(self, interaction: Interaction) -> int: return len(self.client.guilds) async def get_page_content(self, interaction: Interaction, page: int) -> Dict[str, Any]: # this method should return the arguments used for interaction.response.edit_message # e.g. {'content': 'hello'} means the message content will be edited to hello guild = self.client.guilds[page] # this cannot throw a index error because page is between 0 and the guild count return { "content": f"Guild {page+1}/{await self.get_page_count(interaction)}", "embed": ( Embed( title="Guild", colour=Color.green() ) .add_field(name="Name", value=guild.name) .add_field(name="ID", value=str(guild.id)) .add_field(name="Member count", value=str(guild.member_count), inline=False) ), "ephemeral": True } @app_commands.command( name="guilds", description="Show all the guilds" ) async def show_guilds(interaction: Interaction): await interaction.response.send_message( content="The bot guilds", view=await GuildPaginator( interaction.client, interaction.user ).run() )