repo_name
stringlengths
6
97
path
stringlengths
3
341
text
stringlengths
8
1.02M
bradshjg/nameko-cron
nameko_cron/__init__.py
import datetime from enum import Enum import time from logging import getLogger from croniter import croniter from eventlet import Timeout from eventlet.event import Event import pytz from nameko.extensions import Entrypoint _log = getLogger(__name__) class ConcurrencyPolicy(Enum): ALLOW = 'allow' SKIP = 'skip' WAIT = 'wait' class Cron(Entrypoint): def __init__(self, schedule: str, tz: str = None, concurrency: str = ConcurrencyPolicy.WAIT, **kwargs): """ Cron entrypoint. Fires according to a (possibly timezone-aware) cron schedule. If no timezone info is passed, the default is UTC. Set ``concurrency`` to ``ConcurrencyPolicy.ALLOW`` to allow multiple workers to run simultaneously. Set ``concurrency`` to ``ConcurrencyPolicy.SKIP`` to skip lapsed scheduled runs. The default behavior (``ConcurrencyPolicy.WAIT``) is to wait until the running worker completes and immediately spawn another if the schedule has lapsed. Example:: class Service(object): name = "service" @cron(schedule='0 12 * * *', tz='America/Chicago') def ping(self): # method executes every day at noon America/Chicago time print("pong") """ self.schedule = schedule self.tz = tz self.concurrency = concurrency self.should_stop = Event() self.worker_complete = Event() self.gt = None super().__init__(**kwargs) def start(self): _log.debug('starting %s', self) self.gt = self.container.spawn_managed_thread(self._run) def stop(self): _log.debug('stopping %s', self) self.should_stop.send(True) self.gt.wait() def kill(self): _log.debug('killing %s', self) self.gt.kill() def _get_next_interval(self): now_utc = datetime.datetime.now(tz=pytz.UTC) if self.tz: tz = pytz.timezone(self.tz) base = now_utc.astimezone(tz) else: base = now_utc cron_schedule = croniter(self.schedule, base) while True: yield max(cron_schedule.get_next() - time.time(), 0) def _run(self): """ Runs the schedule loop. """ interval = self._get_next_interval() sleep_time = next(interval) while True: # sleep for `sleep_time`, unless `should_stop` fires, in which # case we leave the while loop and stop entirely with Timeout(sleep_time, exception=False): self.should_stop.wait() break self.handle_timer_tick() if self.concurrency != ConcurrencyPolicy.ALLOW: self.worker_complete.wait() self.worker_complete.reset() sleep_time = next(interval) # a sleep time of zero represents that we've elapsed the next start time, so # if the user set the policy to skip, we need to update the interval again. if self.concurrency == ConcurrencyPolicy.SKIP and sleep_time == 0: sleep_time = next(interval) def handle_timer_tick(self): args = () kwargs = {} # Note that we don't catch ContainerBeingKilled here. If that's raised, # there is nothing for us to do anyway. The exception bubbles, and is # caught by :meth:`Container._handle_thread_exited`, though the # triggered `kill` is a no-op, since the container is already # `_being_killed`. self.container.spawn_worker( self, args, kwargs, handle_result=self.handle_result) def handle_result(self, worker_ctx, result, exc_info): # we only care about the worker completion if we're going to be waiting for it. if self.concurrency != ConcurrencyPolicy.ALLOW: self.worker_complete.send() return result, exc_info cron = Cron.decorator
joramwessels/torcs-client
torcs_tournament.py
#! /usr/bin/env python3.5 import logging import os import re import pwd import csv import time import shutil import psutil import pathlib import datetime import subprocess from collections import OrderedDict, abc import elo import yaml from bs4 import BeautifulSoup DROPBOX_DEBUG = logging.DEBUG - 1 logger = logging.getLogger(None if __name__ == '__main__' else __name__) def path_rel_to_dir(path, direcotry): if not os.path.isabs(path): path = os.path.join(direcotry, path) return path def really_running(proc): """Check whether a process is running _and_ isn't a zombie""" return proc.is_running() and proc.status() != psutil.STATUS_ZOMBIE class OrderedLoader(yaml.Loader): def construct_mapping(self, node, deep=False): # self.flatten_mapping(node) return OrderedDict(self.construct_pairs(node, deep)) OrderedLoader.add_constructor( yaml.resolver.BaseResolver.DEFAULT_MAPPING_TAG, OrderedLoader.construct_mapping ) class ParseError(Exception): pass class Player(object): """ Container for player information. Every argument of `start_command` will be formatted using `format(port=<value>)` `start_command` is issued with `working_dir` as working directory and `process_owner` as user. If `process_owner` is None, `token` will be used. The filenames `stdout` and `stderr` are relative to `output_dir`. """ def __init__(self, token, working_dir, rating=None, start_command=['./start.sh', '-p', '{port}'], output_dir='./output/', stdout='./{timestamp}-stdout.txt', stderr='./{timestamp}-stderr.txt', message_file='./current_rating.txt', rating_message="Your current rating is: {rating}", rank_message="You are ranked {rank} out of {total}", process_owner=None): self.token = token self.working_dir = working_dir if rating is not None: self.rating = elo.RATING_CLASS(rating) else: self.init_rating() self.start_command = start_command self.output_dir = path_rel_to_dir(output_dir, self.working_dir) self.stdout = path_rel_to_dir(stdout, self.output_dir) self.stderr = path_rel_to_dir(stderr, self.output_dir) self.message_file = path_rel_to_dir(message_file, self.output_dir) self.rating_message = rating_message self.rank_message = rank_message self.process_owner = process_owner \ if process_owner is not None \ else self.token if not os.path.exists(self.output_dir): os.mkdir(self.output_dir) def __str__(self): return self.__class__.__name__ + "({self.token!r}, " \ "{self.rating!r}" \ ")".format(self=self) def __repr__(self): return self.__class__.__name__ + "(" \ "{self.token!r}, " \ "{self.working_dir!r}, " \ "{self.rating!r}, " \ "{self.start_command!r}, " \ "{self.output_dir!r}, " \ "{self.stdout!r}, " \ "{self.stderr!r}, " \ "{self.message_file!r}, " \ "{self.rating_message!r}, " \ "{self.process_owner!r}" \ ")".format(self=self) def init_rating(self): self.rating = elo.RATING_CLASS(elo.INITIAL) class Rater(object): def __init__(self, players=(), filename=None, ignore_unknown_players=False): self.player_map = {} for player in players: self.add_player(player) self.filename = filename self.ignore_unknown_players = ignore_unknown_players if self.filename is not None and os.path.exists(self.filename): self.read_file() def add_player(self, player): """Add a player to this rater.""" if player.token in self.player_map: raise ValueError( "A token may only be specified once. Token: {}".format( player.token ) ) self.player_map[player.token] = player def filename_check(self, filename=None): if filename is None: if self.filename is None: raise ValueError( "Please specify a filename as argument or assign it to" " `self.filename`." ) else: filename = self.filename return filename def read_file(self, filename=None): filename = self.filename_check(filename) with open(filename) as fd: self.set_ratings(map(self.clean_line, csv.reader(fd))) def set_ratings(self, iterable): tokens = set() for line in iterable: token = line[0] if token in tokens: raise ValueError( "A token may only be specified once. Token: {}".format( token ) ) tokens.add(token) if len(line) > 2: raise ValueError( "No extra information next to a token and the desired " "rating should be specified: {}".format(line) ) if len(line) == 2: if token in self.player_map: self.player_map[token].rating = elo.RATING_CLASS(line[1]) elif not self.ignore_unknown_players: raise ValueError( "Rating specified for unknown player: {}".format(token) ) @staticmethod def clean_line(iterable): li = list(iterable) if len(li) != 1 and len(li) != 2: raise ValueError( "A ratings file should only contain lines with one or two " "values, got {}".format(li) ) if len(li) == 2: try: li[1] = elo.RATING_CLASS(li[1]) except ValueError as error: raise ValueError( "The second value of a rating line should be " "interpretable as {}. I received the following error " "while casting:\n\t{}".format( elo.RATING_CLASS.__name__, error ) ) return li def save_ratings(self, filename=None): """ Save the ratings of all players to a file. If a filename is specified, that file is used, otherwise `self.filename` is used. If neither is specified, a ValueError is raised. """ filename = self.filename_check(filename) logger.info("Saving ratings in {}".format(filename)) with open(filename, 'w') as fd: csv.writer(fd).writerows( sorted( ((p.token, p.rating) for p in self.player_map.values()), key=lambda p: p[1] ) ) @staticmethod def adjust_all(ranking): """ Adjust the ratings of given Players according to the ranked results. In a ranking every player won from all players before it. """ ranking = list(ranking) # Calculate new ratings new_ratings = [ elo.rate( player.rating, [ ((pi < oi), opponent.rating) for oi, opponent in enumerate(ranking) if opponent is not player ] ) for pi, player in enumerate(ranking) ] # Save new ratings for player, rating in zip(ranking, new_ratings): player.rating = rating def restart(self): for player in self.player_map.values(): player.init_rating() class Controller(object): def __init__(self, rater, queue, torcs_config_file, server_stdout='{timestamp}-server_out.txt', server_stderr='{timestamp}-server_err.txt', separate_player_uid=False, set_file_owner=False, set_file_mode=False, rater_backup_filename=None, result_filename_format="{driver} - {base}", timestamp_format='%Y-%m-%d-%H.%M', result_path='~/.torcs/results/', torcs_command=['torcs', '-r', '{config_file}'], driver_to_port=OrderedDict([ ('scr_server 1', 3001), ('scr_server 2', 3002), ('scr_server 3', 3003), ('scr_server 4', 3004), ('scr_server 5', 3005), ('scr_server 6', 3006), ('scr_server 7', 3007), ('scr_server 8', 3008), ('scr_server 9', 3009), ('scr_server 10', 3010), ]), raise_on_too_fast_completion=True, torcs_min_time=1, torcs_child_wait=0.5, shutdown_wait=1, crash_check_wait=0.2, file_mode=0o700): """ Orchestrate the races and save the ratings. When the rating is left out of the ratings file for a token, it is assigned the default rating, which will be saved to the same file when running `save_ratings`. N.B. `~` is only expanded to the user directory in `result_path` at initialisation of the controller. """ self.rater = rater self.queue = queue self.torcs_config_file = torcs_config_file self.server_stdout = server_stdout self.server_stderr = server_stderr self.separate_player_uid = separate_player_uid self.set_file_owner = set_file_owner self.set_file_mode = set_file_mode self.rater_backup_filename = rater_backup_filename self.result_filename_format = result_filename_format self.timestamp_format = timestamp_format self.result_path = os.path.expanduser(result_path) self.torcs_command = torcs_command self.driver_to_port = driver_to_port self.raise_on_too_fast_completion = raise_on_too_fast_completion self.torcs_min_time = torcs_min_time self.torcs_child_wait = torcs_child_wait self.shutdown_wait = shutdown_wait self.crash_check_wait = crash_check_wait self.file_mode = file_mode logger.debug("Result path: {}".format(self.result_path)) # Read drivers from config self.drivers = self.read_lineup(self.torcs_config_file) def timestamp(self): return datetime.datetime.now().strftime(self.timestamp_format) @staticmethod def rank_text(rank): if rank == 0: return '1st' elif rank == 1: return '2nd' elif rank == 2: return '3rd' else: return str(rank + 1) + 'th' @staticmethod def read_ranking(results_file): """ Return a ranked list of driver names read from the given results file. NB. Driver names are _not_ tokens. One should first look up which token corresponds with which driver name. """ with open(results_file) as fd: soup = BeautifulSoup(fd, 'xml') result_soup = soup.find('section', attrs={'name': 'Results'}) rank_soup = result_soup.find('section', attrs={'name': 'Rank'}) ranks = [ ( int(section['name']), section.find('attstr', attrs={'name': 'name'})['val'] ) for section in rank_soup.findAll('section') ] return list(zip(*sorted(ranks)))[1] @staticmethod def read_lineup(torcs_config_file): with open(torcs_config_file) as fd: soup = BeautifulSoup(fd, 'xml') drivers_sec = soup.find('section', attrs={'name': 'Drivers'}) drivers = [] for sec in drivers_sec.findAll('section'): tag, attrs = 'attstr', {'name': 'module'} module = sec.find(tag, attrs=attrs) if module is None: raise ParseError( "Error parsing {file}: expected a {tag} tag with the " "following attributes: {attrs!r}".format( file=torcs_config_file, tag=tag, attrs=attrs ) ) expected = 'scr_server' if module.get('val', Exception()) != expected: raise ParseError( "Error parsing {file}: all drivers are expected to be the " "'{expected}' module.".format( file=torcs_config_file, expected=expected ) ) tag, attrs = 'attnum', {'name': 'idx'} idx = sec.find(tag, attrs=attrs) if idx is None: raise ParseError( "Error parsing {file}: expected a {tag} tag with the " "following attributes: {attrs!r}".format( file=torcs_config_file, tag=tag, attrs=attrs ) ) val = idx.get('val', None) if val is None: raise ParseError( "Error parsing {file}: expected {tag} to have the " "attribute {attr}.".format( file=torcs_config_file, tag=tag, attr='val', ) ) drivers.append((sec['name'], val)) # I now have a list of (rank, id) pairs # Somehow, the number in the name of the scr_server driver is one # larger that the `idx` of the driver. return [ 'scr_server {}'.format(int(idx) + 1) for _, idx in sorted(drivers) ] def restart(self): """Restart the tournament, making all ratings equal.""" self.rater.restart() def race_and_save(self, simulate=False): """ Run a race (see `Controller.race`) and save the ratings. """ self.race(simulate=simulate) self.rater.save_ratings() def race(self, simulate=False): """ Run a race Automatically determine the number of players to be raced and ask the queue which players are next. Race the players, save the results and update the queue. """ players = self.queue.first_n(len(self.drivers)) logger.info("Racing: {}".format(', '.join( repr(player.token) for player in players ))) self.race_once(players, simulate=simulate) self.queue.requeue(players) def race_tokens(self, tokens, simulate=False): return self.race_once( map(self.rater.player_map.get, tokens), simulate=simulate ) def race_once(self, players, simulate=False): """ Run one race with TORCS and the given players. Also make a backup of the ratings if `self.rater_backup_filename` is not None. NB. Please make sure the number of players given matches the specified number of players in the configuration file of this Controller. The output can be found under: <torcs installation directory>/results """ players = list(players) if len(self.drivers) != len(players): raise ValueError( "{nplay} players where given, but {file} specifies {ndriv} " "drivers".format( nplay=len(players), ndriv=len(self.drivers), file=self.torcs_config_file ) ) driver_to_player = OrderedDict(zip(self.drivers, players)) open_files = [] processes = [] try: # Start server server_stdout = open( self.server_stdout.format(timestamp=self.timestamp()), 'w' ) open_files.append(server_stdout) server_stderr = open( self.server_stderr.format(timestamp=self.timestamp()), 'w' ) open_files.append(server_stderr) logger.info("Starting TORCS...") if simulate: logger.warning( "This is a simulation! No child processes are started." ) else: logger.debug( "TORCS config to use: {}".format(self.torcs_config_file) ) config_file = os.path.abspath(self.torcs_config_file) logger.debug("TORCS config to use: {}".format(config_file)) command = list(map( lambda s: s.format( config_file=config_file ), self.torcs_command )) logger.debug("TORCS command to be run: {}".format(command)) server_process = psutil.Popen( command, stdout=server_stdout, stderr=server_stderr, ) processes.append(server_process) # TORCS starts a child process, which doesn't terminate # automatically if `server_process` is terminated or crashes. time.sleep(self.torcs_child_wait) children = server_process.children() logger.debug("TORCS server children: {}".format(children)) processes.extend(children) # Start players logger.info("Starting players...") for driver, player in driver_to_player.items(): stdout = open( player.stdout.format(timestamp=self.timestamp()), 'w' ) open_files.append(stdout) stderr = open( player.stderr.format(timestamp=self.timestamp()), 'w' ) open_files.append(stderr) # Set the ownership of the files if self.set_file_owner: self.change_owner(player) if self.set_file_mode: self.change_mode(player) if simulate: # Always simulate these functions, just to be sure they # work self.get_change_user_fn(player) self.get_player_env(player) elif self.separate_player_uid: processes.append(psutil.Popen( map( lambda s: s.format( port=self.driver_to_port[driver] ), player.start_command ), stdout=stdout, stderr=stderr, preexec_fn=self.get_change_user_fn(player), cwd=player.working_dir, env=self.get_player_env(player) )) else: processes.append(psutil.Popen( map( lambda s: s.format( port=self.driver_to_port[driver] ), player.start_command ), stdout=stdout, stderr=stderr, cwd=player.working_dir, )) logger.debug("Started {}".format(player)) time.sleep(self.crash_check_wait) # Check no one crashed in the mean time for proc in processes: if not really_running(proc): name = proc.name() if hasattr(proc, 'name') else proc raise subprocess.CalledProcessError( proc.poll() if hasattr(proc, 'poll') else 0, list(proc.args) or name ) # Wait for server logger.info("Waiting for TORCS to finish...") start_time = time.time() if not simulate: server_process.wait() end_time = time.time() # Time TORCS ran in seconds diff_time = end_time - start_time if not simulate and diff_time < self.torcs_min_time: logger.warning( "TORCS only ran for {:.2f} seconds".format(diff_time) ) if self.raise_on_too_fast_completion: raise subprocess.SubprocessError( "TORCS only took {:.2f} seconds to complete".format( diff_time ) ) logger.debug("Finished!") # Check exit status of TORCS # However, even if something goes wrong, the exit status is 0, # so I can't know if something went wrong. # logger.debug("really_running(server_process): {}".format( # really_running(server_process) # )) # logger.debug("server_process.returncode: {}".format( # server_process.returncode # )) # if server_process.returncode: # raise subprocess.CalledProcessError( # proc.returncode, # proc.args # ) except: logger.error("An error occurred, trying to stop gracefully...") raise finally: # Exit running processes if not simulate: # Wait a second to give the processes some time time.sleep(self.shutdown_wait) # First be nice for proc in processes: if really_running(proc): logger.info("Terminating {}".format(proc)) proc.terminate() # Wait a second to give the processes some time time.sleep(self.shutdown_wait) # Time's up for proc in processes: if really_running(proc): logger.warning("Killing {}".format(proc)) proc.kill() # Wait a second to give the processes some time time.sleep(self.shutdown_wait) # Double check for proc in processes: if really_running(proc): logger.error( "The following process could not be killed: {}" .format(proc.cmdline()) ) # Close all open files for fd in open_files: logger.debug("Closing {}".format(fd.name)) try: fd.close() except Exception as e: logger.error(e) logger.info("Closed all files and processes!") # Give the players the server output for player in players: shutil.copyfile( server_stdout.name, os.path.join( player.output_dir, os.path.basename(server_stdout.name) ) ) shutil.copyfile( server_stderr.name, os.path.join( player.output_dir, os.path.basename(server_stderr.name) ) ) # End of `finally` clause # Find the correct results file logger.debug("Result path: {}".format(self.result_path)) out_dir = os.path.join( self.result_path, # remove head path and extension '.'.join(os.path.split(self.torcs_config_file)[1].split('.')[:-1]) ) out_base = sorted(os.listdir(out_dir))[-1] out_file = os.path.join( out_dir, out_base ) # Give the players the results file for driver, player in driver_to_player.items(): shutil.copyfile( out_file, os.path.join( player.output_dir, self.result_filename_format.format( driver=driver, base=out_base ) ) ) # Update ratings according to ranking ranked_drivers = self.read_ranking(out_file) self.rater.adjust_all(map(driver_to_player.get, ranked_drivers)) # Make a backup if self.rater_backup_filename is given if self.rater_backup_filename is not None: backup_filename = self.rater_backup_filename.format( timestamp=self.timestamp() ) # logger.info("Backing up ratings in {}".format(backup_filename)) self.rater.save_ratings( backup_filename ) # Tell players their own rating and rank sorted_players = sorted( self.rater.player_map.values(), key=lambda p: p.rating, reverse=True ) total = len(sorted_players) for rank, player in enumerate(sorted_players): with open(player.message_file, 'w') as fd: fd.write(player.rating_message.format(rating=player.rating)) fd.write('\n') fd.write( player.rank_message.format( rank=self.rank_text(rank), total=total ) ) fd.write('\n') def change_owner(self, player): """ Make `player.process_owner` the owner of all files in `player.working_dir` """ pw_record = pwd.getpwnam(player.process_owner) logger.debug( "Changing file ownership for {}".format(player.token) ) for dirpath, _, filenames in os.walk(player.working_dir): # Change directory ownership os.chown(dirpath, pw_record.pw_uid, pw_record.pw_gid) # Change file ownership for filename in filenames: os.chown( os.path.join(dirpath, filename), pw_record.pw_uid, pw_record.pw_gid ) def change_mode(self, player, mode=None): """ Make `player.process_owner` the owner of all files in `player.working_dir` """ if mode is None: mode = self.file_mode logger.debug( "Changing file mode for {}".format(player.token) ) for dirpath, _, filenames in os.walk(player.working_dir): # Change directory mode os.chmod(dirpath, mode) # Change file mode for filename in filenames: os.chmod(os.path.join(dirpath, filename), mode) @staticmethod def get_change_user_fn(player): pw_record = pwd.getpwnam(player.process_owner) def change_user(): logger.debug( "Starting demotion. UID: {uid}, GID: {gid}".format( uid=os.getuid(), gid=os.getgid() ) ) try: logger.debug("Trying to set gid...") os.setgid(pw_record.pw_gid) logger.debug("Trying to set uid...") os.setuid(pw_record.pw_uid) except Exception as e: logger.error(e) raise logger.debug( "Finished demotion. UID: {uid}, GID: {gid}".format( uid=os.getuid(), gid=os.getgid() ) ) return change_user @staticmethod def get_player_env(player): # Info from https://stackoverflow.com/questions/1770209/run-child-processes-as-different-user-from-a-long-running-process/6037494#6037494 # NOQA pw_record = pwd.getpwnam(player.process_owner) env = os.environ.copy() env['LOGNAME'] = env['USER'] = pw_record.pw_name env['HOME'] = pw_record.pw_dir logger.debug("ENV PWD: {}".format(env.get('PWD', None))) env['PWD'] = player.working_dir logger.debug("Set PWD to: {!r}".format(env['PWD'])) logger.debug("PATH: {}".format(env['PATH'])) return env @classmethod def load_config(cls, config_file, extra_config={}): """ Load a controller from the given config file. NB. Only the first layer of `extra_config` is merged, everything else is overwritten, e.g.: original_config = { 'test': { 'test-one': 'hello' 'test-two': { 'test-two-one': 'bye' } } } extra_config = { 'test': { 'test-two': { 'test-two-two': 'override' } 'test-three': 'added' } } results in: config = { 'test': { 'test-one': 'hello' 'test-two': { 'test-two-two': 'override' } 'test-three': 'added' } } """ error_regex = re.compile( r"__init__\(\) got an unexpected keyword argument '(\w+)'" ) with open(config_file) as fd: config = yaml.load(fd, OrderedLoader) for key, value in extra_config.items(): if isinstance(value, abc.Mapping): cur_conf = config.setdefault(key, {}) cur_conf.update(value) else: config[key] = value logger.debug("Config: {}".format(config)) try: rater = cls.load_rater(config) fbq = cls.load_fbq(config, rater.player_map.values()) controller = cls(rater, fbq, **config.get('controller', {})) except TypeError as e: match = error_regex.fullmatch(e.args[0]) if match is not None: config_key = match.groups()[0] logger.debug("Match: {}".format(config_key)) raise ValueError( "Unexpected configuration key in {filename}: {key!r}" .format(filename=config_file, key=config_key) ) from e else: logger.debug("No match...") raise return controller @classmethod def load_rater(cls, config_dic): players = cls.load_players(config_dic) rater = Rater(players, **config_dic.get('rater', {})) return rater @staticmethod def load_players(config_dic): key = 'players' if key not in config_dic: raise ValueError( "No players specified! Expected a {!r} key in the" " configuration file.".format(key) ) players = config_dic[key] if not isinstance(players, abc.Mapping): # If it's not a mapping, I'm assuming I can open it. the_exception = TypeError( "Expected {key!r} to point to a {{token: config}} mapping or" " a path to a .yml file containing a {{token: config}}" " mapping. Instead I found: {players!r}".format( key=key, players=players ) ) fd = None try: fd = open(players) except Exception as e: raise the_exception from e else: players = yaml.load(fd, OrderedLoader) finally: if fd is not None: fd.close() # logger.debug("Closed players config!") if not isinstance(players, abc.Mapping): raise the_exception return [ Player(token, **player_conf) for token, player_conf in players.items() ] @staticmethod def load_fbq(config_dic, players=()): return FileBasedQueue(players, **config_dic.get('queue', {})) class DropboxDisablingController(Controller): def __init__(self, *args, dropbox_start_command=['dropbox', 'start'], dropbox_stop_command=['dropbox', 'stop'], start_dropbox=False, stop_dropbox=False, **kwargs): self.dropbox_start_command = dropbox_start_command self.dropbox_stop_command = dropbox_stop_command self.start_dropbox = start_dropbox self.stop_dropbox = stop_dropbox super(DropboxDisablingController, self).__init__(*args, **kwargs) def race_once(self, *args, **kwargs): """Disable Dropbox before racing and start it again afterwards.""" try: # Try to disable Dropbox # The catch is that the return status of the Dropbox control script # is always 0, even if something went wrong... if self.stop_dropbox: logger.info("Stopping Dropbox...") completed = subprocess.run( self.dropbox_stop_command, stdout=subprocess.PIPE, stderr=subprocess.PIPE ) logger.info("Dropbox says:\n{}".format( completed.stdout.decode() )) del completed # Race return super(DropboxDisablingController, self).race_once( *args, **kwargs ) finally: if self.start_dropbox: # Enable Dropbox logger.info("Starting Dropbox...") # Somehow stderr captures the output of the started Dropbox # daemon. However, capturing it isn't an option because the # daemon doesn't stop, which means the stream will hang. Thus # if you want to see its output, we'll just leave it as is, # otherwise we'll squelch the daemon's output. stderr = None if logger.getEffectiveLevel() <= DROPBOX_DEBUG \ else subprocess.DEVNULL completed = subprocess.run( self.dropbox_start_command, stdout=subprocess.PIPE, stderr=stderr ) logger.info("Dropbox says:\n{}".format(completed.stdout.decode())) del completed class FileBasedQueue(object): """ Queue players according to the last modified time of a specifically named file in their `working_dir`. """ def __init__(self, players, filename='start.sh'): self.filename = filename self.players = list(players) @staticmethod def touch(filename): """ Touch a file. I.E. create it if it does not exist or change the last modified time to the current time if it does. """ logger.debug("Touching: {}".format(filename)) pathlib.Path(filename).touch() logger.debug("Touched!") @staticmethod def get_last_modified(filename): modified_time = os.path.getmtime(filename) logger.debug("Filename: {}".format(filename)) logger.debug("Modified time: {}".format(modified_time)) return modified_time def get_filename(self, player): """Get the full path to the queue file of a player""" return os.path.join( player.working_dir, self.filename ) def first_n(self, n): """ Get the `n` players that are first in line """ return sorted( self.players, key=lambda p: self.get_last_modified(self.get_filename(p)), # reverse=True, )[:n] def requeue(self, players): """ Put the given players at the end of the queue In this case this is done by touching their respective queue files in the order the players are passed. """ for player in players: self.touch(self.get_filename(player)) def log_level_type(string): try: value = int(string) except: value = string return value if __name__ == '__main__': # Parse command line arguments from argparse import ArgumentParser parser = ArgumentParser() parser.add_argument('config_file', help="YAML file with configuration") parser.add_argument('-l', '--level', default='INFO', type=log_level_type, help="Logging level to use") parser.add_argument( '-s', '--simulate', action='store_true', help="Attempts to mimic a full run without starting child processes." " May fail if no old TORCS output files are present in the expected" " directory.") parser.add_argument( '--start-dropbox', action='store_true', help="Start Dropbox again after the race." ) parser.add_argument( '--stop-dropbox', action='store_true', help="Stop Dropbox before the race. Implies --start-dropbox." ) args = parser.parse_args() # Initialise logging logging.basicConfig(level=args.level) extra_config = {} if args.stop_dropbox: control_config = extra_config.setdefault('controller', {}) control_config['stop_dropbox'] = True if args.start_dropbox or args.stop_dropbox: control_config = extra_config.setdefault('controller', {}) control_config['start_dropbox'] = True # Race # controller = Controller.load_config(args.config_file) controller = DropboxDisablingController.load_config( args.config_file, extra_config ) controller.race_and_save(simulate=args.simulate) logger.info("Done!")
joramwessels/torcs-client
driver_utils.py
<reponame>joramwessels/torcs-client # file: driver_utils.py # author: <NAME> # date: 11-12-2017 # dependencies: numpy # description # usage # from sys import stderr from numpy import sign # as stated by Torcs documentation MAX_WHEEL_ROTATION = 21 # confusing polarities STR_R, STR_L = -1, 1 ANG_R, ANG_L = 1, -1 DFC_R, DFC_L = -1, 1 # printing ENABLE_DEBUG_INFO = True PRINT_CYCLE_INTERVAL = 50 # freqency of print output in game cycles PRINT_STATE = True PRINT_COMMAND = False def to_ang(ang): """ Steers towards the road angle Args: ang: The angle of the car with the road Returns: The angle to steer in """ if sign(ang) == ANG_R: return STR_L elif sign(ang) == ANG_L: return STR_R else: return 0 def away_from_ang(ang): """ Steers away from the road angle Args: ang: The angle of the car with the road Returns: The angle to steer in """ return -to_ang(ang) def debug(iter, *args): """ prints debug info to stderr """ if iter % PRINT_CYCLE_INTERVAL == 0: err(iter, *args) def err(iter, *args): if ENABLE_DEBUG_INFO: spc = 6-len(str(iter)) ovwr = 50 - len(' '.join([str(x) for x in args])) print(iter, ' '*spc, *args, ' '*ovwr, file=stderr)
joramwessels/torcs-client
mlp_driver.py
from pytocl.driver import Driver from pytocl.car import State, Command import mlp class MyDriver(Driver): def __init__(self, model_file="mlp_100x100.pt"): mlp.use_cuda = False self.model = mlp.load_model(model_file) self.it = 0 def drive(self, carstate): self.it += 1 x = [carstate.angle, carstate.speed_x, carstate.speed_y, carstate.speed_z] + \ list(carstate.distances_from_edge) + \ [carstate.distance_from_center] pred_y = list(self.model.predict(x).data)[0] command = Command() command.accelerator = pred_y[0] command.brake = pred_y[1] command.steering = pred_y[2] gear_flt = pred_y[3] if self.it > 750 else self.it/250.0 command.gear = min(5, max(1, int(gear_flt + 0.5))) print(self.it,"acc: %.2f, brk: %.2f, ste: %.2f, gea: %.2f" %(command.accelerator, command.brake, command.steering, gear_flt), end='\r') return command
joramwessels/torcs-client
basic_control.py
<gh_stars>1-10 # file: basic_control.py # author: <NAME> # date: # dependencies: # description: # usage: # from math import radians import numpy as np import torch import ffnn_speed import ffnn_steer ANGLES = [90, 75, 60, 45, 30, 20, 15, 10, 5, 0, -5, -10, -15, -20, -30, -45, -60, -75, -90] class BasicControl: def __init__(self, steering_values): """ Short description Complete multiline description of class Args: steering_values: ??? """ self.brake_row = 0 self.speed = ffnn_speed.Speed(10) self.steer = ffnn_steer.Steer(10) self.steer.load_state_dict(torch.load("./steer.data")) self.speed.load_state_dict(torch.load("./ffnn_speed.data")) self.steering_values = steering_values self.alphas = [radians(x) for x in ANGLES] def deal_with_opponents(self, steer_pred, pedal, speed_x, distance_from_center, opponents_new, opponents_delta): """ Description Args: steer_pred: ? pedal: ? speed_x: ? distance_from_center: ? opponents_new: ? opponents_delta: ? Returns: ??? """ # index 18 is in front # index 35 in behind us adjustment = 0.1 # if there are cars infront-left -> move to right if opponents_new[17] < 10 or opponents_new[16] < 10 or opponents_new[15] < 10: print("ADJUSTING SO NOT TO HIT") steer_pred -= adjustment if opponents_new[19] < 10 or opponents_new[20] < 10 or opponents_new[21] < 10: print("ADJUSTING SO NOT TO HIT") # move to left steer_pred += adjustment if opponents_new[18] < 50: # we are on left side -> move right if distance_from_center > 0: steer_pred -= adjustment # o.w. move left else: steer_pred += adjustment if speed_x > 100: # we are getting closer to the car in front (and we can't avoid it). We need to slow down a bit if (opponents_delta[18] < 0 and opponents_new[18] < 20) or (opponents_delta[17] < 0 and opponents_new[17] < 4) or (opponents_delta[19] < 0 and opponents_new[19] < 4): pedal -= 0.1 return steer_pred, pedal def steer_decider(self, carstate): """ Description Args: carstate: The full carstate Returns: Steering angle? """ alpha_index = np.argmax(carstate.distances_from_edge) if is_straight_line(carstate=carstate, radians=self.alphas[alpha_index], factor=self.steering_values[4]): return carstate.angle * 0.5 steering_function = lambda index, offset:\ (self.alphas[index-offset] * carstate.distances_from_edge[index-offset] \ + self.alphas[index+offset] * carstate.distances_from_edge[index+offset]) \ / (carstate.distances_from_edge[index+offset] \ + carstate.distances_from_edge[index-offset]) steer = self.steering_values[0] * self.alphas[alpha_index] for x in range(1, 4): if alpha_index - x > -1 and alpha_index + x < len(self.steering_values): steer += self.steering_values[x]*steering_function(alpha_index, x) return steer def speed_decider(self, carstate, max_speed=120): """ Description Args: carstate: The full carstate max_speed: ??? Returns: ??? """ # we predict speed and map that to pedal x_in = ffnn_speed.carstate_to_variable(carstate) target_speed = self.speed(x_in).data[0] # we limit the speed if target_speed >= max_speed: target_speed = max_speed pedal = 2/(1 + np.exp(carstate.speed_x - target_speed))-1 return pedal def gear_decider(self, carstate): """ Description Args: carstate: The full carstate Returns: The gear to shift to (int) """ gear = carstate.gear rpm = carstate.rpm # we do gears by hand # up if {9500 9500 9500 9500 9000} # down if {4000 6300 7000 7300 7300} if gear == -1: return 1 elif gear == 0: if rpm >= 5000: gear = 1 elif gear == 1: if rpm >= 9500: gear = 2 elif gear == 2: if rpm >= 9500: gear = 3 elif rpm <= 4000: gear = 2 elif gear == 3: if rpm >= 9500: gear = 4 elif rpm <= 6300: gear = 3 elif gear == 4: if rpm >= 9500: gear = 5 elif rpm <= 7000: gear = 3 elif gear == 5: if rpm >= 9000: gear = 6 elif rpm <= 7300: gear = 4 elif gear == 6: if rpm <= 7300: gear = 5 return gear def disambiguate_pedal(self, pedal, accel_cap=0.5, break_cap=0.75, break_max_length=5): """ Description Args: ??? Returns: The break and accelerator command values """ if pedal >= 0.0: accelerator = pedal*accel_cap brake = 0 else: # we need to make sure that we don't break hard enough and not too long self.brake_row += 1 if self.brake_row <= break_max_length: brake = abs(pedal)*break_cap else: self.brake_row = 0 brake = 0 accelerator = 0 return brake, accelerator def is_straight_line(carstate, radians, factor): """ Decides whether ??? is a straight line Args: carstate: The full carstate radians: ??? factor: ??? Returns: A boolean indicating whether ??? """ if abs(carstate.distance_from_center) < 0.75: if radians == 0: return True if carstate.distances_from_edge[9] > 190: return True if carstate.distances_from_edge[9] > factor * carstate.speed_x: return True return False
joramwessels/torcs-client
combined_driver.py
# file: crisis_driver.py # authors: Haukur, <NAME> # date: 07-12-2017 # dependencies: pytocl, basic_control, swarm, crisis_driver # description # usage # from pytocl.driver import Driver from pytocl.car import State, Command from sys import stderr from math import radians from operator import sub from driver_utils import * from basic_control import BasicControl from swarm import FeromoneTrail from crisis_driver import CrisisDriver from mlp import load_model # - NOTE crash detection for swarm only checks for off road # - NOTE collision detection for swarm might be too sensitive # - TODO clear up swarm global parameters # - TODO swarm debug output is piped to stderr ENABLE_SWARM = True ENABLE_CRISIS_DRIVER = True ENABLE_NETWORK = False # Neural network parameters STR_MODELS = ["steering_model_1.pt", "steering_model_2.pt", "steering_model_2b.pt", "steering_model_3.pt", "steering_model_3b.pt"] MODEL_FILENAME = STR_MODELS[4] # swarm metaparameters swarm_pos_int = 50 swarm_spd_int = 20 swarm_spd0 = 0 swarm_spd_n = 20 swarm_expl_int = 40 class Final_Driver(Driver): def __init__(self, steering_values, global_max_speed): """ Short description Multiline description on details and usage Args: steering_values: ??? global_max_speed: ??? """ super(Final_Driver, self).__init__() self.iter = 0 self.basic_control = BasicControl(steering_values) self.back_up_driver = CrisisDriver(logdata=False) self.bad_counter = 0 self.lap_counter = 0 self.last_opponents = [0 for x in range(36)] self.global_max_speed = global_max_speed self.max_speed = global_max_speed self.cummulative_time = 0 if ENABLE_SWARM: self.swarm = FeromoneTrail( swarm_pos_int, swarm_spd_int, swarm_spd0, swarm_spd_n, swarm_expl_int, self.global_max_speed) self.crashed_in_last_frame = False self.contact_in_last_frame = False self.previous_frame_position = 0 if ENABLE_NETWORK: self.steering_model = load_model(MODEL_FILENAME) def drive(self, carstate: State) -> Command: """ Description Args: carstate: All parameters packed in a State object Returns: command: The next move packed in a Command object """ self.iter += 1 self.back_up_driver.update_status(carstate) # trackers self.update_trackers(carstate) if PRINT_STATE:# and (self.iter % PRINT_CYCLE_INTERVAL) == 0: self.print_trackers(carstate, r=True) # crash and collision detection for swarm if ENABLE_SWARM: if self.back_up_driver.needs_help or self.back_up_driver.is_off_road: self.crashed_in_last_frame = True if not self.crashed_in_last_frame: debug(self.iter, "SWARM: crashed") for dist in carstate.opponents: if dist == 0: self.contact_in_last_frame = True # crisis handling if ENABLE_CRISIS_DRIVER: if self.back_up_driver.is_in_control: return self.back_up_driver.drive(carstate) elif self.back_up_driver.needs_help: self.back_up_driver.pass_control(carstate) return self.back_up_driver.drive(carstate) # since the data and python's values differ we need to adjust them try: carstate.angle = radians(carstate.angle) carstate.speed_x = carstate.speed_x*3.6 command = self.make_next_command(carstate) except Exception as e: err(self.iter, str(e)) command = self.back_up_driver.driver.drive(carstate) return command def make_next_command(self, carstate): """ Description Args: carstate: The full carstate object as passed to Driver() Returns: command: The command object to pass back to the server """ # checking in on the swarm position = carstate.distance_from_start position = int(position - (position % self.swarm.pos_int)) new_frame = position > (self.previous_frame_position + self.swarm.pos_int) new_lap = self.previous_frame_position > (position + self.swarm.pos_int) if ENABLE_SWARM and (new_frame or new_lap): self.max_speed = self.swarm.check_in( position, carstate.speed_x, self.crashed_in_last_frame, self.contact_in_last_frame) self.crashed_in_last_frame = False self.contact_in_last_frame = False self.previous_frame_position = position err(self.iter, "SWARM: pos=%i, max_speed=%i" %(position, self.max_speed)) # basic predictions if ENABLE_NETWORK: steer_pred = self.steering_model.predict([carstate.angle, carstate.speed_x] + list(carstate.distances_from_edge) + [carstate.distance_from_center]) steer_pred = steer_pred[0] else: steer_pred = self.basic_control.steer_decider(carstate) gear = self.basic_control.gear_decider(carstate) pedal = self.basic_control.speed_decider(carstate, max_speed=self.max_speed) # making sure we don't drive at people opponents_deltas = list(map(sub, carstate.opponents, self.last_opponents)) steer_pred, pedal = self.basic_control.deal_with_opponents(steer_pred, pedal, carstate.speed_x, carstate.distance_from_center, carstate.opponents, opponents_deltas) # if too fast descelerate to max speed if carstate.speed_x > self.max_speed: pedal = 0.0 err(self.iter, "MAIN: capping speed") # disambiguating pedal with smoothing brake, accel = self.basic_control.disambiguate_pedal(pedal, accel_cap=1.0) # debug output if PRINT_COMMAND and self.iter % PRINT_CYCLE_INTERVAL: print("Executing comand: gear=%.2f, acc=%.2f," %(gear, accel), "break=%.2f, steering=%.2f" %(brake, steer_pred)) # command construction command = Command() command.brake = brake command.accelerator = accel command.steering = steer_pred command.gear = gear if command.steering > 0.10: debug(self.iter, "BASIC: turning left") elif command.steering < -0.10: debug(self.iter, "BASIC: turning right") return command def update_trackers(self, carstate): """ Updates info about the race """ self.iter += 1 if abs(carstate.current_lap_time) < 0.020: self.lap_counter += 1 self.cummulative_time += carstate.last_lap_time + self.cummulative_time def print_trackers(self, carstate, r=False): """ Prints info on the race """ line_end = '\r' if r else '\n' print(" Lap=%i CurLapTime=%.2f dist=%.2f time=%.2f" %(self.lap_counter, carstate.current_lap_time, carstate.distance_raced, self.cummulative_time + carstate.current_lap_time) , end=line_end)
joramwessels/torcs-client
ffnn_steer.py
<reponame>joramwessels/torcs-client<filename>ffnn_steer.py import torch from torch.autograd import Variable import torch.nn as nn import torch.optim as optim import torch.nn.functional as F import time import sys import numpy as np import argparse import driver_support import math from os import listdir from os.path import isfile, join class Steer(nn.Module): def __init__(self, hidden_dimension): super(Steer, self).__init__() n_states = 22 n_actions = 1 self.layer_1 = nn.Linear(n_states, hidden_dimension) self.non_lin = nn.Sigmoid() self.layer_2 = nn.Linear(hidden_dimension, n_actions) def forward(self, inputs): out = self.layer_1(inputs) out = self.non_lin(out) out = self.layer_2(out) return out def carstate_to_variable(carstate): # y=speed, x=angle, distance*19, distToMiddle return Variable(torch.FloatTensor([math.radians(carstate.angle), carstate.speed_x] + list(carstate.distances_from_edge) + [carstate.distance_from_center]), requires_grad=True) def create_model(out_file, training_folder, learning_rate, epochs, hidden_dimension): # Read in the data training = [] for file_in in [join(training_folder, f) for f in listdir(training_folder) if isfile(join(training_folder, f))]: training += list(driver_support.read_lliaw_dataset_steer_angle_speed_dist_middle(file_in)) model = Steer(hidden_dimension) print(model) optimizer = optim.SGD(model.parameters(), lr=learning_rate) loss = nn.MSELoss() for ITER in range(epochs): train_loss = 0.0 start = time.time() for y_true, state in training: # forward pass optimizer.zero_grad() in_state = Variable(torch.FloatTensor(state)) y_pred = model(in_state) y_true = Variable(torch.FloatTensor(y_true)) output = loss(y_pred, y_true) train_loss += output.data[0] # backward pass output.backward() # update weights optimizer.step() print("last prediction made:", y_pred, y_true) print("iter %r: train loss/action=%.4f, time=%.2fs" %(ITER, train_loss/len(training), time.time()-start)) torch.save(model.state_dict(), out_file) def main(): parser = argparse.ArgumentParser() parser.add_argument('--epochs', type=int) parser.add_argument('--hidden', type=int) parser.add_argument('--learn', type=float) parser.add_argument('--in_file', type=str) parser.add_argument('--out_file', type=str) args = parser.parse_args() create_model(args.out_file, args.in_file, args.learn, args.epochs, args.hidden) if __name__ == "__main__": main()
joramwessels/torcs-client
ffnn_gears.py
<filename>ffnn_gears.py import torch from torch.autograd import Variable import torch.nn as nn import torch.optim as optim import torch.nn.functional as F import time import sys import numpy as np import argparse import driver_support from os import listdir from os.path import isfile, join from collections import defaultdict gear_number = 9 # 0 = reverse, 1 = neutral, 2=1st gear, etc. to 7th gear class Gear_switcher(nn.Module): def __init__(self, hidden_dimension): super(Gear_switcher, self).__init__() n_states = 1 + gear_number n_actions = gear_number self.layer_1 = nn.Linear(n_states, hidden_dimension) self.layer_2 = nn.Linear(hidden_dimension, n_actions) def forward(self, inputs): out = self.layer_1(inputs) out = nn.functional.relu(out) out = self.layer_2(out) return out def gear_to_tensor(gear_value): gear_value += 1 return torch.LongTensor([gear_value]) def to_tensor(carstate): #gear, rpm return torch.FloatTensor(driver_support.binerize_input(value=carstate.gear, mapping=get_gear_map()) + [carstate.rpm]) def prediction_to_action(prediction): # the index is the gear index = prediction.data.numpy().argmax() index -= 1 return index def get_gear_map(): gear_to_index_map = dict() for x in range(-1, gear_number - 1): gear_to_index_map[str(x)] = x + 1 return gear_to_index_map def evaluate(model, data): """Evaluate a model on a data set.""" correct = 0.0 for y_true, state in data: y_true = int(y_true[0]) lookup_tensor = Variable(torch.FloatTensor(state)) scores = model(lookup_tensor) action = prediction_to_action(scores) if action == y_true: correct += 1 print("percent correct={}".format(correct/len(data))) def split_data_set(data_set, eval_perc=0.2): total = len(data_set) split = int(total*eval_perc) train = data_set[:split] evaluate = data_set[split:] return train, evaluate def create_model(out_file, training_folder, learning_rate, epochs, hidden_dimension): # Read in the data training = [] for file_in in [join(training_folder, f) for f in listdir(training_folder) if isfile(join(training_folder, f))]: training += list(driver_support.read_lliaw_dataset_gear_gear_rpm_spe(file_in)) model = Gear_switcher(hidden_dimension) training = driver_support.binerize_data_input(data=training, index=0, mapping=get_gear_map()) training, evalu = split_data_set(training) print(model) evaluate(model, evalu) optimizer = optim.Adam(model.parameters(), lr=learning_rate) loss = nn.CrossEntropyLoss() for ITER in range(epochs): train_loss = 0.0 start = time.time() lowest_gear = 10 highest_gear = 0 last_state = None for y_true, state in training: if last_state == None: last_state = state continue correct_gear = int(y_true[0]) optimizer.zero_grad() in_state = Variable(torch.FloatTensor(last_state)) y_pred = model(in_state).view(1, gear_number) y_true = Variable(gear_to_tensor(correct_gear)) #print(y_true, prediction_to_action(y_pred)) output = loss(y_pred, y_true) train_loss += output.data[0] # backward pass output.backward() # update weights optimizer.step() last_state = state print("last prediction made:pred={}, actual={}".format(prediction_to_action(y_pred), y_true)) print("iter %r: train loss/action=%.4f, time=%.2fs" %(ITER, train_loss/len(training), time.time()-start)) evaluate(model, evalu) torch.save(model.state_dict(), out_file) def main(): parser = argparse.ArgumentParser() parser.add_argument('--epochs', type=int) parser.add_argument('--hidden', type=int) parser.add_argument('--learn', type=float) parser.add_argument('--in_file', type=str) parser.add_argument('--out_file', type=str) args = parser.parse_args() create_model(args.out_file, args.in_file, args.learn, args.epochs, args.hidden) if __name__ == "__main__": main()
joramwessels/torcs-client
data_generator.py
import argparse import xml.etree.ElementTree as ET import subprocess parser = argparse.ArgumentParser() parser.add_argument('--drivers', nargs='+', type=str) parser.add_argument('--tracks', nargs='+', type=str) parser.add_argument('--length', type=int) parser.add_argument('--laps', type=int) parser.add_argument('--file', type=str) args = parser.parse_args() drivers = args.drivers tracks = args.tracks length = args.length laps = args.laps file_in = args.file file_out = "quickrace_auto_gen.xml" all_drivers = ["berniw", "berniw3", "damned", "inferno", "lliaw", "tita", "berniw2", "bt", "inferno2", "olethros", "sparkle"] best_drivers = ["lliaw", "inferno", "olethros", "tita"] best_driver = ["lliaw"] all_tracks = ["aalborg", "alpine-1", "alpine-2", "brondehach", "corkscrew", "e-track-1", "e-track-2", "e-track-3", "e-track-4", "e-track-6", "eroad", "forza", "g-track-1", "g-track-2", "g-track-3", "ole-road-1", "ruudskogen", "spring", "street-1", "wheel-1", "wheel-2"] if tracks[0] == "all": tracks = all_tracks if drivers[0] == "best": drivers = best_driver def set_drivers(root, drivers): xml_drivers = root[4] #Drivers section del xml_drivers[3] #a specific driver definition for index, driver in enumerate(drivers): ele = ET.Element("section", attrib={"name": str(index + 1)}) ele.append(ET.Element("attnum", attrib={"name": "idx", "val": str(index + 1)})) ele.append(ET.Element("attstr", attrib={"name": "module", "val": driver})) xml_drivers.append(ele) def set_track(root, track): xml_track = root[1] xml_track[1][0].attrib["val"] = track def write_and_run(file_out, tree, track): # Write back to file tree.write(file_out) # Run TORCS command = "torcs -r ~/ci/torcs-client/" + file_out completed_command = subprocess.run(command, shell=True, check=True) command = "mv /tmp/lliaw_data.csv ./lliaw_"+track+".data" completed_command = subprocess.run(command, shell=True, check=True) # Open original file tree = ET.parse(file_in) root = tree.getroot() set_drivers(root, drivers) for track in tracks: set_track(root, track) write_and_run(file_out, tree, track)
joramwessels/torcs-client
combined_driver_old.py
<reponame>joramwessels/torcs-client<gh_stars>1-10 from pytocl.driver import Driver from pytocl.car import State, Command import torch import time import sys import math import numpy as np from os import listdir from os.path import isfile, join import driver_support from torch.autograd import Variable from operator import sub import ffnn_steer import ffnn_speed class Final_Driver(Driver): def __init__(self, steering_values, global_max_speed): super(Final_Driver, self).__init__() self.steer = ffnn_steer.Steer(10) self.steer.load_state_dict(torch.load("./steer.data")) self.speed = ffnn_speed.Speed(10) self.speed.load_state_dict(torch.load("./ffnn_speed.data")) self.back_up_driver = Driver(logdata=False) self.bad_counter = 0 self.lap_counter = 0 self.brake_row = 0 self.angles = [90, 75, 60, 45, 30, 20, 15, 10, 5, 0, -5, -10, -15, -20, -30, -45, -60, -75, -90] self.alphas = [math.radians(x) for x in self.angles] self.last_opponents = [0 for x in range(36)] self.steering_values = steering_values self.global_max_speed = global_max_speed def update_trackers(self, carstate): if carstate.current_lap_time == 0: self.lap_counter += 1 print("Lap={}".format(self.lap_counter)) print("distance:",carstate.distance_raced) def drive(self, carstate: State) -> Command: self.update_trackers(carstate) if self.in_a_bad_place(carstate): command = self.back_up_driver.drive(carstate) if self.bad_counter >= 600 and is_stuck(carstate): # we try reversing command.gear = -command.gear if command.gear < 0: command.steering = -command.steering command.gear = -1 self.bad_counter = 200 else: # since the data and python's values differ we need to adjust them carstate.angle = math.radians(carstate.angle) carstate.speed_x = carstate.speed_x*3.6 command = self.make_next_command(carstate) # based on target, implement speed/steering manually print("Executing command: gear={}, acc={}, break={}, steering={}".format(command.gear, command.accelerator, command.brake, command.steering)) return command def make_next_command(self, carstate): command = Command() # we switch gears manually gear = self.gear_decider(carstate) # we get the steering prediction steer_pred = self.steer_decider(carstate, steering_values=self.steering_values) # steer_pred = self.steer_decider_nn(carstate) # pedal =[-1;1], combining breaking and accelerating to one variable pedal = self.speed_decider(carstate, max_speed=self.global_max_speed) # make sure we don't drive at people opponents_deltas = list(map(sub, carstate.opponents, self.last_opponents)) steer_pred, pedal = self.deal_with_opponents(steer_pred, pedal, carstate.speed_x, carstate.distance_from_center, carstate.opponents, opponents_deltas) # disambiguate pedal with smoothing brake, accel = self.disambiguate_pedal(pedal, accel_cap=1.0, break_cap=0.75, break_max_length=5) command.brake = brake command.accelerator = accel command.steering = steer_pred command.gear = gear return command def deal_with_opponents(self, steer_pred, pedal, speed, distance_from_center, opponents_new, opponents_delta): # index 18 is in front # index 35 in behind us adjustment = 0.1 # if there are cars infront-left -> move to right if opponents_new[17] < 10 or opponents_new[16] < 10 or opponents_new[15] < 10: print("ADJUSTING SO NOT TO HIT") steer_pred -= adjustment if opponents_new[19] < 10 or opponents_new[20] < 10 or opponents_new[21] < 10: print("ADJUSTING SO NOT TO HIT") # move to left steer_pred += adjustment if opponents_new[18] < 50: # we are on left side -> move right if distance_from_center > 0: steer_pred -= adjustment # o.w. move left else: steer_pred += adjustment if speed > 100: # we are getting closer to the car in front (and we can't avoid it). We need to slow down a bit if (opponents_delta[18] < 0 and opponents_new[18] < 20) or (opponents_delta[17] < 0 and opponents_new[17] < 4) or (opponents_delta[19] < 0 and opponents_new[19] < 4): pedal -= 0.1 return steer_pred, pedal def disambiguate_pedal(self, pedal, accel_cap=0.5, break_cap=0.75, break_max_length=5): if pedal >= 0.0: accelerator = pedal*accel_cap brake = 0 else: # we need to make sure that we don't break hard enough and not too long self.brake_row += 1 if self.brake_row <= break_max_length: brake = abs(pedal)*break_cap else: self.brake_row = 0 brake = 0 accelerator = 0 return brake, accelerator def steer_decider(self, carstate, steering_values): alpha_index = np.argmax(carstate.distances_from_edge) if is_straight_line(carstate=carstate, radians=self.alphas[alpha_index], factor=steering_values[4]): return carstate.angle*0.5 steering_function = lambda index, offset: (self.alphas[index-offset]*carstate.distances_from_edge[index-offset] + self.alphas[index+offset]*carstate.distances_from_edge[index+offset])/(carstate.distances_from_edge[index+offset]+carstate.distances_from_edge[index-offset]) steer = steering_values[0]*self.alphas[alpha_index] for x in range(1, 4): if alpha_index - x > -1 and alpha_index + x < len(steering_values): steer += steering_values[x]*steering_function(alpha_index, x) return steer def speed_decider(self, carstate, max_speed=120): # we predict speed and map that to pedal x_in = ffnn_speed.carstate_to_variable(carstate) target_speed = self.speed(x_in).data[0] # we limit the speed if target_speed >= max_speed: target_speed = max_speed pedal = 2/(1 + np.exp(carstate.speed_x - target_speed))-1 return pedal def gear_decider(self, carstate): gear = carstate.gear rpm = carstate.rpm # we do gears by hand # up if {9500 9500 9500 9500 9000} # down if {4000 6300 7000 7300 7300} if gear == -1: return 1 elif gear == 0: if rpm >= 5000: gear = 1 elif gear == 1: if rpm >= 9500: gear = 2 elif gear == 2: if rpm >= 9500: gear = 3 elif rpm <= 4000: gear = 2 elif gear == 3: if rpm >= 9500: gear = 4 elif rpm <= 6300: gear = 3 elif gear == 4: if rpm >= 9500: gear = 5 elif rpm <= 7000: gear = 3 elif gear == 5: if rpm >= 9000: gear = 6 elif rpm <= 7300: gear = 4 elif gear == 6: if rpm <= 7300: gear = 5 return gear def in_a_bad_place(self, carstate): something_wrong = False if is_offroad(carstate): print("I'm offroad!") something_wrong = True if is_reversed(carstate): print("I'm reversed!") something_wrong = True if is_stuck(carstate): print("I'm stuck!") something_wrong = True if (something_wrong): self.bad_counter += 1 else: self.bad_counter = 0 # if we have been in a bad place for 2 seconds if self.bad_counter >= 100: return True return False def is_straight_line(carstate, radians, factor): if abs(carstate.distance_from_center) < 0.75: if radians == 0: return True if carstate.distances_from_edge[9] > 190: return True if carstate.distances_from_edge[9] > factor * carstate.speed_x: return True return False def is_offroad(carstate): return max(carstate.distances_from_edge) == -1 def is_stuck(carstate): return abs(carstate.speed_x) <= 5 and carstate.current_lap_time >= 10 def is_reversed(carstate): return abs(carstate.angle) >= 90
joramwessels/torcs-client
gru.py
<reponame>joramwessels/torcs-client<gh_stars>1-10 # file: gru.py # author: <NAME> # date: 23-11-2017 # source: https://github.com/spro/practical-pytorch # description: # # Trains a GRU RNN on your training data and saves the weights. # Run by calling "python gru.py <train_data_file> <save_file>". # """ TODO not tested TODO index out of range in GRU.forward() """ from __future__ import unicode_literals, print_function, division from io import open import unicodedata import string import re import random import sys import time import torch import torch.nn as nn from torch.autograd import Variable from torch import optim import torch.nn.functional as F allow_cuda = False use_cuda = torch.cuda.is_available() and allow_cuda teacher_forcing_ratio = 0.5 class GRU(nn.Module): def __init__(self, i_dim, h_dim, o_dim, n_layers=1): """ The encoder turns a continuous input into a continuous output i_dim: The input dimension h_dim: The amount of hidden units o_dim: The output dimension n_layers: The amount of hidden layers (default=1) """ super(GRU, self).__init__() self.n_layers = n_layers self.h_dim = h_dim self.input_layer = nn.Linear(i_dim, h_dim) self.hidden_layer = nn.GRU(i_dim, h_dim, n_layers, batch_first=True) self.output_layer = nn.Linear(h_dim, o_dim) def forward(self, input, hidden): #input = input.view(1, 1, -1) #output = self.input_layer(input) #for i in range(self.n_layers): # output, hidden = self.hidden_layer(output, hidden) print(input.size()) print(hidden.size()) output, hidden = self.hidden_layer(input, hidden) output = self.output_layer(output) return output, hidden def initHidden(self): result = Variable(torch.zeros(self.n_layers, 1, self.h_dim)) if use_cuda: return result.cuda() else: return result def train(input, target, encoder, optimizer, criterion): """ Trains all timesteps for a sequence of parameters input: Sequence of input variables target: Sequence of target variables encoder: The GRU encoder model optimizer: The encoder optimizer object criterion: The loss function """ hidden = encoder.initHidden() optimizer.zero_grad() loss = 0 #for i in range(len(input)): # output, hidden = encoder(input[i], hidden) # loss += criterion(output, target) output, hidden = encoder(input, hidden) loss = criterion(output, target) loss.backward() optimizer.step() return loss.data[0] / target.size()[0] def train_iters(data, encoder, n_iters, print_every=1000, plot_every=100, learning_rate=0.01): """ Trains the model for 'n_iters' epochs and reports on the progress data: The training sequence pairs as a list of Variables of floats, e.g [Variable(FloatTensor(FloatTensor))] encoder: The encoder model n_iters: The amount of sequences to train print_every: The interval for printing the current loss plot_every: The interval for adding a datapoint to the plot learning_rate: The learning rate of the model """ start = time.time() plot_losses = [] print_loss_total = 0 # Reset every print_every plot_loss_total = 0 # Reset every plot_every encoder_optimizer = optim.SGD(encoder.parameters(), lr=learning_rate) criterion = nn.NLLLoss() for iter in range(1, n_iters + 1): training_pair = data[iter - 1] input_variable = training_pair[0] target_variable = training_pair[1] loss = train(input_variable, target_variable, encoder, encoder_optimizer, criterion) print_loss_total += loss plot_loss_total += loss if iter % print_every == 0: print_loss_avg = print_loss_total / print_every print_loss_total = 0 print('%s (%d %d%%) %.4f' % (timeSince(start, iter / n_iters), iter, iter / n_iters * 100, print_loss_avg)) if iter % plot_every == 0: plot_loss_avg = plot_loss_total / plot_every plot_losses.append(plot_loss_avg) plot_loss_total = 0 showPlot(plot_losses) return def train_gru_on(filename, target_indices, save_as, sep=',', skip_first_line=True): """ Trains a GRU RNN on a given data file filename: The name of a CSV file containing the training data target_indices: A list of integers indicating the columns with target variables save_as: Filename of the saved model parameter file sep: The separator used in the CSV file skip_first_line: Boolean indicating labels on the first line of the CSV """ layers = 1 units = 10 iters = 500 print("Reading data file",filename,"...") training_pairs = readTrainingData(filename, target_indices, sep, skip_first_line) print("Creating GRU model with %i layers of %i hidden units" %(layers, units),"...") model = GRU(len(training_pairs[0][0][0]), units, len(target_indices), n_layers=layers) print("Training model on data...") train_iters(training_pairs, model, iters) print("Model trained for %i iterations" %iters) model.save_state_dict(save_as) print("Model saved as",save_as) def readTrainingData(filename, target_indices, sep, skip_first_line): """ Reads the data files to a list of sequence pairs filename: The name of a CSV file containing the training data target_indices: A list of integers indicating the columns with target variables sep: The separator used in the CSV file skip_first_line: Boolean indicating labels on the first line of the CSV """ inputs = [] targets = [] with open(filename) as file: if skip_first_line: next(file) for line in file: params = [float(s) for s in line.strip().split(sep)] if not len(params) == 25: continue inputs.append(params) targets.append([params[i] for i in target_indices]) if use_cuda: inputs = Variable(torch.FloatTensor([inputs])).cuda() targets = Variable(torch.FloatTensor([targets])).cuda() else: inputs = Variable(torch.FloatTensor([inputs])) targets = Variable(torch.FloatTensor([targets])) return [(inputs, targets)] if __name__ == "__main__": train_gru_on(sys.argv[1], [0,1,2], sys.argv[2], sep=',', skip_first_line=True)
joramwessels/torcs-client
run.py
<filename>run.py<gh_stars>1-10 #! /usr/bin/env python3 from pytocl.main import main from my_driver import MyDriver from nn_driver import FFNNDriver if __name__ == '__main__': main(FFNNDriver())
joramwessels/torcs-client
driver_support.py
import typing as t import numpy as np def get_gear(current_rpm, current_gear): if current_rpm >= 8500: current_gear += 1 if current_rpm <= 2000: current_gear -= 1 if current_gear <= 0: current_gear = 1 if current_gear >= 6: current_gear = 6 return current_gear def get_steer(target_pos, actual_pos, angle, epsilon=0.1): error = target_pos - actual_pos angle = angle + error * epsilon steer = angle/180 return steer def get_accel(target_speed, actual_speed): accel = 0 if (target_speed - actual_speed) > 0: accel = (target_speed - actual_speed)/20 if target_speed - actual_speed > 20: accel = 1 return accel def get_break(target_speed, actual_speed): brake = 0 if (target_speed - actual_speed) < 0: brake = -(target_speed - actual_speed)/20 if target_speed - actual_speed < -20: brake = 1 return brake def map_to_gear(prediction): # we represent the reverse gear as index 0 # then gear=1 is just index 1 of the list, etc. index = np.argmax(prediction) if index == 0: return -1 else: return index def binerize_input(value, mapping): value = str(value) index = mapping[value] classes = [0 for x in range(len(mapping))] classes[index] = 1 return classes def binerize_data_input(data, index, mapping): new_data = [] for correct, state in data: value = int(state[index]) new_data.append((correct, state[:index] + binerize_input(value, mapping) + state[:index + 1])) return new_data def read_dataset_stear_speed(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: next(f) # as the file is a csv, we don't want the first line # acc, break, steer, # 1.0, 0.0, -1.6378514771737686E-5, # speed, pos, angle, # -0.0379823,-5.61714E-5,4.30409E-4, # distances... len(distances) = 19 # 5.00028, 5.0778, 5.32202, 5.77526, 6.52976, 7.78305, 10.008, 14.6372, 28.8659, 200.0, # 28.7221, 14.6009, 9.99199, 7.7742, 6.52431, 5.77175, 5.31976, 5.07646, 4.99972 for line in f: # we predict based on steer + speed # based on pos + angle + distances yield ([float(x) for x in line.strip().split(",")[2:4]], [float(x) for x in line.strip().split(",")[4:]]) def read_lliaw_dataset_gear_acc_bre_rpm_spe(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: for line in f: # 0 1 2 3 4 5 6 7 8 9 # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # 10 11 12 13 14-22 23 24-49 # rpm speedx speedy speedz tracksensor1_19 distToMiddle oppSenso1_36 # y=gear, x=accel, break, rpm, speedx line_values = [float(x) for x in line.strip().split(",")[:-1]] yield ([line_values[8]], [line_values[0], line_values[1], line_values[10], line_values[11]]) def read_lliaw_dataset_gear_gear_rpm_spe(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: for line in f: # 0 1 2 3 4 5 6 7 8 9 # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # 10 11 12 13 14-22 23 24-49 # rpm speedx speedy speedz tracksensor1_19 distToMiddle oppSenso1_36 # y=gear, x=gear, rpm line_values = [float(x) for x in line.strip().split(",")[:-1]] yield ([line_values[8]], [line_values[8], line_values[10]]) def read_lliaw_dataset_speed_angle_dist_middle(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: for line in f: # 0 1 2 3 4 5 6 7 8 9 # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # 10 11 12 13 14-32 33 34-59 # rpm speedx speedy speedz tracksensor1_19 distToMiddle oppSenso1_36 # y=speed, x=angle, distance*19, distToMiddle line_values = [float(x) for x in line.strip().split(",")[:-1]] yield ([line_values[11]], [line_values[3], line_values[14], line_values[15], line_values[16], line_values[17], line_values[18], line_values[19], line_values[20], line_values[21], line_values[22], line_values[23], line_values[24], line_values[25], line_values[26], line_values[27], line_values[28], line_values[29], line_values[30], line_values[31], line_values[32], line_values[33]]) def read_lliaw_dataset_steer_angle_speed_dist_middle(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: for line in f: # 0 1 2 3 4 5 6 7 8 9 # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # 10 11 12 13 14-32 33 34-59 # rpm speedx speedy speedz tracksensor1_19 distToMiddle oppSenso1_36 # y=steer, x=angle, speed, distance*19, distToMiddle line_values = [float(x) for x in line.strip().split(",")[:-1]] yield ([line_values[2]], [line_values[3], line_values[11], line_values[14], line_values[15], line_values[16], line_values[17], line_values[18], line_values[19], line_values[20], line_values[21], line_values[22], line_values[23], line_values[24], line_values[25], line_values[26], line_values[27], line_values[28], line_values[29], line_values[30], line_values[31], line_values[32], line_values[33]]) def read_lliaw_dataset_acc_bre_steer_bunch(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: for line in f: # 0 1 2 3 4 5 6 7 8 9 # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # 10 11 12 13 14-32 33 34-59 # rpm speedx speedy speedz tracksensor1_19 distToMiddle oppSenso1_36 # y=accel, break, stear, x=angle, speed, distance*19, distToMiddle line_values = [float(x) for x in line.strip().split(",")[:-1]] yield ([line_values[0],line_values[1], line_values[2]], [line_values[3], line_values[11], line_values[14], line_values[15], line_values[16], line_values[17], line_values[18], line_values[19], line_values[20], line_values[21], line_values[22], line_values[23], line_values[24], line_values[25], line_values[26], line_values[27], line_values[28], line_values[29], line_values[30], line_values[31], line_values[32], line_values[33]]) def read_lliaw_dataset_acc_br_stee_ge(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: # accel break steer angle curLapTime distFromStartLine distRaced fuel gear racepos # rpm speedx speedy speedz tracksensor1_19 distToMiddle for line in f: # we predict based on steer + speed # based on pos + angle + distances yield ([float(x) for x in line.strip().split(",")[0:3] + line.strip().split(",")[8:9]], [float(x) for x in line.strip().split(",")[3:34]]) def read_dataset_acc_break_steer(filename: str) -> t.Iterable[t.Tuple[t.List[float], t.List[float]]]: with open(filename, "r") as f: next(f) # as the file is a csv, we don't want the first line # acc, break, steer, # 1.0, 0.0, -1.6378514771737686E-5, # speed, pos, angle, # -0.0379823,-5.61714E-5,4.30409E-4, # distances... len(distances) = 19 # 5.00028, 5.0778, 5.32202, 5.77526, 6.52976, 7.78305, 10.008, 14.6372, 28.8659, 200.0, # 28.7221, 14.6009, 9.99199, 7.7742, 6.52431, 5.77175, 5.31976, 5.07646, 4.99972 for line in f: # we predict based on steer + speed # based on pos + angle + distances yield ([float(x) for x in line.strip().split(",")[0:3]], [float(x) for x in line.strip().split(",")[3:]])
joramwessels/torcs-client
train.py
<filename>train.py #! /usr/bin/env python3 from pytocl.main import main from combined_driver import Final_Driver steering = [REPLACE_STEERING] max_speed = REPLACE_MAX_SPEED if __name__ == '__main__': main(Final_Driver(steering, max_speed))
joramwessels/torcs-client
ffnn_driver.py
from pytocl.driver import Driver from pytocl.car import State, Command import torch from torch.autograd import Variable import torch.nn as nn import torch.optim as optim import time import sys from os import listdir from os.path import isfile, join import driver_support class FFNN_Driver(Driver): def __init__(self, hidden_dimension, in_file): super(FFNN_Driver, self).__init__() self.model = FFNN(hidden_dimension) self.model.load_state_dict(torch.load(in_file)) self.last_command = Command() self.last_command.accelerator = 1 self.last_command.brake = 0 self.last_command.steering = 0 self.last_command.gear = 1 self.sensor_data = [] def drive(self, carstate: State) -> Command: # translate carstate to tensor for NN x_in = Variable(carstate_to_tensor(carstate)) # get speed/steering target acc_break_steer_prediction = self.model(x_in).data print(acc_break_steer_prediction) # based on target, implement speed/steering manually command = Command() command.accelerator = acc_break_steer_prediction[0] command.brake = acc_break_steer_prediction[1] command.steering = acc_break_steer_prediction[2] command.gear = acc_break_steer_prediction[3] self.last_command = command return command def carstate_to_tensor(carstate: State) -> torch.FloatTensor: # angle curLapTime distFromStartLine distRaced fuel gear racepos # rpm speedx speedy speedz tracksensor1_19 distToMiddle return torch.FloatTensor([carstate.angle, carstate.current_lap_time, carstate.distance_from_start, carstate.distance_raced, carstate.fuel, carstate.gear, carstate.race_position, carstate.rpm, carstate.speed_x, carstate.speed_y, carstate.speed_z] + list(carstate.distances_from_edge) + [carstate.distance_from_center] ) class FFNN(nn.Module): def __init__(self, hidden_dimension): super(FFNN, self).__init__() n_states = 31 # see above n_actions = 4 # acc + break + steering + gear self.layer_1 = nn.Linear(n_states, hidden_dimension) self.non_lin = nn.Sigmoid() self.layer_2 = nn.Linear(hidden_dimension, n_actions) def forward(self, inputs): out = self.layer_1(inputs) out = self.non_lin(out) out = self.layer_2(out) return out def create_model(out_file, training_folder, learning_rate, epochs, hidden_dimension): epochs = int(epochs) learning_rate = float(learning_rate) hidden_dimension = int(hidden_dimension) # Read in the data training = [] for file_in in [join(training_folder, f) for f in listdir(training_folder) if isfile(join(training_folder, f))]: training += list(driver_support.read_lliaw_dataset_acc_br_stee_ge(file_in)) model = FFNN(hidden_dimension) print(model) optimizer = optim.SGD(model.parameters(), lr=learning_rate) # I'm not sure about this loss... loss = nn.MSELoss() for ITER in range(epochs): train_loss = 0.0 start = time.time() for y_true, state in training: # forward pass in_state = Variable(torch.FloatTensor(state)) y_pred = model(in_state) y_true = Variable(torch.FloatTensor(y_true)) output = loss(y_pred, y_true) train_loss += output.data[0] # backward pass optimizer.zero_grad() output.backward() # update weights optimizer.step() print("last prediction made:", y_pred, y_true) print("iter %r: train loss/action=%.4f, time=%.2fs" %(ITER, train_loss/len(training), time.time()-start)) torch.save(model.state_dict(), out_file) def main(): create_model(*sys.argv[1:]) if __name__ == "__main__": main()
joramwessels/torcs-client
swarm.py
<reponame>joramwessels/torcs-client # file: swarm.py # author: <NAME> # date: 05-12-2017 # depedencies: numpy # description: # Handles the swarm technology behind the autonomous TORCS driver. # usage: # Assuming the variables # *distTraveled*, *pos*, *spd*, *crashed*, and *contact*, # # if distTraveled % trail.pos_int == 0: # max_speed = trail.check_in(pos, spd, crashed, contact) # crashed, contact = False, False # from sys import stderr from collections import defaultdict from numbers import Real from os.path import isfile from os import remove import numpy as np SEP = ',' # the separator used in the feromone trail files NAME = '.feromones' # the basic filename for feromone trail files NOMAX = 10000 # returned when an error occurs # - TODO unit test: back_prop, report_result, check_in # - TODO if track is known, or if feromone trail looks like a known track, switch to known trail # - NOTE if they drive close behind each other, they'll always explore the same max_speeds # - NOTE initializing a FeromoneTrail will not read the previous .feromone entries # - NOTE back_prop will stop at the finish line, since the length of the track isn't known class FeromoneTrail: def __init__(self, pos_int, spd_int, spd0, spd_n, expl_int, glb_max, track_id=None): """ FeromoneTrail keeps track of-, and syncs the feromone trail A FeromoneTrail contains a table of known feromones and syncs to a .feromone file in storage to communicate with the swarm. The initialization requires a position interval in meters at which to check and update the feromone trail, a grid of possible max_speed values to explore, an exploration interval to increase the max_speed with when no negative experiences are known, and a global maximum speed to default to when there are no positive experiences. The resulting max_speeds can be lower than the global maximum when this speed resulted in a negative experience. Make sure to choose your speed grid big enough to harbor any speed category you want to learn. Args: pos_int: The interval at which to check feromones in m (int) spd_int: The interval between speed boxes in km/h (int) spd0: The first speed box in km/h (int) spd_n: The amount of speed boxes (int) expl_int: The jump interval (dividable by spd_int) in km/h (int) glb_max: The global max speed that ensures a finish in km/h (int) track_id: The name of the race track if known """ self.pos_int = int(pos_int) self.spd_int = int(spd_int) self.spd0 = int(spd0) self.spd_n = int(spd_n) self.spd_max = (spd0 + spd_n * spd_int) - spd_int self.expl_int = int(expl_int) self.glb_max = int(glb_max) self.prev_pos = 0 self.prev_spd = 0 self.filename = NAME + '_' + track_id if track_id else NAME self.table = defaultdict(lambda: np.zeros(spd_n)) self.leave_feromone(0, 0, 0) if isfile(self.filename): remove(self.filename) def ___str___(self): """ Casts the feromone trail table to a string representation """ return self.__repr__() def __repr__(self): """ Casts the feromone trail table to a string representation """ i = 0 speeds = [str(self.to_speed(i)) for i in range(self.spd_n)] string = "\t " + ' '.join(speeds) + " km/h" + '\n' while str(i) in self.table: string += str(i) + ':\t' + str(self.table[str(i)]) + '\n' i += self.pos_int string += "m\n" return string def to_index(self, spd): """ Converts absolute speed to table index """ return int((spd - self.spd0) // self.spd_int) def to_speed(self, ind): """ Converts table index to absolute speed """ return self.spd0 + ind * self.spd_int def is_on_grid(self, spd): """ Returns True if speed value is on the speed grid """ return not (spd < self.spd0) or (spd > self.spd_max) def write_feromone(self, pos, speed, val): """ Writes a new feromone to the .feromone file Args: pos: The position on the track, CurLapTime (int) speed: The speed that has been tested (int) val: The result of the test (-1, 0, 1) """ file = open(self.filename, 'a') file.write('\n' + SEP.join([str(pos), str(speed), str(val)])) file.close() def read_feromone(self): """ Reads the last feromones and updates it if they're new Returns: List of [pos, speed, val] lists if there are any """ file = open(self.filename, 'r') contents = file.readlines() file.close() i = 1 changes = [] changed = True while changed: if contents[-i].strip() == '': continue feromone = [int(s) for s in contents[-i].strip().split(SEP)] if feromone == self.last_change: break changes.append(feromone) i += 1 if changes: self.last_change = changes[0] return changes def update_table(self, pos, spd, val): """ Updates a newly received feromone in the table Args: pos: The position on the track, CurLapTime (int) spd: The speed that has been tested (int) val: The result of the test (-1, 0, 1) """ index = self.to_index(spd) if val == -1: for i in range(index, self.to_index(self.spd_max) +1): self.table[str(pos)][i] = -1 elif val == 1: for i in range(index, -1, -1): self.table[str(pos)][i] = 1 def next_experiment(self, pos): """ Checks the table for the next best max speed experiment Returns the ideal next max speed to try out, regardless of the current speed of the car. Args: pos: The position on the track, CurLapTime (int) Returns: The next best max speed value (int) """ row = self.table[str(pos)] i1 = find_first(row, 1, rev=True) i2 = find_first(row, -1) i_glb_max = self.to_index(self.glb_max) # if there are no occurences of + above glb_max if i1 == -1 or (i1 < i_glb_max and not row[i_glb_max] == -1): if row[i_glb_max] == -1: i1 = i2 - 1 # last 0 before first - else: i1 = i_glb_max # resort to global max # exploring, value in between known values, or safe value if i2 == -1: spd = min(self.spd_max, self.to_speed(i1) + self.expl_int) index = self.to_index(spd) else: index = i1 + (i2 - i1) // 2 return index * self.spd_int + self.spd0 def leave_feromone(self, pos, spd, val): """ Updates the table and writes the new feromone to the file If an off-grid pos value is passed, it defaults to the last on-grid value Args: pos: The position on the track, CurLapTime (int) spd: The speed that has been tested (int) val: The result of the test (-1, 0, 1) """ self.last_change = [pos, spd, val] self.update_table(pos, spd, val) self.write_feromone(pos, spd, val) def back_prop(self, pos, max_spd): """ Updates previous frames to anticipate this failed *max_spd* Args: pos: The position on the track, CurLapTime (int) max_spd: The max speed that has failed (int) """ while max_spd < self.spd_max and pos > -1: first_minus = find_first(self.table[str(pos)], -1) if self.to_index(max_spd) >= first_minus and first_minus > -1: break self.leave_feromone(pos, max_spd, -1) max_spd = int(breakable_speed(max_spd, self.pos_int)) max_spd -= int(max_spd % self.spd_int) pos -= self.pos_int def get_max_speed(self, pos): """ Updates the feromone table and returns the next max speed If an off-grid pos value is passed, it defaults to the next on-grid value Args: pos: The position on the track, CurLapTime (int) Returns: The next best max speed value (int) """ if not pos % self.pos_int == 0: err("SWARM WARNING: Invalid position:", pos) pos += self.pos_int - (pos % self.pos_int) err(" Defaulted to", pos) change = self.read_feromone() while change: ppos, speed, val = change.pop() self.update_table(ppos, speed, val) max_speed = self.next_experiment(pos) return max_speed def report_result(self, pos, spd, val): """ Updates the feromone trail with the new information Args: pos: The position on the track, CurLapTime (int) spd: The current speed of the car (int) val: The result of the experiment (-1, 0, 1) """ spd -= 1 # such that 160 falls into the category with max_spd=160 max_spd = spd - (spd % self.spd_int) + self.spd_int spd_i = self.to_index(max_spd) if val == -1: self.back_prop(pos, max_spd) elif not self.table[str(pos)][spd_i] == val: self.leave_feromone(pos, max_spd, val) def check_in(self, pos, spd, crashed, contact): """ Called at the start of ever frame to check/update feromones Args: pos: The position on the track, distTraveled in m (num) spd: The current speed of the car in km/h (num) crashed: Indicates a crash or off-track in last frame (bool) contact: Indicates contact with another car in last frame (bool) Returns: The maximum speed for the next frame according to the swarm """ # input verification if not isinstance(pos, Real): err("SWARM ERROR: pos isn't a real number, but:", pos) return NOMAX if not isinstance(spd, Real): err("SWARM ERROR: spd isn't a real number, but:", pos) return NOMAX if spd > self.spd_max: err("SWARM WARNING: Speed is beyond speed grid:", spd) err(" Swarm can't learn from this experience") if not pos % self.pos_int == 0: err("SWARM WARNING: Invalid position:", pos) pos -= pos % self.pos_int err(" Defaulted to: ", pos) pos, spd = int(pos), int(spd) # update if self.is_on_grid(self.prev_spd): if crashed and not contact: self.report_result(self.prev_pos, self.prev_spd, -1) elif not crashed and not contact: self.report_result(self.prev_pos, self.prev_spd, 1) # predict self.prev_pos, self.prev_spd = pos, spd max_speed = self.get_max_speed(pos) return max_speed def err(*args): """ prints to standard error """ print(*args, file=stderr) def find_first(array, val, rev=False): """ Finds the first (or last) occurence of val in array Args: array: The numpy array to evaluate val: The value to find rev: If True, returns the last occurence of val Returns: The index of the first (or last) occurence of val in array, or -1 if the value doesn't appear in array """ ar = np.array(list(array)) # copies the array if rev: ar = np.flip(ar, 0) i = np.argmax(ar==val) if i == 0 and not ar[0] == val: return -1 if rev: i = abs(i - len(ar) + 1) return i def breakable_speed(end_speed, trajectory): """ Computes the max speed that can break to reach *end_speed* Args: end_speed: The speed at the end of the trajectory in km/h (num) trajectory: The distance over which to descelerate in m (num) Returns: The maximum absolute speed at the beginning of the trajectory that ensures a desceleration to *end_speed* """ # The car is about 5m long, it descelerated from 280 to 0 # in about 12-14 times its length, which would be 60-70m. # Assuming a linear decrease in speed, the maximum rate # of desceleration is therefore -280/65 = -4.31 km/h/m. # To be safe, we use half that: -2.15 return trajectory * 2.15 + end_speed
joramwessels/torcs-client
rnn.py
import torch.nn as nn from torch.autograd import Variable import torch import torch.optim as optim import argparse import driver_support import time import sys from os import listdir from os.path import isfile, join class RNNMove(nn.Module): """Container module with an encoder, a recurrent module, and a decoder.""" def __init__(self, rnn_type, input_dimension, hidden_dimension, output_dimension, nlayers, dropout=0.5, tie_weights=False): super(RNNMove, self).__init__() self.drop = nn.Dropout(dropout) nonlinearity = {'RNN_TANH': 'tanh', 'RNN_RELU': 'relu'}[rnn_type] self.encoder = nn.Linear(input_dimension, hidden_dimension) self.rnn = nn.RNN(hidden_dimension, hidden_dimension, nlayers, nonlinearity=nonlinearity, dropout=dropout) self.decoder = nn.Linear(hidden_dimension, output_dimension) self.init_weights() self.rnn_type = rnn_type self.hidden_dimension = hidden_dimension self.nlayers = nlayers def init_weights(self): initrange = 0.1 self.decoder.bias.data.fill_(0) self.decoder.weight.data.uniform_(-initrange, initrange) def forward(self, input, hidden): output = self.encoder(input).view(-1, 1) output, hidden = self.rnn(output, hidden) output = self.drop(output) decoded = self.decoder(output.view(output.size(0)*output.size(1), output.size(2))) return decoded.view(output.size(0), output.size(1), decoded.size(1)), hidden def init_hidden(self, bsz): return nn.Parameter(torch.zeros(self.nlayers, bsz, self.hidden_dimension), requires_grad=True) def split_data_set(data_set, eval_perc=0.2): total = len(data_set) split = int(total*eval_perc) train = data_set[:split] evaluate = data_set[split:] return train, evaluate def create_model(out_file, training_folder, learning_rate, epochs, hidden_dimension): # Read in the data training = [] for file_in in [join(training_folder, f) for f in listdir(training_folder) if isfile(join(training_folder, f))]: training += list(driver_support.read_lliaw_dataset_acc_bre_steer_bunch(file_in)) n_states = 22 n_actions = 3 n_layers = 1 model = RNNMove("RNN_RELU", n_states, hidden_dimension, n_actions, n_layers, dropout=0.5, tie_weights=False) training, evalu = split_data_set(training) print(model) optimizer = optim.Adam(model.parameters(), lr=learning_rate) loss = nn.MSELoss() for ITER in range(epochs): train_loss = 0.0 start = time.time() hidden = model.init_hidden(1) for y_true, state in training: optimizer.zero_grad() in_state = Variable(torch.FloatTensor(state)) print(in_state, hidden) y_pred, hidden = model(in_state, hidden) y_true = Variable(torch.FloatTensor(y_true)) #print(y_true, prediction_to_action(y_pred)) output = loss(y_pred, y_true) train_loss += output.data[0] # backward pass output.backward() # update weights optimizer.step() print("last prediction made:pred={}, actual={}".format(prediction_to_action(y_pred), y_true)) print("iter %r: train loss/action=%.4f, time=%.2fs" %(ITER, train_loss/len(training), time.time()-start)) #evaluate(model, evalu) torch.save(model.state_dict(), out_file) def main(): parser = argparse.ArgumentParser() parser.add_argument('--epochs', type=int) parser.add_argument('--hidden', type=int) parser.add_argument('--learn', type=float) parser.add_argument('--in_file', type=str) parser.add_argument('--out_file', type=str) args = parser.parse_args() create_model(args.out_file, args.in_file, args.learn, args.epochs, args.hidden) if __name__ == "__main__": main()
joramwessels/torcs-client
crisis_driver.py
# file: crisis_driver.py # author: <NAME> # date: 07-12-2017 # dependencies: numpy, pytocl # description # usage # from pytocl.driver import Driver from pytocl.car import State, Command from sys import stderr from numpy import sign from driver_utils import * # margins of error when trying the navigate_to_middle approach CTR_ANG_MARGIN = 10 PRP_ANG_MARGIN = 10 # careful acceleration values during crisis handling ACC = 0.1 REV_ACC = 0.2 OFF_ROAD_ACC = 0.1 OFF_ROAD_REV_ACC = 0.2 # timers (100 = 2 sec) BAD_COUNTER_THRESHOLD = 100 BAD_COUNTER_MANUAL_THRESHOLD = 600 APPROACH_TIMEOUT = 1000 # cross-cycle trackers MAX_ANGLES = 64 # amount of angles to keep track of MAX_ACCELS = 16 # amount of cycles in which you should have moved MAX_GEARS = MAX_ACCELS GEAR_SHIFTING_TRANSITION_CYCLES = 5 # NOTE # steer left = positive # angle left = negative # left of the road = positive class CrisisDriver(Driver): def __init__(self, logdata=True): """ Handles difficult situations in which the car gets stuck """ super(CrisisDriver, self).__init__(logdata=logdata) self.iter = 0 self.driver = Driver(logdata=False) self.is_in_control = False self.approaches = [self.navigate_to_middle, self.original_implementation] self.previous_angles = [] self.previous_accels = [] self.previous_gears = [] self.bad_counter = 0 self.needs_help = False def drive(self, carstate): """ Gets the car out of a difficult situation Tries different approaches and gives each one a time out. A class level variable keeps track of the current approach across game cycles. If the timer runs out, the approach is terminated and the next approach gets initiated. Args: carstate: All parameters packed in a State object Returns: command: The next move packed in a Command object """ command = Command() command.accelerator = 0 self.appr_counter += 1 if self.appr_counter > APPROACH_TIMEOUT: self.next_approach() try: command = self.approaches[self.approach](carstate, command) self.previous_accels.append(command.accelerator) self.previous_gears.append(command.gear) except Exception as e: err(self.iter, "ERROR:", str(e)) return command def pass_control(self, carstate): """ Initializes a new crisis that has not been handled yet Args: carstate: The original carstate """ err(self.iter,"CRISIS: control received") self.is_in_control = True self.approach = 0 self.appr_counter = 0 # check if car in front # check if car behind # check track angle and side of the road # determine reverse or straight ahead def return_control(self): """ Passes control back to the main driver """ err(self.iter,"CRISIS: control returned") self.is_in_control = False self.needs_help = False def next_approach(self): """ Adjusts state to next approach """ self.approach += 1 self.appr_counter = 0 if self.approach >= len(self.approaches): self.approach -= 1 self.return_control() else: err(self.iter,"CRISIS: next approach:", self.approaches[self.approach].__name__) def approach_succesful(self): """ Called when a technique finished executing """ err(self.iter,"CRISIS: approach succesful:", self.approaches[self.approach].__name__) if self.has_problem: self.next_approach() else: self.return_control() def update_status(self, carstate): """ Updates the status of the car regarding its problems Args: carstate: The full carstate """ self.iter += 1 if len(self.previous_angles) >= MAX_ANGLES: self.previous_angles.pop(0) self.previous_angles.append(carstate.angle) if len(self.previous_accels) >= MAX_ACCELS: self.previous_accels.pop(0) self.previous_gears.append(sign(carstate.gear)) if len(self.previous_gears) >= MAX_GEARS: self.previous_gears.pop(0) self.is_on_left_side = sign(carstate.distance_from_center) == DFC_L self.is_on_right_side = not self.is_on_left_side self.faces_left = sign(carstate.angle) == ANG_L self.faces_right = not self.faces_left self.faces_front = abs(carstate.angle) < 90 self.faces_back = not self.faces_front self.faces_middle = self.is_on_left_side == self.faces_right self.is_standing_still = abs(carstate.speed_x) < 0.1 self.has_car_in_front = car_in_front(carstate.opponents) self.has_car_behind = car_behind(carstate.opponents) self.is_blocked = blocked(self.previous_accels, self.previous_gears, carstate.speed_x) self.is_off_road = max(carstate.distances_from_edge) == -1 self.is_reversed = self.faces_back self.is_going_in_circles = going_in_circles(self.previous_angles) self.is_traveling_sideways = abs(carstate.speed_y) > 15 self.has_problem = self.is_off_road or \ self.is_going_in_circles or \ self.is_blocked or \ self.is_reversed or \ self.is_traveling_sideways if self.has_problem: self.bad_counter += 1 if self.bad_counter >= BAD_COUNTER_THRESHOLD: if self.is_off_road: debug(self.iter, " off road") if self.is_going_in_circles: debug(self.iter, " going in circles") if self.is_blocked: debug(self.iter, " blocked") if self.is_reversed: debug(self.iter, " reversed") self.needs_help = True else: self.bad_counter = 0 self.needs_help = False # # Approach Implementations # def original_implementation(self, carstate, command): """ approach 0) Args: carstate: The full carstate as passed down by the server command: The command to adjust """ if not self.is_off_road and abs(carstate.angle) < CTR_ANG_MARGIN: self.approach_succesful() command.gear, carstate.gear = 1, 1 command = self.driver.drive(carstate) # TODO is this legal? is_stuck = abs(carstate.speed_x) <= 5 and carstate.current_lap_time >= 10 #if self.bad_counter >= BAD_COUNTER_MANUAL_THRESHOLD and is_stuck: # we try reversing # command.gear = -command.gear # if command.gear < 0: # command.steering = -command.steering # command.gear = -1 # self.bad_counter = 200 command.accelerator /= 2.0 return command def navigate_to_middle(self, carstate, command): """ Finds it way to the middle of the road by driving in reverse approach 1) reverse towards the road, then once on the road, reverse towards the the middle until facing foward with an angle that's within the margin Args: carstate: The full carstate as passed down by the server command: The command to adjust """ debug(self.iter,"CRISIS: navigate_to_middle") dfc = carstate.distance_from_center if self.is_off_road: dist_spd = 2 if abs(dfc) > 10 else 1 perp_angle = 90 * sign(dfc) if self.faces_middle: debug(self.iter," off road and facing road") diff_with_perp_angle = perp_angle - carstate.angle if not abs(diff_with_perp_angle) < PRP_ANG_MARGIN: command.steering = sign(diff_with_perp_angle) * STR_R command.gear = 1 command.accelerator = OFF_ROAD_ACC * dist_spd else: debug(self.iter," off road and not facing road") diff_with_perp_angle = perp_angle + carstate.angle if not abs(diff_with_perp_angle) < PRP_ANG_MARGIN: command.steering = sign(diff_with_perp_angle) * STR_R command.gear = -1 command.accelerator = OFF_ROAD_REV_ACC * dist_spd else: if abs(carstate.angle) < CTR_ANG_MARGIN: self.approach_succesful() else: if self.faces_middle or abs(carstate.angle) < 50: # TODO globalize debug(self.iter," on road facing middle") command.steering = to_ang(carstate.angle) command.gear = 1 command.accelerator = ACC elif abs(dfc) > 0.5: debug(self.iter," on road not facing middle") command.steering = away_from_ang(carstate.angle) command.gear = -1 command.accelerator = REV_ACC if self.is_blocked: command.gear = 1 command.accelerator *= 2 debug(self.iter, "ang=%.2f, spd=%.2f, dfc=%.2f" %(carstate.angle, carstate.speed_x, carstate.distance_from_center)) debug(self.iter, "ste=%.2f, acc=%.2f, gea=%.2f" %(command.steering, command.accelerator, command.gear)) # when spinning out of control, don't do anything if self.is_traveling_sideways or self.is_going_in_circles: command.acceleration = 0 gstc = GEAR_SHIFTING_TRANSITION_CYCLES # braking when shifting gear so it doesn't continue its course if any(not g == command.gear for g in self.previous_gears[-gstc:]): command.brake = 1 command.acceleration = 0 # faster acceleration after having shifted gears elif any(not g == command.gear for g in self.previous_gears[-gstc*2:-gstc]): command.brake = 0 command.acceleration = 1.0 return command # # Problem Detectors # def car_in_front(opp): """ Returns True if there's a car in front of ours """ return any([o < 1 for o in opp[17:18]]) # TODO are these the 20deg slices in front? def car_behind(opp): """ Returns True if there's a car behind ours """ return any([0 < 0 for o in [opp[0], opp[35]]]) # TODO and these in the back? def going_in_circles(angles): """ Returns True if the car is rapidly going round in circles """ # TODO checks if there is a pattern in the angles that # constitutes rotation. Take care of skip from 180 to -180. return False def blocked(accels, gears, speed): """ Returns True if the car is blocked by something """ same_gear = all([g > 0 for g in gears]) or all([g < 0 for g in gears]) been_accel = all([a > 0 for a in accels]) return been_accel and same_gear and abs(speed) < 5
joramwessels/torcs-client
empty_driver.py
from pytocl.driver import Driver from pytocl.car import State, Command class EmptyDriver(Driver): ...
joramwessels/torcs-client
genetic_approach.py
<filename>genetic_approach.py<gh_stars>1-10 import numpy as np import random import run_torcs import matplotlib.pyplot as plt from multiprocessing import TimeoutError DATA_FILE = "gen.data" PLOT_NAME = "evolutionary.png" ranges = [(-5,5,float),(-5,5,float),(-5,5,float),(-5,5,float),(-5,5,float),(10,360,int)] def mutate_small(gene, ranges, mutation_probability=0.1, increment=0.01): new_gene = [] for index, feature in enumerate(gene): new_feature = feature # with probability we mutate if np.random.random_sample() >= (1 - mutation_probability): total_range = ranges[index][1] - ranges[index][0] change = increment * total_range sign = 1 if np.random.random_sample() >= 0.5: sign = -1 change *= sign new_feature += change # if we are at the maximum, we don't go over it if new_feature > ranges[index][1]: new_feature = ranges[index][1] elif new_feature < ranges[index][0]: new_feature = ranges[index][0] new_gene.append(ranges[index][2](new_feature)) return new_gene def mutate_random(gene, ranges, mutation_probability=0.05): new_gene = [] for index, feature in enumerate(gene): new_feature = feature # with probability we mutate if np.random.random_sample() >= (1 - mutation_probability): new_feature = np.random.uniform(ranges[index][0], ranges[index][1]) new_gene.append(ranges[index][2](new_feature)) return new_gene def evaluate(gene): evaluation = -99999999 client, server = run_torcs.run_on_ea_tracks('scr_server', steering_values=gene[:5], max_speed=int(gene[5]), timeout=5) times = [] distances = [] for track_index, track in enumerate(client): distance = run_torcs.get_distance_covered(track) if distance == -1: print("error, server=",server[track_index]) print("error, client=",client[track_index]) else: distances.append(distance) time_run = run_torcs.get_total_time_covered(track) if time_run != -1: times.append(time_run) # if all runs failed, we just try again. if len(times) == 0: return evaluate(gene) print(distances) print(times) evaluation = sum(list(map((lambda x, y: x/y), distances, times)))/len(client) return evaluation def select(population, evaluations, count): surviving_parents = [] surv_fitness = [] for x in range(count): index = np.argmax(np.array(evaluations)) # we make a copy of the gene surviving_parents.append(population[index][:]) surv_fitness.append(evaluations[index]) del population[index] del evaluations[index] return surviving_parents, surv_fitness def get_random_gene(ranges): gene = [] for low, high,_ in ranges: gene.append(np.random.uniform(low, high)) return gene def terminate(max_generations, generation): if generation > max_generations: print("Maximum generation reached") return True # if maximum - average <= 1: # print("Maximum and average are close") # return True # else: return False def print_generation_values(maximum, average, minimum): print("maximum:", maximum) print("average:", average) print("minimum:", minimum) def print_generation(number): print("Generation={}".format(number)) def print_gene(gene, gene_index, fitness): floats_adjusted = ", ".join(["%.2f"%x for x in gene[:5]]) print("gene_{}: {}, speed={}, fitness={}".format(gene_index, floats_adjusted, int(gene[5]), fitness)) def print_survivors(survivors, evaluations): print("selecting survivors:") for survivor_index, survivor in enumerate(survivors): print_gene(survivor, survivor_index, evaluations[survivor_index]) def run_eval_only(population): evaluation = [evaluate(gene) for gene in population] for gene_index, gene in enumerate(population): print_gene(gene, gene_index, evaluation[gene_index]) def main(population_size, ranges, max_generations=100, survivor_count=5, file_name=DATA_FILE): population = [] # randomly initialized population for index in range(population_size): population.append(get_random_gene(ranges)) generation = 1 print("max_generations={}, population_size={}, survivor_count={}".format(max_generations, population_size, survivor_count)) with open(file_name, "+w") as f: f.writelines("generation,max_fitness\n") while not terminate(max_generations, generation): print_generation(generation) # evaluate population evaluation = [evaluate(gene) for gene in population] # write best gene value to file f.writelines("{},{} \n".format(generation, max(evaluation))) for gene_index, gene in enumerate(population): print_gene(gene, gene_index, evaluation[gene_index]) # select changes the population and evaluation object survivors, surv_fitness = select(population=population, evaluations=evaluation, count=survivor_count) print_survivors(survivors=survivors, evaluations=surv_fitness) population = [] for survivor in survivors: population.append(survivor) for child_i in range(0, int(population_size/survivor_count) - 1): changed = False while not changed: new_gene = mutate_small(gene=survivor, ranges=ranges, mutation_probability=0.1, increment=0.01) new_gene = mutate_random(gene=new_gene, ranges=ranges, mutation_probability=0.05) if new_gene != survivor: changed = True population.append(new_gene) generation += 1 def plot(filename, plot_name): with open(filename) as fl: xs = [] ys = [] next(fl) for line in fl: x, y = line.strip().split(",") xs.append(int(x)) ys.append(float(y)) plt.plot(xs, ys) #plt.show() plt.savefig(plot_name) if __name__ == "__main__": # run_eval_only([[0.21, 1.56, 0.68, 0.53, 1.25, 120], \ # [0.75, 0.75, 0, 0, 1.5, 120], \ # [0.19, 1.56, 0.68, 0.53, 1.25, 120], \ # [0.23, 1.56, 0.68, 0.53, 1.25, 120], \ # [0.23, 1.56, 0.68, 0.53, 1.25, 110], \ # [0.60, 0.80, 0.1, 0.2, 1.5, 120]]) plot(DATA_FILE, PLOT_NAME) #main(population_size=20, ranges=ranges, max_generations=100, survivor_count=4)
joramwessels/torcs-client
mlp.py
<reponame>joramwessels/torcs-client # file: mlp.py # author: <NAME> # source: http://pytorch.org/tutorials/beginner/ # date: 15-11-2017 # description: # Trains a fully connected feed forward neural network with # a variable amount of units and layers. The model can be # trained using CUDA, saved, loaded, and used without CUDA # by calling model.predict([v1, v2, ..., vn]) # TODO Every training batch shows the same sequence of losses # There's no improvement among batches import sys, os import torch from torch.autograd import Variable learning_rate = 5e-7 epochs = 10 layers = 10 units = 10 allow_cuda = False use_cuda = torch.cuda.is_available() and allow_cuda class MLP(torch.nn.Module): def __init__(self, D_inp, D_hid, D_out, layers, x_scale, y_scale): """ Multilayer Perceptron with a variable amount of layers Args: D_inp: Dimension of the input layer D_hid: Dimension of all hidden layers D_out: Dimension of the output layer layers: The total amount of layers (2 means 1 hidden layer) x_scale: The maximum values for all input variables y_scale: The maximum values for all target variables """ super(MLP, self).__init__() self.input_layer = torch.nn.Linear(D_inp, D_hid) self.hidden_layer = torch.nn.Linear(D_hid, D_hid) self.output_layer = torch.nn.Linear(D_hid, D_out) self.layers = layers self.sigmoid = torch.nn.Sigmoid() self.x_scale = x_scale.cuda() if use_cuda else x_scale self.y_scale = y_scale.cuda() if use_cuda else y_scale def forward(self, x): """ Forward propagation: call to predict using current weights Args: x: Input tensor as a 'D_inp'-dimensional torch.autograd.Variable Returns: The predicted target value given this input """ h = self.input_layer(x) #h = self.sigmoid(h) for _ in range(self.layers-1): h = self.hidden_layer(h) #h = self.sigmoid(h) y_pred = self.output_layer(h) return y_pred.cuda() if use_cuda else y_pred def predict(self, x): """ Predicts a single variable, given as a list Args: x: The input variables, given as a normal list Returns: The prediction as a list """ x_var = Variable(torch.FloatTensor([x]), requires_grad=False) if use_cuda: x_norm = x_var.cuda() / self.x_scale else: x_norm = x_var / self.x_scale pred_var = self.forward(x_norm)[0] pred_var = pred_var * self.y_scale return [p.data[0] for p in pred_var] def train_model(x, y, metaparams): """ Trains a fully connected feed forward network Args: x: An autograd Variable with inputs (batch, n_samples, n_var) y: An autograd Variable with targets (batch, n_samples, n_var) metaparams: A dictionary including the fields d_inp: Dimension of the input layer d_hid: Dimension of all hidden layers d_out: Dimension of the output layer layers: The total amount of layers (2 means 1 hidden layer) x_max: The maximum values for all input variables y_max: The maximum values for all target variables Returns: A torch model object """ model = MLP(metaparams['d_inp'], metaparams['d_hid'], metaparams['d_out'], metaparams['layers'], metaparams['x_max'], metaparams['y_max']) if use_cuda: model = model.cuda() criterion = torch.nn.MSELoss(size_average=False) optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate) for t in range(epochs): print("Batch",t) for b in range(len(x)): # Forward y_pred = model(x[b]) loss = criterion(y_pred, y[b]) #print(t, loss.data[0]) print("Training loss:",loss.data[0]) # Backward optimizer.zero_grad() loss.backward() optimizer.step() return model def read_file(filename, tgt_ind, inp_ind, sep=',', skip_first_line=False): """ Reads the CSV training data Args: filename: The name of the CSV data file tgt_ind: A list of indices indicating the targets inp_ind: A list of indices indicating the inputs sep: The separator used in the CSV skip_first_line: Boolean indicating labels on the first line Returns: (x, y) tuple with 2D torch autograd Variables """ x = [] y = [] with open(filename) as file: if skip_first_line: next(file) for line in file: clean_line = line.strip().split(sep) if len(clean_line) > 1: params = [float(s) for s in clean_line if not s == ''] x.append([params[i] for i in inp_ind]) y.append([params[i] for i in tgt_ind]) x = Variable(torch.FloatTensor(x)) y = Variable(torch.FloatTensor(y), requires_grad=False) return (x.cuda(), y.cuda()) if use_cuda else (x, y) def read_all_files(folder, tgt_ind, inp_ind): """ Reads every file into a batch of the data Args: folder: The path to the folder with the data files tgt_ind: A list of indices indicating the targets inp_ind: A list of indices indicating the inputs Returns: Two lists of autograd Variables (batch, n_samples, n_var) """ xs, ys = [], [] for f in os.listdir(folder): file = os.path.join(folder, f) if os.path.isfile(file): x, y = read_file(file, tgt_ind, inp_ind) xs.append(x) ys.append(y) return xs, ys def load_model(filename, cuda=False): """ Loads a model from just the filename Args: filename: The name of the save file cuda: Set to True to make predictions use CUDA Returns: model """ global use_cuda use_cuda = cuda metaparams = torch.load(filename + ".meta") model = MLP(metaparams['d_inp'], metaparams['d_hid'], metaparams['d_out'], metaparams['layers'], metaparams['x_max'], metaparams['y_max']) model.load_state_dict(torch.load(filename, map_location=lambda storage, loc: storage)) return model.cuda() if use_cuda else model def normalize(x, y, metaparams): """ Normalizes the data by dividing all variables by their max Args: x: The x data (batch, n_samples, n_vars) y: The y data (batch, n_samples, n_vars) metaparams: The dictionary of metaparameters containing the max Returns: A tuple with the normalized x, y Variables """ x_max, y_max = metaparams['x_max'], metaparams['y_max'] if use_cuda: x_max, y_max = x_max.cuda(), y_max.cuda() y_max[y_max==0] = 1 # to prevent x_max[x_max==0] = 1 # division by zero for i in range(len(x)): for j in range(len(x[i])): x[i][j] = torch.div(x[i][j], x_max) y[i][j] = torch.div(y[i][j], y_max) return x, y def find_max(x): """ Finds the maximum value for each of the variables Args: x: The list of batches as autograd Variables Returns: An autograd Variable the size of a single datapoint """ x_max = Variable(torch.ones(len(x), len(x[0][0]))) for b in range(len(x)): x_max[b] = torch.max(x[b], 0)[0].data max_v = torch.max(x_max, 0)[0] return max_v def test_on_train_set(model, x, max_pred=500): """ """ for i in range(min(len(x), max_pred)): pred_y = list(model.predict(list(x[i].data)).data) print(i,"ang:%.2f, ste: %.2f" %(x[0].data, pred_y[0]), end='\n') def main(folder, save_as, targets, inputs): x, y = read_all_files(folder, targets, inputs) print("Read %i batches" %len(x)) metaparams = {'d_inp':len(x[0][0]), 'd_hid':units, 'd_out':len(y[0][0]), 'layers':layers, 'x_max':find_max(x), 'y_max':find_max(y)} print("Normalizing data...") x, y = normalize(x, y, metaparams) model = train_model(x[:12], y[:12], metaparams) print("Trained model for %i epochs" %epochs) # Always save as CPU model, cast to CUDA while loading if required torch.save(metaparams, save_as + ".meta") torch.save(model.float().state_dict(), save_as) print("Model saved as",save_as) #test_on_train_set(model, x[0]) return model if __name__ == "__main__": # targets: steerCmd targets = [2] # inputs: angle, speed_x, trackSens(0-18), distToMiddle inputs = [3] + [11] + list(range(14, 34)) model = main(sys.argv[1], sys.argv[2], targets, inputs) data_folder = "C:/Users/Joram/Documents/Studie/torcs-client/train_single/"
joramwessels/torcs-client
run_torcs.py
<filename>run_torcs.py<gh_stars>1-10 import argparse import xml.etree.ElementTree as ET import multiprocessing import subprocess from pytocl.main import main from combined_driver import Final_Driver import os import signal import time parser = argparse.ArgumentParser() parser.add_argument('--drivers', nargs='+', type=str) parser.add_argument('--tracks', nargs='+', type=str) parser.add_argument('--length', type=int) parser.add_argument('--laps', type=int) args = parser.parse_args() drivers = args.drivers tracks = args.tracks length = args.length laps = args.laps file_in = "quickrace.xml" our_driver = ["src_server 1"] all_drivers = ["berniw", "berniw3", "damned", "inferno", "lliaw", "tita", "berniw2", "bt", "inferno2", "olethros", "sparkle"] best_drivers = ["lliaw", "inferno", "olethros", "tita"] best_driver = ["lliaw"] all_tracks = ["aalborg", "alpine-1", "alpine-2", "brondehach", "corkscrew", "e-track-1", "e-track-2", "e-track-3", "e-track-4", "e-track-6", "eroad", "forza", "g-track-1", "g-track-2", "g-track-3", "ole-road-1", "ruudskogen", "spring", "street-1", "wheel-1", "wheel-2"] # if not tracks or tracks[0] == "all": # tracks = all_tracks # # if drivers[0] == "best": # drivers = best_driver # elif drivers[0] == "our": # drivers = our_driver def set_drivers(root, drivers, port_offset): xml_drivers = root[4] #Drivers section del xml_drivers[3] #a specific driver definition for index, driver in enumerate(drivers): ele = ET.Element("section", attrib={"name": str(index + 1)}) ele.append(ET.Element("attnum", attrib={"name": "idx", "val": str(port_offset)})) ele.append(ET.Element("attstr", attrib={"name": "module", "val": driver})) xml_drivers.append(ele) def set_track(root, track): xml_track = root[1] xml_track[1][0].attrib["val"] = track def write_and_run(tree, track, steering_values, max_speed, timeout, port): file_out = "tmp.quickrace.xml" # Write back to file tree.write(file_out) # Run TORCS # translate results # fitness = -running time*100 + distance done # print out result fitness steering_values = ", ".join([str(x) for x in steering_values]) command = "./run_evaluation.sh '{}' {} {} {}".format(steering_values, max_speed, timeout, port) #["./run_evaluation.sh", steering_values, str(max_speed), str(timeout)] completed_command = subprocess.run(command, shell=True, check=True, stdout=subprocess.PIPE) pids = list(completed_command.stdout.decode("utf-8").strip().split()) try: os.kill(pid[0], signal.SIGTERM) except: pass try: os.kill(pid[1], signal.SIGTERM) except: pass return completed_command def run_on_ea_tracks(driver, steering_values, max_speed, timeout): return run_on_tracks(driver, ["ole-road-1", "alpine-1", "forza"], steering_values, max_speed, timeout) def run_on_all_tracks(driver, steering_values, max_speed, timeout): return run_on_tracks(driver, all_tracks, steering_values, max_speed, timeout) def run_on_tracks(driver, tracks, steering_values, max_speed, timeout): # Open original file tree = ET.parse(file_in) root = tree.getroot() client = [] server = [] server_temp_file = "server.out" client_temp_file = "client.out" base_port = 3001 port_offset = 0 for track in tracks: port = base_port + port_offset set_drivers(root, [driver], port_offset) if os.path.isfile(server_temp_file): os.remove(server_temp_file) if os.path.isfile(client_temp_file): os.remove(client_temp_file) time.sleep(1) set_track(root, track) completed_command = write_and_run(tree, track, steering_values, max_speed, timeout, port) with open("server.out") as f: server.append(f.readlines()) with open("client.out") as f: client.append(list(f.readlines())) time.sleep(1) # we cycle through the ports port_offset += 1 if base_port + port_offset > 3009: port_offset = 0 return client, server def get_distance_covered(client_out): distance = -1 for index in range(len(client_out) - 1, 0, -1): if "dist" in client_out[index]: distance = int(float(client_out[index].strip().split()[2].split("=")[1])) break return distance def get_total_time_covered(client_out): time = -1 for index in range(len(client_out) - 1, 0, -1): if "time" in client_out[index]: time = int(float(client_out[index].strip().split()[3].split("=")[1])) break if time == -1: print(client_out) return time
joramwessels/torcs-client
my_driver.py
from pytocl.driver import Driver from pytocl.car import State, Command class MyDriver(Driver): def drive(self, carstate): command = Command() command.steering = carstate.angle / 180 if carstate.angle > 30 or carstate.angle < -30: command.brake = 0.5 command.accelerator = 0 else: command.brake = 0 command.accelerator = 1 command.gear = 1 return command
lankunyao/txt_to_json
main.py
<reponame>lankunyao/txt_to_json<filename>main.py # -*- coding: utf-8 -*- import argparse import os import json from pathlib import Path parser = argparse.ArgumentParser(description="命令行传入参数") parser.add_argument('-t', '--txt', default="SciKG_min_1.0") args = parser.parse_args() txt = args.txt root_dir = os.path.dirname(__file__) def main(): path = os.path.normcase(os.path.join(root_dir, "txt", txt)) print(path) for root, dirs, files in os.walk(path): for name in files: file_dir = os.path.join(path, name) f = open(file_dir, 'r', encoding='utf-8') data = f.read().encode('utf-8') data_dict = json.loads(data) data_json = json.dumps(data_dict, ensure_ascii=False) json_path = os.path.normcase(os.path.join(root_dir, "json", txt)) if not Path(json_path).is_dir(): os.mkdir(json_path) json_path = os.path.normcase(os.path.join(json_path, name[:-3]+"json")) w = open(json_path, 'wb') w.write(data_json.encode('utf-8')) w.close() f.close() if __name__ == '__main__': main()
bactopia/bactopia-parser
bactopia/__init__.py
"""Top-level package for Bactopia.""" __author__ = """<NAME> III""" __email__ = '<EMAIL>' __version__ = '0.1.1' from .parse import parse, get_bactopia_files, parse_bactopia_files
bactopia/bactopia-parser
bactopia/parse.py
""" Bactopia's parser entry-point. Example: bactopia.parse(result_type, filename) """ import errno import os from collections import OrderedDict from typing import Union from . import parsers from .const import RESULT_TYPES, IGNORE_LIST def parse(result_type: str, *files: str) -> Union[list, dict]: """ Use the result type to automatically select the appropriate parsing method for an input. Args: result_type (str): the type of results (e.g. assembly, mlst, qc, etc...) *files (str): one or more input files to be parsed Raises: FileNotFoundError: the input file could not be found ValueError: the result type is not an accepted type Returns: Union[list, dict]: The results parsed for a given input. """ if result_type in RESULT_TYPES: for f in files: if not os.path.exists(f): raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), f) return getattr(parsers, result_type).parse(*files) else: raise ValueError(f"'{result_type}' is not an accepted result type. Accepted types: {', '.join(RESULT_TYPES)}") def parse_genome_size(gs_file: str) -> int: """ Parse genome size from input file Args: gs_file (str): File containing the genome size of the sample Returns: int: genome size """ with open(gs_file, 'rt') as gs_fh: return int(gs_fh.readline().rstrip()) def _is_bactopia_dir(path: str, name: str) -> list: """ Check if a directory contains Bactopia output and any errors. Args: path (str): a path to expected Bactopia results name (str): the name of sample to test Returns: list: 0 (bool): path looks like Bactopia, 1 (list): any errors found """ from .parsers.error import ERROR_TYPES errors = [] is_bactopia = os.path.exists(f"{path}/{name}/{name}-genome-size.txt") for error_type in ERROR_TYPES: filename = f"{path}/{name}/{name}-{error_type}-error.txt" if os.path.exists(filename): is_bactopia = True errors.append(parsers.error.parse(filename)) return [is_bactopia, errors] def get_bactopia_files(path: str, name: str) -> dict: """ Build a list of all parsable Bactopia files. Args: path (str): a path to expected Bactopia results name (str): the name of sample to test Returns: dict: path and info on all parsable Bactopia files """ path = os.path.abspath(os.path.expanduser(os.path.expandvars(path))) is_bactopia, errors = _is_bactopia_dir(path, name) bactopia_files = OrderedDict() bactopia_files['has_errors'] = True if errors else False bactopia_files['errors'] = errors bactopia_files['ignored'] = False bactopia_files['message'] = "" bactopia_files['files'] = OrderedDict() if is_bactopia: if not errors: bactopia_files['genome_size'] = parse_genome_size(f"{path}/{name}/{name}-genome-size.txt") for result_type in RESULT_TYPES: result_key = result_type if result_type == "amr": result_key = "antimicrobial-resistance" elif result_type == "qc": result_key = "quality-control" if result_type not in ['error', 'generic', 'kmers']: bactopia_files['files'][result_key] = getattr(parsers, result_type).get_parsable_list(path, name) else: bactopia_files['ignored'] = True if name not in IGNORE_LIST: bactopia_files['message'] = f"'{path}/{name}' is does not look like Bactopia directory." else: bactopia_files['message'] = f"'{path}/{name}' is on the Bactopia ignore list." return bactopia_files def parse_bactopia_files(path: str, name: str) -> dict: """ Parse all results associated with an input sample. Args: path (str): a path to expected Bactopia results name (str): the name of sample to test Returns: dict: The parsed set of results associated with the input sample """ from bactopia.parsers.qc import is_paired bactopia_files = get_bactopia_files(path, name) bactopia_results = OrderedDict(( ('sample', name), ('genome_size', None), ('is_paired', None), ('has_errors', bactopia_files['has_errors']), ('errors', bactopia_files['errors']), ('has_missing', False), ('missing', []), ('ignored', bactopia_files['ignored']), ('message', bactopia_files['message']), ('results', OrderedDict()) )) if not bactopia_results['has_errors'] and not bactopia_results['ignored']: bactopia_results['genome_size'] = bactopia_files['genome_size'] for result_type, results in bactopia_files['files'].items(): bactopia_results['is_paired'] = is_paired(path, name) bactopia_results['results'][result_type] = OrderedDict() result_key = result_type if result_type == "-_resistance": result_key = "amr" elif result_type == "quality-control": result_key = "qc" for result in results: if result['missing']: if not result['optional']: bactopia_results['has_missing'] = True bactopia_results['missing'].append([result_type, result["files"]]) else: bactopia_results['results'][result_type][result['result_name']] = {} else: bactopia_results['results'][result_type][result['result_name']] = parse(result_key, *result['files']) return bactopia_results def parse_bactopia_directory(path: str) -> list: """ Scan a Bactopia directory and return parsed results. Args: path (str): a path to expected Bactopia results Returns: list: Parsed results for all samples in a Bactopia directory """ results = [] with os.scandir(path) as dirs: for directory in dirs: if directory.name not in IGNORE_LIST: sample = directory.name results.append(parse_bactopia_files(path, sample)) return results
bactopia/bactopia-parser
bactopia/parsers/__init__.py
<reponame>bactopia/bactopia-parser<gh_stars>1-10 import os import glob modules = glob.glob(os.path.join(os.path.dirname(__file__), "*.py")) __all__ = [os.path.basename(f)[:-3] for f in modules if os.path.isfile(f) and not f.endswith('__init__.py')] del modules del glob del os from . import *
bactopia/bactopia-parser
tests/__init__.py
<reponame>bactopia/bactopia-parser<gh_stars>1-10 """Unit test package for bactopia."""
bactopia/bactopia-parser
bactopia/parsers/qc.py
<reponame>bactopia/bactopia-parser """ Parsers for QC (FASTQ) related results. """ import os from .generic import get_file_type, parse_json RESULT_TYPE = 'quality-control' ACCEPTED_FILES = ["final.json", "original.json"] def parse(r1: str, r2: str = None) -> dict: """ Check input file is an accepted file, then select the appropriate parsing method. Args: r1 (str): input file associated with R1 or SE FASTQ r2 (str, optional): input file associated with R2 FASTQ. Defaults to None. Raises: ValueError: summary results to not have a matching origin (e.g. original vs final FASTQ) Returns: dict: parsed results """ filetype = get_file_type(ACCEPTED_FILES, r1) filetype2 = filetype if r2: filetype2 = get_file_type(ACCEPTED_FILES, r2) if r1.endswith(".json"): if r2 and filetype != filetype2: raise ValueError(f"Original and Final QC files were mixed. R1: {filetype}, R2: {filetype2}") return _merge_qc_stats(parse_json(r1), parse_json(r2)) if r2 else parse_json(r1) def _merge_qc_stats(r1: dict, r2: dict) -> dict: """ Merge appropriate metrics (e.g. coverage) for R1 and R2 FASTQs. Args: r1 (dict): parsed metrics associated with R1 FASTQ r2 (dict): parsed metrics associated with R2 FASTQ Returns: dict: the merged FASTQ metrics """ from statistics import mean merged = { 'qc_stats': {}, 'r1_per_base_quality': r1['per_base_quality'], 'r2_per_base_quality': r2['per_base_quality'], 'r1_read_lengths': r1['read_lengths'], 'r2_read_lengths': r2['read_lengths'] } for key in r1['qc_stats']: if key in ['total_bp', 'coverage', 'read_total']: merged['qc_stats'][key] = r1['qc_stats'][key] + r2['qc_stats'][key] if r2 else r1['qc_stats'][key] else: val = mean([r1['qc_stats'][key], r2['qc_stats'][key]]) if r2 else r1['qc_stats'][key] merged['qc_stats'][key] = f'{val:.4f}' if isinstance(val, float) else val return merged def is_paired(path: str, name: str) -> bool: """ Check if in input sample had paired-end or single-end reads Args: path (str): a path to expected Bactopia results name (str): the name of sample to test Raises: ValueError: Processed FASTQ(s) could not be found. Returns: bool: True: reads are paired, False: reads are single-end """ r1 = f"{path}/{name}/quality-control/{name}_R1.fastq.gz" r2 = f"{path}/{name}/quality-control/{name}_R2.fastq.gz" se = f"{path}/{name}/quality-control/{name}.fastq.gz" if os.path.exists(r1) and os.path.exists(r2): return True elif os.path.exists(se): return False else: raise ValueError(f"Processed FASTQs not found in {path}/{name}/quality-control/") def get_parsable_list(path: str, name: str) -> list: """ Generate a list of parsable files. Args: path (str): a path to expected Bactopia results name (str): the name of sample to test Returns: list: information about the status of parsable files """ import os parsable_results = [] for result in ACCEPTED_FILES: result_name = None filename = None r1 = None r2 = None se = None if result.endswith('original.json'): result_name = 'original' r1 = f"{path}/{name}/{RESULT_TYPE}/summary-original/{name}_R1-{result}" r2 = f"{path}/{name}/{RESULT_TYPE}/summary-original/{name}_R2-{result}" se = f"{path}/{name}/{RESULT_TYPE}/summary-original/{name}-{result}" elif result.endswith('final.json'): result_name = 'final' r1 = f"{path}/{name}/{RESULT_TYPE}/summary-final/{name}_R1-{result}" r2 = f"{path}/{name}/{RESULT_TYPE}/summary-final/{name}_R2-{result}" se = f"{path}/{name}/{RESULT_TYPE}/summary-final/{name}-{result}" if (se): if os.path.exists(se): parsable_results.append({ 'result_name': result_name, 'files': [se], 'optional': False, 'missing': False }) else: missing = True if os.path.exists(r1) and os.path.exists(r2): missing = False parsable_results.append({ 'result_name': result_name, 'files': [r1, r2], 'optional': False, 'missing': missing }) return parsable_results
bactopia/bactopia-parser
tests/test_bactopia.py
#!/usr/bin/env python """Tests for `bactopia` package.""" import unittest from bactopia import bactopia class TestBactopia_parser(unittest.TestCase): """Tests for `bactopia` package.""" def setUp(self): """Set up test fixtures, if any.""" def tearDown(self): """Tear down test fixtures, if any.""" def test_000_something(self): """Test something."""
bactopia/bactopia-parser
bactopia/const.py
<reponame>bactopia/bactopia-parser """ Constant values used throughout the Bactopia package. """ RESULT_TYPES = [ "amr", "annotation", "ariba", "assembly", "blast", "error", "generic", "kmers", "mapping", "minmers", "mlst", "qc", "variants" ] IGNORE_LIST = ['.nextflow', '.nextflow.log', 'bactopia-info', 'work', 'bactopia-tools']
vivatoviva/shell
first.py
import math def quadratic(a, b, c): # 首先计算代尔塔 temp = b**2 - 4*a*c if temp < 0: return 0,0 # 两个解为 x1 = (-b + math.sqrt(temp)) / 2*a x2 = (-b - math.sqrt(temp)) / 2*a return x1,x2 pass # 测试: print('quadratic(2, 3, 1) =', quadratic(2, 3, 1)) print('quadratic(1, 3, -4) =', quadratic(1, 3, -4)) if quadratic(2, 3, 1) != (-0.5, -1.0): print('测试失败') elif quadratic(1, 3, -4) != (1.0, -4.0): print('测试失败') else: print('测试成功')
Lornatang/TensorFlow-AE
main.py
# Copyright 2019 ChangyuLiu Authors. All Rights Reserved. # # Licensed under the MIT License. # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # https://opensource.org/licenses/MIT # ============================================================================== """The training loop begins with generator receiving a random seed as input. That seed is used to produce an image. The discriminator is then used to classify real images (drawn from the training set) and fakes images (produced by the generator). The loss is calculated for each of these models, and the gradients are used to update the generator and discriminator. """ from data.datasets import load_data from model.decoder import make_decoder_model from model.encoder import make_encoder_model from util.loss_and_optim import ae_loss, encoder_optimizer, decoder_optimizer from util.save_checkpoints import save_checkpoints from util.generate_and_save_images import generate_and_save_images import tensorflow as tf import time import os import argparse parser = argparse.ArgumentParser() parser.add_argument('--epochs', default=50, type=int, help='Epochs for training.') args = parser.parse_args() print(args) # define model save path save_path = './training_checkpoints' BUFFER_SIZE = 60000 BATCH_SIZE = 128 num_examples_to_generate = 16 noise_dim = 64 # create dir if not os.path.exists(save_path): os.makedirs(save_path) # define random noise seed = tf.random.normal([num_examples_to_generate, noise_dim]) # load dataset train_dataset = load_data(BUFFER_SIZE, BATCH_SIZE) # load network and optim paras decoder = make_decoder_model() decoder_optimizer = decoder_optimizer() encoder = make_encoder_model() encoder_optimizer = encoder_optimizer() checkpoint_dir, checkpoint, checkpoint_prefix = save_checkpoints(decoder, encoder, decoder_optimizer, encoder_optimizer, save_path) # This annotation causes the function to be "compiled". @tf.function def train_step(images): """ break it down into training steps. Args: images: input images. """ with tf.GradientTape() as de_tape, tf.GradientTape() as en_tape: z = encoder(images) _x = decoder(z) ae_of_loss = ae_loss(images, _x) gradients_of_decoder = de_tape.gradient(ae_of_loss, decoder.trainable_variables) gradients_of_encoder = en_tape.gradient(ae_of_loss, encoder.trainable_variables) decoder_optimizer.apply_gradients( zip(gradients_of_decoder, decoder.trainable_variables)) encoder_optimizer.apply_gradients( zip(gradients_of_encoder, encoder.trainable_variables)) def train(dataset, epochs): """ train op Args: dataset: mnist dataset or cifar10 dataset. epochs: number of iterative training. """ for epoch in range(epochs): start = time.time() for image_batch in dataset: train_step(image_batch) # Produce images for the GIF as we go generate_and_save_images(decoder, epoch + 1, seed, save_path) # Save the model every 15 epochs if (epoch + 1) % 15 == 0: checkpoint.save(file_prefix=checkpoint_prefix) print(f'Time for epoch {epoch + 1} is {time.time() - start:.3f} sec.') # Generate after the final epoch generate_and_save_images(decoder, epochs, seed, save_path) if __name__ == '__main__': train(train_dataset, args.epochs)
Lornatang/TensorFlow-AE
util/loss_and_optim.py
# Copyright 2019 ChangyuLiu Authors. All Rights Reserved. # # Licensed under the MIT License. # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # https://opensource.org/licenses/MIT # ============================================================================== """Generate optim loss and Discriminate optim loss""" import tensorflow as tf def ae_loss(noise, fake_output): """ Automatic coding loss function Args: noise: . fake_output: generate pic for use encoder model Returns: tf.reduce_mean. """ return tf.reduce_mean(tf.square(noise - fake_output)) def decoder_optimizer(): """ The training generator optimizes the network. Returns: tf.optimizers.Adam. """ return tf.optimizers.Adam() def encoder_optimizer(): """ The training discriminator optimizes the network. Returns: tf.optimizers.Adam. """ return tf.optimizers.Adam()
Lornatang/TensorFlow-AE
util/save_checkpoints.py
# Copyright 2019 ChangyuLiu Authors. All Rights Reserved. # # Licensed under the MIT License. # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # https://opensource.org/licenses/MIT # ============================================================================== """save model function""" import os import tensorflow as tf def save_checkpoints(generator, discriminator, generator_optimizer, discriminator_optimizer, save_path): """ save gan model Args: generator: generate model. discriminator: discriminate model. generator_optimizer: generate optimizer func. discriminator_optimizer: discriminator optimizer func. save_path: save gan model dir path. Returns: checkpoint path """ checkpoint_dir = save_path checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt") checkpoint = tf.train.Checkpoint(generator_optimizer=generator_optimizer, discriminator_optimizer=discriminator_optimizer, generator=generator, discriminator=discriminator) return checkpoint_dir, checkpoint, checkpoint_prefix
danya02/booru-mirror-unified
danbooru_dump/danbooru_dump_bulkdata_download_missing.py
<filename>danbooru_dump/danbooru_dump_bulkdata_download_missing.py import subprocess import os import sys sys.path.append('..') from database import * import time import random import danbooru_dump_transfer_queue_into_redis_from_find os.chdir('/hugedata/booru/danbooru_temp') def download_content(id): print (['rsync', '--recursive', '--verbose', f'rsync://172.16.31.10:873/danbooru2020/original/0{str(id)[-3:].zfill(3)}/{id}.*', './danbooru2020/original/']) subprocess.run(['rsync', '--recursive', '--verbose', f'rsync://172.16.31.10:873/danbooru2020/original/0{str(id)[-3:].zfill(3)}/{id}.*', './danbooru2020/original/']) while 1: #time.sleep(3) query = Post.select().where(~fn.EXISTS(Content.select().where(Content.post_id == Post.id))).where(Post.board == Imageboard.get(Imageboard.name == 'danbooru')).order_by(random.choice([Post.id, Post.local_id])).limit(200) for post in query: print(post.__data__) download_content(post.local_id) danbooru_dump_transfer_queue_into_redis_from_find.find_transfer()
danya02/booru-mirror-unified
danbooru_dump/danbooru_dump_redis_load_json.py
<filename>danbooru_dump/danbooru_dump_redis_load_json.py import os import json import redis import time info = redis.Redis(db=12) time_start = time.time() count = 0 os.chdir('/hugedata/booru_old/danbooru2020/danbooru2020/metadata/split') for i in sorted(os.listdir('.')): while info.dbsize() > 100000: print('too many files', info.dbsize()) time.sleep(60) i = i.strip() print('read', i) with open(i) as handle: for row in handle: row = row.strip() id = json.loads(row)['id'] info.set(id, row) print('delete', i) os.unlink(i)
danya02/booru-mirror-unified
queue_work.py
<reponame>danya02/booru-mirror-unified from database import * import functools import traceback ENTITY_IMPORT_METHODS = dict() def to_import_entity(imageboard, ent_type): def wrapper(func): @functools.wraps(func) def wrapped(queued_ent): queued_ent.row_locked = True queued_ent.save() try: props = func(queued_ent.entity_local_id, additional_data=queued_ent.additional_data) queued_ent.report_success(and_mark_as_final=props.get('MARK_FINAL', False)) return True, ret_val except: queued_ent.report_error(traceback.print_exc()) traceback.print_exc() return False, None ENTITY_IMPORT_METHODS[imageboard] = ENTITY_IMPORT_METHODS.get(imageboard, dict()) ENTITY_IMPORT_METHODS[imageboard].update({ent_type: wrapped}) return wrapped return wrapper def import_single_entity(): with db.atomic(): task = QueuedImportEntity.tasks_query().get() print('Taken task: ' + str(task)) result, props = ENTITY_IMPORT_METHODS[task.board.name][task.entity_type.name](task) print('Succeded' if result else 'Failed')
danya02/booru-mirror-unified
danbooru_dump/danbooru_dump_import.py
import sys sys.path.append('..') from booru_db import BooruDatabase from database import * import datetime import json import traceback import random import time import logging import redis import collections r = redis.Redis(db=12) bdb = BooruDatabase('danbooru') def parse_date(val): try: v = datetime.datetime.strptime(val, '%Y-%m-%d %H:%M:%S %Z') except: val = val.split(' ') val[1] = val[1] + '0' val = ' '.join(val) v = datetime.datetime.strptime(val, '%Y-%m-%d %H:%M:%S.%f %Z') if v == datetime.datetime(1970, 1, 1, 0, 0): return None return v def insert_row(struct, skip_if_exists=False): print(struct['id']) post = Post() post_existing = bdb.post[ int(struct['id']) ] if skip_if_exists and (post_existing is not None): return None if post_existing: post = post_existing #print(post, post.__data__, flush=True) post.local_id = struct['id'] #print(post.post_created_at, parse_date(struct['created_at'])) #print(post.post_updated_at, parse_date(struct['updated_at'])) #input() post.post_created_at = parse_date(struct['created_at']) post.post_updated_at = parse_date(struct['updated_at']) post.row_updated_at = datetime.datetime.now() post.rating = struct['rating'] post.source = struct['source'] or None post.score = struct['score'] post.parent_local_id = int(struct['parent_id']) or None post.uploaded_by = User.get_or_create(board=bdb.booru, local_id=int(struct['uploader_id']))[0] bdb.post[ int(struct['id']) ] = post post = bdb.post[ int(struct['id']) ] #print(post, post.__data__, flush=True) old_tags = sorted(bdb.tag[post]) new_tags = sorted([tag['name'] for tag in struct['tags']]) if old_tags != new_tags: #print('updating tag set', old_tags, new_tags) bdb.tag[post] = new_tags else: pass #print('skipped tag set') postfavs_existing = set( [i[0] for i in PostFavs.select(User.local_id).join(User).where(PostFavs.post==post).tuples()] ) postfavs_new = set([int(i) for i in struct['favs']]) diff_set = postfavs_existing.symmetric_difference(postfavs_new) if diff_set: #print('fav set difference so updating', postfavs_existing, '->', postfavs_new) fav_users = dict() for i in User.select().where(User.board == bdb.booru).where(User.local_id.in_([int(i) for i in struct['favs']])): fav_users[i.local_id] = i PostFavs.delete().where(PostFavs.post == post).execute() postfavs_list = [] for fav in struct['favs']: fav = int(fav) if fav not in fav_users: fav_users[fav] = User.get_or_create(board=bdb.booru, local_id=int(fav))[0] postfavs_list.append(PostFavs(post=post, user=fav_users[fav])) PostFavs.bulk_create(postfavs_list) danpost = DanbooruPostMetadata.get_or_none(post=post) or DanbooruPostMetadata(post=post) #print(danpost.__data__) danpost.up_score = int(struct['up_score']) danpost.down_score = int(struct['up_score']) danpost.pixiv_id = int(struct['pixiv_id']) or None danpost.approved_by = User.get_or_create(board=bdb.booru, local_id=int(struct['approver_id']))[0] danpost.is_pending = struct['is_pending'] danpost.is_flagged = struct['is_flagged'] danpost.is_deleted = struct['is_deleted'] danpost.is_banned = struct['is_banned'] danpost.is_status_locked = struct['is_status_locked'] danpost.is_note_locked = struct['is_note_locked'] try: danpost.save(force_insert=True) except IntegrityError: danpost.save() imgpost = ImageMetadata.get_or_none(post=post) or ImageMetadata(post=post) #print(imgpost.__data__) imgpost.image_width = int(struct['image_width']) imgpost.image_height = int(struct['image_height']) imgpost.file_size = int(struct['file_size']) imgpost.md5 = struct['md5'] try: imgpost.save(force_insert=True) except IntegrityError: imgpost.save() def insert_row_atomic(*args, **kwargs): with db.atomic(): return insert_row(*args, **kwargs) #insert_row_atomic(json.loads(input())) class CountHandler(logging.Handler): def __init__(self): self.count = 0 super().__init__() def emit(self, record): self.count += 1 def handle(self, record): self.count += 1 def reset(self): self.count = 0 counter = CountHandler() logger = logging.getLogger('peewee') logger.addHandler(counter) logger.setLevel(logging.DEBUG) class DanbooruDumpRow(MyModel): post_id = IntegerField() content = TextField() DANBOORU = Imageboard.get(name='danbooru') POST_ENTITY, _ = EntityType.get_or_create(name='post') when = time.time() #for row in DanbooruDumpRow.select().order_by(fn.Rand()): def infinite_counter(): i = 0 while 1: i += 1 yield i start = time.time() try: start_times = collections.deque(maxlen=50) to_delete = collections.deque() for ind in db.batch_commit(infinite_counter(), 100): # count = DanbooruDumpRow.select(fn.COUNT(DanbooruDumpRow.id)).scalar() # if not count: # print('All done!!!') # break # row = DanbooruDumpRow.select().where(DanbooruDumpRow.id >= random.randint(0, count)).get() start_times.append(time.time()) rid = r.randomkey() row = r.get(rid) # print('selecting row', rid, 'took', time.time()-when, 'seconds and', counter.count, 'queries') when = time.time() insert_row_atomic(json.loads(row)) print('inserting took', time.time()-when, 'seconds and', counter.count, 'queries') when = time.time() ImportedEntity.get_or_create(entity_type=POST_ENTITY, board=DANBOORU, entity_local_id=rid, final=False) to_delete.append(rid) while len(to_delete) > 500: v = to_delete.popleft() print('deleting from redis', v) r.delete(v) # print('removing took', time.time()-when, 'seconds and', counter.count, 'queries') counter.reset() when = time.time() print('Current rate:', len(start_times) / (when - start_times[0])) print() except KeyboardInterrupt: input('Stopped, enter to continue or another ^C to quit')
danya02/booru-mirror-unified
danbooru_dump/danbooru_dump_bulkdata_transfer.py
<filename>danbooru_dump/danbooru_dump_bulkdata_transfer.py import sys sys.path.append('..') from database import * import traceback import os import shutil #import threading import queue import time import magic import redis import functools import concurrent.futures r = redis.Redis() rc = redis.Redis(db=2) #rhashes_mimes = redis.Redis(db=2) BOARD, _ = Imageboard.get_or_create(name='danbooru', base_url='https://danbooru.donmai.us') started_work_at = time.time() jobs_executed = 0 find_files = False executor = concurrent.futures.ThreadPoolExecutor() <EMAIL>(maxsize=8192) def get_post(id): return Post.get_or_none(Post.local_id == id, Post.board == BOARD) #@create_table class FileToImportTMP(MyModel): path = CharField(unique=True) locked = BooleanField(default=False) @functools.lru_cache(maxsize=None) def get_mimetype(name, ext): mimetype_row = MimeType.get_or_none(MimeType.name == name) if mimetype_row is None: mimetype_row = MimeType.create(name=name, ext=ext) return mimetype_row def time_it(fn): def exec(*args, **kwargs): started = time.time() res = fn(*args, **kwargs) return time.time()-started, res return exec def data_import(path, thread=None, select_row_time=None): started_at = time.time() n=6 ext = path.split('.')[-1] read_start = time.time() id = int( path.split('/')[-1].split('.')[0] ) post = get_post(id) if post is None: raise Post.DoesNotExist res = rc.get(path) if res is not None: res = str(res, 'utf-8') sha256, mt_str = res.split(' ', 1) read_time = time.time() - read_start compute_time = 0 else: @time_it def read_data(path=path): data = rc.get(path) if data is None: handle = open(path, 'rb') data = handle.read() handle.close() if data == b'': print('!!!! file at', path, 'is empty!!') return None, None return id, data read_time, a = read_data() id, data = a if id is None: return @time_it def get_mimetype_from_file(data=data): mimetype = magic.from_buffer(data, mime=True) return mimetype mimetype_get_time, mt_str = get_mimetype_from_file() compute_time = mimetype_get_time @time_it def hashit(data=data): return sha256_hash(data) hash_time, sha256 = hashit() compute_time += hash_time if os.path.exists(get_path(sha256)): raise FileExistsError() print(f'id={id}\tread={round(read_time, n)}', end='\t') print(f'compute={round(compute_time, n)}', end='\t') @time_it def db_ops(id=id, mt_str=mt_str, sha256=sha256): mt_row = get_mimetype(mt_str, ext) try: c = Content.get(post=post) c.sha256_current=sha256 c.mimetype_id=mt_row.id c.file_size_current=os.path.getsize(path) c.save() except Content.DoesNotExist: c = Content.create(post=post, sha256_current=sha256, mimetype_id=mt_row.id, file_size_current=os.path.getsize(path)) return mt_row db_ops_time, mt_row = db_ops() @time_it def save(mt_row=mt_row): ensure_dir(sha256) #print(path, '->', IMAGE_DIR+get_path(sha256)) shutil.move(path, IMAGE_DIR+get_path(sha256)+'.'+mt_row.ext) return None save_file_time, _ = save() print(f'db_ops={round(db_ops_time, n)}', f'save={round(save_file_time, n)}', sep='\t', end='\t') ended_at = time.time() global jobs_executed global time_running time_on_job = ended_at - started_at jobs_executed += 1 avg_time = (time.time() - started_work_at) / jobs_executed print(f'time={round(time_on_job, n)}', f'avg={round(avg_time,n)}', f'jobs={jobs_executed}', sep='\t') path = '/hugedata/booru/danbooru2020/danbooru2020/original' count = 0 if __name__ == '__main__': while True: count += 1 if count > 10000: #get_path.cache_clear() count = 0 started_at = time.time() path = r.randomkey() if path is None: print('out of jobs! waiting...') time.sleep(1) continue path = str(path, 'utf8') try: data_import(path, select_row_time=time.time()-started_at) r.delete(path) except FileNotFoundError: traceback.print_exc() r.delete(path) except Post.DoesNotExist: r.delete(path) except: traceback.print_exc() continue
danya02/booru-mirror-unified
danbooru_dump/danbooru_dump_transfer_queue_into_redis_from_find.py
<filename>danbooru_dump/danbooru_dump_transfer_queue_into_redis_from_find.py<gh_stars>0 import sys sys.path.append('..') from database import * import sys import os import redis import time r = redis.Redis() def find_transfer(): os.system('find /hugedata/booru/danbooru_temp -type f -empty -print -delete') f = os.popen('find /hugedata/booru/danbooru_temp -type f') for row in f: row = row.strip() if row.split('/')[-1].startswith('.'): continue print(row, file=sys.stderr) r.set(row, '0') if __name__ == '__main__': while 1: find_transfer() time.sleep(20)
danya02/booru-mirror-unified
mv_hash_tree_files.py
from database import * import os import shutil import time import humanize import datetime import redis import tqdm import traceback queue = redis.Redis(db=5) os.chdir('/hugedata/booru_old/unified/files') # copy from cwd = os.getcwd() TARGET = '/hugedata/booru/' # copy to if input('input Q to enqueue paths, anything else to move them').lower() == 'q': search = os.popen('find -type f') for path in search: path = path.strip() path = cwd + path[1:] print(path) queue.set(path, 0) else: dbsize = queue.dbsize() dbsize_first = dbsize time_started = time.time() while dbsize: current_time = time.time() time_running = current_time - time_started changed_dbsize = dbsize - dbsize_first speed = changed_dbsize / time_running if speed > 0: eta = 'inf' human_eta = 'inf' else: eta = dbsize / (speed or 0.1) eta = -eta eta_td = datetime.timedelta(seconds=eta) human_eta = humanize.precisedelta(eta_td, minimum_unit='seconds') eta = round(eta) print('cursize:', dbsize, 'change:', changed_dbsize, 'speed:', speed, 'it/s', 'eta:', eta, 'seconds or', human_eta) path = queue.randomkey() path = str(path, 'utf8') filename = path.split('/')[-1] ensure_dir(filename) target_path = TARGET + get_path(filename) try: shutil.move(path, target_path) queue.delete(path) except: traceback.print_exc() dbsize = queue.dbsize()
danya02/booru-mirror-unified
booru_db.py
<reponame>danya02/booru-mirror-unified from database import * import functools class BooruDatabase: def __init__(self, booru): if not isinstance(booru, Imageboard): if isinstance(booru, str): booru = Imageboard.get(Imageboard.name == booru) else: booru = Imageboard.get_by_id(booru) else: pass self.booru = booru @property def post(self): return PostDatabase(self.booru) @property def tag(self): return TagDatabase(self.booru) # @property # def comment(self): # return CommentDatabase(self.booru) @property def user(self): return UserDatabase(self.booru) class PostDatabase: def __init__(self, booru): self.booru = booru @functools.lru_cache(maxsize=32) def __getitem__(self, id): return Post.get_or_none(Post.local_id==id, Post.board==self.booru) def __setitem__(self, id, model): if not isinstance(model, Post): raise TypeError model.board = self.booru model.local_id = id try: model.save(force_insert=True) except IntegrityError: model.save() class TagDatabase: def __init__(self, booru): pass @staticmethod def __getitem__(item): if isinstance(item, int) or isinstance(item, Post): if isinstance(item, int): item = Post.select(Post.id).where(Post.id == item).where(Post.board == self.booru).get() return set( [i.name for i in Tag.select(Tag.name).join(PostTag).where(PostTag.post == item)] ) elif isinstance(item, str): return Tag.get_or_none(Tag.name == item) else: raise TypeError @staticmethod @functools.lru_cache(maxsize=1024) def get_tag(name): try: return Tag.get(Tag.name == name) except Tag.DoesNotExist: return Tag.create(name=name) @staticmethod def __setitem__(item, value): if isinstance(item, Post): if isinstance(item, int): item = Post.select(Post.id).where(Post.id == item).where(Post.board == self.booru).get() if not isinstance(value, list): raise TypeError with db.atomic(): subquery_tag_ids = PostTag.select(PostTag.tag_id).where(PostTag.post == item) #TagPostCount.update(post_count = TagPostCount.post_count - 1, changed_at=fn.Now()).where(TagPostCount.board == item.board).where(TagPostCount.tag.in_(subquery_tag_ids)).execute() PostTag.delete().where(PostTag.post == item).execute() post_tags = [] for tagname in value: post_tags.append(PostTag(post=item, tag=TagDatabase.get_tag(tagname))) # TagPostCount.get_or_create(board=item.board, tag=TagDatabase.get_tag(tagname)) try: PostTag.bulk_create(post_tags) except IntegrityError: for i in post_tags: try: i.save(force_insert=True) except IntegrityError: pass #TagPostCount.update(post_count = TagPostCount.post_count + 1, changed_at=fn.Now()).where(TagPostCount.board == item.board).where(TagPostCount.tag.in_(subquery_tag_ids)).execute() else: raise TypeError class UserDatabase: def __init__(self, booru): self.booru = booru def __getitem__(self, item): if isinstance(item, int): return User.get_or_none(User.local_id==item, User.board==self.booru) else: raise TypeError def __setitem__(self, item, value): if not isinstance(item, int): raise TypeError if not isinstance(value, User): raise TypeError value.local_id = item value.board = self.booru try: value.save(force_insert=True) except IntegrityError: value.save()
danya02/booru-mirror-unified
create_thumbnails.py
<reponame>danya02/booru-mirror-unified<filename>create_thumbnails.py from database import * import subprocess import os import datetime import random mimetype_jpg = MimeType.get(MimeType.ext == 'jpg') mimetype_png = MimeType.get(MimeType.ext == 'png') mts = [mimetype_jpg, mimetype_png] mtids = [mt.id for mt in mts] def create_thumbnail(content, skip_existing=True): print(content) if content.mimetype_id not in mtids: return if len(content.thumbnail) and skip_existing: return original_path = IMAGE_DIR + get_path(content.sha256_current) + '.' + content.mimetype.ext thumb_path = IMAGE_DIR + get_thumbnail_path(content.sha256_current) + '.' + content.mimetype.ext old_path = IMAGE_DIR + 'old/' + get_path(content.sha256_current) old_ext_path = IMAGE_DIR + 'old_with_exts/' + get_path(content.sha256_current) + '.' + content.mimetype.ext if not os.path.isfile(thumb_path) or not skip_existing: ensure_dir(content.sha256_current) cmd = ['convert', original_path, '-thumbnail', '256x256', thumb_path] print(cmd) subprocess.run(cmd) size = os.path.getsize(thumb_path) ct = ContentThumbnail.get_or_none(content=content, mimetype=content.mimetype) if ct is None: ct = ContentThumbnail.create(content=content, mimetype=content.mimetype, size=size) else: ct.size = size ct.generated_at = datetime.datetime.now() ct.save() else: print('skipping', content) # try: # open(old_path).close() if os.path.exists(old_path): print('deleted', old_path) os.unlink(old_path) if os.path.exists(old_ext_path): print('deleted', old_ext_path) os.unlink(old_ext_path) # except FileNotFoundError: pass page = 0 iterated = True while iterated: print('============= PAGE', page, '===============') #iterated = False try: for c in Content.select().where(~fn.EXISTS(ContentThumbnail.select().where(ContentThumbnail.content_id==Content.post_id))).join(Post).order_by(Post.local_id).iterator(): try: with db.atomic(): create_thumbnail(c, skip_existing=False) except: continue iterated = True except IntegrityError: continue page += 1
danya02/booru-mirror-unified
database.py
<reponame>danya02/booru-mirror-unified<filename>database.py from peewee import * import peewee import datetime from PIL import Image import hashlib import io import os import random import hashlib IMAGE_DIR = '/hugedata/booru/' DIR_TREE_DEPTH = 2 DIR_TREE_SEGMENT_LEN = 2 #db = SqliteDatabase('test.db', timeout=600) db = MySQLDatabase('unifiedbooru', user='booru', password='<PASSWORD>', host='10.0.0.128') class MediumBlobField(BlobField): field_type = 'MEDIUMBLOB' content_databases = dict() def sha256_hash(data): hasher = hashlib.sha256() hasher.update(data) return hasher.hexdigest() def get_dir(name): path = '' cursor = DIR_TREE_SEGMENT_LEN for ind in range(DIR_TREE_DEPTH): path += name[:cursor] + '/' cursor += DIR_TREE_SEGMENT_LEN return path def get_path(name): return get_dir(name) + name def get_thumbnail_path(name): return 'thumbnails/' + get_dir(name) + name def ensure_dir(name): target_dir = get_dir(name) os.makedirs(IMAGE_DIR + target_dir, exist_ok=True) os.makedirs(IMAGE_DIR + 'thumbnails/' + target_dir, exist_ok=True) class File: @staticmethod def save_file(data, ext, raise_if_exists=True, but_check_for_same=False): name = sha256_hash(data) ensure_dir(name) path_no_ext = IMAGE_DIR + get_path(name) path = path_no_ext + '.' + ext return_value = None if (os.path.isfile(path) or os.path.isfile(path_no_ext)) and raise_if_exists: if not but_check_for_same: raise FileExistsError('file with name', path, '(maybe omitting extension) already exists') else: if os.path.isfile(path): with open(path, 'rb') as handle: exist_data = handle.read() if data == exist_data: return_value = name else: raise FileExistsError('file with name', path, 'already exists and is different from new data') if os.path.isfile(path_no_ext): with open(path_no_ext, 'rb') as handle: exist_data = handle.read() if data == exist_data: print(path_no_ext, 'renamed to', path) os.rename(path_no_ext, path) return_value = name else: raise FileExistsError('file with name', path_no_ext, 'already exists and is different from new data') with open(path, 'wb') as handle: handle.write(data) return name import logging logger = logging.getLogger('peewee') logger.addHandler(logging.StreamHandler()) logger.setLevel(logging.DEBUG) class ForeignKeyField(peewee.ForeignKeyField): @property def field_type(self): return self.rel_field.field_type.replace('BIGAUTO', 'BIGINT').replace('AUTO_INCREMENT', '').replace('AUTO', 'INTEGER') class MyModel(Model): class Meta: database = db legacy_table_names = False def create_table(cls): db.create_tables([cls]) return cls class TinyIntegerField(SmallIntegerField): field_type = 'TINYINT UNSIGNED' class TinyIntegerAutoField(BigAutoField): field_type = 'TINYINT UNSIGNED AUTO_INCREMENT' auto_increment = True class UnsignedSmallIntegerField(IntegerField): field_type = 'SMALLINT UNSIGNED' class SmallIntegerAutoField(BigAutoField): field_type = 'SMALLINT UNSIGNED AUTO_INCREMENT' auto_increment = True class UnsignedIntegerField(IntegerField): field_type = 'INTEGER UNSIGNED' class UnsignedAutoField(AutoField): field_type = 'INTEGER UNSIGNED AUTO_INCREMENT' auto_increment = True @create_table class Imageboard(MyModel): id = TinyIntegerAutoField(primary_key=True) name = CharField(unique=True) base_url = CharField(unique=True) def Board(backref=None, null=None): return ForeignKeyField(Imageboard, index=True, backref=backref, null=null) @create_table class User(MyModel): id = UnsignedAutoField(primary_key=True) board = Board('users') local_id = UnsignedIntegerField(index=True, null=True) username = CharField(null=True) row_created_at = DateTimeField(index=True, default=datetime.datetime.now) row_updated_at = DateTimeField(index=True, default=datetime.datetime.now) class Meta: indexes = ( (('board', 'local_id'), True), ) @create_table class UserJoinDate(MyModel): user = ForeignKeyField(User, primary_key=True) join_date = DateTimeField() @create_table class AccessLevel(MyModel): id = SmallIntegerAutoField(primary_key=True) name = CharField(unique=True) @create_table class UserAccessLevel(MyModel): user = ForeignKeyField(User, primary_key=True) level = ForeignKeyField(AccessLevel) @create_table class Tag(MyModel): id = UnsignedAutoField(primary_key=True) name = CharField(unique=True) created_at = DateTimeField(index=True, default=datetime.datetime.now) @create_table class TagPostCount(MyModel): tag = ForeignKeyField(Tag, index=True) board = Board('tag_counts', null=True) post_count = UnsignedIntegerField(index=True, default=0) changed_at = DateTimeField(index=True, default=datetime.datetime.now) @classmethod def set(cls, tag=None, board=None, post_count=None): if tag is None or post_count is None: raise TypeError('tag and post_count cannot be None') with db.atomic() as tx: board_expr = cls.board == board if board is None: board_expr = cls.board.is_null(True) changed = cls.update(post_count = post_count, changed_at=datetime.datetime.now()).where(board_expr).where(cls.tag == tag).execute() #print(tag, board, post_count, changed) if changed > 1: tx.rollback() raise ValueError('Too many rows changed by query -- wtf?? tag is ', tag, ' board is ', board, ' post_count is ', post_count) elif changed == 1: return else: cls.create(board=board, tag=tag, post_count=post_count, changed_at=datetime.datetime.now()) class Meta: indexes = ( (('board', 'tag'), True), ) @create_table class Type(MyModel): id = TinyIntegerAutoField(primary_key=True) name = CharField(unique=True) created_at = DateTimeField(index=True, default=datetime.datetime.now) @create_table class TagType(MyModel): board = Board('tag_types') tag = ForeignKeyField(Tag, backref='types') type = ForeignKeyField(Type, backref='tags') class Meta: primary_key = CompositeKey('board', 'tag', 'type') @create_table class Post(MyModel): id = UnsignedAutoField(primary_key=True) board = Board('posts') local_id = UnsignedIntegerField(index=True) row_created_at = DateTimeField(default=datetime.datetime.now) row_updated_at = DateTimeField(index=True, default=datetime.datetime.now) post_created_at = DateTimeField(index=True) post_updated_at = DateTimeField(index=True, null=True) uploaded_by = ForeignKeyField(User, null=True) source = TextField(null=True) rating = CharField(max_length=1) score = IntegerField(index=True) parent_local_id = UnsignedIntegerField(null=True) parent = ForeignKeyField('self', null=True) class Meta: indexes = ( (('board', 'local_id'), True), ) @create_table class PostFavs(MyModel): user = ForeignKeyField(User) post = ForeignKeyField(Post) class Meta: indexes = ( (('post', 'user'), True), ) primary_key = CompositeKey('user', 'post') @create_table class ImageMetadata(MyModel): post = ForeignKeyField(Post, primary_key=True) image_width = UnsignedIntegerField(index=True) image_height = UnsignedIntegerField(index=True) file_size = IntegerField(index=True, null=True) md5 = CharField(index=True, max_length=32) @create_table class PreviewSizeInfo(MyModel): post = ForeignKeyField(Post, primary_key=True) sample_width = UnsignedIntegerField() sample_height = UnsignedIntegerField() preview_width = UnsignedSmallIntegerField() preview_height = UnsignedSmallIntegerField() @create_table class ImageURL(MyModel): post = ForeignKeyField(Post, primary_key=True) img_url = CharField() sample_url = CharField() preview_url = CharField() @create_table class Status(MyModel): id = TinyIntegerAutoField(primary_key=True) value = CharField(unique=True) @create_table class PostStatus(MyModel): post = ForeignKeyField(Post, primary_key=True) status = ForeignKeyField(Status) @create_table class DanbooruPostMetadata(MyModel): post = ForeignKeyField(Post, primary_key=True) up_score = IntegerField(null=True) down_score = IntegerField(null=True) props = BitField() is_rating_locked = props.flag() is_status_locked = props.flag() is_pending = props.flag() is_flagged = props.flag() is_deleted = props.flag() is_banned = props.flag() pixiv_id = IntegerField(null=True) last_commented_at = DateTimeField(null=True) last_noted_at = DateTimeField(null=True) approved_by = ForeignKeyField(User) @create_table class MimeType(MyModel): id = TinyIntegerAutoField(primary_key=True) name = CharField(unique=True) ext = CharField(index=True, max_length=16) @create_table class Content(MyModel): post = ForeignKeyField(Post, backref='content', primary_key=True) sha256_current = FixedCharField(unique=True, max_length=64) sha256_when_acquired = FixedCharField(unique=True, null=True, max_length=64) mimetype = ForeignKeyField(MimeType, backref='contents') file_size_current = IntegerField(index=True) file_size_when_acquired = IntegerField(index=True, null=True) we_modified_it = BooleanField(index=True, default=False) @create_table class ContentThumbnail(MyModel): content = ForeignKeyField(Content, backref='thumbnail', primary_key=True) generated_at = DateTimeField(default=datetime.datetime.now, index=True) mimetype = ForeignKeyField(MimeType) size = IntegerField(index=True) class ContentOld(MyModel): post = ForeignKeyField(Post, backref='content') ext = CharField(index=True, null=True, max_length=16) sha256_current = FixedCharField(unique=True, max_length=64) sha256_when_acquired = FixedCharField(unique=True, null=True, max_length=64) mimetype = ForeignKeyField(MimeType, backref='contents') file_size_current = IntegerField(index=True) file_size_when_acquired = IntegerField(index=True, null=True) we_modified_it = BooleanField(index=True, default=False) @create_table class Comment(MyModel): id = UnsignedAutoField(primary_key=True) post = ForeignKeyField(Post, backref='comments') local_id = UnsignedIntegerField(index=True) body = TextField() creator = ForeignKeyField(User, backref='comments') score = IntegerField(index=True, null=True) comment_created_at = DateTimeField() row_created_at = DateTimeField(default=datetime.datetime.now) row_updated_at = DateTimeField(default=datetime.datetime.now) class Meta: indexes = ( (('post', 'local_id'), True), ) @create_table class PostTag(MyModel): tag = ForeignKeyField(Tag, backref='posts') post = ForeignKeyField(Post, backref='tags') class Meta: primary_key = CompositeKey('tag', 'post') @create_table class EntityType(MyModel): id = SmallIntegerAutoField(primary_key=True) name = CharField(unique=True) @create_table class QueuedImportEntity(MyModel): id = UnsignedAutoField(primary_key=True) board = Board('queued_import_entities') entity_type = ForeignKeyField(EntityType) entity_local_id = UnsignedIntegerField() additional_data = TextField(null=True) enqueued_at = DateTimeField(default=datetime.datetime.now, index=True) errors_encountered = UnsignedIntegerField(null=True, index=True) priority = FloatField(default=random.random, index=True) row_locked = BooleanField(index=True) @staticmethod def tasks_query(board=None, type=None, with_less_than_errors=10): query = QueuedImportEntity.select() if board: if isinstance(board, 'str'): board = Imageboard.get(Imageboard.name==board) query = query.where(QueuedImportEntity.board == board) if type: if isinstance(type, str): type = EntityType.get(EntityType.name == type) query = query.where(QueuedImportEntity.type == type) query = query.where(QueuedImportEntity.row_locked == False).where(QueuedImportEntity.errors_encountered <= with_less_than_errors) query = query.order_by(-QueuedImportEntity.priority) return query def report_error(self, error): with db.atomic(): ImportEntityError.create(queued_entity=self, error=error) self.errors_encountered = (self.errors_encountered or 0) + 1 self.priority = random.random() self.row_locked = False self.save() def report_success(self, and_delete_self=True, and_mark_as_final=False): with db.atomic(): imported, did_create = ImportedEntity.get_or_create(board=self.board, entity_type=self.entity_type, entity_local_id=self.entity_local_id, additional_data=self.additional_data) if not did_create: imported.latest_update_at = datetime.datetime.now() imported.save() if and_delete_self: self.delete_instance() class Meta: indexes = ( (('board', 'entity_type', 'entity_local_id'), True), ) @create_table class ImportEntityError(MyModel): id = UnsignedAutoField(primary_key=True) queued_entity = ForeignKeyField(QueuedImportEntity, index=True, on_delete='CASCADE') when = DateTimeField(default=datetime.datetime.now) error = TextField() @create_table class ImportedEntity(MyModel): id = UnsignedAutoField(primary_key=True) board = Board('imported_entities') entity_type = ForeignKeyField(EntityType) entity_local_id = UnsignedIntegerField(index=True) additional_data = TextField(null=True) latest_update_at = DateTimeField(default=datetime.datetime.now, index=True) final = BooleanField() def enqueue_again(self, ignore_final=False): if (self.final and ignore_final) or (not self.final): QueuedImportEntity.create(board=self.board, entity_type=self.entity_type, entity_local_id=self.entity_local_id, additional_data=self.additional_data) else: raise ValueError('not allowed to enqueue a final entity') class Meta: indexes = ( (('board', 'entity_type', 'entity_local_id'), True), ) @create_table class Note(MyModel): id = UnsignedAutoField(primary_key=True) board = Board('notes') local_id = UnsignedIntegerField(index=True) author = ForeignKeyField(User, backref='notes') post = ForeignKeyField(Post, backref='notes') body = TextField() version = SmallIntegerField() note_created_at = DateTimeField() note_updated_at = DateTimeField() row_created_at = DateTimeField(default=datetime.datetime.now) is_active = BooleanField() x = UnsignedIntegerField() y = UnsignedIntegerField() width = UnsignedIntegerField() height = UnsignedIntegerField() class Meta: indexes = ( (('board', 'local_id', 'version'), True), (('board', 'local_id'), False), )
danya02/booru-mirror-unified
transfer_rule34.py
import database from peewee import * import datetime import functools import magic import hashlib import os import shutil BOARD, _ = database.Imageboard.get_or_create(name='rule34', base_url='https://rule34.xxx') olddb = MySQLDatabase('rule34_bk', user='booru', password='<PASSWORD>') class OldModel(Model): deleted = BooleanField(index=True, default=False) # this is used so as not to remove rows from the database but still keep track of those which have been processed. This column must be manually added to all tables. class Meta: database = olddb def confirm_delete(row): print('OK to delete', repr(row), str(row), row.__dict__) return True or input('y/n> ').lower() == 'y' IMAGE_DIR = '/hugedata/booru_old/rule34.xxx-files/' #db = SqliteDatabase(SITE+'.db', timeout=600) db = MySQLDatabase('rule34', user='booru', password='<PASSWORD>', host='10.0.0.2') class File: @staticmethod def get_file_content(name): with open(IMAGE_DIR + name[:2] + '/' + name, 'rb') as handle: return handle.read() @staticmethod def delete_file(name): os.makedirs(IMAGE_DIR + 'deleted/' + name[:2] + '/', exist_ok=True) try: shutil.move(IMAGE_DIR + name[:2] + '/' + name, IMAGE_DIR + 'deleted/' + name[:2] + '/' + name) except FileNotFoundError as e: if not os.path.isfile(IMAGE_DIR + 'deleted/' + name[:2] + '/' + name): raise e class AccessLevel(OldModel): name = CharField(unique=True) def migrate(self): row, _ = database.AccessLevel.get_or_create(name=self.name) return row class User(OldModel): id = IntegerField(primary_key=True, unique=True) username = CharField(null=True) level = ForeignKeyField(AccessLevel, null=True) join_date = DateField(null=True) def migrate(self): row, _ = database.User.get_or_create(board=BOARD, local_id=self.id, username=self.username) if self.join_date: database.UserJoinDate.get_or_create(user=row, join_date=self.join_date) if self.level: database.UserAccessLevel.get_or_create(user=row, level=self.level.migrate()) #count = Post.select(fn.COUNT(1)).where(Post.creator == self).scalar() + Comment.select(fn.COUNT(1)).where(Comment.author == self).scalar() + Note.select(fn.COUNT(1)).where(Note.author == self).scalar() #if count == 0: # if confirm_delete(self): # self.delete_instance() return row class Rating(OldModel): value = CharField(unique=True) class Status(OldModel): value = CharField(unique=True) def migrate(self): row, _ = database.Status.get_or_create(value=self.value) return row class Tag(OldModel): name = CharField(unique=True) def migrate(self): row, _ = database.Tag.get_or_create(name=self.name) return row class Type(OldModel): name = CharField(unique=True) def migrate(self): row, _ = database.Type.get_or_create(name=self.name) return row class TagPostCount(OldModel): tag = ForeignKeyField(Tag, primary_key=True) value = IntegerField() class TagType(OldModel): tag = ForeignKeyField(Tag) type = ForeignKeyField(Type) def migrate(self): type = self.type.migrate() tag = self.tag.migrate() database.TagType.get_or_create(board=BOARD, type=type, tag=tag) if confirm_delete(self): TagType.update(TagType.deleted == True).where(TagType.tag == self.tag).where(TagType.type == self.type).execute() class Meta: primary_key = CompositeKey('tag', 'type') class UnavailablePost(OldModel): id = IntegerField(primary_key=True, unique=True) reason = TextField(null=True) first_detected_at = DateTimeField(default=datetime.datetime.now) class Post(OldModel): id = IntegerField(primary_key=True, unique=True) width = IntegerField() height = IntegerField() url = CharField(unique=True) sample_width = IntegerField() sample_height = IntegerField() sample = CharField() preview_width = IntegerField() preview_height = IntegerField() preview = CharField() md5 = CharField(unique=True) created_at = DateTimeField() changed_at = DateTimeField() score = IntegerField() creator = ForeignKeyField(User) rating = ForeignKeyField(Rating) status = ForeignKeyField(Status) source = CharField() parent = ForeignKeyField('self', backref='children', null=True) def migrate(self): post, _ = database.Post.get_or_create(board=BOARD, local_id=self.id, post_created_at=self.created_at, post_updated_at=self.changed_at, uploaded_by=self.creator.migrate(), source=(self.source or None), rating=self.rating.value, score=self.score, parent_id=None if self.parent is None else self.parent.id) database.ImageMetadata.get_or_create(post=post, image_width=self.width, image_height=self.height, md5=self.md5) database.PreviewSizeInfo.get_or_create(post=post, sample_width=self.sample_width, sample_height=self.sample_height, preview_width=self.preview_width, preview_height=self.preview_height) database.ImageURL.get_or_create(post=post, img_url=self.url, sample_url=self.sample, preview_url=self.preview) database.PostStatus.get_or_create(post=post, status=database.Status.get_or_create(value=self.status.value)[0]) for i in self.comments: i.migrate(post) for i in self.notes: i.migrate(post) tags = [] for i in self.tags: tags.append(i.tag.migrate()) data = [] for i in tags: data.append({'post': post, 'tag': i}) try: database.PostTag.insert_many(data).execute() except: with db.atomic(): for i in self.tags: i.migrate(post_row=post) PostTag.update(deleted=True).where(PostTag.post==self).where(PostTag.tag.in_(tags)).execute() if len(self.content): self.content.get().migrate(post) self.deleted = True self.save() return post class Content(OldModel): post = ForeignKeyField(Post, backref='content') path = CharField(unique=True) original_length = IntegerField() current_length = IntegerField() def migrate(self, post_row=None): data = File.get_file_content(self.path) sha256 = hashlib.sha256(data).hexdigest() mt = magic.from_buffer(data, mime=True) mt_row, _ = database.MimeType.get_or_create(name=mt) ext = self.path.split('.')[-1] row, _ = database.Content.get_or_create(post=post_row or self.post.migrate(), sha256_current=sha256, mimetype=mt_row, file_size_current=len(data), we_modified_it=False) database.File.save_file(data, ext, raise_if_exists=True, but_check_for_same=True) self.deleted = True self.save() File.delete_file(self.path) return row class PostTag(OldModel): tag = ForeignKeyField(Tag, backref='posts') post = ForeignKeyField(Post, backref='tags') class Meta: primary_key = CompositeKey('tag', 'post') def migrate(self, post_row=None): database.PostTag.get_or_create(post=post_row or self.post.migrate(), tag=self.tag.migrate()) # if confirm_delete(self): # PostTag.update(deleted=True).where(PostTag.tag==self.tag).where(PostTag.post==self.post).execute() class Comment(OldModel): author = ForeignKeyField(User, backref='comments') post = ForeignKeyField(Post, backref='comments') id = IntegerField(primary_key=True, unique=True) body = TextField() score = IntegerField() created_at = DateTimeField() def migrate(self, post_row=None): row, _ = database.Comment.get_or_create(post=post_row or self.post.migrate(), local_id=self.id, creator=self.author.migrate(), body=self.body, score=self.score, comment_created_at=self.created_at) self.deleted = True self.save() return row class Note(OldModel): id = IntegerField(primary_key=True, unique=True) author = ForeignKeyField(User, backref='notes') post = ForeignKeyField(Post, backref='notes') body = TextField() created_at = DateTimeField() updated_at = DateTimeField() is_active = BooleanField() version = IntegerField() x = IntegerField() y = IntegerField() width = IntegerField() height = IntegerField() def migrate(self, post_row=None): row, _ = database.Note.get_or_create( board = BOARD, local_id = self.id, author = self.author.migrate(), post = post_row or self.post.migrate(), body = self.body, note_created_at = self.created_at, note_updated_at = self.updated_at, is_active = self.is_active, version = self.version, x = self.x, y = self.y, width = self.width, height = self.height) self.deleted = True self.save() return row page = 1 iterated = True while iterated: iterated = False #for i in Post.select().where(Post.deleted == False).where(Post.status_id.in_([3, 16])).limit(1).iterator(): for i in Content.select().where(Content.deleted == False).limit(1).iterator(): i.migrate() iterated = True page += 1
danya02/booru-mirror-unified
count_post_tags.py
from database import * #boards = list(Imageboard.select()) #last_tag_id = TagPostCount.select(fn.MAX(TagPostCount.tag_id)).scalar() or 1 for tag in Tag.select().join(TagPostCount, JOIN.LEFT_OUTER).where(TagPostCount.tag.is_null(True)).orwhere(TagPostCount.changed_at < (datetime.datetime.now() - datetime.timedelta(days=5))).order_by(Tag.id).iterator(): counts = PostTag.select(Post.board_id, fn.COUNT(1)).join(Post).where(PostTag.tag == tag).group_by(Post.board).tuples() with db.atomic(): sumcount = 0 for board, count in counts: board = Imageboard.get_by_id(board) print(board.name, tag.name, count) sumcount += count #if count != 0: TagPostCount.set(tag=tag, post_count=count, board=board) TagPostCount.set(tag=tag, post_count=sumcount, board=None)
shreyashsrivastava/wagtail
wagtail/admin/tests/test_reports_views.py
import datetime from io import BytesIO from django.conf import settings from django.test import TestCase from django.urls import reverse from django.utils import timezone from openpyxl import load_workbook from wagtail.core.models import Page, PageLogEntry from wagtail.tests.utils import WagtailTestUtils class TestLockedPagesView(TestCase, WagtailTestUtils): def setUp(self): self.user = self.login() def get(self, params={}): return self.client.get(reverse('wagtailadmin_reports:locked_pages'), params) def test_simple(self): response = self.get() self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/reports/locked_pages.html') # Initially there should be no locked pages self.assertContains(response, "No locked pages found.") self.page = Page.objects.first() self.page.locked = True self.page.locked_by = self.user self.page.locked_at = timezone.now() self.page.save() # Now the listing should contain our locked page response = self.get() self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/reports/locked_pages.html') self.assertNotContains(response, "No locked pages found.") self.assertContains(response, self.page.title) def test_csv_export(self): self.page = Page.objects.first() self.page.locked = True self.page.locked_by = self.user if settings.USE_TZ: # 12:00 UTC self.page.locked_at = '2013-02-01T12:00:00.000Z' self.page.latest_revision_created_at = '2013-01-01T12:00:00.000Z' else: # 12:00 in no specific timezone self.page.locked_at = '2013-02-01T12:00:00' self.page.latest_revision_created_at = '2013-01-01T12:00:00' self.page.save() response = self.get(params={'export': 'csv'}) # Check response self.assertEqual(response.status_code, 200) data_lines = response.getvalue().decode().split("\n") self.assertEqual(data_lines[0], 'Title,Updated,Status,Type,Locked At,Locked By\r') if settings.USE_TZ: self.assertEqual(data_lines[1], 'Root,2013-01-01 12:00:00+00:00,live,Page,2013-02-01 12:00:00+00:00,test@email.com\r') else: self.assertEqual(data_lines[1], 'Root,2013-01-01 12:00:00,live,Page,2013-02-01 12:00:00,<EMAIL>\r') def test_xlsx_export(self): self.page = Page.objects.first() self.page.locked = True self.page.locked_by = self.user if settings.USE_TZ: # 12:00 UTC self.page.locked_at = '2013-02-01T12:00:00.000Z' self.page.latest_revision_created_at = '2013-01-01T12:00:00.000Z' else: # 12:00 in no specific timezone self.page.locked_at = '2013-02-01T12:00:00' self.page.latest_revision_created_at = '2013-01-01T12:00:00' self.page.save() response = self.get(params={'export': 'xlsx'}) # Check response - the locked page info should be in it self.assertEqual(response.status_code, 200) workbook_data = response.getvalue() worksheet = load_workbook(filename=BytesIO(workbook_data))['Sheet1'] cell_array = [[cell.value for cell in row] for row in worksheet.rows] self.assertEqual(cell_array[0], ['Title', 'Updated', 'Status', 'Type', 'Locked At', 'Locked By']) self.assertEqual(cell_array[1], ['Root', datetime.datetime(2013, 1, 1, 12, 0), 'live', 'Page', datetime.datetime(2013, 2, 1, 12, 0), '<EMAIL>']) self.assertEqual(len(cell_array), 2) class TestFilteredLockedPagesView(TestCase, WagtailTestUtils): fixtures = ['test.json'] def setUp(self): self.user = self.login() self.unpublished_page = Page.objects.get(url_path='/home/events/tentative-unpublished-event/') self.unpublished_page.locked = True self.unpublished_page.locked_by = self.user self.unpublished_page.locked_at = timezone.now() self.unpublished_page.save() def get(self, params={}): return self.client.get(reverse('wagtailadmin_reports:locked_pages'), params) def test_filter_by_live(self): response = self.get(params={'live': 'true'}) self.assertEqual(response.status_code, 200) self.assertNotContains(response, "Tentative Unpublished Event") self.assertContains(response, "My locked page") response = self.get(params={'live': 'false'}) self.assertEqual(response.status_code, 200) self.assertContains(response, "Tentative Unpublished Event") self.assertNotContains(response, "My locked page") def test_filter_by_user(self): response = self.get(params={'locked_by': self.user.pk}) self.assertEqual(response.status_code, 200) self.assertContains(response, "Tentative Unpublished Event") self.assertNotContains(response, "My locked page") class TestFilteredLogEntriesView(TestCase, WagtailTestUtils): fixtures = ['test.json'] def setUp(self): self.user = self.login() self.home_page = Page.objects.get(url_path='/home/') self.create_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.create') self.edit_log_1 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.edit_log_2 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.edit_log_3 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.create_comment_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.create', data={ 'comment': { 'contentpath': 'title', 'text': 'Foo', } }) self.edit_comment_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.edit', data={ 'comment': { 'contentpath': 'title', 'text': 'Edited', } }) self.create_reply_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.create_reply', data={ 'comment': { 'contentpath': 'title', 'text': 'Foo', } }) def get(self, params={}): return self.client.get(reverse('wagtailadmin_reports:site_history'), params) def assert_log_entries(self, response, expected): actual = set(response.context['object_list']) self.assertSetEqual(actual, set(expected)) def test_unfiltered(self): response = self.get() self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.create_log, self.edit_log_1, self.edit_log_2, self.edit_log_3, self.create_comment_log, self.edit_comment_log, self.create_reply_log, ]) def test_filter_by_action(self): response = self.get(params={'action': 'wagtail.edit'}) self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.edit_log_1, self.edit_log_2, self.edit_log_3, ]) def test_hide_commenting_actions(self): response = self.get(params={'hide_commenting_actions': 'on'}) self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.create_log, self.edit_log_1, self.edit_log_2, self.edit_log_3, ])
shreyashsrivastava/wagtail
wagtail/admin/tests/pages/test_copy_page.py
from django.contrib.auth.models import Group, Permission from django.http import HttpRequest, HttpResponse from django.test import TestCase from django.urls import reverse from wagtail.core.models import GroupPagePermission, Page from wagtail.tests.testapp.models import SimplePage from wagtail.tests.utils import WagtailTestUtils class TestPageCopy(TestCase, WagtailTestUtils): def setUp(self): # Find root page self.root_page = Page.objects.get(id=2) # Create a page self.test_page = self.root_page.add_child(instance=SimplePage( title="Hello world!", slug='hello-world', content="hello", live=True, has_unpublished_changes=False, )) # Create a couple of child pages self.test_child_page = self.test_page.add_child(instance=SimplePage( title="Child page", slug='child-page', content="hello", live=True, has_unpublished_changes=True, )) self.test_unpublished_child_page = self.test_page.add_child(instance=SimplePage( title="Unpublished Child page", slug='unpublished-child-page', content="hello", live=False, has_unpublished_changes=True, )) # Login self.user = self.login() def test_page_copy(self): response = self.client.get(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, ))) # Check response self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/pages/copy.html') # Make sure all fields are in the form self.assertContains(response, "New title") self.assertContains(response, "New slug") self.assertContains(response, "New parent page") self.assertContains(response, "Copy subpages") self.assertContains(response, "Publish copies") self.assertContains(response, "Alias") def test_page_copy_bad_permissions(self): # Remove privileges from user self.user.is_superuser = False self.user.user_permissions.add( Permission.objects.get(content_type__app_label='wagtailadmin', codename='access_admin') ) self.user.save() # Get copy page post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world', 'new_parent_page': str(self.test_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # A user with no page permissions at all should be redirected to the admin home self.assertRedirects(response, reverse('wagtailadmin_home')) # A user with page permissions, but not add permission at the destination, # should receive a form validation error publishers = Group.objects.create(name='Publishers') GroupPagePermission.objects.create( group=publishers, page=self.root_page, permission_type='publish' ) self.user.groups.add(publishers) self.user.save() # Get copy page post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world', 'new_parent_page': str(self.test_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) form = response.context['form'] self.assertFalse(form.is_valid()) self.assertTrue('new_parent_page' in form.errors) def test_page_copy_post(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'publish_copies': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().filter(slug='hello-world-2').first() # Check that the copy exists self.assertNotEqual(page_copy, None) # Check that the copy is not live self.assertFalse(page_copy.live) self.assertTrue(page_copy.has_unpublished_changes) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were not copied self.assertEqual(page_copy.get_children().count(), 0) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_page_copy_post_copy_subpages(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': True, 'publish_copies': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().filter(slug='hello-world-2').first() # Check that the copy exists self.assertNotEqual(page_copy, None) # Check that the copy is not live self.assertFalse(page_copy.live) self.assertTrue(page_copy.has_unpublished_changes) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were copied self.assertEqual(page_copy.get_children().count(), 2) # Check the the child pages # Neither of them should be live child_copy = page_copy.get_children().filter(slug='child-page').first() self.assertNotEqual(child_copy, None) self.assertFalse(child_copy.live) self.assertTrue(child_copy.has_unpublished_changes) unpublished_child_copy = page_copy.get_children().filter(slug='unpublished-child-page').first() self.assertNotEqual(unpublished_child_copy, None) self.assertFalse(unpublished_child_copy.live) self.assertTrue(unpublished_child_copy.has_unpublished_changes) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_page_copy_post_copy_subpages_publish_copies(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': True, 'publish_copies': True, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().filter(slug='hello-world-2').first() # Check that the copy exists self.assertNotEqual(page_copy, None) # Check that the copy is live self.assertTrue(page_copy.live) self.assertFalse(page_copy.has_unpublished_changes) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were copied self.assertEqual(page_copy.get_children().count(), 2) # Check the the child pages # The child_copy should be live but the unpublished_child_copy shouldn't child_copy = page_copy.get_children().filter(slug='child-page').first() self.assertNotEqual(child_copy, None) self.assertTrue(child_copy.live) self.assertTrue(child_copy.has_unpublished_changes) unpublished_child_copy = page_copy.get_children().filter(slug='unpublished-child-page').first() self.assertNotEqual(unpublished_child_copy, None) self.assertFalse(unpublished_child_copy.live) self.assertTrue(unpublished_child_copy.has_unpublished_changes) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_page_copy_post_new_parent(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.test_child_page.id), 'copy_subpages': False, 'publish_copies': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the new parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.test_child_page.id, ))) # Check that the page was copied to the correct place self.assertTrue(Page.objects.filter(slug='hello-world-2').first().get_parent(), self.test_child_page) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_page_copy_post_existing_slug_within_same_parent_page(self): # This tests the existing slug checking on page copy when not changing the parent page # Attempt to copy the page but forget to change the slug post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Should not be redirected (as the save should fail) self.assertEqual(response.status_code, 200) # Check that a form error was raised self.assertFormError( response, 'form', 'new_slug', "This slug is already in use within the context of its parent page \"Welcome to your new Wagtail site!\"" ) def test_page_copy_post_and_subpages_to_same_tree_branch(self): # This tests that a page cannot be copied into itself when copying subpages post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world', 'new_parent_page': str(self.test_child_page.id), 'copy_subpages': True, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id,)), post_data) # Should not be redirected (as the save should fail) self.assertEqual(response.status_code, 200) # Check that a form error was raised self.assertFormError( response, 'form', 'new_parent_page', "You cannot copy a page into itself when copying subpages" ) def test_page_copy_post_existing_slug_to_another_parent_page(self): # This tests the existing slug checking on page copy when changing the parent page # Attempt to copy the page and changed the parent page post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world', 'new_parent_page': str(self.test_child_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.test_child_page.id, ))) def test_page_copy_post_invalid_slug(self): # Attempt to copy the page but set an invalid slug string post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello world!', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Should not be redirected (as the save should fail) self.assertEqual(response.status_code, 200) # Check that a form error was raised self.assertFormError( response, 'form', 'new_slug', "Enter a valid “slug” consisting of Unicode letters, numbers, underscores, or hyphens." ) def test_page_copy_post_valid_unicode_slug(self): post_data = { 'new_title': "Hello wɜːld", 'new_slug': 'hello-wɜːld', 'new_parent_page': str(self.test_page.id), 'copy_subpages': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check response self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.test_page.id, ))) # Get copy page_copy = self.test_page.get_children().filter(slug=post_data['new_slug']).first() # Check that the copy exists with the good slug self.assertNotEqual(page_copy, None) self.assertEqual(page_copy.slug, post_data['new_slug']) def test_page_copy_no_publish_permission(self): # Turn user into an editor who can add pages but not publish them self.user.is_superuser = False self.user.groups.add( Group.objects.get(name="Editors"), ) self.user.save() # Get copy page response = self.client.get(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, ))) # The user should have access to the copy page self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/pages/copy.html') # Make sure the "publish copies" field is hidden self.assertNotContains(response, "Publish copies") def test_page_copy_no_publish_permission_post_copy_subpages_publish_copies(self): # This tests that unprivileged users cannot publish copied pages even if they hack their browser # Turn user into an editor who can add pages but not publish them self.user.is_superuser = False self.user.groups.add( Group.objects.get(name="Editors"), ) self.user.save() # Post post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': True, 'publish_copies': True, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().filter(slug='hello-world-2').first() # Check that the copy exists self.assertNotEqual(page_copy, None) # Check that the copy is not live self.assertFalse(page_copy.live) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were copied self.assertEqual(page_copy.get_children().count(), 2) # Check the the child pages # Neither of them should be live child_copy = page_copy.get_children().filter(slug='child-page').first() self.assertNotEqual(child_copy, None) self.assertFalse(child_copy.live) unpublished_child_copy = page_copy.get_children().filter(slug='unpublished-child-page').first() self.assertNotEqual(unpublished_child_copy, None) self.assertFalse(unpublished_child_copy.live) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_before_copy_page_hook(self): def hook_func(request, page): self.assertIsInstance(request, HttpRequest) self.assertIsInstance(page.specific, SimplePage) return HttpResponse("Overridden!") with self.register_hook('before_copy_page', hook_func): response = self.client.get(reverse('wagtailadmin_pages:copy', args=(self.test_page.id,))) self.assertEqual(response.status_code, 200) self.assertEqual(response.content, b"Overridden!") def test_before_copy_page_hook_post(self): def hook_func(request, page): self.assertIsInstance(request, HttpRequest) self.assertIsInstance(page.specific, SimplePage) return HttpResponse("Overridden!") with self.register_hook('before_copy_page', hook_func): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'publish_copies': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id,)), post_data) self.assertEqual(response.status_code, 200) self.assertEqual(response.content, b"Overridden!") # page should not be copied self.assertFalse(Page.objects.filter(title="Hello world 2").exists()) def test_after_copy_page_hook(self): def hook_func(request, page, new_page): self.assertIsInstance(request, HttpRequest) self.assertIsInstance(page.specific, SimplePage) self.assertIsInstance(new_page.specific, SimplePage) return HttpResponse("Overridden!") with self.register_hook('after_copy_page', hook_func): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'publish_copies': False, 'alias': False, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id,)), post_data) self.assertEqual(response.status_code, 200) self.assertEqual(response.content, b"Overridden!") # page should be copied self.assertTrue(Page.objects.filter(title="Hello world 2").exists()) def test_page_copy_alias_post(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': False, 'publish_copies': False, 'alias': True, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().get(slug='hello-world-2') # Check the copy is an alias of the original self.assertEqual(page_copy.alias_of, self.test_page.page_ptr) # Check that the copy is live # Note: publish_copies is ignored. Alias pages always keep the same state as their original self.assertTrue(page_copy.live) self.assertFalse(page_copy.has_unpublished_changes) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were not copied self.assertEqual(page_copy.get_children().count(), 0) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems') def test_page_copy_alias_post_copy_subpages(self): post_data = { 'new_title': "Hello world 2", 'new_slug': 'hello-world-2', 'new_parent_page': str(self.root_page.id), 'copy_subpages': True, 'publish_copies': False, 'alias': True, } response = self.client.post(reverse('wagtailadmin_pages:copy', args=(self.test_page.id, )), post_data) # Check that the user was redirected to the parents explore page self.assertRedirects(response, reverse('wagtailadmin_explore', args=(self.root_page.id, ))) # Get copy page_copy = self.root_page.get_children().get(slug='hello-world-2') # Check the copy is an alias of the original self.assertEqual(page_copy.alias_of, self.test_page.page_ptr) # Check that the copy is live # Note: publish_copies is ignored. Alias pages always keep the same state as their original self.assertTrue(page_copy.live) self.assertFalse(page_copy.has_unpublished_changes) # Check that the owner of the page is set correctly self.assertEqual(page_copy.owner, self.user) # Check that the children were copied self.assertEqual(page_copy.get_children().count(), 2) # Check the the child pages # Neither of them should be live child_copy = page_copy.get_children().filter(slug='child-page').first() self.assertNotEqual(child_copy, None) self.assertEqual(child_copy.alias_of, self.test_child_page.page_ptr) self.assertTrue(child_copy.live) self.assertFalse(child_copy.has_unpublished_changes) unpublished_child_copy = page_copy.get_children().filter(slug='unpublished-child-page').first() self.assertNotEqual(unpublished_child_copy, None) self.assertEqual(unpublished_child_copy.alias_of, self.test_unpublished_child_page.page_ptr) self.assertFalse(unpublished_child_copy.live) self.assertTrue(unpublished_child_copy.has_unpublished_changes) # treebeard should report no consistency problems with the tree self.assertFalse(any(Page.find_problems()), 'treebeard found consistency problems')
shreyashsrivastava/wagtail
wagtail/admin/urls/reports.py
<reponame>shreyashsrivastava/wagtail<filename>wagtail/admin/urls/reports.py<gh_stars>1000+ from django.urls import path from wagtail.admin.views.reports.audit_logging import LogEntriesView from wagtail.admin.views.reports.locked_pages import LockedPagesView from wagtail.admin.views.reports.workflows import WorkflowTasksView, WorkflowView app_name = 'wagtailadmin_reports' urlpatterns = [ path('locked/', LockedPagesView.as_view(), name='locked_pages'), path('workflow/', WorkflowView.as_view(), name='workflow'), path('workflow_tasks/', WorkflowTasksView.as_view(), name='workflow_tasks'), path('site-history/', LogEntriesView.as_view(), name='site_history'), ]
shreyashsrivastava/wagtail
wagtail/admin/views/reports/base.py
from django.utils.translation import gettext_lazy as _ from django.views.generic.base import TemplateResponseMixin from django.views.generic.list import BaseListView from wagtail.admin.views.mixins import SpreadsheetExportMixin class ReportView(SpreadsheetExportMixin, TemplateResponseMixin, BaseListView): header_icon = "" page_kwarg = "p" template_name = "wagtailadmin/reports/base_report.html" title = "" paginate_by = 50 filterset_class = None def filter_queryset(self, queryset): filters = None if self.filterset_class: filters = self.filterset_class(self.request.GET, queryset=queryset, request=self.request) queryset = filters.qs return filters, queryset def dispatch(self, request, *args, **kwargs): self.is_export = self.request.GET.get("export") in self.FORMATS if self.is_export: self.paginate_by = None return self.as_spreadsheet(self.filter_queryset(self.get_queryset())[1], self.request.GET.get("export")) return super().dispatch(request, *args, **kwargs) def get_context_data(self, *args, object_list=None, **kwargs): queryset = object_list if object_list is not None else self.object_list filters, queryset = self.filter_queryset(queryset) context = super().get_context_data(*args, object_list=queryset, **kwargs) context["title"] = self.title context["header_icon"] = self.header_icon context["filters"] = filters return context class PageReportView(ReportView): template_name = "wagtailadmin/reports/base_page_report.html" export_headings = { "latest_revision_created_at": _("Updated"), "status_string": _("Status"), "content_type.model_class._meta.verbose_name.title": _("Type"), } list_export = [ "title", "latest_revision_created_at", "status_string", "content_type.model_class._meta.verbose_name.title", ]
shreyashsrivastava/wagtail
wagtail/admin/tests/pages/test_workflow_history.py
from django.contrib.auth.models import Permission from django.test import TestCase from django.urls import reverse from wagtail.core.models import Page, PageLogEntry from wagtail.tests.utils import WagtailTestUtils class TestWorkflowHistoryDetail(TestCase, WagtailTestUtils): fixtures = ['test.json'] def setUp(self): self.user = self.create_test_user() self.login(self.user) self.christmas_event = Page.objects.get(url_path='/home/events/christmas/') self.christmas_event.save_revision() workflow = self.christmas_event.get_workflow() self.workflow_state = workflow.start(self.christmas_event, self.user) def test_get_index(self): response = self.client.get( reverse('wagtailadmin_pages:workflow_history', args=[self.christmas_event.id]) ) self.assertEqual(response.status_code, 200) self.assertContains(response, reverse('wagtailadmin_pages:edit', args=[self.christmas_event.id])) self.assertContains(response, reverse('wagtailadmin_pages:workflow_history_detail', args=[self.christmas_event.id, self.workflow_state.id])) def test_get_index_with_bad_permissions(self): # Remove privileges from user self.user.is_superuser = False self.user.user_permissions.add( Permission.objects.get(content_type__app_label='wagtailadmin', codename='access_admin') ) self.user.save() response = self.client.get( reverse('wagtailadmin_pages:workflow_history', args=[self.christmas_event.id]) ) self.assertEqual(response.status_code, 302) def test_get_detail(self): response = self.client.get( reverse('wagtailadmin_pages:workflow_history_detail', args=[self.christmas_event.id, self.workflow_state.id]) ) self.assertEqual(response.status_code, 200) self.assertContains(response, reverse('wagtailadmin_pages:edit', args=[self.christmas_event.id])) self.assertContains(response, reverse('wagtailadmin_pages:workflow_history', args=[self.christmas_event.id])) def test_get_detail_with_bad_permissions(self): # Remove privileges from user self.user.is_superuser = False self.user.user_permissions.add( Permission.objects.get(content_type__app_label='wagtailadmin', codename='access_admin') ) self.user.save() response = self.client.get( reverse('wagtailadmin_pages:workflow_history_detail', args=[self.christmas_event.id, self.workflow_state.id]) ) self.assertEqual(response.status_code, 302) class TestFiltering(TestCase, WagtailTestUtils): fixtures = ['test.json'] def setUp(self): self.user = self.login() self.home_page = Page.objects.get(url_path='/home/') self.create_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.create') self.edit_log_1 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.edit_log_2 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.edit_log_3 = PageLogEntry.objects.log_action(self.home_page, 'wagtail.edit') self.create_comment_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.create', data={ 'comment': { 'contentpath': 'title', 'text': 'Foo', } }) self.edit_comment_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.edit', data={ 'comment': { 'contentpath': 'title', 'text': 'Edited', } }) self.create_reply_log = PageLogEntry.objects.log_action(self.home_page, 'wagtail.comments.create_reply', data={ 'comment': { 'contentpath': 'title', 'text': 'Foo', } }) def get(self, params={}): return self.client.get(reverse('wagtailadmin_reports:site_history'), params) return self.client.get(reverse('wagtailadmin_pages:workflow_history', args=[self.home_page.id]), params) def assert_log_entries(self, response, expected): actual = set(response.context['object_list']) self.assertSetEqual(actual, set(expected)) def test_unfiltered(self): response = self.get() self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.create_log, self.edit_log_1, self.edit_log_2, self.edit_log_3, self.create_comment_log, self.edit_comment_log, self.create_reply_log, ]) def test_filter_by_action(self): response = self.get(params={'action': 'wagtail.edit'}) self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.edit_log_1, self.edit_log_2, self.edit_log_3, ]) def test_hide_commenting_actions(self): response = self.get(params={'hide_commenting_actions': 'on'}) self.assertEqual(response.status_code, 200) self.assert_log_entries(response, [ self.create_log, self.edit_log_1, self.edit_log_2, self.edit_log_3, ])
shreyashsrivastava/wagtail
wagtail/admin/views/reports/__init__.py
from .base import PageReportView, ReportView # noqa
shreyashsrivastava/wagtail
wagtail/admin/tests/test_buttons_hooks.py
<filename>wagtail/admin/tests/test_buttons_hooks.py from django.test import TestCase from django.urls import reverse from django.utils.http import urlencode from wagtail.admin import widgets as wagtailadmin_widgets from wagtail.admin.wagtail_hooks import page_listing_more_buttons from wagtail.core import hooks from wagtail.core.models import Page from wagtail.tests.utils import WagtailTestUtils class PagePerms: def can_move(self): return False def can_copy(self): return False def can_delete(self): return True def can_unpublish(self): return False def can_view_revisions(self): return False class TestButtonsHooks(TestCase, WagtailTestUtils): def setUp(self): self.root_page = Page.objects.get(id=2) self.login() def test_register_page_listing_buttons(self): def page_listing_buttons(page, page_perms, is_parent=False, next_url=None): yield wagtailadmin_widgets.PageListingButton( 'Another useless page listing button', '/custom-url', priority=10 ) with hooks.register_temporarily('register_page_listing_buttons', page_listing_buttons): response = self.client.get( reverse('wagtailadmin_explore', args=(self.root_page.id, )) ) self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_button_with_dropdown.html') self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_buttons.html') self.assertContains(response, 'Another useless page listing button') def test_register_page_listing_more_buttons(self): def page_listing_more_buttons(page, page_perms, is_parent=False, next_url=None): yield wagtailadmin_widgets.Button( 'Another useless button in default "More" dropdown', '/custom-url', priority=10 ) with hooks.register_temporarily('register_page_listing_more_buttons', page_listing_more_buttons): response = self.client.get( reverse('wagtailadmin_explore', args=(self.root_page.id, )) ) self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_button_with_dropdown.html') self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_buttons.html') self.assertContains(response, 'Another useless button in default &quot;More&quot; dropdown') def test_custom_button_with_dropdown(self): def page_custom_listing_buttons(page, page_perms, is_parent=False, next_url=None): yield wagtailadmin_widgets.ButtonWithDropdownFromHook( 'One more more button', hook_name='register_page_listing_one_more_more_buttons', page=page, page_perms=page_perms, is_parent=is_parent, next_url=next_url, attrs={'target': '_blank', 'rel': 'noopener noreferrer'}, priority=50 ) def page_custom_listing_more_buttons(page, page_perms, is_parent=False, next_url=None): yield wagtailadmin_widgets.Button( 'Another useless dropdown button in "One more more button" dropdown', '/custom-url', priority=10 ) with hooks.register_temporarily('register_page_listing_buttons', page_custom_listing_buttons), hooks.register_temporarily('register_page_listing_one_more_more_buttons', page_custom_listing_more_buttons): response = self.client.get( reverse('wagtailadmin_explore', args=(self.root_page.id, )) ) self.assertEqual(response.status_code, 200) self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_button_with_dropdown.html') self.assertTemplateUsed(response, 'wagtailadmin/pages/listing/_buttons.html') self.assertContains(response, 'One more more button') self.assertContains(response, 'Another useless dropdown button in &quot;One more more button&quot; dropdown') def test_delete_button_next_url(self): page_perms = PagePerms() page = self.root_page base_url = reverse('wagtailadmin_pages:delete', args=[page.id]) next_url = "a/random/url/" full_url = base_url + '?' + urlencode({'next': next_url}) # page_listing_more_button generator yields only `Delete button` delete_button = next(page_listing_more_buttons( page, page_perms, is_parent=False, next_url=next_url )) self.assertEqual(delete_button.url, full_url) next_url = reverse('wagtailadmin_explore', args=[page.id]) delete_button = next(page_listing_more_buttons( page, page_perms, is_parent=False, next_url=next_url )) self.assertEqual(delete_button.url, base_url)
shreyashsrivastava/wagtail
wagtail/admin/tests/test_templatetags.py
from datetime import timedelta from unittest import mock from django.conf import settings from django.test import TestCase from django.test.utils import override_settings from django.utils import timezone from freezegun import freeze_time from wagtail.admin.staticfiles import versioned_static from wagtail.admin.templatetags.wagtailadmin_tags import ( avatar_url, notification_static, timesince_last_update, timesince_simple) from wagtail.images.tests.utils import get_test_image_file from wagtail.tests.utils import WagtailTestUtils from wagtail.users.models import UserProfile class TestAvatarTemplateTag(TestCase, WagtailTestUtils): def setUp(self): # Create a user self.test_user = self.create_user( username='testuser', email='<EMAIL>', password='password', ) def test_use_gravatar_by_default(self): url = avatar_url(self.test_user) self.assertIn('www.gravatar.com', url) def test_skip_gravatar_if_no_email(self): self.test_user.email = '' url = avatar_url(self.test_user) self.assertIn('default-user-avatar', url) @override_settings(WAGTAIL_GRAVATAR_PROVIDER_URL='https://robohash.org') def test_custom_gravatar_provider(self): url = avatar_url(self.test_user) self.assertIn('robohash.org', url) @override_settings(WAGTAIL_GRAVATAR_PROVIDER_URL=None) def test_disable_gravatar(self): url = avatar_url(self.test_user) self.assertIn('default-user-avatar', url) def test_uploaded_avatar(self): user_profile = UserProfile.get_for_user(self.test_user) user_profile.avatar = get_test_image_file(filename='custom-avatar.png') user_profile.save() url = avatar_url(self.test_user) self.assertIn('custom-avatar', url) class TestNotificationStaticTemplateTag(TestCase): @override_settings(STATIC_URL='/static/') def test_local_notification_static(self): url = notification_static('wagtailadmin/images/email-header.jpg') self.assertEqual('{}/static/wagtailadmin/images/email-header.jpg'.format(settings.BASE_URL), url) @override_settings(STATIC_URL='/static/', BASE_URL='http://localhost:8000') def test_local_notification_static_baseurl(self): url = notification_static('wagtailadmin/images/email-header.jpg') self.assertEqual('http://localhost:8000/static/wagtailadmin/images/email-header.jpg', url) @override_settings(STATIC_URL='https://s3.amazonaws.com/somebucket/static/', BASE_URL='http://localhost:8000') def test_remote_notification_static(self): url = notification_static('wagtailadmin/images/email-header.jpg') self.assertEqual('https://s3.amazonaws.com/somebucket/static/wagtailadmin/images/email-header.jpg', url) class TestVersionedStatic(TestCase): def test_versioned_static(self): result = versioned_static('wagtailadmin/js/core.js') self.assertRegex(result, r'^/static/wagtailadmin/js/core.js\?v=(\w+)$') @mock.patch('wagtail.admin.staticfiles.static') def test_versioned_static_version_string(self, mock_static): mock_static.return_value = '/static/wagtailadmin/js/core.js?v=123' result = versioned_static('wagtailadmin/js/core.js') self.assertEqual(result, '/static/wagtailadmin/js/core.js?v=123') mock_static.assert_called_once_with('wagtailadmin/js/core.js') def test_versioned_static_absolute_path(self): result = versioned_static('/static/wagtailadmin/js/core.js') self.assertEqual(result, '/static/wagtailadmin/js/core.js') def test_versioned_static_url(self): result = versioned_static('http://example.org/static/wagtailadmin/js/core.js') self.assertEqual(result, 'http://example.org/static/wagtailadmin/js/core.js') @freeze_time("2020-07-01 12:00:00") class TestTimesinceTags(TestCase): def test_timesince_simple(self): now = timezone.now() ts = timesince_simple(now) self.assertEqual(ts, "Just now") ts = timesince_simple(now - timedelta(hours=1, minutes=10)) self.assertEqual(ts, "1\xa0hour ago") ts = timesince_simple(now - timedelta(weeks=2, hours=1, minutes=10)) self.assertEqual(ts, "2\xa0weeks ago") def test_timesince_last_update_today_shows_time(self): dt = timezone.now() - timedelta(hours=1) formatted_time = dt.astimezone(timezone.get_current_timezone()).strftime('%H:%M') timesince = timesince_last_update(dt) self.assertEqual(timesince, formatted_time) # Check prefix output timesince = timesince_last_update(dt, time_prefix='my prefix') self.assertEqual(timesince, 'my prefix {}'.format(formatted_time)) def test_timesince_last_update_before_today_shows_timeago(self): dt = timezone.now() - timedelta(weeks=1, days=2) timesince = timesince_last_update(dt, use_shorthand=False) self.assertEqual(timesince, '1\xa0week, 2\xa0days ago') timesince = timesince_last_update(dt) self.assertEqual(timesince, '1\xa0week ago')
shreyashsrivastava/wagtail
wagtail/admin/filters.py
import django_filters from django.db import models from django.utils.translation import gettext_lazy as _ from django_filters.widgets import SuffixedMultiWidget from wagtail.admin.widgets import AdminDateInput, BooleanButtonSelect, ButtonSelect, FilteredSelect class DateRangePickerWidget(SuffixedMultiWidget): """ A widget allowing a start and end date to be picked. """ template_name = 'wagtailadmin/widgets/daterange_input.html' suffixes = ['after', 'before'] def __init__(self, attrs=None): widgets = (AdminDateInput(attrs={'placeholder': _("Date from")}), AdminDateInput(attrs={'placeholder': _("Date to")})) super().__init__(widgets, attrs) def decompress(self, value): if value: return [value.start, value.stop] return [None, None] class FilteredModelChoiceIterator(django_filters.fields.ModelChoiceIterator): """ A variant of Django's ModelChoiceIterator that, instead of yielding (value, label) tuples, returns (value, label, filter_value) so that FilteredSelect can drop filter_value into the data-filter-value attribute. """ def choice(self, obj): return ( self.field.prepare_value(obj), self.field.label_from_instance(obj), self.field.get_filter_value(obj) ) class FilteredModelChoiceField(django_filters.fields.ModelChoiceField): """ A ModelChoiceField that uses FilteredSelect to dynamically show/hide options based on another ModelChoiceField of related objects; an option will be shown whenever the selected related object is present in the result of filter_accessor for that option. filter_field - the HTML `id` of the related ModelChoiceField filter_accessor - either the name of a relation, property or method on the model instance which returns a queryset of related objects, or a function which accepts the model instance and returns such a queryset. """ widget = FilteredSelect iterator = FilteredModelChoiceIterator def __init__(self, *args, **kwargs): self.filter_accessor = kwargs.pop('filter_accessor') filter_field = kwargs.pop('filter_field') super().__init__(*args, **kwargs) self.widget.filter_field = filter_field def get_filter_value(self, obj): # Use filter_accessor to obtain a queryset of related objects if callable(self.filter_accessor): queryset = self.filter_accessor(obj) else: # treat filter_accessor as a method/property name of obj queryset = getattr(obj, self.filter_accessor) if isinstance(queryset, models.Manager): queryset = queryset.all() elif callable(queryset): queryset = queryset() # Turn this queryset into a list of IDs that will become the 'data-filter-value' used to # filter this listing return queryset.values_list('pk', flat=True) class FilteredModelChoiceFilter(django_filters.ModelChoiceFilter): field_class = FilteredModelChoiceField class WagtailFilterSet(django_filters.FilterSet): @classmethod def filter_for_lookup(cls, field, lookup_type): filter_class, params = super().filter_for_lookup(field, lookup_type) if filter_class == django_filters.ChoiceFilter: params.setdefault('widget', ButtonSelect) params.setdefault('empty_label', _("All")) elif filter_class in [django_filters.DateFilter, django_filters.DateTimeFilter]: params.setdefault('widget', AdminDateInput) elif filter_class == django_filters.DateFromToRangeFilter: params.setdefault('widget', DateRangePickerWidget) elif filter_class == django_filters.BooleanFilter: params.setdefault('widget', BooleanButtonSelect) return filter_class, params
DapperDino/Learning-TensorFlow
BreastCancerDataSVM/breast_cancer_data_svm.py
<filename>BreastCancerDataSVM/breast_cancer_data_svm.py import sklearn as sk from sklearn.datasets import load_breast_cancer from sklearn.svm import SVC from sklearn.metrics import accuracy_score # Load cancer data from dataset. cancer = load_breast_cancer() # Define which data will be on the x and y axis. x = cancer.data y = cancer.target # Use 80% of the data to train with and then test on the remaining 20%. x_train, x_test, y_train, y_test = sk.model_selection.train_test_split( cancer.data, cancer.target, test_size=0.2) # Fit the data using a linear kernal to the classifier. clf = SVC(kernel="linear", C=2, gamma="auto") clf.fit(x_train, y_train) # Predict the y test values using our x test values. y_prediction = clf.predict(x_test) # Store and display the accuracy of this model. accuracy = accuracy_score(y_test, y_prediction) print(accuracy)
DapperDino/Learning-TensorFlow
CarDataKNN/car_data_knn.py
<filename>CarDataKNN/car_data_knn.py import pandas as pd import sklearn as sk from sklearn import preprocessing from sklearn.neighbors import KNeighborsClassifier # Load car data from file. data = pd.read_csv("CarDataKNN/Car.data") # Convert attribute names to int. le = preprocessing.LabelEncoder() buying = le.fit_transform(list(data["buying"])) maint = le.fit_transform(list(data["maint"])) door = le.fit_transform(list(data["door"])) persons = le.fit_transform(list(data["persons"])) lug_boot = le.fit_transform(list(data["lug_boot"])) safety = le.fit_transform(list(data["safety"])) cls = le.fit_transform(list(data["class"])) # Define the label we wish to predict. preict = "class" # Define which data will be on the x and y axis. x = list(zip(buying, maint, door, persons, lug_boot, safety)) y = list(cls) # Use 90% of the data to train with and then test on the remaining 10%. x_train, x_test, y_train, y_test = sk.model_selection.train_test_split( x, y, test_size=0.1) # Create and fit our model. model = KNeighborsClassifier(n_neighbors=9) model.fit(x_train, y_train) # Store and display the accuracy of this model. accuracy = model.score(x_test, y_test) print(accuracy) # Preict the y test values using our x test values. predicted = model.predict(x_test) names = ["unacc", "acc", "good", "vgood"] # Display how accurate each prediction was. for i in range(len(predicted)): print("Predicted: {}\nData: {}\nActual: {}".format( names[predicted[i]], x_test[i], names[y_test[i]])) n = model.kneighbors([x_test[i]], 9) print("N: {}".format(n))
DapperDino/Learning-TensorFlow
StudentDataRegression/student_data_regression.py
import pickle import pandas as pd import numpy as np import sklearn as sk from sklearn import linear_model import matplotlib.pyplot as pyplot from matplotlib import style # Load student data from file. data = pd.read_csv("StudentDataRegression/StudentData.csv", sep=";") # Filter only the data we wish to model. data = data[["G1", "G2", "G3", "studytime", "failures", "absences"]] # Define the label we wish to predict. predict = "G3" # Drop the prediction from the input data since we won't be using it as input data. x = np.array(data.drop([predict], 1)) y = np.array(data[predict]) # Use 90% of the data to train with and then test on the remaining 10%. x_train, x_test, y_train, y_test = sk.model_selection.train_test_split( x, y, test_size=0.1) """ best_accuracy = 0 while(True): #Use 90% of the data to train with and then test on the remaining 10%. x_train, x_test, y_train, y_test = sk.model_selection.train_test_split(x, y, test_size = 0.1) #Create an instance of the Linear Regression class. linear = linear_model.LinearRegression() #Fit the training data to the model. linear.fit(x_train, y_train) accuracy = linear.score(x_test, y_test) #Check to see if this model is more accurate than our previous best. if accuracy > best_accuracy: #Store the new best accuracy to compare future models to. best_accuracy = accuracy print(best_accuracy) #Save the model to a file. with open("StudentDataRegression/student_model.pickle", "wb") as f: pickle.dump(linear, f) """ # Load model from file. pickle_in = open("StudentDataRegression/student_model.pickle", "rb") linear = pickle.load(pickle_in) # Output statistics from the testing. print("Co: {}".format(linear.coef_)) print("Intercept: {}".format(linear.intercept_)) # Output data from the testing. predictions = linear.predict(x_test) for i in range(len(predictions)): print(predictions[i], x_test[i], y_test[i]) # Display data as scatter graph. p = "failures" style.use("ggplot") pyplot.scatter(data[p], data[predict]) pyplot.xlabel(p) pyplot.ylabel("Final Grade") pyplot.show()
OrlandoCano/JavaDevDay2019
data/sessionize_data/transform_data.py
import requests import json import re API_URL = "https://sessionize.com/api/v2/ezuter2b/view/all" res = requests.get(API_URL) json_data = json.loads(res.text) speakers = {} def clean_name(name): regex = re.compile('[^a-zA-Z]') res = regex.sub('', name) return res.lower() # Speakers for i, s in enumerate(json_data['speakers']): new_key = clean_name(s['fullName']) speakers[new_key] = {} speakers[new_key]['bio'] = s['bio'] speakers[new_key]['id'] = new_key speakers[new_key]['title'] = s['tagLine'] speakers[new_key]['company'] = s['tagLine'] speakers[new_key]['order'] = i speakers[new_key]['photoUrl'] = s['profilePicture'] speakers[new_key]['name'] = s['fullName'] speakers[new_key]['featured'] = True speakers[new_key]['socials'] = [] for l in s['links']: to_add = {} to_add['name'] = l['title'] to_add['link'] = l['url'] to_add['icon'] = l['title'].lower() speakers[new_key]['socials'].append(to_add) with open('sessionize_speakers.json', 'w') as f: f.write(json.dumps(speakers))
TeamMacLean/cgr_view
cgr/__init__.py
""" cgr A module for creating, saving and drawing k-mer matrices and Chaos Game Representations (CGRs) of nucleotide sequences Prerequisites ------------- - Jellyfish An external program for counting k-mers. Must be accessible on the path. You can install from conda as follows: .. highlight:: bash .. code-block:: bash conda install -c bioconda jellyfish Quickstart ---------- + Input fasta file, get cgr * one cgr for each entry in the fasta file .. highlight:: python .. code-block:: python cgr.from_fasta("my_seqs.fa", outfile = "my_cgrs", k = 7) * just one cgr with all entries in the fasta file (eg for genomes and contigs) .. highlight:: python .. code-block:: python cgr.from_fasta("my_genome.fa", outfile = "genome_cgr", k = 7, as_single = True) Workflow: --------- 1. make kmer count db in Jellyfish from fasta -> generate cgr from db. 2. optionally merge cgrs into single cgr as separate channels 3. stack all composed cgrs into an array of cgrs 4. save as numpy binary (.npy) files Usage: ------ 1. Import module .. highlight:: python .. code-block:: python import cgr 2. Make kmer count db .. highlight:: python .. code-block:: python cgr.run_jellyfish("test_data/NC_012920.fasta", 11, "11mer.jf") cgr.run_jellyfish("test_data/NC_012920.fasta", 10, "10_mer.jf") 2. Load CGRs from kmer count db .. highlight:: python .. code-block:: python cgr1 = cgr.cgr_matrix("/Users/macleand/Desktop/athal-5-mers.jf") cgr2 = cgr.cgr_matrix("test_data/five_mer.jf") 3. Draw a cgr and save to file * just one cgr, can choose colour (value of 'h') and which channel to put cgr in .. highlight:: python .. code-block:: python cgr.draw_cgr(cgr1, h = 0.64, v = 1.0, out = "my_cgr.png", resize = 1000, main = "s" ) * two cgrs, first in tuple goes in 'h', second goes in 's'. Can set 'v' .. highlight:: python .. code-block:: python cgr.draw_cgr( (cgr1, cgr1), v = 1.0, out = "two_cgrs.png") * three cgrs 'h','s' and 'v' are assigned as order in tuple .. highlight:: python .. code-block:: python cgr.draw_cgr( (cgr1, cgr1, cgr1) ) 4. Save a single cgr into a text file .. highlight:: python .. code-block:: python cgr.save_as_csv(cgr1, file = "out.csv") 5. Join n cgrs into one, extending the number of channels ... .. highlight:: python .. code-block:: python merged_cgr = cgr.join_cgr( (cgr1, cgr2, ... ) ) 6. Write to file (numpy binary) .. highlight:: python .. code-block:: python cgr.save_cgr("my_cgr, merged_cgr ) 7. Input fasta file, get cgr * one cgr for each entry in the fasta file .. highlight:: python .. code-block:: python cgr.from_fasta("my_seqs.fa", outfile = "my_cgrs", k = 7) * just one cgr with all entries in the fasta file (eg for genomes and contigs) .. highlight:: python .. code-block:: python cgr.from_fasta("my_genome.fa", outfile = "genome_cgr", k = 7, as_single = True) """ import os import subprocess import math import numpy import scipy import re import matplotlib.pyplot as plt import skimage.color import skimage.io import skimage.transform import tempfile from Bio import SeqIO from typing import Generator, List, Tuple def estimate_genome_size(fasta: str) -> int: """ Guesses genome size from fasta file size, assumes 1 byte ~= 1 base :param: fasta str -- a fasta file :return: int -- approximate genome size in nucleotides """ return (os.path.getsize(fasta)) def run_jellyfish(fasta: str, k: int, out: str) -> int: """ runs Jellyfish on fasta file using k kmer size, produces Jellyfish db file as side effect. :param: fasta str -- a fasta file :param: k int -- size of kmers to use :param: out str -- file in which to save kmer db :return: int -- return code of Jellyfish subprocess """ genome_size = estimate_genome_size(fasta) cmd = ["jellyfish", "count", "-m", str(k), "-s", str(genome_size), fasta, "-o", out] result = subprocess.run(cmd) return result.returncode def get_kmer_list(jellyfish: str) -> Generator[List, str, None]: """ runs jellyfish dump on a Jellyfish DB. Captures output as a generator stream. Each item returned is a list [kmer: str, count: str] :param: jellyfish str -- a Jellyfish DB file :return: Generator -- a list of [kmer string, times_kmer_seen] """ cmd = ["jellyfish", "dump", "-c", jellyfish] proc = subprocess.Popen(cmd, stdout=subprocess.PIPE) for line in proc.stdout: yield line.decode("utf-8").rstrip().split(" ") proc.wait() proc.stdout.close() def get_grid_size(k: int) -> int: """ returns the grid size (total number of elements for a cgr of k length kmers :param: k int -- the value of k to be used :return: int -- the total number of elements in the grid """ return int(math.sqrt(4 ** k)) def get_coord(kmer: str) -> List[int]: """ given a kmer gets the coordinates of the box position in the cgr grid, returns as list [x,y] of coordinates :param: kmer str -- a string of nucleotides :return: coords [x,y] -- the x,y positions of the nucleotides in the cgr """ grid_size = get_grid_size(len(kmer)) maxx = grid_size maxy = grid_size posx = 1 posy = 1 for char in kmer: if char == "C": posx += (maxx / 2) elif char == "T": posy += (maxy / 2) elif char == "G": posx += (maxx / 2) posy += (maxy / 2) maxx = (maxx / 2) maxy /= 2 return [int(posx) - 1, int(posy) - 1] def get_k(jellyfish: str) -> int: """ asks the jellyfish file what value was used for k :param: jellyfish str -- jellyfish DB file :return: int -- length of k used """ cmd = ["jellyfish", "info", jellyfish] result = subprocess.run(cmd, capture_output=True) r = re.match(r".*count\s-m\s(\d+)", result.stdout.decode("utf-8")) return int(r.group(1)) def get_max_count(jellyfish) -> int: """ estimates the count of the most represented kmer in the jellyfish file by using the last bucket of the :param jellyfish: :return: int estimated count of the most represented kmer """ cmd = ["jellyfish", "histo", jellyfish ] proc = subprocess.Popen(cmd, stdout=subprocess.PIPE) counts = [] for line in proc.stdout: counts.append( line.decode("utf-8").rstrip().split(" ") ) proc.wait() proc.stdout.close() return int(counts[-1][0]) def cgr_matrix(jellyfish: str) -> scipy.sparse.dok_matrix: """ Main function, creates the cgr matrix, a sparse matrix of type scipy.sparse.dok_matrix Runs the cgr process on a jellyfish file and returns a scipy.sparse.dok_matrix object of the CGR with dtype int32 Only observed kmers are represented, absent coordinates mean 0 counts for the kmer at that coordinate. :param: jellyfish str -- jellyfish DB file :return: scipy.sparse.dok_matrix -- sparse matrix of kmer counts """ k = get_k(jellyfish) max_c = get_max_count(jellyfish) dtype_to_use = numpy.uint8 if max_c > 255: dtype_to_use = numpy.uint16 grid_size = get_grid_size(k) cgr_mat = scipy.sparse.dok_matrix((grid_size, grid_size), dtype=dtype_to_use) for kmer, count in get_kmer_list(jellyfish): x, y = get_coord(kmer) cgr_mat[x, y] = count return cgr_mat def join_cgr(cgrs: tuple) -> numpy.ndarray: """ Takes tuple of cgrs of shape (n,n) and returns one stacked array of size (n,n, len(cgrs) ) :param: cgrs tuple -- tuple of cgrs to be joined :return: numpy.ndarray """ return numpy.dstack(cgrs) def save_as_csv(cgr_matrix: scipy.sparse.dok_matrix, file: str = "cgr_matrix.csv", delimiter: str = ",", fmt: str = '%d'): """ Writes simple 1 channel cgr matrix to CSV file. See also numpy.savetxt :param: cgr_matrix scipy.sparse.dok_matrix -- cgr_matrix to save :param: file str -- filename to write to :param: delimiter str -- column separator character :param: fmt str -- text format string :return: None """ numpy.savetxt(file, cgr_matrix.toarray(), delimiter=delimiter, fmt=fmt) def make_blanks_like(a: scipy.sparse.dok_matrix , h: float=1.0, s: float=1.0, v: float=1.0) -> Tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray]: """ returns tuple of numpy.ndarrays with default values of h,s and v of shape of a :param: a scipy.sparse.dok_matrix -- a cgr matrix to make blanks like :param: h float -- the values with which to fill the first numpy.ndarray :param: s float -- the values with which to fill the second numpy.ndarray :param: v float -- the values with which to fill the third numpy.ndarray :return: Tuple of numpy.ndarray """ return numpy.full_like(a, h), numpy.full_like(a, s), numpy.full_like(a, v) def scale_cgr(cgr_matrix: scipy.sparse.dok_matrix) -> scipy.sparse.dok_matrix: """ returns scaled version of cgr_matrix in range 0..1 :param: cgr_matrix scipy.sparse.dok_matrix -- matrix to scale :return: scaled scipy.sparse.dok_matrix """ return (cgr_matrix / max(cgr_matrix.values())).toarray() def blocky_scale(im: numpy.ndarray, nR: int, nC: int) -> numpy.ndarray: """ Upscales an array in preparation for drawing. By default the array is a square with sqrt(k ** 4) wide and high. For many values of k this will be too small to view well on a monitor. This function does a scale operartion that increases the size of the image by simply increasing the pixels in each square. :param: im numpy.ndarray -- the image to be scaled :param: nR int -- the number of height pixels to be in the final image :param: nC int -- the number of width pixels to be in the final image :return: numpy.ndarray -- upscaled image """ nR0 = len(im) # source number of rows nC0 = len(im[0]) # source number of columns return numpy.asarray([[im[int(nR0 * r / nR)][int(nC0 * c / nC)] for c in range(nC)] for r in range(nR)]) def resize_rgb_out(rgb: numpy.ndarray,resize: int) -> numpy.ndarray: """ given an rgb image in one pixel per kmer size, increases size so that the resulting image is resize * resize pixels :param: rgb numpy.ndarray -- an RGB image array :param: resize -- pixel width (and therefore height) of resulting image :return: numpy.ndarray -- resized image with shape (resize, resize) """ r = blocky_scale(rgb[:, :, 0], resize, resize) g = blocky_scale(rgb[:, :, 1], resize, resize) b = blocky_scale(rgb[:, :, 2], resize, resize) return numpy.dstack((r, g, b)) def is_cgr_matrix(obj) -> bool: """returns true if obj is a scipy.sparse.dok.dok_matrix object """ return type(obj) == scipy.sparse.dok.dok_matrix def draw_cgr(cgr_matrices: scipy.sparse.dok_matrix, h: float = 0.8, s: float = 0.5, v: float = 1.0, main: str = "s", show: bool = True, write: bool = True, out: str = "cgr.png", resize: bool = False) -> None: """Draws cgrs to a file. Allows user to set which of up to 3 provided cgr matrices goes in at which of the H, S or V image channels. Typically for one channel, select h to specify the image colour and set cgr as s to change that colour according to counts in cgr. Set v to 1.0 for maximum brightness. :param: cgr_matrices scipy.sparse.dok_matrix or tuple of scipy.sparse.dok_matrix elements, cgrs to be drawn. Tuple provides order for HSV channels of image. :param: h float -- (0..1) value for h channel if not used for cgr data :param: s float -- (0..1) value for s channel if not used for cgr data :param: v float -- (0..1) value for v channel if not used for cgr data :param: main str -- the channel to place the cgr matrix in if a single cgr matrix is passed :param: show bool -- render CGR picture to screen :param: write -- write CGR picture to file :param: out str -- filename to write to :param: resize bool or int -- if False no image resizing is done, if an int image is rescaled to resize pixels width and height :return: None """ if is_cgr_matrix(cgr_matrices): #one channel draw_single_cgr(cgr_matrices, h=h, s=s, v=v, main=main, show = show, write = write, out = out, resize = resize) elif all( [is_cgr_matrix(o) for o in cgr_matrices] ) and len(cgr_matrices) ==2: #all cgr matrices draw_two_cgrs(cgr_matrices, v = v, show = show, write = write, out = out, resize = resize) elif all( [is_cgr_matrix(o) for o in cgr_matrices] ) and len(cgr_matrices) == 3 : draw_three_cgrs(cgr_matrices, show = show, write = write, out = out, resize = resize) else: raise Exception("don't know what to do, cgr_matrices must be one cgr_matrix or a tuple of 2 or 3 cgr_matrices.") def draw_single_cgr(cgr_matrix, h=0.8, s=0.5, v=1.0, main="s", show = True, write = True, out = "cgr.png", resize = False): """ draws a single cgr image, selecting channels and resizing as appropriate :param: cgr_matrix scipy.sparse.dok_matrix to be drawn. :param: h float -- (0..1) value for h channel if not used for cgr data :param: s float -- (0..1) value for s channel if not used for cgr data :param: v float -- (0..1) value for v channel if not used for cgr data :param: main str -- the channel to place the cgr matrix in :param: show bool -- render CGR picture to screen :param: write -- write CGR picture to file :param: out str -- filename to write to :param: resize bool or int -- if False no image resizing is done, if an int image is rescaled to resize pixels width and height :return: None """ scaled = scale_cgr( cgr_matrix ) h_blank, s_blank, v_blank = make_blanks_like(scaled, h,s,v) hsv = None if main == "h": hsv = numpy.dstack((scaled, s_blank, v_blank)) elif main == "s": hsv = numpy.dstack((h_blank, scaled, v_blank)) elif main == "v": hsv = numpy.dstack((h_blank, s_blank, scaled)) rgb = skimage.color.hsv2rgb(hsv) if show: plt.imshow(rgb) plt.show() if write: if resize: rgb = resize_rgb_out(rgb, resize) skimage.io.imsave(out, rgb) def draw_two_cgrs(cgr_matrices, v = 1.0, show = True, write = True, out = "cgr.png", resize = False ): """draws two cgr matrices into a single image. first matrix of tuple becomes h channel, second of tuple becomes v channel :param: cgr_matrices tuple of scipy.sparse.dok_matrix elements, cgrs to be drawn. :param: v float -- (0..1) value for v channel :param: show bool -- render CGR picture to screen :param: write -- write CGR picture to file :param: out str -- filename to write to :param: resize bool or int -- if False no image resizing is done, if an int image is rescaled to resize pixels width and height :return: None """ scaled_l = [scale_cgr(cgrm) for cgrm in cgr_matrices] v_blank = make_blanks_like(scaled_l[0], v=v)[2] hsv_stack = numpy.dstack((scaled_l[0], scaled_l[1], v_blank)) rgb = skimage.color.hsv2rgb(hsv_stack) if show: draw(rgb) if write: write_out(rgb, out, resize) def draw_three_cgrs(cgr_matrices,show = True, write = True, out = "cgr.png", resize = False): """Draws a tuple of 3 cgr matrices as an image :param: cgr_matrices tuple of scipy.sparse.dok_matrix elements, cgrs to be drawn. Tuple provides order for HSV channels of image :param: show bool -- render CGR picture to screen :param: write -- write CGR picture to file :param: out str -- filename to write to :param: resize bool or int -- if False no image resizing is done, if an int image is rescaled to resize pixels width and height :return: None """ scaled_t = (scale_cgr(cgrm) for cgrm in cgr_matrices) hsv_stack = numpy.dstack(scaled_t) rgb = skimage.color.hsv2rgb(hsv_stack) if show: draw(rgb) if write: write_out(rgb, out, resize) def draw(rgb: numpy.ndarray) -> None: """ renders RGB array on the screen. :param: rgb numpy.ndarray -- RGB channel image """ plt.imshow(rgb) plt.show() def write_out(rgb: numpy.ndarray, out: str, resize: int) -> None: """ writes RGB array as image :param rgb: numpy.ndarray -- RGB channel image :param out: str file to write to :param resize: bool or int. If False will not resize, if int will resize image up to that size :return: None """ if resize: rgb = resize_rgb_out(rgb, resize) skimage.io.imsave(out, rgb) def stack_cgrs(cgr_matrices: Tuple) -> numpy.ndarray: """ stacks cgrs of tuple of N numpy.ndarrays of shape (w,h) returns ndarray of ndarrays of shape (w,h,N) :param cgr_matrices: tuple of cgr_matrices :return: numpy.ndarray """ cgr_t = tuple(c.toarray() for c in cgr_matrices) return numpy.stack(cgr_t, axis=-1) def save_cgr(cgr_obj: numpy.ndarray, outfile: str = "cgr") -> None: """ Saves cgr_obj as numpy .npy file. cgr_obj one or more dimensional numpy.ndarray. saves as ndarray not dokmatrix, so can be loaded in regular numpy as collections of cgrs :param cgr_obj: numpy.ndarray constructed cgr_object to save :param outfile: str file :return: None """ numpy.save(outfile, cgr_obj, allow_pickle=True) def load_npy(file: str) -> numpy.ndarray: """ loads numpy .npy file as ndarray. Useful for restoring collections of cgrs but resulting array is not compatible directly with drawing methods here. :param file str -- numpy .npy file to load :return: numpy.ndarray """ return numpy.load(file, allow_pickle=True) def many_seq_record_to_one_cgr(fa_file: str, k: int) -> scipy.sparse.dok_matrix: """ Reads many sequence records in a FASTA file into a single CGR matrix, treating all sequence records as if they are one sequence, EG as if for a genome sequence in Chromosomes. :param fa_file: str FASTA FILE name :param k: int length of k to use :return: scipy.sparse.dok_matrix """ temp_jf = tempfile.NamedTemporaryFile() run_jellyfish(fa_file, k, temp_jf.name) cgr1 = cgr_matrix(temp_jf.name) temp_jf.close() return cgr1 def many_seq_record_to_many_cgr(seq_record: SeqIO.FastaIO, k: int) -> scipy.sparse.dok_matrix: """ :param seq_record: Bio.SeqIO FASTA record :param k: int size of k to use :return: scipy.sparse.dok_matrix """ temp_fa = tempfile.NamedTemporaryFile() temp_jf = tempfile.NamedTemporaryFile() SeqIO.write(seq_record, temp_fa.name, "fasta") run_jellyfish(temp_fa.name, k, temp_jf.name) cgr1 = cgr_matrix(temp_jf.name) temp_fa.close() temp_jf.close() return cgr1 def from_fasta(fasta_file: str, outfile: str = "my_cgrs", as_single: bool=False, k: int = 7) -> None: """ Factory function to load in a FASTA file and generate a binary .npy of CGRs :param fasta_file: str FASTA file to load :param outfile: str outfile to save :param as_single: bool If True treats all entries as single sequence and return one CGR. If False, treats all entries individually and returns many CGR :param k: int length of kmer to use :return: None """ if as_single: cgr1 = many_seq_record_to_one_cgr(fasta_file, k) save_cgr(cgr1.toarray(), outfile=outfile ) else: cgr_t = tuple( many_seq_record_to_many_cgr(seq_record, k) for seq_record in SeqIO.parse(fasta_file, "fasta") ) cgr1 = stack_cgrs(cgr_t) save_cgr(cgr1, outfile = outfile ) # TODO # test new dtype switching cgr matrix function - try using from_fasta
TeamMacLean/cgr_view
test_cgr.py
<reponame>TeamMacLean/cgr_view import unittest import cgr import os from scipy.sparse import dok_matrix import numpy class CGRTest(unittest.TestCase): def test_guess_genome_size(self): self.assertEqual(cgr.estimate_genome_size("test_data/10kfile"),10240) def test_run_jellyfish(self): result = cgr.run_jellyfish("test_data/NC_012920.fasta", 11, "test_data/tmp.jf") os.remove("test_data/tmp.jf") self.assertEqual(result.returncode, 0) def test_get_kmer_list(self): cgr.run_jellyfish("test_data/NC_012920.fasta", 11, "test_data/tmp.jf") counts = [] for i in cgr.get_kmer_list("test_data/tmp.jf"): counts.append(i) self.assertEqual(len(counts), 16438) self.assertEqual(len(counts[0]), 2) os.remove("test_data/tmp.jf") def test_get_coord(self): self.assertEqual(cgr.get_coord("A"), [0,0]) self.assertEqual(cgr.get_coord("G"), [1,1]) self.assertEqual(cgr.get_coord("C"), [1,0]) self.assertEqual(cgr.get_coord("T"), [0,1]) self.assertEqual(cgr.get_coord("CTGA"), [10,6]) def test_get_k(self): self.assertEqual(cgr.get_k("test_data/five_mer.jf"), 5) def test_cgr_matrix(self): m = cgr.cgr_matrix("test_data/five_mer.jf") row1_col1 = m.toarray()[0][0] self.assertEqual(row1_col1, 70) def test_join_cgr(self): a = [ [1,1,1], [1,1,1] ] a = dok_matrix(a) b = [ [2,2,2], [2,2,2] ] b = dok_matrix(b) c = numpy.array( [ [ [1,2], [1,2], [1,2] ], [ [1,2], [1,2], [1,2] ] ] ) d = cgr.join_cgr(a,b) self.assertEqual(d.tolist(), c.tolist() ) def test_pad_two_channel_to_three(self): a = numpy.array( [ [ [1,2], [1,2], [1,2] ], [ [1,2], [1,2], [1,2] ] ] ) b = numpy.array( [ [ [1,2,0], [1,2,0], [1,2,0] ], [ [1,2,0], [1,2,0], [1,2,0] ] ]) c = cgr.pad_two_channel_to_three(a) self.assertEqual(c.tolist(), b.tolist() ) if __name__ == '__main__': unittest.main()
TeamMacLean/cgr_view
setup.py
<gh_stars>0 import setuptools with open("README.md", "r") as fh: long_description = fh.read() setuptools.setup( name='cgr_view', version='0.0.1dev2', url='https://github.com/TeamMacLean/cgr_view', packages=setuptools.find_packages(), long_description_content_type="text/markdown", license='LICENSE.txt', author='<NAME>', author_email='<EMAIL>', description='A package for rendering Chaos Game Representations of DNA sequences', python_requires='>=3.7', install_requires = [ "numpy >= 1.17", "matplotlib >= 3.1.0", "scipy >= 1.3.1", "pillow >= 6.2.1", "scikit-image >= 0.15.0", "biopython >= 1.7" ] )
cazalaa/connectedhomeip
src/controller/python/chip-repl.py
<reponame>cazalaa/connectedhomeip #!/usr/bin/env python # # Copyright (c) 2021 Project CHIP Authors # All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # from IPython import embed import chip def main(): # The chip import at the top level will be visible in the ipython REPL. embed(header = ''' Welcome to the CHIP python REPL utilty. Usage examples: ######## List available BLE adapters ######### import chip.ble print(chip.ble.GetAdapters()) '''.strip()) if __name__ == "__main__": main()
cazalaa/connectedhomeip
src/controller/python/chip/ble/__init__.py
<filename>src/controller/python/chip/ble/__init__.py # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """BLE-related functionality within CHIP""" import sys import platform import chip.native import ctypes from typing import List from ctypes import c_bool, c_void_p, c_char_p, c_uint from dataclasses import dataclass @dataclass class AdapterInfo: index: int address: str name: str alias: str powered_on: bool def _GetBleLibraryHandle() -> ctypes.CDLL: """ Get the native library handle with BLE method initialization. Retreives the CHIP native library handle and attaches signatures to native methods. """ handle = chip.native.GetLibraryHandle() # Uses one of the type decorators as an indicator for everything being # initialized. Native methods default to c_int return types if handle.pychip_ble_adapter_list_new.restype != c_void_p: setter = chip.native.NativeLibraryHandleMethodArguments(handle) setter.Set('pychip_ble_adapter_list_new', c_void_p, []) setter.Set('pychip_ble_adapter_list_next', c_bool, [c_void_p]) setter.Set('pychip_ble_adapter_list_get_index', c_uint, [c_void_p]) setter.Set('pychip_ble_adapter_list_get_address', c_char_p, [c_void_p]) setter.Set('pychip_ble_adapter_list_get_alias', c_char_p, [c_void_p]) setter.Set('pychip_ble_adapter_list_get_name', c_char_p, [c_void_p]) setter.Set('pychip_ble_adapter_list_is_powered', c_bool, [c_void_p]) setter.Set('pychip_ble_adapter_list_delete', None, [c_void_p]) return handle def GetAdapters() -> List[AdapterInfo]: """Get a list of BLE adapters available on the system. """ handle = _GetBleLibraryHandle() result = [] nativeList = handle.pychip_ble_adapter_list_new() if nativeList == 0: raise Exception('Failed to get BLE adapter list') try: while handle.pychip_ble_adapter_list_next(nativeList): result.append( AdapterInfo( index=handle.pychip_ble_adapter_list_get_index(nativeList), address=handle.pychip_ble_adapter_list_get_address(nativeList).decode('utf8'), name=handle.pychip_ble_adapter_list_get_name(nativeList).decode('utf8'), alias=handle.pychip_ble_adapter_list_get_alias(nativeList).decode('utf8'), powered_on=handle.pychip_ble_adapter_list_is_powered(nativeList), )) finally: handle.pychip_ble_adapter_list_delete(nativeList) return result __all__ = [ 'GetAdapters', ]
xyb/python-xxhash
tests/test_xxh3_64.py
<filename>tests/test_xxh3_64.py from __future__ import print_function import os import sys import unittest import random import xxhash class TestXXH(unittest.TestCase): def test_xxh3_64(self): self.assertEqual(xxhash.xxh3_64('a').intdigest(), 16629034431890738719) self.assertEqual(xxhash.xxh3_64('a', 0).intdigest(), 16629034431890738719) self.assertEqual(xxhash.xxh3_64('a', 1).intdigest(), 15201566949650179872) self.assertEqual(xxhash.xxh3_64('a', 2**64-1).intdigest(), 4875116479388997462) def test_xxh3_64_intdigest(self): self.assertEqual(xxhash.xxh3_64_intdigest('a'), 16629034431890738719) self.assertEqual(xxhash.xxh3_64_intdigest('a', 0), 16629034431890738719) self.assertEqual(xxhash.xxh3_64_intdigest('a', 1), 15201566949650179872) self.assertEqual(xxhash.xxh3_64_intdigest('a', 2**64-1), 4875116479388997462) def test_xxh3_64_update(self): x = xxhash.xxh3_64() x.update('a') self.assertEqual(xxhash.xxh3_64('a').digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('a'), x.digest()) x.update('b') self.assertEqual(xxhash.xxh3_64('ab').digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('ab'), x.digest()) x.update('c') self.assertEqual(xxhash.xxh3_64('abc').digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('abc'), x.digest()) seed = random.randint(0, 2**64) x = xxhash.xxh3_64(seed=seed) x.update('a') self.assertEqual(xxhash.xxh3_64('a', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('a', seed), x.digest()) x.update('b') self.assertEqual(xxhash.xxh3_64('ab', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('ab', seed), x.digest()) x.update('c') self.assertEqual(xxhash.xxh3_64('abc', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_64_digest('abc', seed), x.digest()) def test_xxh3_64_reset(self): x = xxhash.xxh3_64() h = x.intdigest() x.update('x' * 10240) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_64_seed_reset(self): seed = random.randint(0, 2**64-1) x = xxhash.xxh3_64(seed=seed) h = x.intdigest() x.update('x' * 10240) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_64_reset_more(self): x = xxhash.xxh3_64() h = x.intdigest() for i in range(random.randint(100, 200)): x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(i)) x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(100)) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_64_seed_reset_more(self): seed = random.randint(0, 2**64-1) x = xxhash.xxh3_64(seed=seed) h = x.intdigest() for i in range(random.randint(100, 200)): x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(i)) x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(100)) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_64_copy(self): a = xxhash.xxh3_64() a.update('xxhash') b = a.copy() self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) b.update('xxhash') self.assertNotEqual(a.digest(), b.digest()) self.assertNotEqual(a.intdigest(), b.intdigest()) self.assertNotEqual(a.hexdigest(), b.hexdigest()) a.update('xxhash') self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) def test_xxh3_64_overflow(self): s = 'I want an unsigned 64-bit seed!' a = xxhash.xxh3_64(s, seed=0) b = xxhash.xxh3_64(s, seed=2**64) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=0)) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=2**64)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=0)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=2**64)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=0)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=2**64)) a = xxhash.xxh3_64(s, seed=1) b = xxhash.xxh3_64(s, seed=2**64+1) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=1)) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=2**64+1)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=1)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=2**64+1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=2**64+1)) a = xxhash.xxh3_64(s, seed=2**65-1) b = xxhash.xxh3_64(s, seed=2**66-1) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=2**65-1)) self.assertEqual(a.intdigest(), xxhash.xxh3_64_intdigest(s, seed=2**66-1)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=2**65-1)) self.assertEqual(a.digest(), xxhash.xxh3_64_digest(s, seed=2**66-1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=2**65-1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_64_hexdigest(s, seed=2**66-1)) if __name__ == '__main__': unittest.main()
xyb/python-xxhash
tests/test_xxh3_128.py
import os import unittest import random import xxhash class TestXXH(unittest.TestCase): def test_xxh3_128(self): self.assertEqual(xxhash.xxh3_128('a').intdigest(), 225219434562328483135862406050043285023) self.assertEqual(xxhash.xxh3_128('a', 0).intdigest(), 225219434562328483135862406050043285023) self.assertEqual(xxhash.xxh3_128('a', 1).intdigest(), 337425133163118381928709500770786453280) self.assertEqual(xxhash.xxh3_128('a', 2**64-1).intdigest(), 198297796855923085494266857744987477846) def test_xxh3_128_intdigest(self): self.assertEqual(xxhash.xxh3_128_intdigest('a'), 225219434562328483135862406050043285023) self.assertEqual(xxhash.xxh3_128_intdigest('a', 0), 225219434562328483135862406050043285023) self.assertEqual(xxhash.xxh3_128_intdigest('a', 1), 337425133163118381928709500770786453280) self.assertEqual(xxhash.xxh3_128_intdigest('a', 2**64-1), 198297796855923085494266857744987477846) def test_xxh3_128_update(self): x = xxhash.xxh3_128() x.update('a') self.assertEqual(xxhash.xxh3_128('a').digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('a'), x.digest()) x.update('b') self.assertEqual(xxhash.xxh3_128('ab').digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('ab'), x.digest()) x.update('c') self.assertEqual(xxhash.xxh3_128('abc').digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('abc'), x.digest()) seed = random.randint(0, 2**64) x = xxhash.xxh3_128(seed=seed) x.update('a') self.assertEqual(xxhash.xxh3_128('a', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('a', seed), x.digest()) x.update('b') self.assertEqual(xxhash.xxh3_128('ab', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('ab', seed), x.digest()) x.update('c') self.assertEqual(xxhash.xxh3_128('abc', seed).digest(), x.digest()) self.assertEqual(xxhash.xxh3_128_digest('abc', seed), x.digest()) def test_xxh3_128_reset(self): x = xxhash.xxh3_128() h = x.intdigest() x.update('x' * 10240) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_128_seed_reset(self): seed = random.randint(0, 2**64-1) x = xxhash.xxh3_128(seed=seed) h = x.intdigest() x.update('x' * 10240) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_128_reset_more(self): x = xxhash.xxh3_128() h = x.intdigest() for i in range(random.randint(100, 200)): x.reset() for i in range(10, 1000): x.update(os.urandom(i)) x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(100)) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_128_seed_reset_more(self): seed = random.randint(0, 2**64-1) x = xxhash.xxh3_128(seed=seed) h = x.intdigest() for i in range(random.randint(100, 200)): x.reset() for i in range(10, 1000): x.update(os.urandom(i)) x.reset() self.assertEqual(h, x.intdigest()) for i in range(10, 1000): x.update(os.urandom(100)) x.reset() self.assertEqual(h, x.intdigest()) def test_xxh3_128_copy(self): a = xxhash.xxh3_128() a.update('xxhash') b = a.copy() self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) b.update('xxhash') self.assertNotEqual(a.digest(), b.digest()) self.assertNotEqual(a.intdigest(), b.intdigest()) self.assertNotEqual(a.hexdigest(), b.hexdigest()) a.update('xxhash') self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) def test_xxh3_128_overflow(self): s = 'I want an unsigned 64-bit seed!' a = xxhash.xxh3_128(s, seed=0) b = xxhash.xxh3_128(s, seed=2**64) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=0)) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=2**64)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=0)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=2**64)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=0)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=2**64)) a = xxhash.xxh3_128(s, seed=1) b = xxhash.xxh3_128(s, seed=2**64+1) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=1)) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=2**64+1)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=1)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=2**64+1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=2**64+1)) a = xxhash.xxh3_128(s, seed=2**65-1) b = xxhash.xxh3_128(s, seed=2**66-1) self.assertEqual(a.seed, b.seed) self.assertEqual(a.intdigest(), b.intdigest()) self.assertEqual(a.hexdigest(), b.hexdigest()) self.assertEqual(a.digest(), b.digest()) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=2**65-1)) self.assertEqual(a.intdigest(), xxhash.xxh3_128_intdigest(s, seed=2**66-1)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=2**65-1)) self.assertEqual(a.digest(), xxhash.xxh3_128_digest(s, seed=2**66-1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=2**65-1)) self.assertEqual(a.hexdigest(), xxhash.xxh3_128_hexdigest(s, seed=2**66-1)) if __name__ == '__main__': unittest.main()
taipahuchu/language-Identification-
code/batcher.py
"""This file holds the Dataset class, which helps with the loading and organizing of the training data.""" import argparse import collections import gzip import itertools import numpy as np import re import sys np.random.seed(666) def LoadData(filename, mode='train', model='tweet'): """Load data stored in tweetlid format. (i.e. tab-separated tweetid, language, tweet) Partitioning between train/dev/eval is done by the last digit of the id number for each training example. Digits 2 through 9 are used for training and 1 is used as a dev set. Currently, the eval set is never loaded and confusingly to load the dev set you have to use 'eval' for the mode argument. This function splits the tweet into a list of units. The level of splitting is controlled by the model argument. Args: filename: where to get the data mode: train or eval or all unit: word, tweets or chars Returns: tuple of sentences, labels and ids """ ids, labels, sentences = [], [], [] with gzip.open(filename, 'r') as f: for line in f: tweetid, lang, tweet = line.split('\t') idx = int(tweetid) % 10 # use this to partition data if mode == 'train' and idx < 2: continue if mode == 'eval+final' and idx > 2: continue if mode == 'eval' and idx != 1: continue if mode == 'final' and idx != 0: continue ids.append(tweetid) # split used to handle code switching tweets labels.append(re.split(r'\+|/', lang)) # The { and } brackets are used for start/end symbols if model in ['word', 'tweet']: #split on whitespace to get words sentences.append(['{'] + [unicode(x_.decode('utf8')) for x_ in tweet.split()] + ['}']) elif model=='char': #include full tweet as single unicode string (list of length 3) sentences.append([u'{'] + [unicode(tweet.decode('utf8'))] + [u'}']) else: msg = 'Invalid unit type <{0}> for tokenizing tweet'.format(model) raise ValueError(msg) print '{0} examples loaded'.format(len(sentences)) return sentences, labels, ids class Dataset(object): def __init__(self, batch_size, preshuffle=True, name='unnamed'): """Init the dataset object. Args: batch_size: size of mini-batch preshuffle: should the order be scrambled before the first epoch name: optional name for the dataset """ self._sentences = [] self._labels = [] self._ids = [] self.dataset_weights = [] self._lines = [] self.name = name self.batch_size = batch_size self.preshuffle = preshuffle def ReadData(self, filename, mode, modeltype, weight=1.0): d = LoadData(filename, mode, modeltype) self.AddDataSource(d, weight=weight) def AddDataSource(self, data, weight=1.0): sentences, labels, ids = data self._sentences.append(sentences) self._labels.append(labels) self.dataset_weights.append(weight) self._ids.append(ids) def GetSentences(self): return itertools.chain(*self._sentences) def GetIds(self): return [x for x in itertools.chain(*self._ids)] def GetLabelSet(self): label_set = set() for d in self._labels: label_set.update([x for x in itertools.chain(*d)]) return label_set def Prepare(self,in_vocab,out_vocab,und_label='und',ignore_categories=[]): # Add a dummy dataset to make the batch size evenly divide the number # of sentences. batch_size = self.batch_size total_sentences = sum([len(x) for x in self._sentences]) r = total_sentences % batch_size if r > 0: n = batch_size - r self.AddDataSource(([list('{dummy}')] * n, [[und_label]] * n, [0] * n), weight=0.0) sentences = list(itertools.chain(*self._sentences)) labels = list(itertools.chain(*self._labels)) self.example_weights = [] for i in xrange(len(self.dataset_weights)): w = self.dataset_weights[i] for _ in xrange(len(self._sentences[i])): self.example_weights.append(w) self.example_weights = np.array(self.example_weights) self.seq_lens = np.array([len(x) for x in sentences]) self.max_sequence_len = self.seq_lens.max() self.batch_size = batch_size self.current_idx = 0 self.sentences = self.GetNumberLines(sentences, in_vocab, self.max_sequence_len) self.labels = np.zeros((len(labels), len(out_vocab))) for i, w in enumerate(labels): for w_ in w: self.labels[i, out_vocab[w_]] = 1.0 self.labels[i, :] /= self.labels[i, :].sum() # class weights # There is a hack to only use the examples with weight 1 as a way # to prevent wikipedia from dominating the weights. counts = self.labels[self.example_weights == 1, :].sum(axis=0) self.w = 1.0/(1.0 + counts) self.w /= self.w.mean() # scale the class weights to reasonable values self.N = len(sentences) if self.preshuffle: self._Permute() @staticmethod def GetNumberLines(lines, vocab, pad_length): """Convert list of words to matrix of word ids.""" out = [] for line in lines: if len(line) < pad_length: line += ['}'] * (pad_length - len(line)) out.append([vocab[w] for w in line]) return np.array(out) def GetNumBatches(self): """Returns num batches per epoch.""" return self.N / self.batch_size def _Permute(self): """Shuffle the training data.""" s = np.arange(self.N) np.random.shuffle(s) self.sentences = self.sentences[s, :] self.seq_lens = self.seq_lens[s] self.labels = self.labels[s, :] self.example_weights = self.example_weights[s] def GetNextBatch(self): if self.current_idx + self.batch_size > self.N: self.current_idx = 0 self._Permute() idx = range(self.current_idx, self.current_idx + self.batch_size) self.current_idx += self.batch_size return (self.sentences[idx, :], self.seq_lens[idx], self.labels[idx, :], self.example_weights[idx]) # print some dataset statistics if __name__ == '__main__': parser = argparse.ArgumentParser() parser.add_argument('data') args = parser.parse_args() _, labels, _ = LoadData(args.data, 'all') total = len(labels) print 'total sentences: {0}'.format(total) unique_labels = len(set([tuple(x) for x in labels])) print 'unique labels: {0}'.format(unique_labels) counts = collections.Counter([tuple(s) for s in labels]) for lang in sorted(counts.keys()): print '{0}\t{1}\t{2:.1f}'.format(' '.join(lang), counts[lang], 100 * counts[lang] / float(total))
taipahuchu/language-Identification-
code/models.py
<reponame>taipahuchu/language-Identification- import tensorflow as tf import numpy as np class BaseModel(object): """Holds code shared between all the different model variants.""" def __init__(self, batch_size, max_sequence_len, out_vocab_size, c2v, dropout_keep_prob=0.0): self._batch_size = batch_size self._dropout_keep_prob = dropout_keep_prob self._out_vocab_size = out_vocab_size self.x = tf.placeholder(tf.int32, [batch_size, max_sequence_len], name='x') self.y = tf.placeholder(tf.float32, [batch_size, out_vocab_size], name='y') # The bidirectional rnn code requires seq_lens as int64 self.seq_lens = tf.placeholder(tf.int64, [batch_size], name='seq_lens') self.example_weights = tf.placeholder(tf.float32, [batch_size], name='example_weights') embeddings = c2v.GetEmbeddings(self.x) self._inputs = [tf.squeeze(input_, [1]) for input_ in tf.split(1, max_sequence_len, embeddings)] # Need to prepare a mask to zero out the padding symbols. # Make a batch_size x max_sequence_len matrix where each # row contains the length repeated max_sequence_len times. lengths_transposed = tf.expand_dims(tf.to_int32(self.seq_lens), 1) lengths_tiled = tf.tile(lengths_transposed, [1, max_sequence_len]) # Make a matrix where each row contains [0, 1, ..., max_sequence_len] r = tf.range(0, max_sequence_len, 1) range_row = tf.expand_dims(r, 0) range_tiled = tf.tile(range_row, [batch_size, 1]) # Use the logical operations to create a mask indicator = tf.less(range_tiled, lengths_tiled) sz = [batch_size, max_sequence_len] self._mask = tf.select(indicator, tf.ones(sz), tf.zeros(sz)) def _DoPredictions(self, in_size, mats, class_weights=None): """Takes in an array of states and calculates predictions. Get the cross-entropy for each example in the vector self._xent. Args: in_size: size of the hidden state vectors mats: list of hidden state vectors """ pred_mat = tf.get_variable('pred_mat', [in_size, self._out_vocab_size]) pred_bias = tf.get_variable('pred_bias', [self._out_vocab_size]) # Make a prediction on every word. def GetWordPred(o_): logits = tf.nn.xw_plus_b(o_, pred_mat, pred_bias) return tf.nn.softmax(logits) self.preds_by_word = tf.pack([GetWordPred(o_) for o_ in mats]) self.cs = self._mask / tf.reduce_sum(self._mask, 1, keep_dims=True) # The final prediction is the average of the predictions for each word # weighted by the individual confidence/utility scores. preds_weighted = tf.mul(tf.reshape(tf.transpose(self.cs), [-1, 1]), tf.reshape(self.preds_by_word, [-1, self._out_vocab_size])) preds_weighted_reshaped = tf.reshape(preds_weighted, self.preds_by_word.get_shape()) self.probs = tf.reduce_sum(preds_weighted_reshaped, 0) self._xent = _SafeXEnt(self.y, self.probs, class_weights=class_weights) class WordAvgModel(BaseModel): #formerly SimpleModel """A bag of word /predictions/.""" def __init__(self, out_vocab_size=None, batch_size=10, model_params=None, c2v=None, max_sequence_len=None, dropout_keep_prob=None, weights=None): super(WordAvgModel, self).__init__(batch_size, max_sequence_len, out_vocab_size, c2v) super(WordAvgModel, self)._DoPredictions(c2v.embedding_dims, self._inputs) self.cost = tf.reduce_mean(self.example_weights * self._xent) class WordSeqModel(BaseModel): """A bag of word embeddings.""" def __init__(self, out_vocab_size=None, batch_size=10, model_params=None, c2v=None, max_sequence_len=None, dropout_keep_prob=None, weights=None): super(WordSeqModel, self).__init__(batch_size, max_sequence_len, out_vocab_size, c2v) in_size = self._inputs[0].get_shape()[1].value # Also, output confidence scores at every word. confidence_mat = tf.get_variable('confidence_mat', [in_size, 1]) confidence_scores = tf.concat(1, [tf.matmul(o_, confidence_mat) for o_ in self._inputs]) # dropout on confidence_scores random_tensor = (1.0 - self._dropout_keep_prob + tf.random_uniform(tf.shape(confidence_scores))) binary_tensor = -50.0 * tf.floor(random_tensor) csshape = confidence_scores.get_shape() self.cs = tf.nn.softmax(tf.constant(1.0, shape=csshape)) # The final prediction is the average of the predictions for each word # weighted by the individual confidence/utility scores. wvs = tf.pack(self._inputs) wvs_weighted = tf.mul(tf.reshape(tf.transpose(self.cs), [-1, 1]), tf.reshape(wvs, [-1, in_size])) wvs_weighted_reshaped = tf.reshape(wvs_weighted, wvs.get_shape()) wvsum = tf.reduce_sum(wvs_weighted_reshaped,0) pred_mat = tf.get_variable('pred_mat', [in_size, self._out_vocab_size]) pred_bias = tf.get_variable('pred_bias', [self._out_vocab_size]) # Make a prediction for each tweet. def GetWordPred(o_): logits = tf.nn.xw_plus_b(o_, pred_mat, pred_bias) return tf.nn.softmax(logits) preds = GetWordPred(wvsum) z = tf.tile(tf.reshape(tf.reduce_sum(preds,1),[-1,1]), [1, out_vocab_size]) self.preds, self.z = preds, z self.probs = tf.div(preds, z) #normalize self.unweighted_xent = _SafeXEnt(self.y, self.probs) self._xent = _SafeXEnt(self.y, self.probs, class_weights=weights) self.cost = tf.reduce_mean(self.example_weights * self._xent) class TweetSeqModel(BaseModel): #formerly SeqModel """Single layer LSTM on top of the word embeddings. Lang id predictions are done on each word and then combined via a weighted average. """ def __init__(self, out_vocab_size=None, batch_size=10, model_params=None, c2v=None, max_sequence_len=None, dropout_keep_prob=None, weights=None): """Initialize the TweetSeqModel Args: out_vocab_size: how many languages we are predicting batch_size: minibatch size model_params: dictionary of other model parameters c2v: char2vec class instance max_sequence_len: length of all the input sequences dropout_keep_prob: dropout probability indicator weights: class weights """ hidden_size = model_params['model_hidden_size'] proj_size = model_params['model_proj_size'] # optional, can be None super(TweetSeqModel, self).__init__(batch_size, max_sequence_len, out_vocab_size, c2v, dropout_keep_prob) weights = tf.constant(weights, dtype=tf.float32, name='class_weights') def GetCell(): """Creates an LSTM cell with dropout.""" c = tf.nn.rnn_cell.LSTMCell(hidden_size, use_peepholes=model_params['peepholes'], num_proj=proj_size) if dropout_keep_prob is not None: c = tf.nn.rnn_cell.DropoutWrapper(c, input_keep_prob=dropout_keep_prob) return c # Create the bi-directional LSTM with tf.variable_scope('wordrnn'): with tf.variable_scope('fw'): cell_fw = GetCell() with tf.variable_scope('bw'): cell_bw = GetCell() rnnout, _, _ = tf.nn.bidirectional_rnn(cell_fw, cell_bw, self._inputs, dtype=tf.float32, sequence_length=self.seq_lens) if proj_size: out_size = 2 * proj_size else: out_size = 2 * hidden_size super(TweetSeqModel, self)._DoPredictions(out_size, rnnout, class_weights=weights) self.cost = tf.reduce_mean(self.example_weights * self._xent) class CharSeqModel(object): #formerly TweetSeqModel """ Treats each document (tweet) as a single "word," which is fed through c2v, and the output "embedding" sized to be a vector of language predictions. """ def __init__(self, out_vocab_size=None, batch_size=10, model_params=None, c2v=None, max_sequence_len=None, dropout_keep_prob=None, weights=None): self.params = model_params self._out_vocab_size = out_vocab_size # num. of languages self.weights = tf.constant(weights, dtype=tf.float32, name='class_weights') with tf.variable_scope("tweetff"): hidden = tf.get_variable("ff_hidden", [c2v.embedding_dims, out_vocab_size]) bias = tf.get_variable('ff_bias', [out_vocab_size]) #probably useless. at least I don't want to use it self.seq_lens = tf.placeholder(tf.int64, [batch_size], name='seq_lens') self.x = tf.placeholder(tf.int32, [batch_size, max_sequence_len], name='x') self.y = tf.placeholder(tf.float32, [batch_size, out_vocab_size], name='y') self.example_weights = tf.placeholder(tf.float32, [batch_size], name='example_weights') # get one 'word' embedding for the full tweet tweet_embedding = c2v.GetEmbeddings(self.x)[:,1,:] logits = tf.nn.xw_plus_b(tweet_embedding, hidden, bias) self.probs = tf.nn.softmax(logits) self._xent = tf.nn.softmax_cross_entropy_with_logits(logits, self.y) self.cost = tf.reduce_mean(self.example_weights * self._xent) class WordLevelModel(object): """ Model to evaluate on word-level predictions Args: batch_size: minibatch size model_params: dictionary of other model parameters c2v: char2vec class instance max_sequence_len: length of all the input/output sequences out_vocab_size: how many languages we are predicting dropout_keep_prob: dropout probability indicator weights: class weights """ def __init__(self, batch_size, model_params, c2v, max_sequence_len, out_vocab_size, dropout_keep_prob=0.0, weights=None): self._batch_size = batch_size self._dropout_keep_prob = dropout_keep_prob self._out_vocab_size = out_vocab_size self.x = tf.placeholder(tf.int32, [batch_size, max_sequence_len], name='x') self.y = tf.placeholder(tf.float32, [batch_size, max_sequence_len, out_vocab_size], name='y') # The bidirectional rnn code requires seq_lens as int64 self.seq_lens = tf.placeholder(tf.int64, [batch_size], name='seq_lens') self.example_weights = tf.placeholder(tf.float32, [batch_size], name='example_weights') embeddings = c2v.GetEmbeddings(self.x) self._inputs = [tf.squeeze(input_, [1]) for input_ in tf.split(1, max_sequence_len, embeddings)] # Need to prepare a mask to zero out the padding symbols. # Make a batch_size x max_sequence_len matrix where each # row contains the length repeated max_sequence_len times. lengths_transposed = tf.expand_dims(tf.to_int32(self.seq_lens), 1) lengths_tiled = tf.tile(lengths_transposed, [1, max_sequence_len]) # Make a matrix where each row contains [0, 1, ..., max_sequence_len] r = tf.range(0, max_sequence_len, 1) range_row = tf.expand_dims(r, 0) range_tiled = tf.tile(range_row, [batch_size, 1]) self.lengths_transposed = lengths_transposed self.lengths_tiled = lengths_tiled self.range_row = range_row self.range_tiled = range_tiled # Use the logical operations to create a mask indicator = tf.less(range_tiled, lengths_tiled+1) #i.e. where seq len is less than index trim = np.ones(indicator.get_shape()) trim[:,0] = 0 #ignore start symbol indicator = tf.logical_and(indicator, trim.astype(bool)) self.indicator = indicator sz = [batch_size, max_sequence_len] self._mask = tf.select(indicator, tf.ones(sz), tf.zeros(sz)) #-------------------------------# self.weights = tf.constant(weights, dtype=tf.float32, name='class_weights') hidden_size = model_params['model_hidden_size'] proj_size = model_params['model_proj_size'] # optional, can be None def GetCell(): """Creates an LSTM cell with dropout.""" c = tf.nn.rnn_cell.LSTMCell(hidden_size, use_peepholes=model_params['peepholes'], num_proj=proj_size) if dropout_keep_prob is not None: c = tf.nn.rnn_cell.DropoutWrapper(c, input_keep_prob=dropout_keep_prob) return c # Create the bi-directional LSTM with tf.variable_scope('wordrnn'): with tf.variable_scope('fw'): cell_fw = GetCell() with tf.variable_scope('bw'): cell_bw = GetCell() rnnout, _, _ = tf.nn.bidirectional_rnn(cell_fw, cell_bw, self._inputs, dtype=tf.float32, sequence_length=self.seq_lens) if proj_size: out_size = 2 * proj_size else: out_size = 2 * hidden_size self._DoPredictions(out_size, rnnout, self.weights) self.cost = tf.reduce_mean(self.example_weights * self._xent) def _DoPredictions(self, in_size, mats, class_weights=None): """Takes in an array of states and calculates predictions. Get the cross-entropy for each example in the vector self._xent. Args: in_size: size of the hidden state vectors mats: list of hidden state vectors """ pred_mat = tf.get_variable('pred_mat', [in_size, self._out_vocab_size]) pred_bias = tf.get_variable('pred_bias', [self._out_vocab_size]) # Make a prediction on every word. def GetWordPred(o_): logits = tf.nn.xw_plus_b(o_, pred_mat, pred_bias) return tf.nn.softmax(logits) #self.preds_by_word1 = tf.pack([GetWordPred(o_) for o_ in mats]) #self.preds_by_word = tf.reshape(self.preds_by_word1, self.y.get_shape()) #self.probs = tf.mul(tf.expand_dims(self._mask,2), self.preds_by_word) self.preds_by_word = tf.pack([GetWordPred(o_) for o_ in mats]) self.preds_by_instance = tf.pack([self.preds_by_word[:,i,:] for i in range(self.preds_by_word.get_shape()[1])]) self.probs = tf.mul(tf.expand_dims(self._mask,2), self.preds_by_instance) self._xent = _SafeXEnt(self.y, self.probs, class_weights=class_weights, sumd=[1,2]) def _SafeXEnt(y, probs, eps=0.0001, class_weights=None, sumd=[1]): """Version of cross entropy loss that should not produce NaNs. If the predicted proability for the true class is near zero then when taking the log it can produce a NaN, which ruins everything. This function ensures each probability is at least eps and no more than one before taking the log. Args: y: matrix of true probabilities same size as probs probs: matrix of probabilities for the minibatch eps: value to clip the probabilities at class_weights: vector of relative weights to be assigned to each class sumd: dimensions along which to sum the x-ent matrix Returns: cross entropy loss for each example in the minibatch """ adjusted_probs = tf.clip_by_value(probs, eps, 1.0 - eps) xent_mat = -y * tf.log(adjusted_probs) if class_weights is not None: xent_mat *= class_weights return tf.reduce_sum(xent_mat, sumd) def _SafeNegEntropy(probs, batch_size, eps=0.0001): """Computes negative entropy in a way that will not overflow.""" adjusted_probs = tf.clip_by_value(probs, eps, 1.0 - eps) entropy = tf.mul(probs, tf.log(adjusted_probs)) return tf.reduce_sum(entropy) / batch_size
taipahuchu/language-Identification-
code/char2vec.py
"""Models to create word embeddings from character sequences. Use BasicEmbedding for traditional word embeddings or use the CharLSTM or CharCNN implementations of char2vec. """ import numpy as np import tensorflow as tf import util initializer = tf.random_uniform_initializer(-0.1, 0.1) class BasicEmbedding(object): """ Baseline (traditional) word embeddings to compare against c2v. If building a model that uses a concatenation of traditional word embeddings with those generated by char2vec then it can be useful to use dropout on the traditional embeddings to avoid favoring them too much during early training. """ def __init__(self, model_params, vocab_size=None, dropout_keep_prob=None): self.dropout_keep_prob = dropout_keep_prob self.embedding_dims = model_params['word_embed_dims'] self.word_embeddings = tf.get_variable("word_embeddings", [vocab_size, self.embedding_dims], initializer=initializer) def GetEmbeddings(self, x): """Looks up some embeddings from the embedding table. Args: x: matrix of word ids to look up Returns: word embedding vectors for the given ids. """ e = tf.nn.embedding_lookup(self.word_embeddings, x) if self.dropout_keep_prob: e = tf.nn.dropout(e, self.dropout_keep_prob) return e def SaveVariables(self, *nargs): pass class Char2Vec(object): """Maps character sequences to word embeddings.""" def __init__(self, char_vocab, max_sequence_len=15): """Initialize the Char2Vec model. Args: char_vocab: vocab class instance for the character vocabulary max_sequence_len: length of the longest word """ self.max_sequence_len = max_sequence_len self.char_vocab = char_vocab self._vocab_size = char_vocab.vocab_size # Placeholder for the character sequences in the form of an # n x k array where n is the number of words and k is the # length of the longest word. Characters are encoded as ints. self.words_as_chars = tf.placeholder(tf.int32, [None, max_sequence_len], name='words_as_chars') def GetEmbeddings(self, x): return tf.nn.embedding_lookup(self.word_embeddings, x) def MakeMat(self, word_list, pad_len=None): """Make a matrix to hold the character sequences in. Special start and end tokens are added to the beggining and end of each word. Args: word_list: A list of strings pad_len: Pad all character sequences to this length. If a word is longer than the pad_len it will be truncated. Returns: Array containing character sequences and a vector of sequence lengths. """ if not pad_len: pad_len = self.max_sequence_len # make the padded char mat the_words = [] word_lengths = [] for word in word_list: word_idx = [self.char_vocab[c] for c in util.Graphemes(word)] word_idx = ([self.char_vocab['<S>']] + word_idx[:pad_len-2] + [self.char_vocab['</S>']]) if len(word_idx) < pad_len: word_idx += [self.char_vocab['</S>']] * (pad_len - len(word_idx)) the_words.append(word_idx) word_lengths.append(min(pad_len, len(word)+2)) the_words = np.array(the_words) word_lengths = np.array(word_lengths) return the_words, word_lengths @staticmethod def GetBatchVocab(words): batch_vocab = np.unique(words) words_remapped = np.copy(words) for i in xrange(len(batch_vocab)): np.place(words_remapped, words==batch_vocab[i], i) return batch_vocab, words_remapped class CharLSTM(Char2Vec): """ LSTM implementation of char2vec. The model is a two layer deep bi-LSTM. Dropout is used as a regularizer on the 2nd layer. The output word embeddings are available in the c2v.word_embeddings variable. The required placeholders are words_as_chars and seq_lens, which contain the words in the form of padded character sequences and the length of each word respectively. The batch_dim placeholder must be supplied. """ def __init__(self, char_vocab, model_params, max_sequence_len=15, dropout_keep_prob=None): super(CharLSTM, self).__init__(char_vocab, max_sequence_len) char_embed_dims = np.log(len(char_vocab)) + 1 word_embed_dims = model_params['word_embed_dims'] self.embedding_dims = word_embed_dims layer1_hidden_size = model_params['c2v_layer1_hidden_size'] layer1_out_size = model_params['c2v_layer1_out_size'] self.hidden_size = model_params['c2v_layer2_hidden_size'] # Placeholder for the word length vector. self.seq_lens = tf.placeholder(tf.int64, [None], name='seq_lens') # Placeholder for the number of words (n) in the minibatch. self.batch_dim = tf.placeholder(tf.int32, name='char_batch_dim') # The following variables define the model. with tf.variable_scope('c2v'): # continuous space character embeddings self.embedding = tf.get_variable("embedding", [self._vocab_size, char_embed_dims], initializer=initializer) def GetCell(hidden_size, num_proj=None, use_peepholes=False): """Helper function to make LSTM cells.""" layer = LSTMCell(hidden_size, num_proj=num_proj, use_peepholes=use_peepholes) layer = tf.nn.rnn.rnn_cell.DropoutWrapper(layer, output_keep_prob=dropout_keep_prob, input_keep_prob=dropout_keep_prob) return layer # This is the 1st bi-LSTM layer. layer1_fw = GetCell(layer1_hidden_size, layer1_out_size, model_params['peepholes']) layer1_bw = GetCell(layer1_hidden_size, layer1_out_size, model_params['peepholes']) # This is the 2nd layer, also a bi-LSTM. The input size is twice # the size of the output size from layer one because the concatenation # of the layer one outputs is the layer 2 input. if layer1_out_size: # Check if proj layer is enabled layer2_input_size = 2 * layer1_out_size else: layer2_input_size = 2 * layer1_hidden_size layer2_fw = LSTMCell(self.hidden_size, use_peepholes=model_params['peepholes']) layer2_bw = LSTMCell(self.hidden_size, use_peepholes=model_params['peepholes']) # The final embeddings is the output from the layer2 LSTM multiplied # by this matrix. One purpose of this matrix is to scale the layer2 # output to match the specified number of dimentions for the word # embedding. out_mat = tf.get_variable('out_mat', [2 * self.hidden_size, word_embed_dims], initializer=initializer) # z should be a tensor of dimensions batch_sz x word_len x embed_dims. z = tf.nn.embedding_lookup(self.embedding, self.words_as_chars) # Each entry in this list is a matrix of dim batch_sz x embed_dims. # There is one entry per timestep and one character is processed per # timestep. inputs = [tf.squeeze(input_) for input_ in tf.split(1, max_sequence_len, z)] for _i in inputs: # newest version of tf needs help with shape inference _i.set_shape((None, char_embed_dims)) # Feed the inputs through a bidirectional LSTM. Output is a list of # word_len tensors with dim batch_sz x 2 * hidden_sz. out1, _, _ = tf.nn.rnn.bidirectional_rnn(layer1_fw, layer1_bw, inputs, dtype=tf.float32, sequence_length=self.seq_lens) # For the 2nd bi-LSTM layer we won't use the bidirectional_rnn # wrapper. This is because we only want to save the last output # from the forward direction and the first output from the backward # direction. It is a little tricky to grab these states because they # are in different positions for each word in the minibatch. batch_range = tf.range(self.batch_dim) # The slices variables keeps track of which position the appropriate # outputs are for each word in the minibatch. slices = self.batch_dim * tf.to_int32(self.seq_lens-1) + batch_range with tf.variable_scope('fw'): outputs_forward, _ = tf.nn.rnn.rnn(layer2_fw, out1, dtype=tf.float32) out_forward = self._GetLastOutput(outputs_forward, slices) with tf.variable_scope('bw'): # Reverse the sequences before processing with the backwards LSTM. out1_bw = reverse_seq(out1, self.seq_lens) outputs_backward, _ = tf.nn.rnn.rnn(layer2_bw, out1_bw, dtype=tf.float32) out_backward = self._GetLastOutput(outputs_backward, slices) # This is the concatenation of the output from the two directions. out = tf.concat(1, [out_forward, out_backward]) # Project to the proper output dimension. This is a dimensionality # reduction step. self.word_embeddings = tf.matmul(out, out_mat) def _GetLastOutput(self, outputs, slices): """Helper function to pull out the last output for each word.""" reshaped = tf.reshape(tf.pack(outputs), [-1, self.hidden_size]) return tf.gather(reshaped, slices) class CharCNN(Char2Vec): """CNN implementation of char2vec. The model uses two layers of convolution. The second one is followed by a max pooling operation. After that there is a resnet layer. """ def __init__(self, char_vocab, model_params, max_sequence_len=15, dropout_keep_prob=None): super(CharCNN, self).__init__(char_vocab, max_sequence_len) char_embed_dims = int(np.log(len(char_vocab))) + 1 layer1_out_size = model_params['c2v_layer1_out_size'] hidden_size = model_params['c2v_layer2_hidden_size'] word_embed_dims = model_params['word_embed_dims'] # The following variables define the model. with tf.variable_scope('c2v'): # continuous space character embeddings self.embedding = tf.get_variable("embedding", [self._vocab_size, char_embed_dims], initializer=initializer) the_filter, filter_b = MakeFilter(3, char_embed_dims, layer1_out_size, 'filt') # z is a tensor of dimensions batch_sz x word_len x embed_dims. z = tf.nn.embedding_lookup(self.embedding, self.words_as_chars) z_expanded = tf.expand_dims(z, -1) conv = tf.nn.conv2d(z_expanded, the_filter, strides=[1, 1, 1, 1], padding='VALID' ) h = tf.nn.relu(tf.nn.bias_add(tf.squeeze(conv), filter_b)) h.set_shape((None, max_sequence_len - 2, layer1_out_size)) if dropout_keep_prob is not None: h_expanded = tf.nn.dropout(tf.expand_dims(h, -1), dropout_keep_prob) else: h_expanded = tf.expand_dims(h, -1) pools = [] filter_sizes = range(3,6) for width in filter_sizes: f, f_bias = MakeFilter(width, layer1_out_size, hidden_size, 'filter_w{0}'.format(width)) conv2 = tf.nn.conv2d(h_expanded, f, strides=[1, 1, 1, 1], padding='VALID') h2 = tf.nn.relu(tf.nn.bias_add(conv2, f_bias)) pooled = tf.nn.max_pool(h2, ksize=[1, max_sequence_len-1-width, 1, 1], strides=[1, 1, 1, 1], padding='VALID') pools.append(pooled) if width == 3: # debugging self.hh = tf.squeeze(pooled) self.hidx = tf.argmax(h2, 1) pooled = tf.squeeze(tf.concat(3, pools), [1,2]) # resnet layer https://arxiv.org/abs/1512.03385 sz = len(filter_sizes) * hidden_size t_mat = tf.get_variable('t_mat', [sz, sz]) t_bias = tf.Variable(tf.constant(0.1, shape=[sz]), name='t_bias') t = tf.nn.relu(tf.matmul(pooled, t_mat) + t_bias) self.word_embeddings = t + pooled self.embedding_dims = sz def MakeFilter(width, in_size, num_filters, name): filter_sz = [width, in_size, 1, num_filters] filter_b = tf.Variable(tf.constant(0.1, shape=[num_filters]), name='{0}_bias'.format(name)) the_filter = tf.get_variable(name, filter_sz) return the_filter, filter_b def reverse_seq(input_seq, lengths): """Reverse a list of Tensors up to specified lengths. Args: input_seq: Sequence of seq_len tensors of dimension (batch_size, depth) lengths: A tensor of dimension batch_size, containing lengths for each sequence in the batch. If "None" is specified, simply reverses the list. Returns: time-reversed sequence """ for input_ in input_seq: input_.set_shape(input_.get_shape().with_rank(2)) # Join into (time, batch_size, depth) s_joined = tf.pack(input_seq) # Reverse along dimension 0 s_reversed = tf.reverse_sequence(s_joined, lengths, 0, 1) # Split again into list result = tf.unpack(s_reversed) return result
taipahuchu/language-Identification-
code/util.py
# Misc. helper functions go in this file. import collections import numpy as np import os import random import sys # Need to use these three lines to import pyplot with a good font import matplotlib matplotlib.rc('font', family='DejaVu Sans') from matplotlib import pyplot def Metrics(preds, labs, show=True): """Print precision, recall and F1 for each language. Assumes a single language per example, i.e. no code switching. Args: preds: list of predictions labs: list of labels show: flag to toggle printing """ all_langs = set(preds + labs) preds = np.array(preds) labs = np.array(labs) label_totals = collections.Counter(labs) pred_totals = collections.Counter(preds) confusion_matrix = collections.Counter(zip(preds, labs)) num_correct = 0 for lang in all_langs: num_correct += confusion_matrix[(lang, lang)] acc = num_correct / float(len(preds)) print 'accuracy = {0:.3f}'.format(acc) if show: print ' Lang Prec. Rec. F1' print '------------------------------' scores = [] fmt_str = ' {0:6} {1:6.2f} {2:6.2f} {3:6.2f}' for lang in sorted(all_langs): idx = preds == lang total = max(1.0, pred_totals[lang]) precision = 100.0 * confusion_matrix[(lang, lang)] / total idx = labs == lang total = max(1.0, label_totals[lang]) recall = 100.0 * confusion_matrix[(lang, lang)] / total if precision + recall == 0.0: f1 = 0.0 else: f1 = 2.0 * precision * recall / (precision + recall) scores.append([precision, recall, f1]) if show: print fmt_str.format(lang, precision, recall, f1) totals = np.array(scores).mean(axis=0) if show: print '------------------------------' print fmt_str.format('Total:', totals[0], totals[1], totals[2]) return totals[2] def ConfusionMat(preds, labs): """Plot and show a confusion matrix. Args: preds: list of predicted labels labs: list of true labels """ all_langs = set(preds + labs) # this is the set of all possible labels num_langs = len(all_langs) # create a mapping from labels to id numbers lookup = dict(zip(sorted(all_langs), range(num_langs))) # make the counts for the confusion matrix counts = np.zeros((num_langs, num_langs)) for p, l in zip(preds, labs): counts[lookup[p], lookup[l]] += 1 # plot a colormap using log scale pyplot.imshow(np.log(counts+1.0), interpolation='none') # plot the text labels for i in xrange(num_langs): for j in xrange(num_langs): pyplot.text(i, j, str(int(counts[i, j])), color='white', horizontalalignment='center') # take care of the axes pyplot.xticks(range(num_langs), sorted(all_langs)) pyplot.yticks(range(num_langs), sorted(all_langs)) pyplot.xlabel('Prediction') pyplot.ylabel('True Label') pyplot.show() def GetColor(percent): """Returns an RGB color scale to represent a given percentile.""" z = int(percent * 512) z = 255 - min(255, z) hexz = hex(z)[2:] if len(hexz) == 1: hexz = '0' + hexz elif len(hexz) == 0: hexz = '00' color = '#ff{0}{0}'.format(hexz) return color def PrintParams(param_list, handle=sys.stdout.write): """Print the names of the parameters and their sizes. Args: param_list: list of tensors handle: where to write the param sizes to """ handle('NETWORK SIZE REPORT\n') param_count = 0 fmt_str = '{0: <25}\t{1: >12}\t{2: >12,}\n' for p in param_list: shape = p.get_shape() shape_str = 'x'.join([str(x.value) for x in shape]) handle(fmt_str.format(p.name, shape_str, np.prod(shape).value)) param_count += np.prod(shape).value handle(''.join(['-'] * 60)) handle('\n') handle(fmt_str.format('total', '', param_count)) if handle==sys.stdout.write: sys.stdout.flush() def GetProj(feat_mat): """Projects a feature matrix into 2 dimensions using PCA.""" m = feat_mat.mean(axis=0) feat_mat -= m covmat = feat_mat.T.dot(feat_mat) m, v = np.linalg.eig(covmat) m /= m.sum() v = v[:, 0:2] proj = feat_mat.dot(v) return proj def PlotText(pts, labels): pyplot.plot(pts[:, 0], pts[:, 1], 'x') for i in xrange(pts.shape[0]): pyplot.text(pts[i, 0], pts[i, 1], labels[i]) pyplot.show() def Graphemes(s): """ Given a string return a list of graphemes. Args: s the input string Returns: A list of graphemes. """ graphemes = [] current = [] if type(s) == unicode: s = s.encode('utf8') for c in s: val = ord(c) & 0xC0 if val == 128: # this is a continuation current.append(c) else: # this is a new grapheme if len(current) > 0: graphemes.append(''.join(current)) current = [] if val < 128: graphemes.append(c) # single byte grapheme else: current.append(c) # multi-byte grapheme if len(current) > 0: graphemes.append(''.join(current)) return graphemes def Bytes(s): if type(s) == unicode: s = s.encode('utf8') z = s.encode('hex') return [z[2*i:2*i+2] for i in range(len(z)/2)] def GetLangName(code): """Convert ISO lang codes to full language names. Args: code: an ISO language code as a string Returns: name of language or ISO code if language name not available """ names = { 'am': 'amharic', 'ar': 'arabic', 'bg': 'bulgarian', 'bn': 'bengali', 'bo': 'tibetan', 'bs': 'bosnian', 'ca': 'catalan', 'ckb': 'kurdish', 'cs': 'czech', 'cy': 'welsh', 'da': 'danish', 'de': 'german', 'dv': 'divehi', 'el': 'greek', 'en': 'english', 'es': 'spanish', 'et': 'estonian', 'eu': 'basque', 'fa': 'persian', 'fi': 'finnish', 'fr': 'french', 'gl': 'galician', 'gu': 'gujarati', 'he': 'hebrew', 'hi': 'hindi', 'hi-Latn': 'hindi-latin', 'hr': 'croatian', 'ht': 'haitian', 'hu': 'hungarian', 'hy': 'armenian', 'id': 'indonesian', 'is': 'icelandic', 'it': 'italian', 'ja': 'japanese', 'ka': 'georgian', 'km': 'cambodian', 'kn': 'kannada', 'ko': 'korean', 'lo': 'lao', 'lt': 'lithuanian', 'lv': 'latvian', 'ml': 'malayalam', 'mr': 'marathi', 'my': 'burmese', 'ne': 'nepali', 'nl': 'dutch', 'no': 'norwegian', 'or': 'oriya', 'pa': 'punjabi', 'pl': 'polish', 'ps': 'pashto', 'pt': 'portuguese', 'ro': 'romanian', 'ru': 'russian', 'sd': 'sindhi', 'si': 'sinhala', 'sk': 'slovak', 'sl': 'slovene', 'sr': 'serbian', 'sv': 'swedish', 'ta': 'tamil', 'te': 'telugu', 'th': 'thai', 'tl': 'tagalog', 'tr': 'turkish', 'ug': 'uighur', 'uk': 'ukranian', 'ur': 'urdu', 'vi': 'vietnamese', 'zh-CN': 'chinese', 'zh-TW': 'taiwanese' } if code in names: return names[code] return code
taipahuchu/language-Identification-
code/vocab.py
<reponame>taipahuchu/language-Identification- import argparse import collections import numpy as np import pickle import re class Vocab(object): def _normalize(self, word): if re.match(ur'^<.*>$', word): return word if self.specialcase: newword = [] prev_is_lower = True for letter in word: if letter.isupper(): if prev_is_lower: newword.append('~') prev_is_lower = False else: prev_is_lower = True newword.append(letter.lower()) word = ''.join(newword) if self.lowercase: word = word.lower() if self.numreplace: word = re.sub(ur'\d', '#', word) return word def __init__(self, tokenset, unk_symbol='<UNK>', lowercase=False, numreplace=False): self.lowercase = lowercase self.numreplace = numreplace self.specialcase = False tokenset = set([self._normalize(w) for w in tokenset]) self.vocab_size = len(tokenset) self.unk_symbol = unk_symbol self.word_to_idx = dict(zip(sorted(tokenset), range(self.vocab_size))) self.idx_to_word = dict(zip(self.word_to_idx.values(), self.word_to_idx.keys())) @staticmethod def Load(filename): with open(filename, 'rb') as f: v = pickle.load(f) return v @classmethod def MakeFromData(cls, lines, min_count, unk_symbol='<UNK>', max_length=None, no_special_syms=False, normalize=False): lowercase=False numreplace=False if normalize: lowercase=True numreplace=True token_counts = collections.Counter() for line in lines: token_counts.update(line) tokenset = set() for word in token_counts: if max_length and len(word) > max_length: continue if token_counts[word] >= min_count: tokenset.add(word) if not no_special_syms: tokenset.add(unk_symbol) tokenset.add('<S>') tokenset.add('</S>') return cls(tokenset, unk_symbol=unk_symbol, lowercase=lowercase, numreplace=numreplace) @classmethod def ByteVocab(cls): """Creates a vocab that has a token for each possible byte. It's useful to have a fixed byte vocab so that the subset of bytes that form the vocab is not dependent on the dataset being used. Thus, the learned byte embeddings can be reused on different datasets. """ c = '0123456789abcdef' tokens = ['<S>', '</S>'] for i in c: for j in c: tokens.append(i + j) return cls(tokens) @classmethod def LoadFromTextFile(cls, filename, unk_symbol='<UNK>'): tokens = [] with open(filename, 'r') as f: for line in f: line = line.strip() tokens.append(line) return cls(tokens, unk_symbol=unk_symbol) def GetWords(self): """Get a list of words in the vocabulary.""" return self.word_to_idx.keys() def LookupIdx(self, token): token = self._normalize(token) if token in self.word_to_idx: return self.word_to_idx[token] return self.word_to_idx.get(self.unk_symbol, None) def __contains__(self, key): key = self._normalize(key) return key in self.word_to_idx def __getitem__(self, key): """If key is an int lookup word by id, if key is a word then lookup id.""" if type(key) == int or type(key) == np.int64: return self.idx_to_word[key] return self.LookupIdx(key) def __iter__(self): word_list = [self.idx_to_word[x] for x in xrange(self.vocab_size)] return word_list.__iter__() def __len__(self): return self.vocab_size def Save(self, filename): if filename.endswith('.pickle'): with open(filename, 'wb') as f: pickle.dump(self, f) elif filename.endswith('.txt'): with open(filename, 'w') as f: for i in range(self.vocab_size): f.write('{0}\n'.format(self.idx_to_word[i])) if __name__ == '__main__': """Print the contents of the vocabulary.""" parser = argparse.ArgumentParser() parser.add_argument('filename') args = parser.parse_args() v = Vocab.Load(args.filename) for i in v.GetWords(): print i
taipahuchu/language-Identification-
code/getembeddings.py
""" Utility for printing nearest neighbors to a given word in embedding space. Uses the char2vec portion of a trained model and the tweetlid vocabulary. Nearest neighbors are given based on cosine similarity. """ import sys reload(sys) sys.setdefaultencoding('utf8') import argparse import json import numpy as np import os import tensorflow as tf from batcher import Dataset from char2vec import CharCNN as Char2Vec from vocab import Vocab parser = argparse.ArgumentParser() parser.add_argument('expdir') args = parser.parse_args() config = tf.ConfigProto(inter_op_parallelism_threads=10, intra_op_parallelism_threads=10) dataset = Dataset(10, preshuffle=False) dataset.ReadData('../data/tweetlid/training.tsv.gz', 'all', 'tweet') input_vocab = Vocab.MakeFromData(dataset.GetSentences(), min_count=1) char_vocab = Vocab.Load(os.path.join(args.expdir, 'char_vocab.pickle')) max_word_len = max([len(x) for x in input_vocab.GetWords()]) + 2 print 'max word len {0}'.format(max_word_len) with open(os.path.join(args.expdir, 'model_params.json'), 'r') as f: model_params = json.load(f) c2v = Char2Vec(char_vocab, model_params, max_sequence_len=max_word_len) the_words, word_lengths = c2v.MakeMat(input_vocab, pad_len=max_word_len) saver = tf.train.Saver(tf.all_variables()) session = tf.Session(config=config) saver.restore(session, os.path.join(args.expdir, 'model.bin')) embeds = tf.nn.l2_normalize(c2v.word_embeddings, 1) out = session.run([embeds], {c2v.words_as_chars: the_words})[0] while True: print 'please input a word:' user_word = raw_input() user_chars, _ = c2v.MakeMat([user_word, 'DUMMY_WORD'], pad_len=max_word_len) user_embed = session.run([embeds], {c2v.words_as_chars: user_chars})[0][0, :] sims = np.matmul(out, user_embed) idx = np.argsort(sims) for i in range(1, 21): index = idx[-i] score = sims[index] print '{0} {1:.3f} {2}'.format(i, score, input_vocab[index])
taipahuchu/language-Identification-
code/langid.py
import sys reload(sys) # need to reload to set default encoding sys.setdefaultencoding('utf8') import argparse import collections import json import logging import numpy as np import os import shutil import tensorflow as tf import util from char2vec import CharCNN as Char2Vec from batcher import Dataset from vocab import Vocab from models import WordSeqModel, CharSeqModel, TweetSeqModel, WordLevelModel parser = argparse.ArgumentParser() parser.add_argument('expdir') parser.add_argument('--start', help='init parameters from saved model') parser.add_argument('--mode', choices=['train', 'debug', 'eval', 'final', 'apply'], default='train') parser.add_argument('--data', default='../data/smallwiki.tsv.gz') parser.add_argument('--params', default='default_params.json', help='load hyperparams from json file') parser.add_argument('--model', choices=['word', 'char', 'tweet'], help='pass "word", "char", or "tweet" to use '\ 'WordSeqModel, CharSeqModel or TweetSeqModel', default='tweet') #default is hierarchical model args = parser.parse_args() if not os.path.exists(args.expdir): os.mkdir(args.expdir) mode = args.mode if mode == 'train': logging.basicConfig(filename=os.path.join(args.expdir, 'logfile.txt'), level=logging.INFO) config = tf.ConfigProto(inter_op_parallelism_threads=10, intra_op_parallelism_threads=10) tf.set_random_seed(666) baseline = False batch_size = 25 dataset = Dataset(batch_size, preshuffle=mode=='train') und_symbol='und' dataset.ReadData(args.data, mode, args.model) # Make the input vocabulary (words that appear in data) if baseline: # The baseline is to use fixed word embeddings. if mode == 'train': # The input vocab is fixed during training. input_vocab = Vocab.MakeFromData(dataset.GetSentences(), min_count=2) input_vocab.Save(os.path.join(args.expdir, 'input_vocab.pickle')) else: # During testing we need to load the saved input vocab. input_vocab = Vocab.Load(os.path.join(args.expdir, 'input_vocab.pickle')) else: # The open vocabulary can be regenerated with each run. min_count = 1 if mode == 'debug': min_count = 10 # When visualizing word embeddings hide rare words maxlens = {'word':40, 'char':150, 'tweet':40} input_vocab = Vocab.MakeFromData(dataset.GetSentences(), min_count=min_count, max_length=maxlens[args.model]) if mode == 'train': # Make the character vocabulary if args.start: shutil.copyfile(os.path.join(args.start, 'char_vocab.pickle'), os.path.join(args.expdir, 'char_vocab.pickle')) char_vocab = Vocab.Load(os.path.join(args.expdir, 'char_vocab.pickle')) with open(os.path.join(args.start, 'model_params.json'), 'r') as f: model_params = json.load(f) else: x = [util.Graphemes(w) for w in input_vocab.GetWords()] char_vocab = Vocab.MakeFromData(x, min_count=2) char_vocab.Save(os.path.join(args.expdir, 'char_vocab.pickle')) with open(args.params, 'r') as f: model_params = json.load(f) with open(os.path.join(args.expdir, 'model_params.json'), 'w') as f: json.dump(model_params, f) else: # eval or debug mode char_vocab = Vocab.Load(os.path.join(args.expdir, 'char_vocab.pickle')) with open(os.path.join(args.expdir, 'model_params.json'), 'r') as f: model_params = json.load(f) # Make the output vocab (the set of possible languages to predict) output_vocab_filename = os.path.join(args.expdir, 'out_vocab.pickle') if mode == 'train': labels = [[x] for x in dataset.GetLabelSet()] if [und_symbol] not in labels: labels += [[und_symbol]] output_vocab = Vocab.MakeFromData(labels, min_count=1, no_special_syms=True) output_vocab.Save(output_vocab_filename) else: output_vocab = Vocab.Load(output_vocab_filename) ignore_categories = ['mixed', 'ambiguous', 'fw', 'unk', '{', '}'] dataset.Prepare(input_vocab, output_vocab, und_symbol, ignore_categories) #Set other hyperparameters and create model max_word_len = max([len(x) for x in input_vocab.GetWords()]) + 2 print 'max word len {0}'.format(max_word_len) dropout_keep_prob = tf.placeholder_with_default(1.0, (), name='keep_prob') if baseline: c2v = BasicEmbedding(model_params, vocab_size=len(input_vocab)) else: c2v = Char2Vec(char_vocab, model_params, max_sequence_len=max_word_len, dropout_keep_prob=dropout_keep_prob) the_words, word_lengths = c2v.MakeMat(input_vocab, pad_len=max_word_len) models = {'word': WordSeqModel, 'char': CharSeqModel, 'tweet': TweetSeqModel} if args.data == 'codeswitch': models['tweet'] = WordLevelModel model = models[args.model](batch_size=batch_size, model_params=model_params, max_sequence_len=dataset.max_sequence_len, dropout_keep_prob=dropout_keep_prob, out_vocab_size=len(output_vocab), weights=dataset.w, c2v=c2v) saver = tf.train.Saver(tf.all_variables()) session = tf.Session(config=config) def Apply(expdir): saver.restore(session, os.path.join(expdir, 'model.bin')) results = collections.defaultdict(list) for _ in xrange(dataset.GetNumBatches()): words, seqlen, labs, _ = dataset.GetNextBatch() batch_data = MakeFeedDict(words, seqlen, labs) cvocab = char_vocab batch_vocab, words_remapped = Char2Vec.GetBatchVocab(words) charseqs = the_words[batch_vocab] hh, hidx = session.run([c2v.hh, c2v.hidx], batch_data) for word_i in range(len(batch_vocab)): for filter_i in range(hh.shape[-1]): activation = hh[word_i, filter_i] if activation == 0.0: continue loc = hidx[word_i, 0, filter_i] char_seq = charseqs[word_i, loc:loc+4] unit = ''.join([cvocab[i] for i in char_seq]) results[filter_i].append(('{0:.1f}'.format(activation), unit)) for filtnum in results: dedup = collections.Counter(results[filtnum]) z = sorted(dedup.keys(), key=lambda x: -float(x[0])) with open('filters/{0}.txt'.format(filtnum), 'w') as f: for i in range(len(z)): first = '{0} {1} {2}\n'.format(z[i][0], z[i][1], dedup[z[i]]) f.write(first) def Eval(expdir): """Evaluates on dev data. Writes results to a results.tsv file in the expdir for use in the scoring script. Args: expdir: path to experiment directory """ if args.data == 'codeswitch': return EvalPerWord(expdir) saver.restore(session, os.path.join(expdir, 'model.bin')) all_preds, all_labs = [], [] for _ in xrange(dataset.GetNumBatches()): words, seqlen, labs, weights = dataset.GetNextBatch() batch_data = MakeFeedDict(words, seqlen, labs) if args.model in ["word", "tweet"]: model_vars = [model.probs, model.preds_by_word] probs, pp = session.run(model_vars, batch_data) elif args.model in ["char"]: probs, pp = session.run([model.probs, model.preds], batch_data) idx = weights != 0 all_preds += [output_vocab[p] for p in np.argmax(probs[idx, :], axis=1)] all_labs += [output_vocab[p] for p in np.argmax(labs[idx, :], axis=1)] util.Metrics(all_preds, all_labs) # This output file is in the format needed to score for TweetLID ids = dataset.GetIds() with open(os.path.join(expdir, 'results.tsv'), 'w') as f: for idnum, p in zip(ids, all_preds): f.write('{0}\t{1}\n'.format(idnum, p)) def Train(expdir): logging.info('Input Vocab Size: {0}'.format(len(input_vocab))) logging.info('Char Vocab Size: {0}'.format(len(char_vocab))) tvars = tf.trainable_variables() grads, _ = tf.clip_by_global_norm(tf.gradients(model.cost, tvars), 5.0) optimizer = tf.train.AdamOptimizer(0.001) train_op = optimizer.apply_gradients(zip(grads, tvars)) session.run(tf.initialize_all_variables()) if args.start: saver.restore(session, os.path.join(args.start, 'model.bin')) util.PrintParams(tf.trainable_variables(), handle=logging.info) maxitr = model_params.get('num_training_iters', 80001) print "Training for {} iterations...".format(maxitr) for idx in xrange(maxitr): if args.data == 'codeswitch': words, seqlen, labs, ws, lines = dataset.GetNextBatch() else: words, seqlen, labs, ws = dataset.GetNextBatch() batch_data = MakeFeedDict(words, seqlen, labs, ws) if idx % 25 == 0: probs = session.run([model.probs], batch_data)[0] s = [input_vocab[i] for i in words[0, :seqlen[0]]] print ' '.join(s) if args.data != 'codeswitch': print "Predicted:", GetTopPreds(probs[0, :]) print "Actual:", GetTopPreds(labs[0,:]) else: print "Predict:", GetTopWordLevelPreds(probs[0,:,:], seqlen[0]) print "Actual: ", GetTopWordLevelPreds(labs[0,:,:], seqlen[0]) cost, _ = session.run([model.cost, train_op], batch_data) logging.info({'iter': idx, 'cost': float(cost)}) if idx == maxitr-1 or (idx % 2000 == 0 and idx > 0): saver.save(session, os.path.join(expdir, 'model.bin')) def Debug(expdir): """Plots language and word embeddings from saved model.""" saver.restore(session, os.path.join(expdir, 'model.bin')) # Plot the language embeddings z = [x for x in tf.trainable_variables() if 'pred_mat' in x.name][0] zz = z.eval(session) c = util.GetProj(zz.T) lang_names = [util.GetLangName(output_vocab[i]) for i in xrange(len(output_vocab))] util.PlotText(c, lang_names) # plot some word embeddings batch_data = {c2v.words_as_chars: the_words} word_embeddings = session.run([c2v.word_embeddings], batch_data)[0] c = util.GetProj(word_embeddings) util.PlotText(c, input_vocab) def GetTopPreds(probs): top_preds = [] for ii in reversed(np.argsort(probs)): if probs[ii] > 0.05: top_preds.append('{0}: {1:.1f}'.format(output_vocab[ii], 100.0 * probs[ii])) return top_preds def MakeFeedDict(words, seqlen, labs, ws=None): """Create the feed dict to process each batch. All the inputs should be from the GetNextBatch command. Args: words: matrix of word ids seqlen: vector of sequence lengths labs: target matrix ws: per-example weights Returns: dictionary to be used as feed dict. """ batch_data = { model.seq_lens: seqlen, model.x: words, } if mode == 'train': batch_data[model.y] = labs batch_data[model.example_weights] = ws batch_data[dropout_keep_prob] = model_params['dropout_keep_prob'] if not baseline: batch_vocab, words_remapped = Char2Vec.GetBatchVocab(words) batch_data.update({ c2v.words_as_chars: the_words[batch_vocab, :], model.x: words_remapped }) if hasattr(c2v, 'seq_lens'): batch_data.update({ c2v.seq_lens: word_lengths[batch_vocab], c2v.batch_dim: len(batch_vocab) }) return batch_data funcs = {'train': Train, # Train the model. 'debug': Debug, # Plot some graphs. 'apply': Apply, # Apply model to new data. 'eval': Eval, # Evaluate on dev set. 'final': Eval} # Evaluate on eval set. funcs[mode](args.expdir)
nickromano/django-daily-digest
leather/ticks/score.py
<gh_stars>1-10 #!/usr/bin/env python from decimal import Decimal, ROUND_CEILING, ROUND_FLOOR import math import sys from leather.ticks.base import Ticker from leather.utils import isclose # Shorthand ZERO = Decimal("0") TEN = Decimal("10") #: Normalized intervals to be tested for ticks INTERVALS = [ Decimal("0.1"), Decimal("0.15"), Decimal("0.2"), Decimal("0.25"), Decimal("0.5"), Decimal("1.0"), ] #: The default number of ticks to produce DEFAULT_TICKS = 5 #: The minimum length of a viable tick sequence MIN_TICK_COUNT = 4 #: The maximum length of a viable tick sequence MAX_TICK_COUNT = 10 #: Most preferred tick intervals BEST_INTERVALS = [Decimal("0.1"), Decimal("1.0")] #: Least preferred tick intervals WORST_INTERVALS = [Decimal("0.15")] class ScoreTicker(Ticker): """ Attempt to find an optimal series of ticks by generating many possible sequences and scoring them based on several criteria. Only the best tick sequence is returned. Based an algorithm described by <NAME>: http://austinclemens.com/blog/2016/01/09/an-algorithm-for-creating-a-graphs-axes/ See :meth:`.ScoreTicker.score` for scoring implementation. :param domain_min: Minimum value of the data series. :param domain_max: Maximum value of the data series. """ def __init__(self, domain_min, domain_max): self._domain_min = domain_min self._domain_max = domain_max self._ticks = self._find_ticks() self._min = self._ticks[0] self._max = self._ticks[-1] @property def ticks(self): return self._ticks @property def min(self): return self._min @property def max(self): return self._max def _find_ticks(self): """ Implements the tick-finding algorithm. """ force_zero = self._domain_min < ZERO and self._domain_max > ZERO interval_guess = abs(self._domain_max - self._domain_min) / (DEFAULT_TICKS - 1) magnitude = interval_guess.log10().to_integral_exact(rounding=ROUND_CEILING) candidate_intervals = [] for interval in INTERVALS: candidate_intervals.append((interval, interval * pow(TEN, magnitude))) candidate_intervals.append((interval, interval * pow(TEN, magnitude - 1))) candidate_intervals.append((interval, interval * pow(TEN, magnitude + 1))) candidate_ticks = [] for base_interval, interval in candidate_intervals: ticks = [] if force_zero: min_steps = (abs(self._domain_min) / interval).to_integral_exact( rounding=ROUND_CEILING ) ticks.append(self._round_tick(-min_steps * interval)) else: ticks.append( self._round_tick( (self._domain_min / interval).to_integral_exact( rounding=ROUND_FLOOR ) * interval ) ) tick_num = 1 while ticks[tick_num - 1] < self._domain_max: t = self._round_tick(ticks[0] + (interval * tick_num)) ticks.append(t) tick_num += 1 # Throw out sequences that are too short or too long if len(ticks) < MIN_TICK_COUNT or len(ticks) > MAX_TICK_COUNT: continue candidate_ticks.append( { "base_interval": base_interval, "interval": interval, "ticks": ticks, "score": self._score(base_interval, interval, ticks), } ) # Order by best score, using number of ticks as a tie-breaker best = sorted( candidate_ticks, key=lambda c: (c["score"]["total"], len(c["ticks"])) ) return best[0]["ticks"] def _score(self, base_interval, interval, ticks): """ Score a given tick sequence based on several criteria. This method returns discrete scoring components for easier debugging. """ s = {"pct_waste": 0, "interval_penalty": 0, "len_penalty": 0, "total": 0} # Penalty for wasted scale space waste = (self._domain_min - ticks[0]) + (ticks[-1] - self._domain_max) pct_waste = waste / (self._domain_max - self._domain_min) s["pct_waste"] = pow(10, pct_waste) # Penalty for choosing less optimal tick intervals if base_interval in BEST_INTERVALS: pass elif base_interval in WORST_INTERVALS: s["interval_penalty"] = 2 else: s["interval_penalty"] = 1 # Penalty for too many ticks if len(ticks) > 5: s["len_penalty"] = len(ticks) - 5 s["total"] = s["pct_waste"] + s["interval_penalty"] + s["len_penalty"] return s def _round_tick(self, t): """ Round a tick to 0-3 decimal places, if the remaining digits do not appear to be significant. """ for r in range(0, 4): exp = pow(Decimal(10), Decimal(-r)) quantized = t.quantize(exp) if isclose(t, quantized): return quantized return t
nickromano/django-daily-digest
daily_digest/tests.py
import random from datetime import date, datetime, timedelta import mock import pytz from django.contrib.auth.models import User from django.core.management import call_command from django.test import TestCase from django.test.client import Client from project.photos.models import PhotoUpload from .config import DailyDigestRequiredFieldException, daily_digest_config, load_config from .utils import ( EmailMultiAlternatives, send_daily_digest, series_data_for_model, series_labels, ) class MockEmailMultiAlternatives(mock.Mock): subject = "" def __init__(self, subject, text_content, from_email, to): self.subject = subject self.text_content = text_content self.from_email = from_email self.to = to class DailyDigestTestCase(TestCase): @mock.patch( "daily_digest.utils.current_time_naive", return_value=datetime(2018, 1, 10, 8, 0), ) @mock.patch.object(EmailMultiAlternatives, "send") @mock.patch.object(EmailMultiAlternatives, "attach") @mock.patch.object(EmailMultiAlternatives, "attach_alternative") @mock.patch.object(EmailMultiAlternatives, "__init__", return_value=None) def test_send_daily_digest( self, mock_init_email, mock_attach_html, mock_add_attachment, *_ ): send_daily_digest() subject, text_content, from_email, to_emails = mock_init_email.call_args_list[ 0 ][0] self.assertEqual(subject, "Daily Digest - 01/10/18") self.assertEqual(text_content, "Daily Digest") self.assertEqual(from_email, "<EMAIL>") self.assertEqual(list(to_emails), []) self.assertEqual( mock_attach_html.call_args_list[0], mock.call( '<!DOCTYPE html>\n<html>\n<head>\n\t<title></title>\n\t<meta name="viewport" content="width=device-width, initial-scale=1.0">\n\n\t</head>\n<body style="font-family:-apple-system, BlinkMacSystemFont, sans-serif">\n\t\n\n\n<h3 style="font-weight:500; margin-bottom:0">New Users</h3>\n\n\n\t<img class="chart" width="100%" src="cid:new-users" style="max-width:480px">\n\n<br>\n\n<h3 style="font-weight:500; margin-bottom:0">Photo Uploads</h3>\n\n\n\t<img class="chart" width="100%" src="cid:photo-uploads" style="max-width:480px">\n\n<br>\n\n\n\n</body>\n</html>', "text/html", ), ) self.assertEqual( mock_add_attachment.call_args_list[0][0][0].get("Content-ID"), "<new-users>" ) @mock.patch("daily_digest.management.commands.send_daily_digest.send_daily_digest") def test_send_daily_digest_command(self, mock_send_digest): call_command("send_daily_digest") self.assertEqual(mock_send_digest.called, True) def test_daily_digest_preview(self): user = User.objects.create(username="test", is_superuser=True, is_staff=True) user.set_password("<PASSWORD>") user.save() client = Client() client.login(username="test", password="<PASSWORD>") response = client.get("/admin/daily-digest-preview/") self.assertEqual(response.status_code, 200) def test_daily_digest_preview_requires_staff(self): user = User.objects.create(username="test") user.set_password("<PASSWORD>") user.save() client = Client() client.login(username="test", password="<PASSWORD>") response = client.get("/admin/daily-digest-preview/") self.assertEqual(response.status_code, 302) @mock.patch( "daily_digest.utils.current_time_naive", return_value=datetime(2018, 1, 10, 16, 0), ) # 8am LA def test_series_data_generated_for_current_period(self, *_): timezone = pytz.timezone("America/Los_Angeles") User.objects.filter(username__startswith="test").delete() PhotoUpload.objects.all().delete() los_angeles_timezone = pytz.timezone("America/Los_Angeles") today = los_angeles_timezone.localize(datetime(2018, 1, 10, 16, 0)) for days in [-1, 0, 1, 2, 3, 4, 5, 6, 7]: for _ in range(0, days + 3): User.objects.create( username="test-{}".format(random.randint(0, 100000000)), date_joined=today - timedelta(days=days), ) queryset = User.objects.all() data, total_count = series_data_for_model(queryset, "date_joined", timezone) self.assertEqual(total_count, 42) self.maxDiff = None self.assertEqual( data, [ [date(2018, 1, 4), 9], [date(2018, 1, 5), 8], [date(2018, 1, 6), 7], [date(2018, 1, 7), 6], [date(2018, 1, 8), 5], [date(2018, 1, 9), 4], [date(2018, 1, 10), 3], ], ) @mock.patch( "daily_digest.utils.current_time_naive", return_value=datetime(2018, 1, 10, 16, 0), ) # 8am LA def test_series_data_generated_for_previous_period(self, *_): timezone = pytz.timezone("America/Los_Angeles") User.objects.filter(username__startswith="test").delete() PhotoUpload.objects.all().delete() los_angeles_timezone = pytz.timezone("America/Los_Angeles") today = los_angeles_timezone.localize(datetime(2018, 1, 10, 16, 0)) for days in [6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]: for _ in range(0, days): User.objects.create( username="test-{}".format(random.randint(0, 100000000)), date_joined=today - timedelta(days=days), ) queryset = User.objects.all() data, total_count = series_data_for_model( queryset, "date_joined", timezone, prev_period=True ) self.assertEqual(total_count, 70) self.maxDiff = None self.assertEqual( data, [ [date(2018, 1, 4), 13], [date(2018, 1, 5), 12], [date(2018, 1, 6), 11], [date(2018, 1, 7), 10], [date(2018, 1, 8), 9], [date(2018, 1, 9), 8], [date(2018, 1, 10), 7], ], ) def test_series_labels(self): current_label, prev_period_label = series_labels(55, 100) self.assertEqual(current_label, "Last 7 Days (55) -45%") self.assertEqual(prev_period_label, "Previous Period (100)") def test_series_labels_with_no_prev_period_data(self): current_label, prev_period_label = series_labels(55, 0) self.assertEqual(current_label, "Last 7 Days (55)") self.assertEqual(prev_period_label, "Previous Period (0)") def test_config_missing_required_field(self): config_missing_fields = { "charts": [ { # 'title': 'New Users' Missing field "model": "django.contrib.auth.models.User", "date_field": "date_joined", } ] } with self.settings(DAILY_DIGEST_CONFIG=config_missing_fields): with self.assertRaises(DailyDigestRequiredFieldException): load_config() def test_config_passing_model_path(self): config_missing_fields = { "charts": [ { "title": "New Users", "model": "django.contrib.auth.models.User", "date_field": "date_joined", } ] } with self.settings(DAILY_DIGEST_CONFIG=config_missing_fields): load_config() self.assertEqual(daily_digest_config.chart_configs[0].model, User) def test_config_passing_app_label(self): config_missing_fields = { "charts": [ { "title": "New Users", "app_label": "auth", "model": "user", "date_field": "date_joined", } ] } with self.settings(DAILY_DIGEST_CONFIG=config_missing_fields): load_config() self.assertEqual(daily_digest_config.chart_configs[0].model, User)
nickromano/django-daily-digest
leather/svg.py
<reponame>nickromano/django-daily-digest #!/usr/bin/env python """ Helpers for working with SVG. """ import xml.etree.ElementTree as ET import six HEADER = ( '<?xml version="1.0" standalone="no"?>\n' + '<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"\n' + '"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">\n' ) def stringify(root): """ Convert an SVG XML tree to a unicode string. """ if six.PY3: return ET.tostring(root, encoding="unicode") else: return ET.tostring(root, encoding="utf-8") def save(f, root): """ Save an SVG XML tree to a file. """ f.write(HEADER) f.write(stringify(root)) def translate(x, y): """ Generate an SVG transform statement representing a simple translation. """ return "translate(%i %i)" % (x, y) def rotate(deg, x, y): """ Generate an SVG transform statement representing rotation around a given point. """ return "rotate(%i %i %i)" % (deg, x, y)
nickromano/django-daily-digest
daily_digest/apps.py
<reponame>nickromano/django-daily-digest # -*- coding: utf-8 -*- from django.apps import AppConfig class DailyDigestConfig(AppConfig): name = "daily_digest"
nickromano/django-daily-digest
leather/ticks/score_time.py
<reponame>nickromano/django-daily-digest<filename>leather/ticks/score_time.py<gh_stars>1-10 #!/usr/bin/env python from datetime import date, datetime, timedelta from decimal import Decimal from functools import partial import math import sys from leather.ticks.score import ScoreTicker from leather import utils #: The default number of ticks to produce DEFAULT_TICKS = 5 #: The minimum length of a viable tick sequence MIN_TICK_COUNT = 4 #: The maximum length of a viable tick sequence MAX_TICK_COUNT = 10 #: The minimum units of the interval needed to use that interval ("4 years") MIN_UNITS = 4 #: The possible intervals as (to_function, from_function, overlap_tick_formatter, simple_tick_formatter) INTERVALS = [ (utils.to_year_count, utils.from_year_count, None, "%Y"), (utils.to_month_count, utils.from_month_count, "%Y-%m", "%m"), (utils.to_day_count, utils.from_day_count, "%m-%d", "%d"), (utils.to_hour_count, utils.from_hour_count, "%d-%H", "%H"), (utils.to_minute_count, utils.from_minute_count, "%H:%M", "%M"), (utils.to_second_count, utils.from_second_count, "%H:%M:%S", "%S"), (utils.to_microsecond_count, utils.from_microsecond_count, "%S-%f", "%f"), ] class ScoreTimeTicker(ScoreTicker): """ A variation on :class:`.ScoreTicker` that generates sequences of dates or datetimes. :param domain_min: Minimum value of the data series. :param domain_max: Maximum value of the data series. """ def __init__(self, domain_min, domain_max): self._domain_min = domain_min self._domain_max = domain_max if isinstance(self._domain_min, datetime): self._type = datetime else: self._type = date # Identify appropriate interval unit self._to_unit = None self._from_unit = None self._fmt = None previous_delta = 0 for to_func, from_func, overlap_fmt, simple_fmt in INTERVALS: delta = to_func(self._domain_max) - to_func(self._domain_min) if delta >= MIN_UNITS or to_func is utils.to_microsecond_count: self._to_unit = to_func self._from_unit = partial(from_func, t=self._type) if previous_delta >= 1: self._fmt = overlap_fmt else: self._fmt = simple_fmt break previous_delta = delta # Compute unit min and max self._unit_min = self._to_unit(self._domain_min) self._unit_max = self._to_unit(self._domain_max) if (self._domain_max - self._from_unit(self._unit_max)).total_seconds() > 0: self._unit_max += 1 self._ticks = self._find_ticks() self._min = self._ticks[0] self._max = self._ticks[-1] def _find_ticks(self): """ Implements the tick-finding algorithm. """ delta = self._unit_max - self._unit_min interval_guess = int(math.ceil(delta / (DEFAULT_TICKS - 1))) candidate_intervals = [] candidate_intervals.append(interval_guess) candidate_intervals.append(interval_guess - 1) candidate_intervals.append(interval_guess + 1) if 0 in candidate_intervals: candidate_intervals.remove(0) candidate_ticks = [] for interval in candidate_intervals: ticks = [] ticks.append(int(math.floor((self._unit_min / interval))) * interval) tick_num = 1 while ticks[tick_num - 1] < self._unit_max: t = ticks[0] + (interval * tick_num) ticks.append(t) tick_num += 1 # Throw out sequences that are too short or too long if len(ticks) < MIN_TICK_COUNT or len(ticks) > MAX_TICK_COUNT: continue candidate_ticks.append( { "interval": interval, "ticks": ticks, "score": self._score(interval, ticks), } ) # Order by best score, using number of ticks as a tie-breaker best = sorted( candidate_ticks, key=lambda c: (c["score"]["total"], len(c["ticks"])) ) ticks = best[0]["ticks"] return [self._from_unit(t) for t in ticks] def _score(self, interval, ticks): """ Score a given tick sequence based on several criteria. This method returns discrete scoring components for easier debugging. """ s = {"pct_waste": 0, "interval_penalty": 0, "len_penalty": 0, "total": 0} # Penalty for wasted scale space waste = (self._unit_min - ticks[0]) + (ticks[-1] - self._unit_max) pct_waste = waste / (self._unit_max - self._unit_min) s["pct_waste"] = pow(10, pct_waste) # Penalty for too many ticks if len(ticks) > 5: s["len_penalty"] = len(ticks) - 5 s["total"] = s["pct_waste"] + s["interval_penalty"] + s["len_penalty"] return s def format_tick(self, tick): """ Format a tick using the inferred time formatting. """ return tick.strftime(self._fmt)
nickromano/django-daily-digest
daily_digest/urls.py
<reponame>nickromano/django-daily-digest from django.conf.urls import url from .views import preview_daily_digest urlpatterns = [ url( r"admin/daily-digest-preview/$", preview_daily_digest, name="preview-daily-digest", ), ]
nickromano/django-daily-digest
project/photos/models.py
from django.contrib.auth.models import User from django.db import models class PhotoUpload(models.Model): user = models.ForeignKey( User, related_name="photo_uploads", on_delete=models.CASCADE ) created = models.DateTimeField(auto_now_add=True)
nickromano/django-daily-digest
leather/scales/ordinal.py
<reponame>nickromano/django-daily-digest<filename>leather/scales/ordinal.py<gh_stars>1-10 #!/usr/bin/env python from decimal import Decimal from leather.scales.base import Scale class Ordinal(Scale): """ A scale that maps individual values (e.g. strings) to a range. """ def __init__(self, domain): self._domain = domain def contains(self, v): """ Return :code:`True` if a given value is contained within this scale's displayed domain. """ return v in self._domain def project(self, value, range_min, range_max): """ Project a value in this scale's domain to a target range. """ range_min = Decimal(range_min) range_max = Decimal(range_max) segments = len(self._domain) segment_size = (range_max - range_min) / segments try: pos = ( range_min + (self._domain.index(value) * segment_size) + (segment_size / 2) ) except ValueError: raise ValueError( 'Value "%s" is not present in Ordinal scale domain' % value ) return pos def project_interval(self, value, range_min, range_max): """ Project a value in this scale's domain to an interval in the target range. This is used for places :class:`.Bars` and :class:`.Columns`. """ range_min = Decimal(range_min) range_max = Decimal(range_max) segments = len(self._domain) segment_size = (range_max - range_min) / segments gap = segment_size / Decimal(20) try: a = range_min + (self._domain.index(value) * segment_size) + gap b = range_min + ((self._domain.index(value) + 1) * segment_size) - gap except ValueError: raise ValueError( 'Value "%s" is not present in Ordinal scale domain' % value ) return (a, b) def ticks(self): """ Generate a series of ticks for this scale. """ return self._domain
nickromano/django-daily-digest
daily_digest/config.py
from collections import namedtuple import importlib import pytz from django.conf import settings from django.contrib.contenttypes.models import ContentType from django.utils.text import slugify class DailyDigestConfigurationException(Exception): pass class DailyDigestRequiredFieldException(DailyDigestConfigurationException): pass def import_class_at_path(path): module_name, class_name = path.rsplit(".", 1) parser_module = importlib.import_module(module_name) parser_class = getattr(parser_module, class_name) return parser_class DailyDigestConfig = namedtuple( "DailyDigestConfig", ["title", "from_email", "to", "timezone", "exclude_today", "chart_configs"], ) def load_config(): global daily_digest_config _config = getattr(settings, "DAILY_DIGEST_CONFIG", {}) DailyDigestChart = namedtuple( "DailyDigestChart", ["title", "slug", "model", "date_field", "filter_kwargs", "distinct_by"], ) _chart_configs = [] for chart in _config.get("charts", []): for key in ["title", "model", "date_field"]: if not chart.get(key): raise DailyDigestRequiredFieldException( "Missing required field {} for chart config.".format(key) ) if chart.get("app_label"): # Allow the user to pass an app_label and model content_type = ContentType.objects.get( app_label=chart["app_label"], model=chart["model"] ) model = content_type.model_class() else: model = import_class_at_path(chart["model"]) chart_config = DailyDigestChart( title=chart["title"], slug=slugify(chart["title"]), model=model, date_field=chart["date_field"], filter_kwargs=chart.get("filter_kwargs", {}), distinct_by=chart.get("distinct_by"), ) _chart_configs.append(chart_config) daily_digest_config = DailyDigestConfig( title=_config.get("title", "Daily Digest"), from_email=_config.get("from_email", settings.DEFAULT_FROM_EMAIL), to=_config.get("to", settings.ADMINS), timezone=pytz.timezone(_config["timezone"]) if _config.get("timezone") else pytz.UTC, exclude_today=_config.get("exclude_today", False), chart_configs=_chart_configs, ) load_config()
nickromano/django-daily-digest
daily_digest/utils.py
# -*- coding: utf-8 -*- from datetime import datetime, timedelta import leather import pytz from cairosvg import svg2png from django.core.mail import EmailMultiAlternatives from django.template.loader import get_template from email.mime.image import MIMEImage from premailer import transform from .config import daily_digest_config leather.theme.background_color = "#ffffff" leather.theme.title_font_family = "Helvetica" leather.theme.legend_font_family = "Helvetica" leather.theme.tick_font_family = "Helvetica" PRIMARY_STROKE_COLOR = "#4383CC" PREV_PERIOD_STROKE_COLOR = "#B4CDEB" EMAIL_TIME_PERIOD = 7 def current_time_naive(): return datetime.now() def series_labels(series_1_count, series_2_count): if series_2_count > 0: percent_change = int( round((float(series_1_count - series_2_count) / series_2_count) * 100, 0) ) if percent_change > 0: percent_change = "+{}%".format(percent_change) else: percent_change = "{}%".format(percent_change) else: percent_change = "" current_label = "Last {} Days ({}) {}".format( EMAIL_TIME_PERIOD, series_1_count, percent_change ) prev_period_label = "Previous Period ({})".format(series_2_count) return current_label.strip(), prev_period_label.strip() def series_data_for_model( queryset, field, timezone, prev_period=False, exclude_today=False ): start_of_today = timezone.localize( current_time_naive().replace(hour=0, minute=0, second=0) ) start_of_today = start_of_today.astimezone(pytz.UTC) if exclude_today: start_of_today = start_of_today - timedelta(days=1) start_of_today += timedelta(days=1) # End of today if prev_period: period_end = start_of_today - timedelta(days=EMAIL_TIME_PERIOD) # End of today period_start = period_end - timedelta(days=EMAIL_TIME_PERIOD) else: period_end = start_of_today period_start = period_end - timedelta(days=EMAIL_TIME_PERIOD) filters = { "{}__gte".format(field): period_start, "{}__lt".format(field): period_end, } series_data = queryset.filter(**filters).values_list(field, flat=True) series_data = list(map(lambda x: x.astimezone(timezone).date(), series_data)) grouped_by_date = [] period_start_local = period_start.astimezone(timezone) for i in range(0, EMAIL_TIME_PERIOD): day = (period_start_local + timedelta(days=i)).date() count = 0 for user_date in series_data: if user_date == day: count += 1 if prev_period: day = day + timedelta( days=EMAIL_TIME_PERIOD ) # offset date to show both series on one chart grouped_by_date.append([day, count]) return grouped_by_date, len(series_data) def svg_data_for_query(queryset, field, chart_name, timezone, exclude_today=False): data, total_count = series_data_for_model( queryset, field, timezone, exclude_today=exclude_today ) data_prev_period, total_count_prev_period = series_data_for_model( queryset, field, timezone, prev_period=True, exclude_today=exclude_today ) # Show every date on the x axis x_axis_ticks = [] for item in data: x_axis_ticks.append(item[0]) chart = leather.Chart() chart.add_x_axis(ticks=x_axis_ticks) # Start at 0 - Turn this into an option y_max = max([item[1] for item in data]) y_max_prev_period = max([item[1] for item in data_prev_period]) chart.add_y_scale(0, max([y_max, y_max_prev_period])) current_label, prev_period_label = series_labels( total_count, total_count_prev_period ) chart.add_line(data, name=current_label, stroke_color=PRIMARY_STROKE_COLOR) chart.add_line( data_prev_period, name=prev_period_label, stroke_color=PREV_PERIOD_STROKE_COLOR, stroke_dasharray="5", ) chart.to_svg("/tmp/{}.svg".format(chart_name), width=480, height=240) with open("/tmp/{}.svg".format(chart_name)) as svgfile: svg_data = svgfile.read() svg2png( url="/tmp/{}.svg".format(chart_name), write_to="/tmp/{}.png".format(chart_name), scale=2, ) return svg_data def charts_data_for_config(chart_format="svg"): charts_data = [] for chart_config in daily_digest_config.chart_configs: title = chart_config.title date_field = chart_config.date_field filter_kwargs = chart_config.filter_kwargs queryset = chart_config.model.objects.filter(**filter_kwargs) if chart_config.distinct_by: queryset = queryset.distinct(chart_config.distinct_by).order_by( chart_config.distinct_by ) result = { "title": title, "svg_data": svg_data_for_query( queryset, date_field, chart_config.slug, timezone=daily_digest_config.timezone, exclude_today=daily_digest_config.exclude_today, ), } if chart_format == "png": del result["svg_data"] result["slug"] = chart_config.slug charts_data.append(result) return charts_data def send_daily_digest(): title = daily_digest_config.title today = daily_digest_config.timezone.localize(current_time_naive()) subject = "{} - {}".format(title, today.strftime("%x")) text_content = title context = {"charts": charts_data_for_config(chart_format="png")} html_template = get_template("daily_digest/email.html") html_content = html_template.render(context) # Inline all css html_content = transform(html_content) msg = EmailMultiAlternatives( subject, text_content, daily_digest_config.from_email, daily_digest_config.to ) msg.attach_alternative(html_content, "text/html") for chart_config in daily_digest_config.chart_configs: with open("/tmp/{}.png".format(chart_config.slug), "rb") as image_file: msg_image = MIMEImage(image_file.read()) msg_image.add_header("Content-ID", "<{}>".format(chart_config.slug)) msg.attach(msg_image) msg.send(fail_silently=False)
nickromano/django-daily-digest
testrunner.py
<reponame>nickromano/django-daily-digest<filename>testrunner.py import os import sys import django from django.conf import settings from django.test.utils import get_runner os.environ["DJANGO_SETTINGS_MODULE"] = "project.settings" test_dir = os.path.dirname(__file__) sys.path.insert(0, test_dir) def runtests(): TestRunner = get_runner(settings) # noqa test_runner = TestRunner(verbosity=1, interactive=True) if hasattr(django, "setup"): django.setup() failures = test_runner.run_tests(["daily_digest"]) sys.exit(bool(failures)) if __name__ == "__main__": runtests()
nickromano/django-daily-digest
project/settings.py
import os import sys BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__))) PROJECT_ROOT = os.path.dirname(os.path.abspath(__file__)) UNIT_TESTING = "test" in sys.argv DEBUG = True ALLOWED_HOSTS = [] SECRET_KEY = "test" LANGUAGE_CODE = "en-us" TIME_ZONE = "UTC" USE_I18N = True USE_L10N = True USE_TZ = True DATABASES = { "default": { "ENGINE": "django.db.backends.sqlite3", "NAME": "db.sqlite3" if UNIT_TESTING else "db.sqlite3", } } MIDDLEWARE = [ "django.middleware.security.SecurityMiddleware", "django.contrib.sessions.middleware.SessionMiddleware", "django.middleware.common.CommonMiddleware", "django.middleware.csrf.CsrfViewMiddleware", "django.contrib.auth.middleware.AuthenticationMiddleware", "django.contrib.messages.middleware.MessageMiddleware", "django.middleware.clickjacking.XFrameOptionsMiddleware", ] MIDDLEWARE_CLASSES = MIDDLEWARE # Django 1.8 support INSTALLED_APPS = [ "django.contrib.admin", "django.contrib.auth", "django.contrib.sessions", "django.contrib.contenttypes", "django.contrib.staticfiles", "django.contrib.messages", # Used for examples "project.photos", "daily_digest", ] ROOT_URLCONF = "project.urls" TEMPLATES = [ { "BACKEND": "django.template.backends.django.DjangoTemplates", "DIRS": [], "APP_DIRS": True, "OPTIONS": { "context_processors": [ "django.template.context_processors.debug", "django.template.context_processors.request", "django.contrib.auth.context_processors.auth", "django.contrib.messages.context_processors.messages", ] }, }, ] LOGIN_URL = "/admin/" STATIC_URL = "/static/" DAILY_DIGEST_CONFIG = { "title": "Daily Digest", "from_email": "<EMAIL>", "timezone": "America/Los_Angeles", "exclude_today": False, "charts": [ { "title": "New Users", "model": "django.contrib.auth.models.User", "date_field": "date_joined", "filter_kwargs": {"is_active": True}, }, { "title": "Photo Uploads", "model": "project.photos.models.PhotoUpload", "date_field": "created", }, ], }
nickromano/django-daily-digest
daily_digest/__init__.py
<reponame>nickromano/django-daily-digest<filename>daily_digest/__init__.py __version__ = "0.0.3" default_app_config = "daily_digest.apps.DailyDigestConfig" # noqa
nickromano/django-daily-digest
leather/scales/linear.py
#!/usr/bin/env python from decimal import Decimal from leather.scales.base import Scale from leather.ticks.score import ScoreTicker class Linear(Scale): """ A scale that linearly maps values from a domain to a range. :param domain_min: The minimum value of the input domain. :param domain_max: The maximum value of the input domain. """ def __init__(self, domain_min, domain_max): if domain_min > domain_max: raise ValueError("Inverted domains are not currently supported.") elif domain_min == domain_max: # Default to unit scale self._data_min = Decimal(0) self._data_max = Decimal(1) else: self._data_min = Decimal(domain_min) self._data_max = Decimal(domain_max) self._ticker = ScoreTicker(self._data_min, self._data_max) def contains(self, v): """ Return :code:`True` if a given value is contained within this scale's domain. """ return self._data_min <= v <= self._data_max def project(self, value, range_min, range_max): """ Project a value in this scale's domain to a target range. """ value = Decimal(value) range_min = Decimal(range_min) range_max = Decimal(range_max) pos = (value - self._ticker.min) / (self._ticker.max - self._ticker.min) return ((range_max - range_min) * pos) + range_min def project_interval(self, value, range_min, range_max): """ Project a value in this scale's domain to an interval in the target range. This is used for places :class:`.Bars` and :class:`.Columns`. """ raise NotImplementedError def ticks(self): """ Generate a series of ticks for this scale. """ return self._ticker.ticks
nickromano/django-daily-digest
leather/chart.py
#!/usr/bin/env python from copy import copy import os import xml.etree.ElementTree as ET import six from leather.axis import Axis from leather.data_types import Date, DateTime from leather.scales import Scale, Linear, Temporal from leather.series import Series, CategorySeries from leather.shapes import Bars, Columns, Dots, Line import leather.svg as svg from leather import theme from leather.utils import X, Y, DIMENSION_NAMES, Box, IPythonSVG, warn class Chart(object): """ Container for all chart types. :param title: An optional title that will be rendered at the top of the chart. """ def __init__(self, title=None): self._title = title self._series_colors = theme.default_series_colors self._layers = [] self._types = [None, None] self._scales = [None, None] self._axes = [None, None] def _palette(self): """ Return a generator for series colors. """ return (color for color in self._series_colors) def set_x_scale(self, scale): """ Set the X :class:`.Scale` for this chart. """ self._scales[X] = scale def set_y_scale(self, scale): """ See :meth:`.Chart.set_x_scale`. """ self._scales[Y] = scale def add_x_scale(self, domain_min, domain_max): """ Create and add a :class:`.Scale`. If the provided domain values are :class:`date` or :class:`datetime` then a :class:`.Temporal` scale will be created, otherwise it will :class:`.Linear`. If you want to set a custom scale class use :meth:`.Chart.set_x_scale` instead. """ scale_type = Linear if isinstance(domain_min, Date.types) or isinstance(domain_min, DateTime.types): scale_type = Temporal self.set_x_scale(scale_type(domain_min, domain_max)) def add_y_scale(self, domain_min, domain_max): """ See :meth:`.Chart.add_x_scale`. """ scale_type = Linear if isinstance(domain_min, Date.types) or isinstance(domain_min, DateTime.types): scale_type = Temporal self.set_y_scale(scale_type(domain_min, domain_max)) def set_x_axis(self, axis): """ Set an :class:`.Axis` class for this chart. """ self._axes[X] = axis def set_y_axis(self, axis): """ See :meth:`.Chart.set_x_axis`. """ self._axes[Y] = axis def add_x_axis(self, ticks=None, tick_formatter=None, name=None): """ Create and add an X :class:`.Axis`. If you want to set a custom axis class use :meth:`.Chart.set_x_axis` instead. """ self._axes[X] = Axis(ticks, tick_formatter, name) def add_y_axis(self, ticks=None, tick_formatter=None, name=None): """ See :meth:`.Chart.add_x_axis`. """ self._axes[Y] = Axis(ticks, tick_formatter, name) def add_series(self, series, shape): """ Add a data :class:`.Series` to the chart. The data types of the new series must be consistent with any series that have already been added. There are several shortcuts for adding different types of data series. See :meth:`.Chart.add_bars`, :meth:`.Chart.add_columns`, :meth:`.Chart.add_dots`, and :meth:`.Chart.add_line`. """ if self._layers and isinstance(self._layers[0][0], CategorySeries): raise RuntimeError( "Additional series can not be added to a chart with a CategorySeries." ) if isinstance(series, CategorySeries): self._types = series._types else: for dim in [X, Y]: if not self._types[dim]: self._types[dim] = series._types[dim] elif series._types[dim] is not self._types[dim]: raise TypeError( "Can't mix axis-data types: %s and %s" % (series._types[dim], self._types[dim]) ) shape.validate_series(series) self._layers.append((series, shape)) def add_bars(self, data, x=None, y=None, name=None, fill_color=None): """ Create and add a :class:`.Series` rendered with :class:`.Bars`. Note that when creating bars in this way the order of the series data will be reversed so that the first item in the series is displayed as the top-most bar in the graphic. If you don't want this to happen use :meth:`.Chart.add_series` instead. """ self.add_series( Series(list(reversed(data)), x=x, y=y, name=name), Bars(fill_color) ) def add_columns(self, data, x=None, y=None, name=None, fill_color=None): """ Create and add a :class:`.Series` rendered with :class:`.Columns`. """ self.add_series(Series(data, x=x, y=y, name=name), Columns(fill_color)) def add_dots(self, data, x=None, y=None, name=None, fill_color=None, radius=None): """ Create and add a :class:`.Series` rendered with :class:`.Dots`. """ self.add_series(Series(data, x=x, y=y, name=name), Dots(fill_color, radius)) def add_line( self, data, x=None, y=None, name=None, stroke_color=None, width=None, stroke_dasharray=None, ): """ Create and add a :class:`.Series` rendered with :class:`.Line`. """ self.add_series( Series(data, x=x, y=y, name=name), Line(stroke_color, width, stroke_dasharray), ) def _validate_dimension(self, dimension): """ Validates that the given scale and axis are valid for the data that has been added to this chart. If a scale or axis has not been set, generates automated ones. """ scale = self._scales[dimension] axis = self._axes[dimension] if not scale: scale = Scale.infer(self._layers, dimension, self._types[dimension]) else: for series, shape in self._layers: if not scale.contains(series.min(dimension)) or not scale.contains( series.max(dimension) ): d = DIMENSION_NAMES[dimension] warn( "Data contains values outside %s scale domain. All data points may not be visible on the chart." % d ) # Only display once per axis break if not axis: axis = Axis() return (scale, axis) def to_svg_group(self, width=None, height=None): """ Render this chart to an SVG group element. This can then be placed inside an :code:`<svg>` tag to make a complete SVG graphic. See :meth:`.Chart.to_svg` for arguments. """ width = width or theme.default_width height = height or theme.default_height if not self._layers: raise ValueError( "You must add at least one series to the chart before rendering." ) if isinstance(theme.margin, float): default_margin = width * theme.margin margin = Box( top=default_margin, right=default_margin, bottom=default_margin, left=default_margin, ) elif isinstance(margin, int): margin = Box(margin, margin, margin, margin) elif not isinstance(margin, Box): margin = Box(*margin) # Root / background root_group = ET.Element("g") root_group.append( ET.Element( "rect", x=six.text_type(0), y=six.text_type(0), width=six.text_type(width), height=six.text_type(height), fill=theme.background_color, ) ) # Margins margin_group = ET.Element("g") margin_group.set("transform", svg.translate(margin.left, margin.top)) margin_width = width - (margin.left + margin.right) margin_height = height - (margin.top + margin.bottom) root_group.append(margin_group) # Header header_group = ET.Element("g") header_margin = 0 if self._title: label = ET.Element( "text", x=six.text_type(0), y=six.text_type(0), fill=theme.title_color ) label.set("font-family", theme.title_font_family) label.set("font-size", six.text_type(theme.title_font_size)) label.text = six.text_type(self._title) header_group.append(label) header_margin += theme.title_font_char_height + theme.title_gap # Legend if len(self._layers) > 1 or isinstance(self._layers[0][0], CategorySeries): legend_group = ET.Element("g") legend_group.set("transform", svg.translate(0, header_margin)) indent = 0 rows = 1 palette = self._palette() for series, shape in self._layers: for item_group, item_width in shape.legend_to_svg(series, palette): if indent + item_width > width: indent = 0 rows += 1 y = (rows - 1) * (theme.legend_font_char_height + theme.legend_gap) item_group.set("transform", svg.translate(indent, y)) indent += item_width legend_group.append(item_group) legend_height = rows * (theme.legend_font_char_height + theme.legend_gap) header_margin += legend_height header_group.append(legend_group) margin_group.append(header_group) # Body body_group = ET.Element("g") body_group.set("transform", svg.translate(0, header_margin)) body_width = margin_width body_height = margin_height - header_margin margin_group.append(body_group) # Axes x_scale, x_axis = self._validate_dimension(X) y_scale, y_axis = self._validate_dimension(Y) bottom_margin = x_axis.estimate_label_margin(x_scale, "bottom") left_margin = y_axis.estimate_label_margin(y_scale, "left") canvas_width = body_width - left_margin canvas_height = body_height - bottom_margin axes_group = ET.Element("g") axes_group.set("transform", svg.translate(left_margin, 0)) axes_group.append(x_axis.to_svg(canvas_width, canvas_height, x_scale, "bottom")) axes_group.append(y_axis.to_svg(canvas_width, canvas_height, y_scale, "left")) header_group.set("transform", svg.translate(left_margin, 0)) body_group.append(axes_group) # Series series_group = ET.Element("g") palette = self._palette() for series, shape in self._layers: series_group.append( shape.to_svg( canvas_width, canvas_height, x_scale, y_scale, series, palette ) ) axes_group.append(series_group) return root_group def to_svg(self, path=None, width=None, height=None): """ Render this chart to an SVG document. The :code:`width` and :code:`height` are specified in SVG's "unitless" units, however, it is usually convenient to specify them as though they were pixels. :param path: Filepath or file-like object to write to. If omitted then the SVG will be returned as a string. If running within IPython, then this will return a SVG object to be displayed. :param width: The output width, in SVG user units. Defaults to :data:`.theme.default_chart_width`. :param height: The output height, in SVG user units. Defaults to :data:`.theme.default_chart_height`. """ width = width or theme.default_chart_width height = height or theme.default_chart_height root = ET.Element( "svg", width=six.text_type(width), height=six.text_type(height), version="1.1", xmlns="http://www.w3.org/2000/svg", ) group = self.to_svg_group(width, height) root.append(group) svg_text = svg.stringify(root) close = True if path: f = None try: if hasattr(path, "write"): f = path close = False else: dirpath = os.path.dirname(path) if dirpath and not os.path.exists(dirpath): os.makedirs(dirpath) f = open(path, "w") f.write(svg.HEADER) f.write(svg_text) finally: if close and f is not None: f.close() else: return IPythonSVG(svg_text)
nickromano/django-daily-digest
leather/axis.py
#!/usr/bin/env python import xml.etree.ElementTree as ET import six from leather import svg from leather import theme class Axis(object): """ A horizontal or vertical chart axis. :param ticks: Instead of inferring tick values from the data, use exactly this sequence of ticks values. These will still be passed to the :code:`tick_formatter`. :param tick_formatter: An optional :func:`.tick_format_function`. """ def __init__(self, ticks=None, tick_formatter=None, name=None): self._ticks = ticks self._tick_formatter = tick_formatter self._name = six.text_type(name) if name is not None else None def _estimate_left_tick_width(self, scale): """ Estimate the y axis space used by tick labels. """ tick_values = self._ticks or scale.ticks() tick_count = len(tick_values) tick_formatter = self._tick_formatter or scale.format_tick max_len = 0 for i, value in enumerate(tick_values): max_len = max(max_len, len(tick_formatter(value, i, tick_count))) return max_len * theme.tick_font_char_width def estimate_label_margin(self, scale, orient): """ Estimate the space needed for the tick labels. """ margin = 0 if orient == "left": margin += self._estimate_left_tick_width(scale) + (theme.tick_size * 2) elif orient == "bottom": margin += theme.tick_font_char_height + (theme.tick_size * 2) if self._name: margin += theme.axis_title_font_char_height + theme.axis_title_gap return margin def to_svg(self, width, height, scale, orient): """ Render this axis to SVG elements. """ group = ET.Element("g") group.set("class", "axis " + orient) # Axis title if self._name is not None: if orient == "left": title_x = -( self._estimate_left_tick_width(scale) + theme.axis_title_gap ) title_y = height / 2 dy = "" transform = svg.rotate(270, title_x, title_y) elif orient == "bottom": title_x = width / 2 title_y = ( height + theme.tick_font_char_height + (theme.tick_size * 2) + theme.axis_title_gap ) dy = "1em" transform = "" title = ET.Element( "text", x=six.text_type(title_x), y=six.text_type(title_y), dy=dy, fill=theme.axis_title_color, transform=transform, ) title.set("text-anchor", "middle") title.set("font-family", theme.axis_title_font_family) title.text = self._name group.append(title) # Ticks if orient == "left": label_x = -(theme.tick_size * 2) x1 = -theme.tick_size x2 = width range_min = height range_max = 0 elif orient == "bottom": label_y = height + (theme.tick_size * 2) y1 = 0 y2 = height + theme.tick_size range_min = 0 range_max = width tick_values = self._ticks or scale.ticks() tick_count = len(tick_values) tick_formatter = self._tick_formatter or scale.format_tick zero_tick_group = None for i, value in enumerate(tick_values): # Tick group tick_group = ET.Element("g") tick_group.set("class", "tick") if value == 0: zero_tick_group = tick_group else: group.append(tick_group) # Tick line projected_value = scale.project(value, range_min, range_max) if value == 0: tick_color = theme.zero_color else: tick_color = theme.tick_color if orient == "left": y1 = projected_value y2 = projected_value elif orient == "bottom": x1 = projected_value x2 = projected_value tick = ET.Element( "line", x1=six.text_type(x1), y1=six.text_type(y1), x2=six.text_type(x2), y2=six.text_type(y2), stroke=tick_color, ) tick.set("stroke-width", six.text_type(theme.tick_width)) tick_group.append(tick) # Tick label if orient == "left": x = label_x y = projected_value dy = "0.32em" text_anchor = "end" elif orient == "bottom": x = projected_value y = label_y dy = "1em" text_anchor = "middle" label = ET.Element( "text", x=six.text_type(x), y=six.text_type(y), dy=dy, fill=theme.label_color, ) label.set("text-anchor", text_anchor) label.set("font-family", theme.tick_font_family) value = tick_formatter(value, i, tick_count) label.text = six.text_type(value) tick_group.append(label) if zero_tick_group is not None: group.append(zero_tick_group) return group def tick_format_function(value, index, tick_count): """ This example shows how to define a function to format tick values for display. :param x: The value to be formatted. :param index: The index of the tick. :param tick_count: The total number of ticks being displayed. :returns: A stringified tick value for display. """ return six.text_type(value)
nickromano/django-daily-digest
daily_digest/management/commands/send_daily_digest.py
from django.core.management.base import BaseCommand from daily_digest.utils import send_daily_digest class Command(BaseCommand): help = "Send daily digest email." def success_message(self, message): if hasattr(self.style, "SUCCESS"): self.stdout.write(self.style.SUCCESS(message)) else: # Django 1.8 self.stdout.write(self.style.MIGRATE_SUCCESS(message)) def handle(self, *args, **options): send_daily_digest() self.success_message("Sent daily digest")
nickromano/django-daily-digest
leather/data_types.py
#!/usr/bin/env python from datetime import date, datetime from decimal import Decimal import six class DataType(object): """ Base class for :class:`.Series` data types. """ @classmethod def infer(cls, v): for t in [DateTime, Date, Number, Text]: if isinstance(v, t.types): return t raise TypeError("No data type available for %s" % type(v)) class Date(DataType): """ Data representing dates. """ types = (date,) class DateTime(DataType): """ Data representing dates with times. """ types = (datetime,) class Number(DataType): """ Data representing numbers. """ types = (int, float, Decimal) class Text(DataType): """ Data representing text/strings. """ types = six.string_types
nickromano/django-daily-digest
daily_digest/views.py
# -*- coding: utf-8 -*- from django.contrib.admin.views.decorators import staff_member_required from django.shortcuts import render from .config import daily_digest_config from .utils import charts_data_for_config @staff_member_required def preview_daily_digest(request): context = {"title": daily_digest_config.title, "charts": charts_data_for_config()} return render(request, "daily_digest/email.html", context)
nickromano/django-daily-digest
leather/ticks/base.py
<reponame>nickromano/django-daily-digest<gh_stars>1-10 #!/usr/bin/env python class Ticker(object): """ Base class for ticker implementations. """ @property def ticks(self): raise NotImplementedError @property def min(self): raise NotImplementedError @property def max(self): raise NotImplementedError
nickromano/django-daily-digest
leather/lattice.py
<gh_stars>1-10 #!/usr/bin/env python from leather.axis import Axis from leather.chart import Chart from leather.data_types import Date, DateTime from leather.grid import Grid from leather.scales import Scale, Linear from leather.series import Series from leather.shapes import Line from leather import theme from leather.utils import X, Y class Lattice(object): """ A grid of charts with synchronized shapes, scales, and axes. Lattice only supports graphing a single series of data. :param shape: An instance of :class:`.Shape` to use to render all series. Defaults to :class:`.Line` if not specified. """ def __init__(self, shape=None): self._shape = shape or Line() self._series = [] self._types = [None, None] self._scales = [None, None] self._axes = [None, None] def set_x_scale(self, scale): """ Set the X :class:`.Scale` for this lattice. """ self._scales[X] = scale def set_y_scale(self, scale): """ See :meth:`.Lattice.set_x_scale`. """ self._scales[Y] = scale def add_x_scale(self, domain_min, domain_max): """ Create and add a :class:`.Scale`. If the provided domain values are :class:`date` or :class:`datetime` then a :class:`.Temporal` scale will be created, otherwise it will :class:`.Linear`. If you want to set a custom scale class use :meth:`.Lattice.set_x_scale` instead. """ scale_type = Linear if isinstance(domain_min, Date.types) or isinstance(domain_min, DateTime.types): scale_type = Temporal self.set_x_scale(scale_type(domain_min, domain_max)) def add_y_scale(self, domain_min, domain_max): """ See :meth:`.Lattice.add_x_scale`. """ scale_type = Linear if isinstance(domain_min, Date.types) or isinstance(domain_min, DateTime.types): scale_type = Temporal self.set_y_scale(scale_type(domain_min, domain_max)) def set_x_axis(self, axis): """ Set an :class:`.Axis` class for this lattice. """ self._axes[X] = axis def set_y_axis(self, axis): """ See :meth:`.Lattice.set_x_axis`. """ self._axes[Y] = axis def add_x_axis(self, ticks=None, tick_formatter=None, name=None): """ Create and add an X :class:`.Axis`. If you want to set a custom axis class use :meth:`.Lattice.set_x_axis` instead. """ self._axes[X] = Axis(ticks=ticks, tick_formatter=tick_formatter, name=name) def add_y_axis(self, ticks=None, tick_formatter=None, name=None): """ See :meth:`.Lattice.add_x_axis`. """ self._axes[Y] = Axis(ticks=ticks, tick_formatter=tick_formatter, name=name) def add_one(self, data, x=None, y=None, title=None): """ Add a data series to this lattice. :param data: A sequence of data suitable for constructing a :class:`.Series`, or a sequence of such objects. :param x: See :class:`.Series`. :param y: See :class:`.Series`. :param title: A title to render above this chart. """ series = Series(data, x=x, y=y, name=title) for dimension in [X, Y]: if self._types[dimension]: if series._types[dimension] is not self._types[dimension]: raise TypeError("All data series must have the same data types.") else: self._types[dimension] = series._types[dimension] self._shape.validate_series(series) self._series.append(series) def add_many(self, data, x=None, y=None, titles=None): """ Same as :meth:`.Lattice.add_one` except :code:`data` is a list of data series to be added simultaneously. See :meth:`.Lattice.add_one` for other arguments. Note that :code:`titles` is a sequence of titles that must be the same length as :code:`data`. """ for i, d in enumerate(data): title = titles[i] if titles else None self.add_one(d, x=x, y=y, title=title) def to_svg(self, path=None, width=None, height=None): """ Render the lattice to an SVG. See :class:`.Grid` for additional documentation. """ layers = [(s, self._shape) for s in self._series] if not self._scales[X]: self._scales[X] = Scale.infer(layers, X, self._types[X]) if not self._scales[Y]: self._scales[Y] = Scale.infer(layers, Y, self._types[Y]) if not self._axes[X]: self._axes[X] = Axis() if not self._axes[Y]: self._axes[Y] = Axis() grid = Grid() for i, series in enumerate(self._series): chart = Chart(title=series.name) chart.set_x_scale(self._scales[X]) chart.set_y_scale(self._scales[Y]) chart.set_x_axis(self._axes[X]) chart.set_y_axis(self._axes[Y]) chart.add_series(series, self._shape) grid.add_one(chart) return grid.to_svg(path, width, height)
Nixon-Aguado/Feature-Extraction-and-Image-Processing-Book-Examples
ExamplesPython_3.6/Chapter9/CamShift.py
<filename>ExamplesPython_3.6/Chapter9/CamShift.py ''' Feature Extraction and Image Processing <NAME> & <NAME> http://www.southampton.ac.uk/~msn/book/ Chapter 9 CamShift: Tracks a region by implementing the CamShift technique. Similar to the MEanShift, but it uses backprojection to update the region positions and size ''' # Set module functions from ImageUtilities import imageReadRGB, showImageRGB, createImageF from ImageRegionsUtilities import densityHistogram, colourFeature, meanShift, backProjection,backProjectionImage,regionSize # Math and iteration from math import exp from timeit import itertools ''' Parameters: pathToDir = Input image directory imageNames = Input image names initialPos = position of the region [column, row] size = Size of the region [column, row] sigma = weight control ''' pathToDir = "../../Images/Chapter9/Input/" imageNames = ["frame1.bmp", "frame2.bmp", "frame3.bmp", "frame4.bmp", "frame5.bmp", "frame6.bmp"] histoSize = 64 initialPos = [100, 60] sizeReg = [12, 18] sigma = 6.0 # Region position and sizes in each frame positions = [ ] positions.append(initialPos) sizes = [ ] sizes.append(sizeReg) # Read image inputImage, width, height = imageReadRGB(pathToDir + imageNames[0]) #showImageRGB(inputImage) # Density and back projection of the region to track q = densityHistogram(inputImage, positions[0], sizeReg, sigma, histoSize) backProjImage = backProjectionImage(inputImage, q, histoSize) #plot3DHistogram(q) # For each frame numImages = len(imageNames) for frameNum in range(1, numImages): # Read next frame and estimate the position by using meanshift currentImage, _, _ = imageReadRGB(pathToDir + imageNames[frameNum]) newPos = meanShift(currentImage, q, sizeReg, sigma, histoSize, positions[frameNum-1]) # Back project and use the projections to determine the new position and size newBackProjImage = backProjectionImage(currentImage, q, histoSize) pos,newSize = regionSize(backProjImage, newBackProjImage, \ positions[frameNum-1], newPos, sizeReg) positions.append(pos) sizes.append(newSize) # Update density and image inputImage = currentImage sizeReg = newSize backProjImage = newBackProjImage #print(positions) #print(sizes) # Show results for frameNum in range(0, numImages): image, _, _ = imageReadRGB(pathToDir + imageNames[frameNum]) p = positions[frameNum] s = sizes[frameNum] borderDistance = [s[0] -5, s[1] -5] for x, y in itertools.product(range(p[0]-s[0], p[0]+s[0]), \ range(p[1]-s[1], p[1]+s[1])): if abs(x-p[0]) > borderDistance[0] or abs(y-p[1]) > borderDistance[1]: image[y,x] = [20, 20, 80] showImageRGB(image)
Nixon-Aguado/Feature-Extraction-and-Image-Processing-Book-Examples
ExamplesPython_3.6/Chapter10/Reprojection.py
''' Feature Extraction and Image Processing <NAME> & <NAME> http://www.southampton.ac.uk/~msn/book/ Chapter 10 Reprojection: Compute a projection from seven corresponding image and 3D points and re-project the image to create a new view of the scene ''' # Set module functions from ImageUtilities import imageReadRGB, imageReadL, showImageRGB, createImageRGB from GeometricUtilities import projectionCubePoints, computeProjection, getPointColours, fillImageColours # Math from math import sin, cos ''' Parameters: pathToDir = Input image directory imageName = Input image name maskName = Mask image name ''' pathToDir = "../../Images/Chapter10/Input/" imageName = "cube1.png" maskName = "mask1.png" # Read image data inputImage, width, height = imageReadRGB(pathToDir + imageName) maskImage, width, height = imageReadL(pathToDir + maskName) showImageRGB(inputImage) centreX, centreY = width/2, height/2 # Corresponding points in the cube image and 3D world pts = [[131-centreX,378-centreY],[110-centreX,188-centreY], [200-centreX,70-centreY],[412-centreX,100-centreY], [410-centreX,285-centreY],[349-centreX,418-centreY], [345-centreX,220-centreY]] q = [[0,0,1],[0,1,1], [0,1,0],[1,1,0], [1,0,0],[1,0,1], [1,1,1]] # Obtain the projection p = computeProjection(pts,q) # Get the image position of the 3D cube points npts = 100 xy = projectionCubePoints(npts, p, centreX, centreY) # Get the colour of the points colours = getPointColours(xy, maskImage, inputImage) # Transform the q points and store in qt qT = [ ] angY = .3 angX = -.2 for pointNum in range(0,len(q)): s = [q[pointNum][0]-.5, q[pointNum][1]-.5, q[pointNum][2]-.5] rx = .5 + cos(angY)*s[0] + sin(angY)*s[2] ry = .5 + sin(angX)*sin(angY)*s[0] + cos(angX)*s[1] - sin(angX)*cos(angY)*s[2] rz = .5 - cos(angX)*sin(angY)*s[0] + sin(angX)*s[1] + cos(angX)*cos(angY)*s[2] qT.append([rx,ry,rz]) # Get the projection of the transformed points p = computeProjection(pts,qT) # The position of the cube points according to the projection of the transformed data xy = projectionCubePoints(npts, p, centreX, centreY) # Use the colours of the original image and the points of the transformed projection to generate an image tImage = createImageRGB(width, height) fillImageColours(colours, xy, tImage) showImageRGB(tImage)
Nixon-Aguado/Feature-Extraction-and-Image-Processing-Book-Examples
ExamplesPython_3.6/Chapter5/InvariantGeneralizedHoughTransform.py
''' Feature Extraction and Image Processing <NAME> & <NAME> http://www.southampton.ac.uk/~msn/book/ Chapter 5 InvariantGeneralizedHoughTransform: Shape detection by the invariant generalized Hough transform ''' # Set module functions from ImageUtilities import imageReadL, showImageF, showImageL, createImageF from ImageOperatorsUtilities import applyCannyEdgeDetector from ImageRegionsUtilities import computeReferencePoint from PlotUtilities import plot3DHistogram # Math and iteration from math import pi, tan, sqrt, atan from timeit import itertools ''' Parameters: pathToDir = Input image directory imageName = Input image name gaussianKernelSize = Gaussian kernel size. Filter noise sobelKernelSize = Sobel kernel size. Edge detection upperT = Upper threshold lowerT = Lower threshold numEntries = Size of the R table minimaDistPoints = To avoid to use close pairs of points maxDistPoints = To gather evidence in local regions ''' pathToDir = "../../Images/Chapter5/Input/" imageName = "Works.png" templateName = "TemplateWorks.png" gaussianKernelSize = 5 sobelKernelSize = 3 upperT = 0.4 lowerT = 0.3 numEntries = 90 minimaDistPoints = 20 maxDistPoints = 200 # Defines the direction of the line used to find pairs of points alpha = pi/2.0 # Read image into array and show templateImage, widthTemplate, heightTemplate = imageReadL(pathToDir + templateName) inputImage, width, height = imageReadL(pathToDir + imageName) showImageL(templateImage) showImageL(inputImage) # Compute edges magnitudeTemplate, angleTemplate = applyCannyEdgeDetector(templateImage, gaussianKernelSize, sobelKernelSize, upperT, lowerT) magnitude, angle = applyCannyEdgeDetector(inputImage, gaussianKernelSize, sobelKernelSize, upperT, lowerT) showImageF(magnitudeTemplate) showImageF(magnitude) # Compute reference point in the template. Template centre refPoint, edgePoints = computeReferencePoint(magnitudeTemplate) # Find the pairs of points in the template according to alpha angle pairPoints = [] numPts = len(edgePoints) for p in range(0, numPts): y1, x1 = (edgePoints[p])[0], (edgePoints[p])[1] # We are looking for two points along the line with slope m m = tan(angleTemplate[y1,x1] - alpha) if m>-1 and m<1: if x1 < refPoint[1]: xi, xf, step = x1 + minimaDistPoints,widthTemplate, 1 else: xi, xf, step = x1 - minimaDistPoints, 0, -1 for x2 in range(xi, xf, step): y2 = int(y1 - m * (x2 - x1)) if y2 > 0 and y2 < heightTemplate and magnitudeTemplate[y2,x2] != 0: pairPoints.append((y2,x2,y1,x1)) break else: m = 1.0/m if y1 < refPoint[0]: yi, yf, step = y1 + minimaDistPoints,heightTemplate, 1 else: yi, yf, step = y1 - minimaDistPoints, 0, -1 for y2 in range(yi, yf, step): x2 = int(x1 - m * (y2 - y1)) if x2 > 0 and x2 < widthTemplate and magnitudeTemplate[y2,x2] != 0: pairPoints.append((y2,x2,y1,x1)) break # Build table (k,c) from each pair of points rTable = [[] for entryIndex in range(numEntries)] deltaAngle = pi / (numEntries - 1.0) numPairs = len(pairPoints) for pair in range(0, numPairs): y2, x2 = (pairPoints[pair])[0], (pairPoints[pair])[1] y1, x1 = (pairPoints[pair])[2], (pairPoints[pair])[3] # Compute beta phi1, phi2 = tan(-angleTemplate[y1,x1]), tan(-angleTemplate[y2,x2]) if 1.0+phi1*phi2 != 0: beta = atan((phi1-phi2)/(1.0+phi1*phi2)) else: beta=1.57 # Compute k if x1- refPoint[1] !=0: m = atan(-float(y1-refPoint[0])/(x1-refPoint[1])) else: m =1.57 k = angleTemplate[y1,x1] - m # Scale distCentre = sqrt((y1-refPoint[0])*(y1-refPoint[0]) + (x1-refPoint[1])*(x1-refPoint[1])) distPoints = sqrt((y2-y1)*(y2-y1) + (x2-x1)*(x2-x1)) c = distCentre/distPoints # Insert in the table. The angle is in the interval -pi/2 to pi/2 entryIndex = int((beta+(pi/2.0))/deltaAngle) entry = rTable[entryIndex] entry.append((k,c)) # Gather evidence for the location accumulator = createImageF(width, height) for x1,y1 in itertools.product(range(0, width), range(0, height)): if magnitude[y1,x1] != 0: # Looking for potential the second points along a line secondPoints = [ ] m = tan(angle[y1,x1] - alpha) if m>-1 and m<1: for delta in range(minimaDistPoints, maxDistPoints): x2 = min(x1 + delta, width-1) y2 = int(y1 - m * (x2 - x1)) if y2 > 0 and y2 < height and magnitude[y2,x2] != 0: secondPoints.append((y2,x2)) break for delta in range(minimaDistPoints, maxDistPoints): x2 = max(0, x1 - delta) y2 = int(y1 - m * (x2 - x1)) if y2 > 0 and y2 < height and magnitude[y2,x2] != 0: secondPoints.append((y2,x2)) break else: m = 1.0/m for delta in range(minimaDistPoints, maxDistPoints): y2 = min(y1 + delta, height-1) x2 = int(x1 - m * (y2 - y1)) if x2 > 0 and x2 < width and magnitude[y2,x2] != 0: secondPoints.append((y2,x2)) break for delta in range(minimaDistPoints, maxDistPoints): y2 = max(0, y1 - delta) x2 = int(x1 - m * (y2 - y1)) if x2 > 0 and x2 < width and magnitude[y2,x2] != 0: secondPoints.append((y2,x2)) break # Gather evidence numPts = len(secondPoints) > 0 for ptIndex in range(0, numPts): secondPoint = secondPoints[ptIndex] y2, x2 = secondPoint[0], secondPoint[1] distPoints = sqrt((y2-y1)*(y2-y1) + (x2-x1)*(x2-x1)) # Compute beta phi1, phi2 = tan(-angle[y1,x1]), tan(-angle[y2,x2]) if 1.0+phi1*phi2 != 0: beta = atan((phi1-phi2)/(1.0+phi1*phi2)) else: beta=1.57 # Find entry in table entryIndex = int((beta+(pi/2.0))/deltaAngle) row = rTable[entryIndex] numEntriesinRow = len(row) for kIndex in range(0, numEntriesinRow): k, c = (row[kIndex])[0], (row[kIndex])[1] distCentre = c * distPoints m = tan(angle[y1,x1] - k) if m>-1 and m<1: for x in range(0, width): y = int(y1 - m * (x - x1)) d = sqrt((x-x1)*(x-x1)+(y-y1)*(y-y1)) if y > 0 and y < height and abs(d-distCentre) < 3: accumulator[y,x] += 3.0 - abs(d-distCentre) else: m = 1.0/m for y in range(0, height): x = int(x1 - m * (y - y1)) d = sqrt((x-x1)*(x-x1)+(y-y1)*(y-y1)) if x > 0 and x < width and abs(d-distCentre) < 3: accumulator[y,x] += 3.0 - abs(d-distCentre) # Plot accumulator plot3DHistogram(accumulator)
Nixon-Aguado/Feature-Extraction-and-Image-Processing-Book-Examples
ExamplesPython_3.6/Chapter4/HarrisOperator.py
''' Feature Extraction and Image Processing <NAME> & <NAME> http://www.southampton.ac.uk/~msn/book/ Chapter 4 HarrisOperator: Compute corners by using the Harris operator or minimal direction ''' # Set module functions from ImageUtilities import imageReadL, createImageF, showImageF, showImageL from ImageOperatorsUtilities import applyCannyEdgeDetector # Iteration from timeit import itertools ''' Parameters: pathToDir = Input image directory imageName = Input image name GaussianKernelSize = Gaussian kernel size. Filter noise sobelKernelSize = Sobel kernel size. Edge detection upperT = Upper threshold lowerT = Lower threshold kernelSize = Size of the kernel k = Second term constant op = H for Harris operator M for minimum direction ''' pathToDir = "../../Images/Chapter4/Input/" imageName = "Shapes.png" GaussianKernelSize = 7 sobelKernelSize = 3 upperT = 0.4 lowerT = 0.2 kernelSize = 9 k = .02 op = "H" # Read image into array inputImage, width, height = imageReadL(pathToDir + imageName) # Show input image showImageL(inputImage) # We apply Canny to obtain the edges from the image # but also need the results of the Sobel operator (Gradient) magnitude, angle, mX, mY = applyCannyEdgeDetector(inputImage, GaussianKernelSize, sobelKernelSize, upperT, lowerT, True) \ # The center of the kernel kernelCentre = int((kernelSize - 1) / 2) # Compute curvature curvature = createImageF(width, height) for x,y in itertools.product(range(0, width), range(0, height)): # If it is an edge if magnitude[y,x] > 0: A, B, C = 0.0, 0.0, 0.0 for wx,wy in itertools.product(range(0, kernelSize), range(0, kernelSize)): posY = y + wy - kernelCentre posX = x + wx - kernelCentre if posY > -1 and posY < height and posX > -1 and posX < width: A += mX[posY,posX] * mX[posY,posX] B += mY[posY,posX] * mY[posY,posX] C += mX[posY,posX] * mY[posY,posX] if op == "H": curvature[y,x] = (A * B) - (C * C) - (k * ((A+B) * (A+B))) if op == "M": d = mX[y,x] * mX[y,x] + mY[y,x] * mY[y,x] if d != 0.0: curvature[y,x] = (A * mY[y,x] * mY[y,x] - \ 2.0 * C * mX[y,x] * mY[y,x] + \ B * mX[y,x] * mX[y,x]) / d showImageF(curvature)
Nixon-Aguado/Feature-Extraction-and-Image-Processing-Book-Examples
ExamplesPython_3.6/Chapter3/Dilation.py
<gh_stars>10-100 ''' Feature Extraction and Image Processing <NAME> & <NAME> http://www.southampton.ac.uk/~msn/book/ Chapter 3 Dilation: Dilation morphological filter ''' # Set module functions from ImageUtilities import imageReadL, showImageL, createImageL # Iteration from timeit import itertools ''' Parameters: pathToDir = Input image directory imageName = Input image name kernelSize = Size of the kernel ''' pathToDir = "../../Images/Chapter3/Input/" imageName = "Logs.png" kernelSize = 5 # Read image into array inputImage, width, height = imageReadL(pathToDir + imageName) # Show input image showImageL(inputImage) # Create Kernel kernelCentre = int((kernelSize - 1) / 2) kernelImage = createImageL(kernelSize, kernelSize) # Set the pixels of a flat kernel for x in range(0, kernelSize): for y in range(0, kernelSize): kernelImage[y,x] = 1 # Create images to store the result outputImage = createImageL(width, height) # Apply kernel kernelCentre = int((kernelSize - 1) / 2) for x,y in itertools.product(range(0, width), range(0, height)): maxValue = 0; for wx,wy in itertools.product(range(0, kernelSize), range(0, kernelSize)): posY = y + wy - kernelCentre posX = x + wx - kernelCentre if posY > -1 and posY < height and posX > -1 and posX < width: sub = float(inputImage[posY,posX]) - kernelImage[wy, wx] if sub > 0 and sub > maxValue: maxValue = sub outputImage[y,x] = maxValue # Show output image showImageL(outputImage)