wikiq_util.py

   1 import sys
   2 import re
   3 from subprocess import Popen, PIPE
   4 from collections import deque
   5 from hashlib import sha1
   6 from deltas.tokenizers import wikitext_split
   7 from mwxml import Dump
   8 import mwpersistence
   9 import mwreverts
  10 from urllib.parse import quote
  11 from deltas import SequenceMatcher
  12
  13 TO_ENCODE = ('title', 'editor')
  14 PERSISTENCE_RADIUS=7
  15
  16 def calculate_persistence(tokens_added):
  17     return(sum([(len(x.revisions)-1) for x in tokens_added]),
  18            len(tokens_added))
  19
  20 class WikiqIterator():
  21     def __init__(self, fh, collapse_user=False):
  22         self.fh = fh
  23         self.collapse_user = collapse_user
  24         self.mwiterator = Dump.from_file(self.fh)
  25         self.namespace_map = { ns.id : ns.name for ns in
  26                                self.mwiterator.site_info.namespaces }
  27         self.__pages = self.load_pages()
  28
  29     def load_pages(self):
  30         for page in self.mwiterator:
  31             yield WikiqPage(page,
  32                             namespace_map = self.namespace_map,
  33                             collapse_user=self.collapse_user)
  34
  35     def __iter__(self):
  36         return self.__pages
  37
  38     def __next__(self):
  39         return next(self._pages)
  40
  41
  42 class WikiqPage():
  43     __slots__ = ('id', 'title', 'namespace', 'redirect',
  44                  'restrictions', 'mwpage', '__revisions',
  45                  'collapse_user')
  46
  47     def __init__(self, page, namespace_map, collapse_user=False):
  48         self.id = page.id
  49         self.namespace = page.namespace
  50         if page.namespace != 0:
  51             self.title = ':'.join([namespace_map[page.namespace], page.title])
  52         else:
  53             self.title = page.title
  54         self.restrictions = page.restrictions
  55         self.collapse_user = collapse_user
  56         self.mwpage = page
  57         self.__revisions = self.rev_list()
  58
  59     def rev_list(self):
  60         # Outline for how we want to handle collapse_user=True
  61         # iteration   rev.user   prev_rev.user   add prev_rev?
  62         #         0          A            None           Never
  63         #         1          A               A           False
  64         #         2          B               A            True
  65         #         3          A               B            True
  66         #         4          A               A           False
  67         # Post-loop                          A          Always
  68         for i, rev in enumerate(self.mwpage):
  69             # never yield the first time
  70             if i == 0:
  71                 if self.collapse_user:
  72                     collapsed_revs = 1
  73                     rev.collapsed_revs = collapsed_revs
  74
  75             else:
  76                 if self.collapse_user:
  77                     # yield if this is the last edit in a seq by a user and reset
  78                     # also yield if we do know who the user is
  79
  80                     if rev.deleted.user or prev_rev.deleted.user:
  81                         yield prev_rev
  82                         collapsed_revs = 1
  83                         rev.collapsed_revs = collapsed_revs
  84
  85                     elif not rev.user.text == prev_rev.user.text:
  86                         yield prev_rev
  87                         collapsed_revs = 1
  88                         rev.collapsed_revs = collapsed_revs
  89                     # otherwise, add one to the counter
  90                     else:
  91                         collapsed_revs += 1
  92                         rev.collapsed_revs = collapsed_revs
  93                 # if collapse_user is false, we always yield
  94                 else:
  95                     yield prev_rev
  96
  97             prev_rev = rev
  98
  99         # also yield the final time
 100         yield prev_rev
 101
 102     def __iter__(self):
 103         return self.__revisions
 104
 105     def __next__(self):
 106         return next(self.__revisions)
 107
 108 class WikiqParser():
 109
 110     def __init__(self, input_file, output_file, collapse_user=False, persist=False, urlencode=False, persist_legacy=False):
 111
 112         self.input_file = input_file
 113         self.output_file = output_file
 114         self.collapse_user = collapse_user
 115         self.persist = persist
 116         self.persist_legacy = persist_legacy
 117         self.printed_header = False
 118         self.namespaces = []
 119         self.urlencode = urlencode
 120
 121     def __get_namespace_from_title(self, title):
 122         default_ns = None
 123
 124         for ns in self.namespaces:
 125             # skip if the namespace is not defined
 126             if ns == None:
 127                 default_ns = self.namespaces[ns]
 128                 continue
 129
 130             if title.startswith(ns + ":"):
 131                 return self.namespaces[ns]
 132
 133         # if we've made it this far with no matches, we return the default namespace
 134         return default_ns
 135
 136     def process(self):
 137
 138         # create a regex that creates the output filename
 139         # output_filename = re.sub(r'^.*/(enwiki\-\d+)\-.*p(\d+)p.*$',
 140         #                         r'output/wikiq-\1-\2.tsv',
 141         #                         input_filename)
 142
 143         # Construct dump file iterator
 144         dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user)
 145
 146         # extract list of namspaces
 147         self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces}
 148
 149         page_count = 0
 150         rev_count = 0
 151
 152
 153         # Iterate through pages
 154         for page in dump:
 155             rev_detector = mwreverts.Detector()
 156
 157             if self.persist or self.persist_legacy:
 158                 window = deque(maxlen=PERSISTENCE_RADIUS)
 159
 160                 if not self.persist_legacy:
 161                     state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
 162                                                     revert_radius=PERSISTENCE_RADIUS)
 163
 164                 else:
 165                     from mw.lib import persistence
 166                     state = persistence.State()
 167
 168             # Iterate through a page's revisions
 169             for rev in page:
 170
 171                 rev_data = {'revid' : rev.id,
 172                             'date_time' : rev.timestamp.strftime('%Y-%m-%d %H:%M:%S'),
 173                             'articleid' : page.id,
 174                             'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
 175                             'title' : '"' + page.title + '"',
 176                             'namespace' : page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title),
 177                             'deleted' : "TRUE" if rev.deleted.text else "FALSE" }
 178
 179                 # if revisions are deleted, /many/ things will be missing
 180                 if rev.deleted.text:
 181                     rev_data['text_chars'] = ""
 182                     rev_data['sha1'] = ""
 183                     rev_data['revert'] = ""
 184                     rev_data['reverteds'] = ""
 185
 186                 else:
 187                     # rev.text can be None if the page has no text
 188                     if not rev.text:
 189                         rev.text = ""
 190                     # if text exists, we'll check for a sha1 and generate one otherwise
 191
 192                     if rev.sha1:
 193                         text_sha1 = rev.sha1
 194                     else:
 195
 196                         text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
 197
 198                     rev_data['sha1'] = text_sha1
 199
 200                     # TODO rev.bytes doesn't work.. looks like a bug
 201                     rev_data['text_chars'] = len(rev.text)
 202
 203                     # generate revert data
 204                     revert = rev_detector.process(text_sha1, rev.id)
 205
 206                     if revert:
 207                         rev_data['revert'] = "TRUE"
 208                         rev_data['reverteds'] = '"' + ",".join([str(x) for x in revert.reverteds]) + '"'
 209                     else:
 210                         rev_data['revert'] = "FALSE"
 211                         rev_data['reverteds'] = ""
 212
 213                 # if the fact that the edit was minor can be hidden, this might be an issue
 214                 rev_data['minor'] = "TRUE" if rev.minor else "FALSE"
 215
 216                 if not rev.deleted.user:
 217                     # wrap user-defined editors in quotes for fread
 218                     rev_data['editor'] = '"' + rev.user.text + '"'
 219                     rev_data['anon'] = "TRUE" if rev.user.id == None else "FALSE"
 220
 221                 else:
 222                     rev_data['anon'] = ""
 223                     rev_data['editor'] = ""
 224
 225                 #if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
 226                 #    redirect = True
 227                 #else:
 228                 #    redirect = False
 229
 230                 #TODO missing: additions_size deletions_size
 231
 232                 # if collapse user was on, lets run that
 233                 if self.collapse_user:
 234                     rev_data['collapsed_revs'] = rev.collapsed_revs
 235
 236                 if self.persist or self.persist_legacy:
 237                     if rev.deleted.text:
 238
 239                         for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]:
 240                             old_rev_data[k] = None
 241                     else:
 242
 243                         if not self.persist_legacy:
 244                             _, tokens_added, tokens_removed = state.update(rev.text, rev.id)
 245
 246                         else:
 247                             _, tokens_added, tokens_removed = state.process(rev.text, rev.id, text_sha1)
 248
 249                         window.append((rev.id, rev_data, tokens_added, tokens_removed))
 250
 251                         if len(window) == PERSISTENCE_RADIUS:
 252                             old_rev_id, old_rev_data, old_tokens_added, old_tokens_removed = window[0]
 253
 254                             num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
 255
 256                             old_rev_data["token_revs"] = num_token_revs
 257                             old_rev_data["tokens_added"] = num_tokens
 258                             old_rev_data["tokens_removed"] = len(old_tokens_removed)
 259                             old_rev_data["tokens_window"] = PERSISTENCE_RADIUS-1
 260
 261                             self.print_rev_data(old_rev_data)
 262
 263                 else:
 264                     self.print_rev_data(rev_data)
 265
 266                 rev_count += 1
 267
 268             if self.persist or self.persist_legacy:
 269                 # print out metadata for the last RADIUS revisions
 270                 for i, item in enumerate(window):
 271                     # if the window was full, we've already printed item 0
 272                     if len(window) == PERSISTENCE_RADIUS and i == 0:
 273                         continue
 274
 275                     rev_id, rev_data, tokens_added, tokens_removed = item
 276                     num_token_revs, num_tokens = calculate_persistence(tokens_added)
 277
 278                     rev_data["token_revs"] = num_token_revs
 279                     rev_data["tokens_added"] = num_tokens
 280                     rev_data["tokens_removed"] = len(tokens_removed)
 281                     rev_data["tokens_window"] = len(window)-(i+1)
 282
 283                     self.print_rev_data(rev_data)
 284
 285             page_count += 1
 286
 287         print("Done: %s revisions and %s pages." % (rev_count, page_count),
 288               file=sys.stderr)
 289
 290     def print_rev_data(self, rev_data):
 291         # if it's the first time through, print the header
 292         if self.urlencode:
 293             for field in TO_ENCODE:
 294                 rev_data[field] = quote(str(rev_data[field]))
 295
 296         if not self.printed_header:
 297             print("\t".join([str(k) for k in sorted(rev_data.keys())]), file=self.output_file)
 298             self.printed_header = True
 299
 300         print("\t".join([str(v) for k, v in sorted(rev_data.items())]), file=self.output_file)
 301
 302
 303 def open_input_file(input_filename):
 304     if re.match(r'.*\.7z$', input_filename):
 305         cmd = ["7za", "x", "-so", input_filename, '*']
 306     elif re.match(r'.*\.gz$', input_filename):
 307         cmd = ["zcat", input_filename]
 308     elif re.match(r'.*\.bz2$', input_filename):
 309         cmd = ["bzcat", "-dk", input_filename]
 310
 311     try:
 312         input_file = Popen(cmd, stdout=PIPE).stdout
 313     except NameError:
 314         input_file = open(input_filename, 'r')
 315
 316     return input_file
 317
 318 def open_output_file(input_filename):
 319     # create a regex that creates the output filename
 320     output_filename = re.sub(r'\.(7z|gz|bz2)?$', '', input_filename)
 321     output_filename = re.sub(r'\.xml', '', output_filename)
 322     output_filename = output_filename + ".tsv"
 323     output_file = open(output_filename, "w")
 324
 325     return output_file