wikiq

   1 #!/usr/bin/env python3
   2
   3 # original wikiq headers are: title articleid revid date_time anon
   4 # editor editor_id minor text_size text_entropy text_md5 reversion
   5 # additions_size deletions_size
   6
   7 import argparse
   8 import sys
   9 import os, os.path
  10 import re
  11 from datetime import datetime
  12
  13 from subprocess import Popen, PIPE
  14 from collections import deque
  15 from hashlib import sha1
  16
  17 from mwxml import Dump
  18
  19 from deltas.tokenizers import wikitext_split
  20 import mwpersistence
  21 import mwreverts
  22 from urllib.parse import quote
  23 TO_ENCODE = ('title', 'editor')
  24 PERSISTENCE_RADIUS=7
  25 from deltas import SequenceMatcher
  26 from deltas import SegmentMatcher
  27
  28 import dataclasses as dc
  29 from dataclasses import dataclass, make_dataclass
  30 import pyarrow as pa
  31 import pyarrow.parquet as pq
  32
  33 class PersistMethod:
  34     none = 0
  35     sequence = 1
  36     segment = 2
  37     legacy = 3
  38
  39 def calculate_persistence(tokens_added):
  40     return(sum([(len(x.revisions)-1) for x in tokens_added]),
  41            len(tokens_added))
  42
  43 class WikiqIterator():
  44     def __init__(self, fh, collapse_user=False):
  45         self.fh = fh
  46         self.collapse_user = collapse_user
  47         self.mwiterator = Dump.from_file(self.fh)
  48         self.namespace_map = { ns.id : ns.name for ns in
  49                                self.mwiterator.site_info.namespaces }
  50         self.__pages = self.load_pages()
  51
  52     def load_pages(self):
  53         for page in self.mwiterator:
  54             yield WikiqPage(page,
  55                             namespace_map = self.namespace_map,
  56                             collapse_user=self.collapse_user)
  57
  58     def __iter__(self):
  59         return self.__pages
  60
  61     def __next__(self):
  62         return next(self._pages)
  63
  64 class WikiqPage():
  65     __slots__ = ('id', 'title', 'namespace', 'redirect',
  66                  'restrictions', 'mwpage', '__revisions',
  67                  'collapse_user')
  68
  69     def __init__(self, page, namespace_map, collapse_user=False):
  70         self.id = page.id
  71         self.namespace = page.namespace
  72         # following mwxml, we assume namespace 0 in cases where
  73         # page.namespace is inconsistent with namespace_map
  74         if page.namespace not in namespace_map:
  75             self.title = page.title
  76             page.namespace = 0
  77         if page.namespace != 0:
  78             self.title = ':'.join([namespace_map[page.namespace], page.title])
  79         else:
  80             self.title = page.title
  81         self.restrictions = page.restrictions
  82         self.collapse_user = collapse_user
  83         self.mwpage = page
  84         self.__revisions = self.rev_list()
  85
  86     def rev_list(self):
  87         # Outline for how we want to handle collapse_user=True
  88         # iteration   rev.user   prev_rev.user   add prev_rev?
  89         #         0          A            None           Never
  90         #         1          A               A           False
  91         #         2          B               A            True
  92         #         3          A               B            True
  93         #         4          A               A           False
  94         # Post-loop                          A          Always
  95         for i, rev in enumerate(self.mwpage):
  96             # never yield the first time
  97             if i == 0:
  98                 if self.collapse_user:
  99                     collapsed_revs = 1
 100                     rev.collapsed_revs = collapsed_revs
 101
 102             else:
 103                 if self.collapse_user:
 104                     # yield if this is the last edit in a seq by a user and reset
 105                     # also yield if we do know who the user is
 106
 107                     if rev.deleted.user or prev_rev.deleted.user:
 108                         yield prev_rev
 109                         collapsed_revs = 1
 110                         rev.collapsed_revs = collapsed_revs
 111
 112                     elif not rev.user.text == prev_rev.user.text:
 113                         yield prev_rev
 114                         collapsed_revs = 1
 115                         rev.collapsed_revs = collapsed_revs
 116                     # otherwise, add one to the counter
 117                     else:
 118                         collapsed_revs += 1
 119                         rev.collapsed_revs = collapsed_revs
 120                 # if collapse_user is false, we always yield
 121                 else:
 122                     yield prev_rev
 123
 124             prev_rev = rev
 125
 126         # also yield the final time
 127         yield prev_rev
 128
 129     def __iter__(self):
 130         return self.__revisions
 131
 132     def __next__(self):
 133         return next(self.__revisions)
 134
 135
 136 class RegexPair(object):
 137     def __init__(self, pattern, label):
 138         self.pattern = re.compile(pattern)
 139         self.label = label
 140         self.has_groups = bool(self.pattern.groupindex)
 141         if self.has_groups:
 142             self.capture_groups = list(self.pattern.groupindex.keys())
 143
 144     def _make_key(self, cap_group):
 145         return ("{}_{}".format(self.label, cap_group))
 146
 147     def matchmake(self, content, rev_data):
 148
 149         temp_dict = {}
 150         # if there are named capture groups in the regex
 151         if self.has_groups:
 152
 153             # if there are matches of some sort in this revision content, fill the lists for each cap_group
 154             if self.pattern.search(content) is not None:
 155                 m = self.pattern.finditer(content)
 156                 matchobjects = list(m)
 157
 158                 for cap_group in self.capture_groups:
 159                     key = self._make_key(cap_group)
 160                     temp_list = []
 161                     for match in matchobjects:
 162                         # we only want to add the match for the capture group if the match is not None
 163                         if match.group(cap_group) != None:
 164                             temp_list.append(match.group(cap_group))
 165
 166                     # if temp_list of matches is empty just make that column None
 167                     if len(temp_list)==0:
 168                         temp_dict[key] = None
 169                     # else we put in the list we made in the for-loop above
 170                     else:
 171                         temp_dict[key] = ', '.join(temp_list)
 172
 173             # there are no matches at all in this revision content, we default values to None
 174             else:
 175                 for cap_group in self.capture_groups:
 176                     key = self._make_key(cap_group)
 177                     temp_dict[key] = None
 178
 179         # there are no capture groups, we just search for all the matches of the regex
 180         else:
 181             #given that there are matches to be made
 182             if type(content) in(str, bytes):
 183                 if self.pattern.search(content) is not None:
 184                     m = self.pattern.findall(content)
 185                     temp_dict[self.label] = ', '.join(m)
 186                 else:
 187                     temp_dict[self.label] = None
 188
 189         # update rev_data with our new columns
 190         for k, v in temp_dict.items():
 191             setattr(rev_data, k, v)
 192
 193         return rev_data
 194
 195 @dataclass()
 196 class RevDataBase():
 197     revid: int
 198     date_time: datetime
 199     articleid: int
 200     editorid: int
 201     title: str
 202     namespace: int
 203     deleted: bool
 204     text_chars: int = None
 205     revert: bool = None
 206     reverteds: list[int] = None
 207     sha1: str = None
 208     minor: bool = None
 209     editor: str = None
 210     anon: bool = None
 211
 212     pa_schema_fields = [
 213         pa.field("revid", pa.int64()),
 214         pa.field("date_time",pa.timestamp('ms')),
 215         pa.field("articleid",pa.int64()),
 216         pa.field("editorid",pa.int64()),
 217         pa.field("title",pa.string()),
 218         pa.field("namespace",pa.int32()),
 219         pa.field("deleted",pa.bool_()),
 220         pa.field("test_chars",pa.int32()),
 221         pa.field("revert",pa.bool_()),
 222         pa.field("reverteds",pa.list_(pa.int64())),
 223         pa.field("sha1",pa.string()),
 224         pa.field("minor",pa.bool_()),
 225         pa.field("editor",pa.string()),
 226         pa.field("anon",pa.bool_())
 227     ]
 228
 229     def to_pyarrow(self):
 230         return dc.astuple(self)
 231
 232     def to_tsv_row(self):
 233
 234         row = []
 235         for f in self.fields():
 236             val = getattr(self, f.name)
 237             if getattr(self, f.name) is None:
 238                 row.append("")
 239             elif f.type == bool:
 240                 row.append("TRUE" if val else "FALSE")
 241
 242             elif f.type == datetime:
 243                 row.append(val.strftime('%Y-%m-%d %H:%M:%S'))
 244
 245             elif f.name in {'editor','title'}:
 246                 s = '"' + val + '"'
 247                 if f.name in TO_ENCODE:
 248                     row.append(quote(str(val)))
 249
 250             elif f.type == list[int]:
 251                 row.append('"' + ",".join([str(x) for x in val]) + '"')
 252
 253             elif f.type == str:
 254                 if f.name in TO_ENCODE:
 255                     row.append(quote(str(val)))
 256             else:
 257                 row.append(val)
 258
 259         return '\t'.join(row)
 260
 261     # def __init__(revid: int,
 262     #              date_time: datetime,
 263     #              articleid: int,
 264     #              editorid: int,
 265     #              title: str,
 266     #              namespace: int,
 267     #              deleted: bool,
 268     #              test_chars: int,
 269     #              revert: bool,
 270     #              reverteds: list[bool],
 271     #              sha1: str,
 272     #              minor: bool,
 273     #              editor: str,
 274     #              anon: bool):
 275
 276
 277
 278 @dataclass()
 279 class RevDataCollapse(RevDataBase):
 280     collapsed_revs:int = None
 281     pa_collapsed_revs_schema = pa.field('collapsed_revs',pa.int64())
 282     pa_schema_fields = RevDataBase.pa_schema_fields + [pa_collapsed_revs_schema]
 283
 284 @dataclass()
 285 class RevDataPersistence(RevDataBase):
 286     token_revs:int = None
 287     tokens_added:int = None
 288     tokens_removed:int = None
 289     tokens_window:int = None
 290
 291     pa_persistence_schema_fields = [
 292         pa.field("token_revs", pa.int64()),
 293         pa.field("tokens_added", pa.int64()),
 294         pa.field("tokens_removed", pa.int64()),
 295         pa.field("tokens_window", pa.int64())]
 296
 297     pa_schema_fields = RevDataBase.pa_schema_fields  + pa_persistence_schema_fields
 298
 299 @dataclass()
 300 class RevDataCollapsePersistence(RevDataCollapse, RevDataPersistence):
 301     pa_schema_fields = RevDataCollapse.pa_schema_fields + RevDataPersistence.pa_persistence_schema_fields
 302
 303 class WikiqParser():
 304     def __init__(self, input_file, output_file, regex_match_revision, regex_match_comment, regex_revision_label, regex_comment_label, collapse_user=False, persist=None, urlencode=False, namespaces = None, revert_radius=15, output_parquet=True, parquet_buffer_size=2000):
 305         """
 306         Parameters:
 307            persist : what persistence method to use. Takes a PersistMethod value
 308         """
 309         self.input_file = input_file
 310
 311         self.collapse_user = collapse_user
 312         self.persist = persist
 313         self.namespaces = []
 314         self.urlencode = urlencode
 315         self.revert_radius = revert_radius
 316
 317         if namespaces is not None:
 318             self.namespace_filter = set(namespaces)
 319         else:
 320             self.namespace_filter = None
 321
 322         self.regex_schemas = []
 323         self.regex_revision_pairs = self.make_matchmake_pairs(regex_match_revision, regex_revision_label)
 324         self.regex_comment_pairs = self.make_matchmake_pairs(regex_match_comment, regex_comment_label)
 325
 326         if self.collapse_user is True:
 327             if self.persist == PersistMethod.none:
 328                 revdata_type = RevDataCollapse
 329             else:
 330                 revdata_type = RevDataCollapsePersistence
 331         elif self.persist != PersistMethod.none:
 332             revdata_type = RevDataPersistence
 333         else:
 334             revdata_type = RevDataBase
 335
 336         regex_fields = [(field.name, list[str], dc.field(default=None)) for field in self.regex_schemas]
 337
 338         self.revdata_type = make_dataclass('RevData_Parser',
 339                                            fields=regex_fields,
 340                                            bases=(revdata_type,))
 341
 342         self.revdata_type.pa_schema_fields = revdata_type.pa_schema_fields + self.regex_schemas
 343
 344
 345         # print(list(map(lambda d: d.name, dc.fields(self.revdata_type))))
 346         # print(self.revdata_type.pa_schema_fields)
 347
 348         if output_parquet is True:
 349             self.output_parquet = True
 350             self.pq_writer = None
 351             self.output_file = output_file
 352             self.parquet_buffer = []
 353             self.parquet_buffer_size = parquet_buffer_size
 354         else:
 355             self.output_file = open(output_file,'w')
 356
 357
 358     def make_matchmake_pairs(self, patterns, labels):
 359         if (patterns is not None and labels is not None) and \
 360            (len(patterns) == len(labels)):
 361             result = []
 362             for pattern, label in zip(patterns, labels):
 363                 result.append(RegexPair(pattern, label))
 364                 self.regex_schemas.append(pa.field(label, pa.list_(pa.string())))
 365
 366             return result
 367         elif (patterns is None and labels is None):
 368             return []
 369         else:
 370             sys.exit('Each regular expression *must* come with a corresponding label and vice versa.')
 371
 372     def matchmake(self, rev, rev_data):
 373         rev_data = self.matchmake_revision(rev.text, rev_data)
 374         rev_data = self.matchmake_comment(rev.comment, rev_data)
 375         return rev_data
 376
 377     def matchmake_revision(self, text, rev_data):
 378          return self.matchmake_pairs(text, rev_data, self.regex_revision_pairs)
 379
 380     def matchmake_comment(self, comment, rev_data):
 381         return self.matchmake_pairs(comment, rev_data, self.regex_comment_pairs)
 382
 383     def matchmake_pairs(self, text, rev_data, pairs):
 384         for pair in pairs:
 385             rev_data = pair.matchmake(text, rev_data)
 386         return rev_data
 387
 388     def __get_namespace_from_title(self, title):
 389         default_ns = None
 390
 391         for ns in self.namespaces:
 392             # skip if the namespace is not defined
 393             if ns == None:
 394                 default_ns = self.namespaces[ns]
 395                 continue
 396
 397             if title.startswith(ns + ":"):
 398                 return self.namespaces[ns]
 399
 400         # if we've made it this far with no matches, we return the default namespace
 401         return default_ns
 402
 403
 404     def process(self):
 405
 406         # create a regex that creates the output filename
 407         # output_filename = re.sub(r'^.*/(enwiki\-\d+)\-.*p(\d+)p.*$',
 408         #                         r'output/wikiq-\1-\2.tsv',
 409         #                         input_filename)
 410
 411         # Construct dump file iterator
 412         dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user)
 413
 414         # extract list of namspaces
 415         self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces}
 416
 417         page_count = 0
 418         rev_count = 0
 419
 420
 421         # Iterate through pages
 422         for page in dump:
 423             namespace = page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title)
 424
 425             # skip namespaces not in the filter
 426             if self.namespace_filter is not None:
 427                 if namespace not in self.namespace_filter:
 428                     continue
 429
 430             rev_detector = mwreverts.Detector(radius = self.revert_radius)
 431
 432             if self.persist != PersistMethod.none:
 433                 window = deque(maxlen=PERSISTENCE_RADIUS)
 434
 435                 if self.persist == PersistMethod.sequence:
 436                     state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
 437                                                     revert_radius=PERSISTENCE_RADIUS)
 438
 439                 elif self.persist == PersistMethod.segment:
 440                     state = mwpersistence.DiffState(SegmentMatcher(tokenizer = wikitext_split),
 441                                                     revert_radius=PERSISTENCE_RADIUS)
 442
 443                 # self.persist == PersistMethod.legacy
 444                 else:
 445                     from mw.lib import persistence
 446                     state = persistence.State()
 447
 448             # Iterate through a page's revisions
 449             for rev in page:
 450
 451                 rev_data = self.revdata_type(revid = rev.id,
 452                                              date_time = datetime.fromtimestamp(rev.timestamp.unix()),
 453                                              articleid = page.id,
 454                                              editorid = "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
 455                                              title =  page.title,
 456                                              deleted = rev.deleted.text,
 457                                              namespace = namespace
 458                                              )
 459
 460                 rev_data = self.matchmake(rev, rev_data)
 461
 462                 if not rev.deleted.text:
 463                     # rev.text can be None if the page has no text
 464                     if not rev.text:
 465                         rev.text = ""
 466                     # if text exists, we'll check for a sha1 and generate one otherwise
 467
 468                     if rev.sha1:
 469                         text_sha1 = rev.sha1
 470                     else:
 471
 472                         text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
 473
 474                     rev_data.sha1 = text_sha1
 475
 476                     # TODO rev.bytes doesn't work.. looks like a bug
 477                     rev_data.text_chars = len(rev.text)
 478
 479                     # generate revert data
 480                     revert = rev_detector.process(text_sha1, rev.id)
 481
 482                     if revert:
 483                         rev_data.revert = True
 484                         rev_data.reverteds = revert.reverteds
 485                     else:
 486                         rev_data.revert = False
 487
 488                 # if the fact that the edit was minor can be hidden, this might be an issue
 489                 rev_data.minor = rev.minor
 490
 491                 if not rev.deleted.user:
 492                     # wrap user-defined editors in quotes for fread
 493                     rev_data.editor = rev.user.text
 494                     rev_data.anon = rev.user.id == None
 495
 496                 #if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
 497                 #    redirect = True
 498                 #else:
 499                 #    redirect = False
 500
 501                 #TODO missing: additions_size deletions_size
 502
 503                 # if collapse user was on, lets run that
 504                 if self.collapse_user:
 505                     rev_data.collapsed_revs = rev.collapsed_revs
 506
 507                 if self.persist != PersistMethod.none:
 508
 509                     if not rev.deleted.text:
 510
 511                         if self.persist != PersistMethod.legacy:
 512                             _, tokens_added, tokens_removed = state.update(rev.text, rev.id)
 513
 514                         else:
 515                             _, tokens_added, tokens_removed = state.process(rev.text, rev.id, text_sha1)
 516
 517                         window.append((rev.id, rev_data, tokens_added, tokens_removed))
 518
 519                         if len(window) == PERSISTENCE_RADIUS:
 520                             old_rev_id, old_rev_data, old_tokens_added, old_tokens_removed = window[0]
 521
 522                             num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
 523
 524                             rev_data.token_revs = num_token_revs
 525                             rev_data.tokens_added = num_tokens
 526                             rev_data.tokens_removed = len(old_tokens_removed)
 527                             rev_data.tokens_window = PERSISTENCE_RADIUS-1
 528
 529                             self.print_rev_data(rev_data)
 530
 531                 else:
 532                     self.print_rev_data(rev_data)
 533
 534                 rev_count += 1
 535
 536             if self.persist != PersistMethod.none:
 537                 # print out metadata for the last RADIUS revisions
 538                 for i, item in enumerate(window):
 539                     # if the window was full, we've already printed item 0
 540                     if len(window) == PERSISTENCE_RADIUS and i == 0:
 541                         continue
 542
 543                     rev_id, rev_data, tokens_added, tokens_removed = item
 544                     num_token_revs, num_tokens = calculate_persistence(tokens_added)
 545
 546                     rev_data.token_revs = num_token_revs
 547                     rev_data.tokens_added = num_tokens
 548                     rev_data.tokens_removed = len(tokens_removed)
 549                     rev_data.tokens_window = len(window)-(i+1)
 550                     self.print_rev_data(rev_data)
 551
 552             page_count += 1
 553
 554         print("Done: %s revisions and %s pages." % (rev_count, page_count),
 555               file=sys.stderr)
 556
 557         if self.output_parquet is True:
 558             self.flush_parquet_buffer()
 559             self.pq_writer.close()
 560
 561         else:
 562             output_file.close()
 563
 564
 565     def write_parquet_row(self, rev_data):
 566         padata = rev_data.to_pyarrow()
 567         self.parquet_buffer.append(padata)
 568
 569         if len(self.parquet_buffer) >= self.parquet_buffer_size:
 570             self.flush_parquet_buffer()
 571
 572
 573     def flush_parquet_buffer(self):
 574         schema = pa.schema(self.revdata_type.pa_schema_fields)
 575
 576         def row_to_col(rg, types):
 577             cols = []
 578             first = rg[0]
 579             for col in first:
 580                 cols.append([col])
 581
 582             for row in rg[1:]:
 583                 for j in range(len(cols)):
 584                     cols[j].append(row[j])
 585
 586             arrays = []
 587             for col, typ in zip(cols, types):
 588                 arrays.append(pa.array(col, typ))
 589             return arrays
 590
 591         outtable = pa.Table.from_arrays(row_to_col(self.parquet_buffer, schema.types), schema=schema)
 592         if self.pq_writer is None:
 593             self.pq_writer = pq.ParquetWriter(self.output_file, schema, flavor='spark')
 594
 595         self.pq_writer.write_table(outtable)
 596         self.parquet_buffer = []
 597
 598     def print_rev_data(self, rev_data):
 599         if self.output_parquet is False:
 600             printfunc = self.write_tsv_row
 601         else:
 602             printfunc = self.write_parquet_row
 603
 604         printfunc(rev_data)
 605
 606     def write_tsv_row(self, rev_data):
 607         line = rev_data.to_tsv_line()
 608         print(line, file=self.output_file)
 609
 610
 611 def open_input_file(input_filename):
 612     if re.match(r'.*\.7z$', input_filename):
 613         cmd = ["7za", "x", "-so", input_filename, "*.xml"]
 614     elif re.match(r'.*\.gz$', input_filename):
 615         cmd = ["zcat", input_filename]
 616     elif re.match(r'.*\.bz2$', input_filename):
 617         cmd = ["bzcat", "-dk", input_filename]
 618
 619     try:
 620         input_file = Popen(cmd, stdout=PIPE).stdout
 621     except NameError:
 622         input_file = open(input_filename, 'r')
 623
 624     return input_file
 625
 626 def get_output_filename(input_filename, parquet = False):
 627     output_filename = re.sub(r'\.(7z|gz|bz2)?$', '', input_filename)
 628     output_filename = re.sub(r'\.xml', '', output_filename)
 629     if parquet is False:
 630         output_filename = output_filename + ".tsv"
 631     else:
 632         output_filename = output_filename + ".parquet"
 633     return output_filename
 634
 635 def open_output_file(input_filename):
 636     # create a regex that creates the output filename
 637     output_filename = get_output_filename(input_filename, parquet = False)
 638     output_file = open(output_filename, "w")
 639     return output_file
 640
 641 parser = argparse.ArgumentParser(description='Parse MediaWiki XML database dumps into tab delimitted data.')
 642
 643 # arguments for the input direction
 644 parser.add_argument('dumpfiles', metavar="DUMPFILE", nargs="*", type=str,
 645                     help="Filename of the compressed or uncompressed XML database dump. If absent, we'll look for content on stdin and output on stdout.")
 646
 647 parser.add_argument('-o', '--output-dir', metavar='DIR', dest='output_dir', type=str, nargs=1,
 648                     help="Directory for output files. If it ends with .parquet output will be in parquet format.")
 649
 650 parser.add_argument('-s', '--stdout', dest="stdout", action="store_true",
 651                     help="Write output to standard out (do not create dump file)")
 652
 653 parser.add_argument('--collapse-user', dest="collapse_user", action="store_true",
 654                     help="Operate only on the final revision made by user a user within all sequences of consecutive edits made by a user. This can be useful for addressing issues with text persistence measures.")
 655
 656 parser.add_argument('-p', '--persistence', dest="persist", default=None, const='', type=str, choices = ['','segment','sequence','legacy'], nargs='?',
 657                     help="Compute and report measures of content persistent: (1) persistent token revisions, (2) tokens added, and (3) number of revision used in computing the first measure. This may by slow.  The defualt is -p=sequence, which uses the same algorithm as in the past, but with improvements to wikitext parsing. Use -p=legacy for old behavior used in older research projects. Use -p=segment for advanced persistence calculation method that is robust to content moves, but prone to bugs, and slower.")
 658
 659 parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true",
 660                     help="Output url encoded text strings. This works around some data issues like newlines in editor names. In the future it may be used to output other text data.")
 661
 662 parser.add_argument('-n', '--namespace-include', dest="namespace_filter", type=int, action='append',
 663                     help="Id number of namspace to include. Can be specified more than once.")
 664
 665 parser.add_argument('-rr',
 666                     '--revert-radius',
 667                     dest="revert_radius",
 668                     type=int,
 669                     action='store',
 670                     default=15,
 671                     help="Number of edits to check when looking for reverts (default: 15)")
 672
 673 parser.add_argument('-RP', '--revision-pattern', dest="regex_match_revision", default=None, type=str, action='append',
 674                     help="The regular expression to search for in revision text. The regex must be surrounded by quotes.")
 675
 676 parser.add_argument('-RPl', '--revision-pattern-label', dest="regex_revision_label", default=None, type=str, action='append',
 677                     help="The label for the outputted column based on matching the regex in revision text.")
 678
 679 parser.add_argument('-CP', '--comment-pattern', dest="regex_match_comment", default=None, type=str, action='append',
 680                     help="The regular expression to search for in comments of revisions.")
 681
 682 parser.add_argument('-CPl', '--comment-pattern-label', dest="regex_comment_label", default=None, type=str, action='append',
 683                     help="The label for the outputted column based on matching the regex in comments.")
 684
 685 args = parser.parse_args()
 686
 687
 688
 689 # set persistence method
 690
 691 if args.persist is None:
 692     persist = PersistMethod.none
 693 elif args.persist == "segment":
 694     persist = PersistMethod.segment
 695 elif args.persist == "legacy":
 696     persist = PersistMethod.legacy
 697 else:
 698     persist = PersistMethod.sequence
 699
 700 if args.namespace_filter is not None:
 701     namespaces = args.namespace_filter
 702 else:
 703     namespaces = None
 704
 705 if len(args.dumpfiles) > 0:
 706     output_parquet = False
 707     for filename in args.dumpfiles:
 708         input_file = open_input_file(filename)
 709
 710         # open directory for output
 711         if args.output_dir:
 712             output_dir = args.output_dir[0]
 713         else:
 714             output_dir = "."
 715
 716         if output_dir.endswith(".parquet"):
 717             output_parquet = True
 718
 719         print("Processing file: %s" % filename, file=sys.stderr)
 720
 721         if args.stdout:
 722             output_file = sys.stdout
 723         else:
 724             filename = os.path.join(output_dir, os.path.basename(filename))
 725             output_file = get_output_filename(filename, parquet = output_parquet)
 726
 727         wikiq = WikiqParser(input_file,
 728                             output_file,
 729                             collapse_user=args.collapse_user,
 730                             persist=persist,
 731                             urlencode=args.urlencode,
 732                             namespaces=namespaces,
 733                             revert_radius=args.revert_radius,
 734                             regex_match_revision = args.regex_match_revision,
 735                             regex_revision_label = args.regex_revision_label,
 736                             regex_match_comment = args.regex_match_comment,
 737                             regex_comment_label = args.regex_comment_label,
 738                             output_parquet=output_parquet)
 739
 740         print(wikiq.output_parquet)
 741         wikiq.process()
 742
 743         # close things
 744         input_file.close()
 745
 746 else:
 747     wikiq = WikiqParser(sys.stdin,
 748                         sys.stdout,
 749                         collapse_user=args.collapse_user,
 750                         persist=persist,
 751                         #persist_legacy=args.persist_legacy,
 752                         urlencode=args.urlencode,
 753                         namespaces=namespaces,
 754                         revert_radius=args.revert_radius,
 755                         regex_match_revision = args.regex_match_revision,
 756                         regex_revision_label = args.regex_revision_label,
 757                         regex_match_comment = args.regex_match_comment,
 758                         regex_comment_label = args.regex_comment_label)
 759
 760     wikiq.process()
 761
 762 # stop_words = "a,able,about,across,after,all,almost,also,am,among,an,and,any,are,as,at,be,because,been,but,by,can,cannot,could,dear,did,do,does,either,else,ever,every,for,from,get,got,had,has,have,he,her,hers,him,his,how,however,i,if,in,into,is,it,its,just,least,let,like,likely,may,me,might,most,must,my,neither,no,nor,not,of,off,often,on,only,or,other,our,own,rather,said,say,says,she,should,since,so,some,than,that,the,their,them,then,there,these,they,this,tis,to,too,twas,us,wants,was,we,were,what,when,where,which,while,who,whom,why,will,with,would,yet,you,your"
 763 # stop_words = stop_words.split(",")