refactor into src-layout package.
This commit is contained in:
Nathan TeBlunthuis
702
src/wikiq/__init__.py
Executable file
702
src/wikiq/__init__.py
Executable file
@@ -0,0 +1,702 @@
|
||||
#!/usr/bin/env python3
|
||||
|
||||
# original wikiq headers are: title articleid revid date_time anon
|
||||
# editor editor_id minor text_size text_entropy text_md5 reversion
|
||||
# additions_size deletions_size
|
||||
|
||||
import argparse
|
||||
import sys
|
||||
import os.path
|
||||
import re
|
||||
from io import TextIOWrapper
|
||||
from itertools import groupby
|
||||
|
||||
from subprocess import Popen, PIPE
|
||||
from collections import deque
|
||||
from hashlib import sha1
|
||||
from typing import Any, IO, TextIO, Generator, Union
|
||||
|
||||
import mwxml
|
||||
from mwxml import Dump
|
||||
|
||||
from deltas.tokenizers import wikitext_split
|
||||
import mwpersistence
|
||||
import mwreverts
|
||||
|
||||
import wikiq.tables as tables
|
||||
from wikiq.tables import RevisionTable
|
||||
from wikiq.wiki_diff_matcher import WikiDiffMatcher
|
||||
|
||||
TO_ENCODE = ('title', 'editor')
|
||||
PERSISTENCE_RADIUS = 7
|
||||
from deltas import SequenceMatcher, SegmentMatcher
|
||||
|
||||
import pyarrow as pa
|
||||
import pyarrow.parquet as pq
|
||||
import pyarrow.csv as pacsv
|
||||
|
||||
class PersistMethod:
|
||||
none = 0
|
||||
sequence = 1
|
||||
segment = 2
|
||||
legacy = 3
|
||||
wikidiff2 = 4
|
||||
|
||||
def calculate_persistence(tokens_added):
|
||||
return (sum([(len(x.revisions) - 1) for x in tokens_added]),
|
||||
len(tokens_added))
|
||||
|
||||
|
||||
def fix_hex_digests(revs: list[mwxml.Revision]) -> list[mwxml.Revision]:
|
||||
i = 0
|
||||
for rev in revs:
|
||||
if rev.text is None:
|
||||
rev.text = ""
|
||||
if not rev.sha1 and not rev.deleted.text:
|
||||
rev.sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
|
||||
revs[i] = rev
|
||||
i += 1
|
||||
return revs
|
||||
|
||||
|
||||
class WikiqIterator:
|
||||
def __init__(self, fh, collapse_user=False):
|
||||
self.fh = fh
|
||||
self.collapse_user = collapse_user
|
||||
self.mwiterator = Dump.from_file(self.fh)
|
||||
self.namespace_map = {ns.id: ns.name for ns in
|
||||
self.mwiterator.site_info.namespaces}
|
||||
self.__pages: Generator[WikiqPage] = self.load_pages()
|
||||
|
||||
def load_pages(self):
|
||||
for page in self.mwiterator:
|
||||
yield WikiqPage(page,
|
||||
namespace_map=self.namespace_map,
|
||||
collapse_user=self.collapse_user)
|
||||
|
||||
def __iter__(self):
|
||||
return self.__pages
|
||||
|
||||
def __next__(self):
|
||||
return next(self.__pages)
|
||||
|
||||
|
||||
class WikiqPage:
|
||||
__slots__ = ('id', 'redirect',
|
||||
'restrictions', 'mwpage', '__revisions',
|
||||
'collapse_user')
|
||||
|
||||
def __init__(self, page, namespace_map, collapse_user=False):
|
||||
self.id = page.id
|
||||
# following mwxml, we assume namespace 0 in cases where
|
||||
# page.namespace is inconsistent with namespace_map
|
||||
if page.namespace not in namespace_map:
|
||||
page.namespace = 0
|
||||
if page.namespace != 0:
|
||||
page.title = ':'.join([namespace_map[page.namespace], page.title])
|
||||
self.restrictions = page.restrictions
|
||||
self.collapse_user = collapse_user
|
||||
self.mwpage = page
|
||||
self.__revisions: Generator[list[mwxml.Revision]] = self.rev_list()
|
||||
|
||||
@staticmethod
|
||||
def user_text(rev) -> Union[str, None]:
|
||||
return None if rev.deleted.user else rev.user.text
|
||||
|
||||
def rev_list(self):
|
||||
# Outline for how we want to handle collapse_user=True
|
||||
# iteration rev.user prev_rev.user add prev_rev?
|
||||
# 0 A None Never
|
||||
# 1 A A False
|
||||
# 2 B A True
|
||||
# 3 A B True
|
||||
# 4 A A False
|
||||
# Post-loop A Always
|
||||
|
||||
if not self.collapse_user:
|
||||
for rev in self.mwpage:
|
||||
yield [rev]
|
||||
return
|
||||
|
||||
for _, revs in groupby(self.mwpage, self.user_text):
|
||||
# All revisions are either from the same user, or this is a single
|
||||
# revision where the user is missing.
|
||||
yield list(revs)
|
||||
|
||||
def __iter__(self):
|
||||
return self.__revisions
|
||||
|
||||
def __next__(self):
|
||||
return next(self.__revisions)
|
||||
|
||||
|
||||
"""
|
||||
A RegexPair is defined by a regular expression (pattern) and a label.
|
||||
The pattern can include capture groups. If it does then each capture group will have a resulting column in the output.
|
||||
If the pattern does not include a capture group, then only one output column will result.
|
||||
"""
|
||||
|
||||
class RegexPair(object):
|
||||
def __init__(self, pattern, label):
|
||||
self.pattern = re.compile(pattern)
|
||||
self.label = label
|
||||
self.has_groups = bool(self.pattern.groupindex)
|
||||
if self.has_groups:
|
||||
self.capture_groups = list(self.pattern.groupindex.keys())
|
||||
|
||||
def get_pyarrow_fields(self):
|
||||
if self.has_groups:
|
||||
fields = [pa.field(self._make_key(cap_group), pa.string())
|
||||
for cap_group in self.capture_groups]
|
||||
else:
|
||||
fields = [pa.field(self.label, pa.string())]
|
||||
|
||||
return fields
|
||||
|
||||
def _make_key(self, cap_group):
|
||||
return "{}_{}".format(self.label, cap_group)
|
||||
|
||||
def matchmake(self, content: str) -> dict:
|
||||
temp_dict = {}
|
||||
# if there are named capture groups in the regex
|
||||
if self.has_groups:
|
||||
|
||||
# if there are matches of some sort in this revision content, fill the lists for each cap_group
|
||||
if self.pattern.search(content) is not None:
|
||||
m = self.pattern.finditer(content)
|
||||
matchobjects = list(m)
|
||||
|
||||
for cap_group in self.capture_groups:
|
||||
key = self._make_key(cap_group)
|
||||
temp_list = []
|
||||
for match in matchobjects:
|
||||
# we only want to add the match for the capture group if the match is not None
|
||||
if match.group(cap_group) is not None:
|
||||
temp_list.append(match.group(cap_group))
|
||||
|
||||
# if temp_list of matches is empty just make that column None
|
||||
if len(temp_list) == 0:
|
||||
temp_dict[key] = None
|
||||
# else we put in the list we made in the for-loop above
|
||||
else:
|
||||
temp_dict[key] = ', '.join(temp_list)
|
||||
|
||||
# there are no matches at all in this revision content, we default values to None
|
||||
else:
|
||||
for cap_group in self.capture_groups:
|
||||
key = self._make_key(cap_group)
|
||||
temp_dict[key] = None
|
||||
|
||||
# there are no capture groups, we just search for all the matches of the regex
|
||||
else:
|
||||
# given that there are matches to be made
|
||||
if type(content) in (str, bytes):
|
||||
if self.pattern.search(content) is not None:
|
||||
m = self.pattern.findall(content)
|
||||
temp_dict[self.label] = ', '.join(m)
|
||||
else:
|
||||
temp_dict[self.label] = None
|
||||
|
||||
return temp_dict
|
||||
|
||||
|
||||
class WikiqParser:
|
||||
def __init__(self,
|
||||
input_file: Union[TextIOWrapper, IO[Any], IO[bytes]],
|
||||
output_file: Union[TextIO, str],
|
||||
regex_match_revision: list[str],
|
||||
regex_match_comment: list[str],
|
||||
regex_revision_label: list[str],
|
||||
regex_comment_label: list[str],
|
||||
text: bool = False,
|
||||
diff: bool = False,
|
||||
collapse_user: bool = False,
|
||||
persist: int = None,
|
||||
namespaces: Union[list[int], None] = None,
|
||||
revert_radius: int = 15,
|
||||
output_parquet: bool = True,
|
||||
parquet_buffer_size: int = 2000
|
||||
):
|
||||
|
||||
"""
|
||||
Parameters:
|
||||
persist : what persistence method to use. Takes a PersistMethod value
|
||||
"""
|
||||
self.input_file = input_file
|
||||
|
||||
self.collapse_user: bool = collapse_user
|
||||
self.persist: int = persist
|
||||
self.namespaces = []
|
||||
self.revert_radius = revert_radius
|
||||
self.diff = diff
|
||||
self.text = text
|
||||
if namespaces is not None:
|
||||
self.namespace_filter = set(namespaces)
|
||||
else:
|
||||
self.namespace_filter = None
|
||||
|
||||
self.regex_schemas = []
|
||||
self.regex_revision_pairs: list[RegexPair] = self.make_matchmake_pairs(regex_match_revision,
|
||||
regex_revision_label)
|
||||
self.regex_comment_pairs: list[RegexPair] = self.make_matchmake_pairs(regex_match_comment, regex_comment_label)
|
||||
|
||||
# here we initialize the variables we need for output.
|
||||
if output_parquet is True:
|
||||
self.output_parquet = True
|
||||
self.pq_writer = None
|
||||
self.output_file = output_file
|
||||
self.parquet_buffer = []
|
||||
self.parquet_buffer_size = parquet_buffer_size
|
||||
else:
|
||||
self.print_header = True
|
||||
if output_file == sys.stdout.buffer:
|
||||
|
||||
self.output_file = output_file
|
||||
else:
|
||||
self.output_file = open(output_file, 'wb')
|
||||
self.output_parquet = False
|
||||
|
||||
def make_matchmake_pairs(self, patterns, labels) -> list[RegexPair]:
|
||||
if (patterns is not None and labels is not None) and \
|
||||
(len(patterns) == len(labels)):
|
||||
result: list[RegexPair] = []
|
||||
for pattern, label in zip(patterns, labels):
|
||||
rp = RegexPair(pattern, label)
|
||||
result.append(rp)
|
||||
self.regex_schemas = self.regex_schemas + rp.get_pyarrow_fields()
|
||||
return result
|
||||
elif (patterns is None) and (labels is None):
|
||||
return []
|
||||
else:
|
||||
sys.exit('Each regular expression *must* come with a corresponding label and vice versa.')
|
||||
|
||||
def matchmake_revision(self, rev: mwxml.Revision):
|
||||
result = self.matchmake_text(rev.text)
|
||||
for k, v in self.matchmake_comment(rev.comment).items():
|
||||
result[k] = v
|
||||
return result
|
||||
|
||||
def matchmake_text(self, text: str):
|
||||
return self.matchmake_pairs(text, self.regex_revision_pairs)
|
||||
|
||||
def matchmake_comment(self, comment: str):
|
||||
return self.matchmake_pairs(comment, self.regex_comment_pairs)
|
||||
|
||||
@staticmethod
|
||||
def matchmake_pairs(text, pairs):
|
||||
result = {}
|
||||
for pair in pairs:
|
||||
for k, v in pair.matchmake(text).items():
|
||||
result[k] = v
|
||||
return result
|
||||
|
||||
def __get_namespace_from_title(self, title):
|
||||
default_ns = None
|
||||
|
||||
for ns in self.namespaces:
|
||||
# skip if the namespace is not defined
|
||||
if ns is None:
|
||||
default_ns = self.namespaces[ns]
|
||||
continue
|
||||
|
||||
if title.startswith(ns + ":"):
|
||||
return self.namespaces[ns]
|
||||
|
||||
# if we've made it this far with no matches, we return the default namespace
|
||||
return default_ns
|
||||
|
||||
def process(self):
|
||||
|
||||
# create a regex that creates the output filename
|
||||
# output_filename = re.sub(r'^.*/(enwiki\-\d+)\-.*p(\d+)p.*$',
|
||||
# r'output/wikiq-\1-\2.tsv',
|
||||
# input_filename)
|
||||
|
||||
# Construct dump file iterator
|
||||
dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user)
|
||||
|
||||
reverts_column = tables.RevisionReverts()
|
||||
|
||||
table = RevisionTable([
|
||||
tables.RevisionId(),
|
||||
tables.RevisionTimestamp(),
|
||||
tables.RevisionArticleId(),
|
||||
tables.RevisionPageTitle(),
|
||||
tables.RevisionNamespace(),
|
||||
tables.RevisionDeleted(),
|
||||
tables.RevisionEditorId(),
|
||||
tables.RevisionEditSummary(),
|
||||
tables.RevisionTextChars(),
|
||||
reverts_column,
|
||||
tables.RevisionSha1(),
|
||||
tables.RevisionIsMinor(),
|
||||
tables.RevisionEditorText(),
|
||||
tables.RevisionIsAnon(),
|
||||
])
|
||||
|
||||
if self.text:
|
||||
table.columns.append(tables.RevisionText())
|
||||
|
||||
if self.collapse_user:
|
||||
table.columns.append(tables.RevisionCollapsed())
|
||||
|
||||
# extract list of namespaces
|
||||
self.namespaces = {ns.name: ns.id for ns in dump.mwiterator.site_info.namespaces}
|
||||
|
||||
page_count = 0
|
||||
rev_count = 0
|
||||
|
||||
writer: Union[pq.ParquetWriter, pacsv.CSVWriter]
|
||||
|
||||
schema = table.schema()
|
||||
schema = schema.append(pa.field('revert', pa.bool_(), nullable=True))
|
||||
|
||||
if self.diff:
|
||||
from wikiq.diff_pyarrow_schema import diff_field
|
||||
schema = schema.append(diff_field)
|
||||
|
||||
# Add regex fields to the schema.
|
||||
for pair in self.regex_revision_pairs:
|
||||
for field in pair.get_pyarrow_fields():
|
||||
schema = schema.append(field)
|
||||
|
||||
for pair in self.regex_comment_pairs:
|
||||
for field in pair.get_pyarrow_fields():
|
||||
schema = schema.append(field)
|
||||
|
||||
if self.persist != PersistMethod.none:
|
||||
table.columns.append(tables.RevisionText())
|
||||
schema = schema.append(pa.field('token_revs', pa.int64(), nullable=True))
|
||||
schema = schema.append(pa.field('tokens_added', pa.int64(), nullable=True))
|
||||
schema = schema.append(pa.field('tokens_removed', pa.int64(), nullable=True))
|
||||
schema = schema.append(pa.field('tokens_window', pa.int64(), nullable=True))
|
||||
|
||||
if self.output_parquet:
|
||||
pageid_sortingcol = pq.SortingColumn(schema.get_field_index('pageid'))
|
||||
revid_sortingcol = pq.SortingColumn(schema.get_field_index('pageid'))
|
||||
writer = pq.ParquetWriter(self.output_file, schema, flavor='spark', sorting_columns=[pageid_sortingcol, revid_sortingcol])
|
||||
else:
|
||||
writer = pacsv.CSVWriter(self.output_file, schema, write_options=pacsv.WriteOptions(delimiter='\t'))
|
||||
|
||||
regex_matches = {}
|
||||
|
||||
# Iterate through pages
|
||||
for page in dump:
|
||||
revision_texts = []
|
||||
|
||||
# skip namespaces not in the filter
|
||||
if self.namespace_filter is not None:
|
||||
if page.mwpage.namespace not in self.namespace_filter:
|
||||
continue
|
||||
|
||||
# Disable detecting reverts if radius is 0.
|
||||
if self.revert_radius > 0:
|
||||
reverts_column.rev_detector = mwreverts.Detector(radius=self.revert_radius)
|
||||
else:
|
||||
reverts_column.rev_detector = None
|
||||
|
||||
# Iterate through a page's revisions
|
||||
for revs in page:
|
||||
# Revisions may or may not be grouped into lists of contiguous revisions by the
|
||||
# same user. We call these "edit sessions". Otherwise revs is a list containing
|
||||
# exactly one revision.
|
||||
revs = list(revs)
|
||||
revs = fix_hex_digests(revs)
|
||||
|
||||
table.add(page.mwpage, revs)
|
||||
|
||||
# if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
|
||||
# redirect = True
|
||||
# else:
|
||||
# redirect = False
|
||||
|
||||
# TODO missing: additions_size deletions_size
|
||||
|
||||
rev_count += 1
|
||||
|
||||
# Get the last revision in the edit session.
|
||||
rev = revs[-1]
|
||||
regex_dict = self.matchmake_revision(rev)
|
||||
for k, v in regex_dict.items():
|
||||
if regex_matches.get(k) is None:
|
||||
regex_matches[k] = []
|
||||
regex_matches[k].append(v)
|
||||
|
||||
revision_texts.append(rev.text)
|
||||
|
||||
wikidiff_matcher = None
|
||||
if self.diff or self.persist == PersistMethod.wikidiff2:
|
||||
wikidiff_matcher = WikiDiffMatcher(revision_texts,
|
||||
tokenizer=wikitext_split,
|
||||
)
|
||||
|
||||
# Collect the set of revisions currently buffered in the table so we can run multi-revision functions on them.
|
||||
row_buffer = table.pop()
|
||||
|
||||
if self.diff:
|
||||
row_buffer['diff'] = [[entry for entry in wikidiff_matcher.diffs[i]['diff'] if entry['type'] != 0 ] for i in range(len(revision_texts))]
|
||||
|
||||
is_revert_column: list[Union[bool, None]] = []
|
||||
for r, d in zip(row_buffer['reverteds'], row_buffer['deleted']):
|
||||
if self.revert_radius == 0 or d:
|
||||
is_revert_column.append(None)
|
||||
else:
|
||||
is_revert_column.append(r is not None)
|
||||
|
||||
row_buffer['revert'] = is_revert_column
|
||||
|
||||
for k, v in regex_matches.items():
|
||||
row_buffer[k] = v
|
||||
regex_matches = {}
|
||||
|
||||
|
||||
if self.persist != PersistMethod.none:
|
||||
window = deque(maxlen=PERSISTENCE_RADIUS)
|
||||
|
||||
row_buffer['token_revs'] = []
|
||||
row_buffer['tokens_added'] = []
|
||||
row_buffer['tokens_removed'] = []
|
||||
row_buffer['tokens_window'] = []
|
||||
|
||||
if self.persist == PersistMethod.sequence:
|
||||
state = mwpersistence.DiffState(SequenceMatcher(tokenizer=wikitext_split),
|
||||
revert_radius=PERSISTENCE_RADIUS)
|
||||
elif self.persist == PersistMethod.segment:
|
||||
state = mwpersistence.DiffState(SegmentMatcher(tokenizer=wikitext_split),
|
||||
revert_radius=PERSISTENCE_RADIUS)
|
||||
elif self.persist == PersistMethod.wikidiff2:
|
||||
|
||||
state = mwpersistence.DiffState(wikidiff_matcher,
|
||||
revert_radius=PERSISTENCE_RADIUS)
|
||||
else:
|
||||
from mw.lib import persistence
|
||||
state = persistence.State()
|
||||
|
||||
for idx, text in enumerate(row_buffer['text']):
|
||||
rev_id = row_buffer['revid'][idx]
|
||||
if self.persist != PersistMethod.legacy:
|
||||
_, tokens_added, tokens_removed = state.update(text, rev_id)
|
||||
else:
|
||||
_, tokens_added, tokens_removed = state.process(text, rev_id)
|
||||
|
||||
window.append((rev_id, tokens_added, tokens_removed))
|
||||
|
||||
if len(window) == PERSISTENCE_RADIUS:
|
||||
old_rev_id, old_tokens_added, old_tokens_removed = window.popleft()
|
||||
num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
|
||||
|
||||
row_buffer['token_revs'].append(num_token_revs)
|
||||
row_buffer['tokens_added'].append(num_tokens)
|
||||
row_buffer['tokens_removed'].append(len(old_tokens_removed))
|
||||
row_buffer['tokens_window'].append(PERSISTENCE_RADIUS - 1)
|
||||
|
||||
# print out metadata for the last RADIUS revisions
|
||||
for i, item in enumerate(window):
|
||||
# if the window was full, we've already printed item 0
|
||||
if len(window) == PERSISTENCE_RADIUS and i == 0:
|
||||
continue
|
||||
|
||||
rev_id, tokens_added, tokens_removed = item
|
||||
num_token_revs, num_tokens = calculate_persistence(tokens_added)
|
||||
|
||||
row_buffer['token_revs'].append(num_token_revs)
|
||||
row_buffer['tokens_added'].append(num_tokens)
|
||||
row_buffer['tokens_removed'].append(len(tokens_removed))
|
||||
row_buffer['tokens_window'].append(len(window) - (i + 1))
|
||||
|
||||
if not self.text:
|
||||
del row_buffer['text']
|
||||
|
||||
writer.write(pa.table(row_buffer, schema=schema))
|
||||
|
||||
page_count += 1
|
||||
|
||||
print("Done: %s revisions and %s pages." % (rev_count, page_count),
|
||||
file=sys.stderr)
|
||||
|
||||
writer.close()
|
||||
|
||||
def match_archive_suffix(input_filename):
|
||||
if re.match(r'.*\.7z$', input_filename):
|
||||
cmd = ["7za", "x", "-so", input_filename]
|
||||
elif re.match(r'.*\.gz$', input_filename):
|
||||
cmd = ["zcat", input_filename]
|
||||
elif re.match(r'.*\.bz2$', input_filename):
|
||||
cmd = ["bzcat", "-dk", input_filename]
|
||||
else:
|
||||
raise ValueError("Unrecognized file type: %s" % input_filename)
|
||||
return cmd
|
||||
|
||||
|
||||
def open_input_file(input_filename, fandom_2020=False):
|
||||
cmd = match_archive_suffix(input_filename)
|
||||
if fandom_2020:
|
||||
cmd.append("*.xml")
|
||||
try:
|
||||
return Popen(cmd, stdout=PIPE).stdout
|
||||
except NameError:
|
||||
return open(input_filename, 'r')
|
||||
|
||||
|
||||
def get_output_filename(input_filename, parquet=False) -> str:
|
||||
output_filename = re.sub(r'\.(7z|gz|bz2)?$', '', input_filename)
|
||||
output_filename = re.sub(r'\.xml', '', output_filename)
|
||||
if parquet is False:
|
||||
output_filename = output_filename + ".tsv"
|
||||
else:
|
||||
output_filename = output_filename + ".parquet"
|
||||
return output_filename
|
||||
|
||||
|
||||
def open_output_file(input_filename):
|
||||
# create a regex that creates the output filename
|
||||
output_filename = get_output_filename(input_filename, parquet=False)
|
||||
output_file = open(output_filename, "w")
|
||||
return output_file
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description='Parse MediaWiki XML database dumps into tab delimited data.')
|
||||
|
||||
# arguments for the input direction
|
||||
parser.add_argument('dumpfiles', metavar="DUMPFILE", nargs="*", type=str,
|
||||
help="Filename of the compressed or uncompressed XML database dump. If absent, we'll look for content on stdin and output on stdout.")
|
||||
|
||||
parser.add_argument('-o', '--output', metavar='OUTPUT', dest='output', type=str, nargs=1,
|
||||
help="Directory for output files. If it ends with .parquet output will be in parquet format.")
|
||||
|
||||
parser.add_argument('-s', '--stdout', dest="stdout", action="store_true",
|
||||
help="Write output to standard out (do not create dump file)")
|
||||
|
||||
parser.add_argument('--collapse-user', dest="collapse_user", action="store_true",
|
||||
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.")
|
||||
|
||||
parser.add_argument('-p', '--persistence', dest="persist", default=None, const='', type=str,
|
||||
choices=['', 'wikidiff2', 'segment', 'sequence', 'legacy'], nargs='?',
|
||||
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 default is no persistence. -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. -p=segment attempts advanced persistence calculation method that is robust to content moves, but prone to bugs, and slower. -p=wikidiff2 is like segment, but uses the wikidiff2 algorithm, which (should be) faster and more robust.")
|
||||
|
||||
parser.add_argument('-n', '--namespace-include', dest="namespace_filter", type=int, action='append',
|
||||
help="Id number of namespace to include. Can be specified more than once.")
|
||||
|
||||
parser.add_argument('-rr',
|
||||
'--revert-radius',
|
||||
dest="revert_radius",
|
||||
type=int,
|
||||
action='store',
|
||||
default=15,
|
||||
help="Number of edits to check when looking for reverts (default: 15)")
|
||||
|
||||
parser.add_argument('-RP', '--revision-pattern', dest="regex_match_revision", default=None, type=str,
|
||||
action='append',
|
||||
help="The regular expression to search for in revision text. The regex must be surrounded by quotes.")
|
||||
|
||||
parser.add_argument('-RPl', '--revision-pattern-label', dest="regex_revision_label", default=None, type=str,
|
||||
action='append',
|
||||
help="The label for the outputted column based on matching the regex in revision text.")
|
||||
|
||||
parser.add_argument('-CP', '--comment-pattern', dest="regex_match_comment", default=None, type=str, action='append',
|
||||
help="The regular expression to search for in comments of revisions.")
|
||||
|
||||
parser.add_argument('-CPl', '--comment-pattern-label', dest="regex_comment_label", default=None, type=str,
|
||||
action='append',
|
||||
help="The label for the outputted column based on matching the regex in comments.")
|
||||
|
||||
parser.add_argument('-d', '--diff', dest="diff", default=False,
|
||||
action='store_true',
|
||||
help="Output a diff structure for each revision with information about changed or moved lines.")
|
||||
|
||||
parser.add_argument('-t', '--text', dest="text", default=False,
|
||||
action='store_true',
|
||||
help="Output the text of the revision.")
|
||||
|
||||
parser.add_argument('--fandom-2020', dest="fandom_2020",
|
||||
action='store_true',
|
||||
help="Whether the archive is from the fandom 2020 dumps by Wikiteam. These dumps can have multiple .xml files in their archives.")
|
||||
|
||||
args = parser.parse_args()
|
||||
|
||||
# set persistence method
|
||||
|
||||
if args.persist is None:
|
||||
persist = PersistMethod.none
|
||||
elif args.persist == "segment":
|
||||
persist = PersistMethod.segment
|
||||
elif args.persist == "legacy":
|
||||
persist = PersistMethod.legacy
|
||||
elif args.persist == "wikidiff2":
|
||||
persist = PersistMethod.wikidiff2
|
||||
else:
|
||||
persist = PersistMethod.sequence
|
||||
|
||||
if args.namespace_filter is not None:
|
||||
namespaces = args.namespace_filter
|
||||
else:
|
||||
namespaces = None
|
||||
|
||||
print(args, file=sys.stderr)
|
||||
if len(args.dumpfiles) > 0:
|
||||
for filename in args.dumpfiles:
|
||||
input_file = open_input_file(filename, args.fandom_2020)
|
||||
|
||||
# open directory for output
|
||||
if args.output:
|
||||
output = args.output[0]
|
||||
else:
|
||||
output = "."
|
||||
|
||||
output_parquet = output.endswith(".parquet")
|
||||
|
||||
print("Processing file: %s" % filename, file=sys.stderr)
|
||||
|
||||
if args.stdout:
|
||||
# Parquet libraries need a binary output, so just sys.stdout doesn't work.
|
||||
output_file = sys.stdout.buffer
|
||||
elif os.path.isdir(output) or output_parquet:
|
||||
filename = os.path.join(output, os.path.basename(filename))
|
||||
output_file = get_output_filename(filename, parquet=output_parquet)
|
||||
else:
|
||||
output_file = output
|
||||
|
||||
wikiq = WikiqParser(input_file,
|
||||
output_file,
|
||||
collapse_user=args.collapse_user,
|
||||
persist=persist,
|
||||
namespaces=namespaces,
|
||||
revert_radius=args.revert_radius,
|
||||
regex_match_revision=args.regex_match_revision,
|
||||
regex_revision_label=args.regex_revision_label,
|
||||
regex_match_comment=args.regex_match_comment,
|
||||
regex_comment_label=args.regex_comment_label,
|
||||
text=args.text,
|
||||
diff=args.diff,
|
||||
output_parquet=output_parquet,
|
||||
)
|
||||
|
||||
wikiq.process()
|
||||
|
||||
# close things
|
||||
input_file.close()
|
||||
|
||||
else:
|
||||
wikiq = WikiqParser(sys.stdin,
|
||||
sys.stdout,
|
||||
collapse_user=args.collapse_user,
|
||||
persist=persist,
|
||||
# persist_legacy=args.persist_legacy,
|
||||
namespaces=namespaces,
|
||||
revert_radius=args.revert_radius,
|
||||
regex_match_revision=args.regex_match_revision,
|
||||
regex_revision_label=args.regex_revision_label,
|
||||
regex_match_comment=args.regex_match_comment,
|
||||
regex_comment_label=args.regex_comment_label,
|
||||
diff=args.diff,
|
||||
text=args.text)
|
||||
|
||||
wikiq.process()
|
||||
|
||||
# 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"
|
||||
# stop_words = stop_words.split(",")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
33
src/wikiq/diff_pyarrow_schema.py
Normal file
33
src/wikiq/diff_pyarrow_schema.py
Normal file
@@ -0,0 +1,33 @@
|
||||
import pyarrow as pa
|
||||
|
||||
# Schema for the `highlightRanges` object, an array of which can be nested in a diff object.
|
||||
highlight_range_struct = pa.struct([
|
||||
pa.field('start', pa.int64(), nullable=False, metadata={'description': 'Where the highlighted text should start, in bytes.'}),
|
||||
pa.field('length', pa.int64(), nullable=False, metadata={'description': 'The length of the highlighted section, in bytes.'}),
|
||||
pa.field('type', pa.int64(), nullable=False, metadata={'description': 'The type of highlight (0: addition, 1: deletion).'})
|
||||
])
|
||||
|
||||
# Schema for the `moveInfo` object, which can be nested in a diff object.
|
||||
move_info_struct = pa.struct([
|
||||
pa.field('id', pa.string(), nullable=False, metadata={'description': 'The ID of the paragraph.'}),
|
||||
pa.field('linkId', pa.string(), nullable=False, metadata={'description': 'The ID of the corresponding paragraph.'}),
|
||||
pa.field('linkDirection', pa.int64(), nullable=False, metadata={'description': 'Visual indicator of the relationship (0: lower, 1: higher).'})
|
||||
])
|
||||
|
||||
# Schema for the `offset` object, which is required in a diff object.
|
||||
offset_struct = pa.struct([
|
||||
pa.field('from', pa.int64(), nullable=True, metadata={'description': 'The first byte of the line in the `from` revision.'}),
|
||||
pa.field('to', pa.int64(), nullable=True, metadata={'description': 'The first byte of the line in the `to` revision.'})
|
||||
])
|
||||
|
||||
# The final schema for the entire structure.
|
||||
diff_field = pa.field('diff', pa.list_(
|
||||
pa.struct([
|
||||
pa.field('type', pa.int64(), nullable=False, metadata={'description': 'The type of change (0: context, 1: addition, 2: deletion, etc.).'}),
|
||||
pa.field('lineNumber', pa.int64(), nullable=True, metadata={'description': 'The line number of the change based on the `to` revision.'}),
|
||||
pa.field('text', pa.string(), nullable=False, metadata={'description': 'The text of the line.'}),
|
||||
pa.field('highlightRanges', pa.list_(highlight_range_struct), nullable=True, metadata={'description': 'Highlights to visually represent changes.'}),
|
||||
pa.field('moveInfo', move_info_struct, nullable=True, metadata={'description': 'Visual indicators for paragraph location changes.'}),
|
||||
pa.field('offset', offset_struct, nullable=False, metadata={'description': 'The location of the line in bytes from the beginning of the page.'})
|
||||
])
|
||||
))
|
||||
228
src/wikiq/tables.py
Normal file
228
src/wikiq/tables.py
Normal file
@@ -0,0 +1,228 @@
|
||||
import sys
|
||||
from abc import abstractmethod, ABC
|
||||
from datetime import datetime, timezone
|
||||
from hashlib import sha1
|
||||
from typing import Generic, TypeVar, Union
|
||||
|
||||
import mwreverts
|
||||
import mwtypes
|
||||
import mwxml
|
||||
|
||||
import pyarrow as pa
|
||||
|
||||
T = TypeVar('T')
|
||||
|
||||
|
||||
class RevisionField(ABC, Generic[T]):
|
||||
def __init__(self):
|
||||
self.data: list[T] = []
|
||||
|
||||
"""
|
||||
Abstract type which represents a field in a table of page revisions.
|
||||
"""
|
||||
|
||||
@property
|
||||
@abstractmethod
|
||||
def field(self) -> pa.Field:
|
||||
pass
|
||||
|
||||
@abstractmethod
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> T:
|
||||
"""
|
||||
:param page: The page for this set of revisions.
|
||||
:param revisions: The set of revisions to compute the field from.
|
||||
Revisions are passed in chronological order, so use revisions[-1] to
|
||||
access the most recent revision in the set.
|
||||
|
||||
Implementations of extract should handle the case where revisions is
|
||||
either a single revision (collapse-user=FALSE), or a full edit session
|
||||
of contiguous edits by the same user (collapse-user=TRUE).
|
||||
"""
|
||||
pass
|
||||
|
||||
def add(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> None:
|
||||
self.data.append(self.extract(page, revisions))
|
||||
|
||||
def pop(self) -> list[T]:
|
||||
data = self.data
|
||||
self.data = []
|
||||
return data
|
||||
|
||||
|
||||
class RevisionTable:
|
||||
columns: list[RevisionField]
|
||||
|
||||
def __init__(self, columns: list[RevisionField]):
|
||||
self.columns = columns
|
||||
|
||||
def add(self, page: mwtypes.Page, revisions: list[mwxml.Revision]):
|
||||
for column in self.columns:
|
||||
column.add(page=page, revisions=revisions)
|
||||
|
||||
def schema(self) -> pa.Schema:
|
||||
return pa.schema([c.field for c in self.columns])
|
||||
|
||||
def pop(self) -> dict:
|
||||
data = {}
|
||||
for column in self.columns:
|
||||
data[column.field.name] = column.pop()
|
||||
|
||||
return data
|
||||
|
||||
|
||||
class RevisionId(RevisionField[int]):
|
||||
field = pa.field("revid", pa.int64())
|
||||
|
||||
def extract(self, _: mwtypes.Page, revisions: list[mwxml.Revision]) -> int:
|
||||
revision = revisions[-1]
|
||||
return revision.id
|
||||
|
||||
|
||||
class RevisionTimestamp(RevisionField[datetime]):
|
||||
field = pa.field("date_time", pa.timestamp('s'))
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> datetime:
|
||||
revision = revisions[-1]
|
||||
return revision.timestamp
|
||||
|
||||
|
||||
class RevisionArticleId(RevisionField[int]):
|
||||
field = pa.field("articleid", pa.int64())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> int:
|
||||
return page.id
|
||||
|
||||
|
||||
class RevisionEditorId(RevisionField[Union[int, None]]):
|
||||
field = pa.field("editorid", pa.int64(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[int, None]:
|
||||
revision = revisions[-1]
|
||||
if revision.deleted.user:
|
||||
return None
|
||||
|
||||
return revision.user.id
|
||||
|
||||
|
||||
class RevisionEditSummary(RevisionField[Union[str, None]]):
|
||||
field = pa.field("edit_summary", pa.string(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[str, None]:
|
||||
revision = revisions[-1]
|
||||
|
||||
return revision.comment
|
||||
|
||||
class RevisionIsAnon(RevisionField[Union[bool, None]]):
|
||||
field = pa.field("anon", pa.bool_(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[bool, None]:
|
||||
revision = revisions[-1]
|
||||
if revision.deleted.user:
|
||||
return None
|
||||
|
||||
return revision.user.id is None
|
||||
|
||||
|
||||
class RevisionEditorText(RevisionField[Union[str, None]]):
|
||||
field = pa.field("editor", pa.string(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[str, None]:
|
||||
revision = revisions[-1]
|
||||
if revision.deleted.user:
|
||||
return None
|
||||
|
||||
return revision.user.text
|
||||
|
||||
|
||||
class RevisionPageTitle(RevisionField[str]):
|
||||
field = pa.field("title", pa.string())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> str:
|
||||
return page.title
|
||||
|
||||
|
||||
class RevisionDeleted(RevisionField[bool]):
|
||||
field = pa.field("deleted", pa.bool_())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> bool:
|
||||
revision = revisions[-1]
|
||||
return revision.deleted.text
|
||||
|
||||
|
||||
class RevisionNamespace(RevisionField[int]):
|
||||
field = pa.field("namespace", pa.int32())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> int:
|
||||
return page.namespace
|
||||
|
||||
|
||||
class RevisionSha1(RevisionField[str]):
|
||||
field = pa.field("sha1", pa.string())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> str:
|
||||
revision = revisions[-1]
|
||||
return revision.sha1
|
||||
|
||||
|
||||
class RevisionTextChars(RevisionField[Union[int, None]]):
|
||||
field = pa.field("text_chars", pa.int32(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[int, None]:
|
||||
revision = revisions[-1]
|
||||
if not revision.deleted.text:
|
||||
return len(revision.text)
|
||||
|
||||
return None
|
||||
|
||||
|
||||
class RevisionText(RevisionField[str]):
|
||||
field = pa.field("text", pa.string())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> str:
|
||||
revision = revisions[-1]
|
||||
return revision.text
|
||||
|
||||
|
||||
class RevisionIsMinor(RevisionField[bool]):
|
||||
field = pa.field("minor", pa.bool_())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> bool:
|
||||
revision = revisions[-1]
|
||||
return revision.minor
|
||||
|
||||
|
||||
class RevisionReverts(RevisionField[Union[str, None]]):
|
||||
def __init__(self):
|
||||
super().__init__()
|
||||
self.rev_detector: Union[mwreverts.Detector, None] = None
|
||||
|
||||
field = pa.field("reverteds", pa.string(), nullable=True)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> Union[str, None]:
|
||||
if self.rev_detector is None:
|
||||
return None
|
||||
|
||||
revision = revisions[-1]
|
||||
if revision.deleted.text:
|
||||
return None
|
||||
|
||||
revert = self.rev_detector.process(revision.sha1, revision.id)
|
||||
if revert is None:
|
||||
return None
|
||||
|
||||
return ",".join([str(s) for s in revert.reverteds])
|
||||
|
||||
|
||||
class RevisionCollapsed(RevisionField[int]):
|
||||
field = pa.field("collapsed_revs", pa.int64())
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> int:
|
||||
return len(revisions)
|
||||
|
||||
|
||||
class RevisionText(RevisionField[str]):
|
||||
field = pa.field("text", pa.string(), nullable=False)
|
||||
|
||||
def extract(self, page: mwtypes.Page, revisions: list[mwxml.Revision]) -> str:
|
||||
revision = revisions[-1]
|
||||
return revision.text
|
||||
451
src/wikiq/wiki_diff_matcher.py
Normal file
451
src/wikiq/wiki_diff_matcher.py
Normal file
@@ -0,0 +1,451 @@
|
||||
import json
|
||||
import sys
|
||||
from collections import namedtuple
|
||||
from itertools import chain
|
||||
from typing import Dict, Generator, List, Optional, Tuple
|
||||
|
||||
from deltas import (Delete, DiffEngine, Equal, Insert, Operation,
|
||||
RegexTokenizer, Token, tokenizers)
|
||||
from mwpersistence import Token
|
||||
from sortedcontainers import SortedDict
|
||||
|
||||
TOKENIZER = tokenizers.wikitext_split
|
||||
import pywikidiff2
|
||||
|
||||
|
||||
class DiffToOperationMap:
|
||||
def __init__(self, diff, tokenizer):
|
||||
self.tokenizer = tokenizer
|
||||
self.from_par_move_dict = {}
|
||||
self.to_par_move_dict = {}
|
||||
self.highlights_without_offset = []
|
||||
self.diff = diff
|
||||
# we need to keep track of the bytes of line numbers to recover when wikidiff2 loses offsets.
|
||||
self.to_linenumber_bytes_map: SortedDict[int, int] = SortedDict()
|
||||
self.from_linenumber_bytes_map: SortedDict[int, int] = SortedDict()
|
||||
|
||||
def tokenize(self, bytes):
|
||||
return self.tokenizer.tokenize(bytes.decode("utf-8"), token_class=Token)
|
||||
|
||||
def to_operations(self):
|
||||
for entry in self.diff["diff"]:
|
||||
# add back the newline
|
||||
entry["text"] += "\n"
|
||||
text = entry["text"]
|
||||
offset = entry["offset"]
|
||||
# this is the first byte of the line in the 'from' revision.
|
||||
from_start_line = entry["offset"]["from"]
|
||||
# this is the first byte of the line in the 'to' revision.
|
||||
to_start_line = entry["offset"]["to"]
|
||||
|
||||
if entry["type"] == 0:
|
||||
yield from self.doEqual(entry)
|
||||
|
||||
# a line included in the 'to' revision, but not in the 'from' revision
|
||||
elif entry["type"] == 1:
|
||||
yield from self.doInsert(entry)
|
||||
|
||||
# a line included in the 'from' revision, but not in the 'to' revision
|
||||
elif entry["type"] == 2:
|
||||
yield from self.doDelete(entry)
|
||||
|
||||
elif entry["type"] == 3:
|
||||
# sometimes, for some reason we don't have a 'to' index here. we'll save these for later
|
||||
if entry["offset"]["to"] is None:
|
||||
self.highlights_without_offset.append(entry)
|
||||
else:
|
||||
yield from self.doHighlightRange(entry)
|
||||
|
||||
elif entry["type"] == 4:
|
||||
linkId = entry["moveInfo"]["linkId"]
|
||||
|
||||
if linkId in self.to_par_move_dict:
|
||||
yield from self.doParMove(entry, self.to_par_move_dict.pop(linkId))
|
||||
else:
|
||||
self.from_par_move_dict[entry["moveInfo"]["id"]] = entry
|
||||
|
||||
elif entry["type"] == 5:
|
||||
linkId = entry["moveInfo"]["linkId"]
|
||||
if linkId in self.from_par_move_dict:
|
||||
yield from self.doParMove(
|
||||
self.from_par_move_dict.pop(linkId), entry
|
||||
)
|
||||
else:
|
||||
self.to_par_move_dict[entry["moveInfo"]["id"]] = entry
|
||||
else:
|
||||
# The 'type' isn't one of the known
|
||||
raise ValueError(d)
|
||||
|
||||
# now we should be able to apply highlights
|
||||
|
||||
for entry in self.highlights_without_offset:
|
||||
yield from self.doHighlightRange(entry)
|
||||
|
||||
if len(self.from_par_move_dict) > 0 or len(self.to_par_move_dict) > 0:
|
||||
print("PROBLEM! Unmatched parmoves!")
|
||||
print(self.from_par_move_dict)
|
||||
print(self.to_par_move_dict)
|
||||
# We can try to match them:
|
||||
for lkey in self.from_par_move_dict.keys():
|
||||
for rkey in self.to_par_move_dict.keys():
|
||||
from_diff = self.from_par_move_dict[lkey]
|
||||
to_diff = self.to_par_move_dict[rkey]
|
||||
if self.match_parmoves_exact(from_diff, to_diff):
|
||||
yield from self.doParMove(from_diff, to_diff)
|
||||
del self.to_par_move_dict[lkey]
|
||||
del self.from_par_move_dict[rkey]
|
||||
break
|
||||
|
||||
# if len(self.from_par_move_dict) > 0 or len(self.to_par_move_dict) > 0:
|
||||
# print("Couldn't find exact matches for all parmoves!")
|
||||
# # we couldn't find all the matches via exact match
|
||||
# # let's try matching based on line number instead
|
||||
# lkeys_to_remove = []
|
||||
# for lkey, from_diff in self.from_par_move_dict.items():
|
||||
# from_linenum = from_diff["moveInfo"]["linkId"].split("_")[2]
|
||||
# rkey_to_remove = None
|
||||
# for rkey, to_diff in self.to_par_move_dict.items():
|
||||
# to_linenum = rkey.split("_")[2]
|
||||
# if from_linenum == to_linenum:
|
||||
# print("Matching on line number")
|
||||
# yield from self.doParMove(from_diff, to_diff)
|
||||
# rkey_to_remove = rkey
|
||||
# lkeys_to_remove.append(lkey)
|
||||
# break
|
||||
# if rkey_to_remove is not None:
|
||||
# del self.to_par_move_dict[rkey_to_remove]
|
||||
# for lkey in lkeys_to_remove:
|
||||
# del self.from_par_move_dict[lkey]
|
||||
|
||||
# if len(self.from_par_move_dict) > 0 or len(self.to_par_move_dict) > 0:
|
||||
# print("Couldn't find exact matches for all parmoves!")
|
||||
# # we couldn't find all the matches via exact match or line number
|
||||
# # let's try matching based on opIndex instead
|
||||
# lkeys_to_remove = []
|
||||
# for lkey, from_diff in self.from_par_move_dict.items():
|
||||
# rkey_to_remove = None
|
||||
# from_idx = from_diff["moveInfo"]["linkId"].split("_")[1]
|
||||
# for rkey, to_diff in self.to_par_move_dict.items():
|
||||
# to_idx = rkey.split("_")[1]
|
||||
# print(from_idx)
|
||||
# print(to_idx)
|
||||
# if from_idx == to_idx:
|
||||
# yield from self.doParMove(from_diff, to_diff)
|
||||
# rkey_to_remove = rkey
|
||||
# lkeys_to_remove.append(lkey)
|
||||
# if rkey_to_remove is not None:
|
||||
# del self.to_par_move_dict[rkey_to_remove]
|
||||
# for lkey in lkeys_to_remove:
|
||||
# del self.from_par_move_dict[lkey]
|
||||
|
||||
# we couldn't find matches. treat type 4 as removal and type 5 as highlight.
|
||||
for from_diff in self.from_par_move_dict.values():
|
||||
yield from self.doDelete(from_diff)
|
||||
|
||||
# only we don't know the from index; we assume its already handled.
|
||||
for to_diff in self.to_par_move_dict.values():
|
||||
offset["from"] = 0
|
||||
offset["to"] = None
|
||||
diffops = self.doHighlightRange(
|
||||
{
|
||||
"text": to_diff["text"],
|
||||
"highlightRanges": to_diff["highlightRanges"],
|
||||
"offset": offset,
|
||||
"lineNumber": to_diff["lineNumber"],
|
||||
}
|
||||
)
|
||||
diffops = [
|
||||
(type(op)(None, None, op.b1, op.b2), [], bseq)
|
||||
for op, _, bseq in diffops
|
||||
if isinstance(op, Insert) or isinstance(op, Equal)
|
||||
]
|
||||
yield from diffops
|
||||
|
||||
def match_parmoves_exact(self, from_diff, to_diff):
|
||||
ops, from_tokens, to_tokens = list(zip(*self.doParMove(from_diff, to_diff)))
|
||||
from_text = "".join(chain.from_iterable(from_tokens))
|
||||
# we know they match if we apply the highlight ranges and the "from" tokens equal the lhs tokens.
|
||||
if from_text == from_diff["text"]:
|
||||
print("MATCH FOUND")
|
||||
return True
|
||||
else:
|
||||
print("NO MATCH")
|
||||
print(len(from_text))
|
||||
print(len(from_diff["text"]))
|
||||
return False
|
||||
|
||||
# mwpersistence expects differences to be represented in order from the
|
||||
# result's perspective ("to"), not the previous text. Thus, if a line
|
||||
# is moved earlier then its insertion should appear before its deletion.
|
||||
# As a rule of thumb, the "to" segments should be non-overlapping and
|
||||
# strictly increasing, while the "from" segments should merely be
|
||||
# non-overlapping.
|
||||
|
||||
def doEqual(self, entry):
|
||||
equal_segment, offset, lineNumber = (
|
||||
entry["text"],
|
||||
entry["offset"],
|
||||
entry["lineNumber"],
|
||||
)
|
||||
if isinstance(equal_segment, str):
|
||||
equal_bytes = equal_segment.encode()
|
||||
elif isinstance(equal_segment, bytes):
|
||||
equal_bytes = equal_segment
|
||||
else:
|
||||
raise ValueError(equal_segment)
|
||||
|
||||
self.from_linenumber_bytes_map[lineNumber] = offset["from"] + len(equal_bytes)
|
||||
self.to_linenumber_bytes_map[lineNumber] = offset["to"] + len(equal_bytes)
|
||||
|
||||
tokens = self.tokenize(equal_bytes)
|
||||
n_tokens = len(tokens)
|
||||
yield (
|
||||
Equal(
|
||||
offset["from"],
|
||||
None,
|
||||
offset["to"],
|
||||
None,
|
||||
),
|
||||
tokens,
|
||||
tokens,
|
||||
)
|
||||
|
||||
def doInsert(self, entry):
|
||||
insert_segment, offset, lineNumber = (
|
||||
entry["text"],
|
||||
entry["offset"],
|
||||
entry["lineNumber"],
|
||||
)
|
||||
if isinstance(insert_segment, str):
|
||||
insert_bytes = insert_segment.encode()
|
||||
elif isinstance(insert_segment, bytes):
|
||||
insert_bytes = insert_segment
|
||||
else:
|
||||
raise ValueError(insert_segment)
|
||||
tokens = self.tokenize(insert_bytes)
|
||||
self.to_linenumber_bytes_map[lineNumber] = offset["to"] + len(insert_bytes)
|
||||
yield (
|
||||
Insert(
|
||||
None,
|
||||
None,
|
||||
offset["to"],
|
||||
None,
|
||||
),
|
||||
[],
|
||||
tokens,
|
||||
)
|
||||
|
||||
def doDelete(self, entry):
|
||||
delete_segment, offset, lineNumber = (
|
||||
entry["text"],
|
||||
entry["offset"],
|
||||
entry.get("lineNumber", None),
|
||||
)
|
||||
if isinstance(delete_segment, str):
|
||||
delete_bytes = delete_segment.encode()
|
||||
elif isinstance(delete_segment, bytes):
|
||||
delete_bytes = delete_segment
|
||||
else:
|
||||
raise ValueError(delete_segment)
|
||||
tokens = self.tokenize(delete_bytes)
|
||||
if lineNumber is not None:
|
||||
self.from_linenumber_bytes_map[lineNumber] = offset["from"] + len(
|
||||
delete_bytes
|
||||
)
|
||||
|
||||
yield (
|
||||
Delete(offset["from"], None, None, None),
|
||||
tokens,
|
||||
[],
|
||||
)
|
||||
|
||||
def doHighlightRange(self, entry):
|
||||
highlight_text, highlightRanges, offset, lineNumber = (
|
||||
entry["text"],
|
||||
entry["highlightRanges"],
|
||||
entry["offset"],
|
||||
entry["lineNumber"],
|
||||
)
|
||||
|
||||
# The text field is an overlapping mix of both the from and to,
|
||||
# so we need to handle it highlight-by-highlight.
|
||||
# there can be gaps between highlight segments.
|
||||
# for instance, if a word is deleted from the middle of a line.
|
||||
# we need to track that.
|
||||
highlight_bytes = highlight_text.encode()
|
||||
highlight_end = 0
|
||||
|
||||
# it's possible for offset['to'] to be null.
|
||||
# we can get it from the line number?
|
||||
# this bit is a little hacky as it deals with ideosyncratic wikidiff2 behavior
|
||||
if offset["to"] is None:
|
||||
# if the line already exists, we insert before it.
|
||||
if lineNumber in self.to_linenumber_bytes_map:
|
||||
keyidx = self.to_linenumber_bytes_map.bisect_left(lineNumber) - 1
|
||||
else:
|
||||
keyidx = self.to_linenumber_bytes_map.bisect_right(lineNumber) - 1
|
||||
key = None
|
||||
if keyidx == -1:
|
||||
offset["to"] = 0
|
||||
elif len(self.to_linenumber_bytes_map.keys()) > 0:
|
||||
key = self.to_linenumber_bytes_map.keys()[keyidx]
|
||||
else:
|
||||
key = 0
|
||||
if key is not None:
|
||||
offset["to"] = self.to_linenumber_bytes_map.get(key, 0)
|
||||
|
||||
highlight_offset = offset
|
||||
# note that diffs are token-level, but the indexes are byte-level
|
||||
|
||||
for highlightRange in highlightRanges:
|
||||
highlight_start = highlightRange["start"]
|
||||
# equal bytes in between highlights
|
||||
if highlight_start > highlight_end:
|
||||
equal_bytes = highlight_bytes[highlight_end:highlight_start]
|
||||
n_equal_bytes = len(equal_bytes)
|
||||
|
||||
yield from self.doEqual(
|
||||
{
|
||||
"text": equal_bytes,
|
||||
"offset": highlight_offset,
|
||||
"lineNumber": lineNumber,
|
||||
}
|
||||
)
|
||||
highlight_offset["from"] += n_equal_bytes
|
||||
highlight_offset["to"] += n_equal_bytes
|
||||
|
||||
# handle highlighted insert / delete
|
||||
highlight_end = highlight_start + highlightRange["length"]
|
||||
range_bytes = highlight_bytes[highlight_start:highlight_end]
|
||||
n_range_bytes = len(range_bytes)
|
||||
|
||||
if highlightRange["type"] == 0:
|
||||
yield from self.doInsert(
|
||||
{
|
||||
"text": range_bytes,
|
||||
"offset": highlight_offset,
|
||||
"lineNumber": lineNumber,
|
||||
}
|
||||
)
|
||||
highlight_offset["to"] += n_range_bytes
|
||||
elif highlightRange["type"] == 1:
|
||||
yield from self.doDelete(
|
||||
{
|
||||
"text": range_bytes,
|
||||
"offset": highlight_offset,
|
||||
"lineNumber": lineNumber,
|
||||
}
|
||||
)
|
||||
highlight_offset["from"] += n_range_bytes
|
||||
else:
|
||||
raise Exception(entry)
|
||||
|
||||
# handle the rest of the line which is equal
|
||||
if highlight_end < len(highlight_bytes):
|
||||
range_bytes = highlight_bytes[highlight_end:]
|
||||
yield from self.doEqual(
|
||||
{
|
||||
"text": range_bytes,
|
||||
"offset": highlight_offset,
|
||||
"lineNumber": lineNumber,
|
||||
}
|
||||
)
|
||||
|
||||
def doParMove(self, from_diff, to_diff):
|
||||
from_byte_start = from_diff["offset"]["from"]
|
||||
to_byte_start = to_diff["offset"]["to"]
|
||||
offset = {"from": from_byte_start, "to": to_byte_start}
|
||||
yield from self.doHighlightRange(
|
||||
{
|
||||
"text": to_diff["text"],
|
||||
"highlightRanges": to_diff["highlightRanges"],
|
||||
"offset": offset,
|
||||
"lineNumber": to_diff["lineNumber"],
|
||||
}
|
||||
)
|
||||
|
||||
|
||||
class WikiDiffMatcher:
|
||||
def __init__(
|
||||
self,
|
||||
texts: list[str] = None,
|
||||
tokenizer: Optional[RegexTokenizer] = None,
|
||||
):
|
||||
differ = pywikidiff2.pywikidiff2(
|
||||
numContextLines=1000000, moved_paragraph_detection_cutoff=200000
|
||||
)
|
||||
# Pre-compute diffs to reduce traffic overhead.
|
||||
self.diffs = [json.loads(diff) for diff in differ.inline_json_diff_sequence(list(texts))]
|
||||
self.tokenizer = tokenizer or TOKENIZER
|
||||
|
||||
class Processor(DiffEngine.Processor):
|
||||
def __init__(self, texts, tokenizer=None):
|
||||
self.diffs = iter(texts)
|
||||
self.tokenizer = tokenizer or TOKENIZER
|
||||
self.last_tokens = []
|
||||
self.previous_text = ""
|
||||
|
||||
def update(self, last_tokens):
|
||||
self.last_tokens = last_tokens
|
||||
|
||||
def process(self, text, token_class=None):
|
||||
# The diff has already been computed, but we need to incrementally
|
||||
# retrieve it to recreate the behavior DiffState expects.
|
||||
diff = next(self.diffs)
|
||||
diffToOperationsMapper = DiffToOperationMap(diff, self.tokenizer)
|
||||
|
||||
diffops = list(diffToOperationsMapper.to_operations())
|
||||
|
||||
# this happens when revisions are actually equal.
|
||||
if len(diffops) == 0:
|
||||
self.last_tokens = self.tokenizer.tokenize(text, token_class=Token)
|
||||
ops = [Equal(0, len(self.last_tokens), 0, len(self.last_tokens))]
|
||||
return ops, self.last_tokens, self.last_tokens
|
||||
|
||||
# we get back the byte indices; now we transform to token indices
|
||||
|
||||
diffops.sort(
|
||||
key=lambda t: (t[0].a1 if t[0].a1 is not None else 1e32, t[0].b1)
|
||||
)
|
||||
aorder_ops = []
|
||||
token_offset = 0
|
||||
_, aseq, _ = list(zip(*diffops))
|
||||
|
||||
for op, tokens, _ in diffops:
|
||||
a1 = token_offset
|
||||
if isinstance(op, Equal) or isinstance(op, Delete):
|
||||
token_offset += len(tokens)
|
||||
a2 = token_offset
|
||||
aorder_ops.append(type(op)(a1, a2, op.b1, op.b1))
|
||||
else:
|
||||
aorder_ops.append(Insert(a1, a1, op.b1, op.b1))
|
||||
|
||||
_, aseq, bseq = zip(*diffops)
|
||||
diffops = list(zip(aorder_ops, aseq, bseq))
|
||||
diffops.sort(
|
||||
key=lambda t: (t[0].b1 if t[0].b1 is not None else 1e32, t[0].a1)
|
||||
)
|
||||
_, _, bseq = list(zip(*diffops))
|
||||
border_ops = []
|
||||
token_offset = 0
|
||||
for op, _, tokens in diffops:
|
||||
b1 = token_offset
|
||||
if isinstance(op, Equal) or isinstance(op, Insert):
|
||||
token_offset += len(tokens)
|
||||
b2 = token_offset
|
||||
border_ops.append(type(op)(op.a1, op.a2, b1, b2))
|
||||
else:
|
||||
border_ops.append(type(op)(op.a1, op.a2, b1, b1))
|
||||
|
||||
self.previous_text = text
|
||||
|
||||
self.last_tokens = list(chain.from_iterable(aseq))
|
||||
tokens = list(chain.from_iterable(bseq))
|
||||
return border_ops, self.last_tokens, tokens
|
||||
|
||||
def processor(self, *args, **kwargs):
|
||||
return self.Processor(self.diffs, self.tokenizer)
|
||||
|
||||
def process(self):
|
||||
# DiffState checks for this method even though it is not called.
|
||||
raise Exception("Unnecessary implementation")
|
||||
Reference in New Issue
Block a user