collective/mediawiki_dump_tools
16
0
Fork 0
Code Issues Pull Requests Activity

Remove unused dependencies and fix spacing

Browse Source 
The "mw" and "numpy" dependencies were unneeded.

Spaces and tabs were inconsistently used.
They are now used consistently, changes via auto-formatter.

Signed-off-by: Will Beason <willbeason@gmail.com>
This commit is contained in:
Will Beason 2025-05-26 14:15:01 -05:00
parent 4804ecc4b3
commit 9c5bf577e6
2 changed files with 134 additions and 113 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
requirements.txt
View File

@ -1,4 +1,3 @@
apeek==0.1.1
attrs==25.3.0 attrs==25.3.0
certifi==2025.4.26 certifi==2025.4.26
charset-normalizer==3.4.2 charset-normalizer==3.4.2
@ -11,14 +10,12 @@ jsonable==0.3.1
jsonschema==4.23.0 jsonschema==4.23.0
jsonschema-specifications==2025.4.1 jsonschema-specifications==2025.4.1
mediawiki-utilities==0.4.18 mediawiki-utilities==0.4.18
mw==0.4.0
mwcli==0.0.3 mwcli==0.0.3
mwdiffs==0.0.2 mwdiffs==0.0.2
mwpersistence==0.2.4 mwpersistence==0.2.4
mwreverts==0.1.5 mwreverts==0.1.5
mwtypes==0.4.0 mwtypes==0.4.0
mwxml==0.3.6 mwxml==0.3.6
numpy==1.26.4
pandas==2.2.3 pandas==2.2.3
para==0.0.8 para==0.0.8
parsimonious==0.10.0 parsimonious==0.10.0

244
wikiq
View File

@ -6,9 +6,9 @@
import argparse import argparse
import sys import sys
import os, os.path import os.path
import re import re
from datetime import datetime,timezone from datetime import datetime, timezone
from subprocess import Popen, PIPE from subprocess import Popen, PIPE
from collections import deque from collections import deque
@ -20,8 +20,9 @@ from deltas.tokenizers import wikitext_split
import mwpersistence import mwpersistence
import mwreverts import mwreverts
from urllib.parse import quote from urllib.parse import quote
TO_ENCODE = ('title', 'editor') TO_ENCODE = ('title', 'editor')
PERSISTENCE_RADIUS=7 PERSISTENCE_RADIUS = 7
from deltas import SequenceMatcher from deltas import SequenceMatcher
from deltas import SegmentMatcher from deltas import SegmentMatcher
@ -30,42 +31,46 @@ from dataclasses import dataclass
import pyarrow as pa import pyarrow as pa
import pyarrow.parquet as pq import pyarrow.parquet as pq
class PersistMethod: class PersistMethod:
none = 0 none = 0
sequence = 1 sequence = 1
segment = 2 segment = 2
legacy = 3 legacy = 3
def calculate_persistence(tokens_added): def calculate_persistence(tokens_added):
return(sum([(len(x.revisions)-1) for x in tokens_added]), return (sum([(len(x.revisions) - 1) for x in tokens_added]),
len(tokens_added)) len(tokens_added))
class WikiqIterator(): class WikiqIterator():
def __init__(self, fh, collapse_user=False): def __init__(self, fh, collapse_user=False):
self.fh = fh self.fh = fh
self.collapse_user = collapse_user self.collapse_user = collapse_user
self.mwiterator = Dump.from_file(self.fh) self.mwiterator = Dump.from_file(self.fh)
self.namespace_map = { ns.id : ns.name for ns in self.namespace_map = {ns.id: ns.name for ns in
self.mwiterator.site_info.namespaces } self.mwiterator.site_info.namespaces}
self.__pages = self.load_pages() self.__pages = self.load_pages()
def load_pages(self): def load_pages(self):
for page in self.mwiterator: for page in self.mwiterator:
yield WikiqPage(page, yield WikiqPage(page,
namespace_map = self.namespace_map, namespace_map=self.namespace_map,
collapse_user=self.collapse_user) collapse_user=self.collapse_user)
def __iter__(self): def __iter__(self):
return self.__pages return self.__pages
def __next__(self): def __next__(self):
return next(self._pages) return next(self.__pages)
class WikiqPage(): class WikiqPage():
__slots__ = ('id', 'title', 'namespace', 'redirect', __slots__ = ('id', 'title', 'namespace', 'redirect',
'restrictions', 'mwpage', '__revisions', 'restrictions', 'mwpage', '__revisions',
'collapse_user') 'collapse_user')
def __init__(self, page, namespace_map, collapse_user=False): def __init__(self, page, namespace_map, collapse_user=False):
self.id = page.id self.id = page.id
self.namespace = page.namespace self.namespace = page.namespace
@ -95,7 +100,7 @@ class WikiqPage():
for i, rev in enumerate(self.mwpage): for i, rev in enumerate(self.mwpage):
# never yield the first time # never yield the first time
if i == 0: if i == 0:
if self.collapse_user: if self.collapse_user:
collapsed_revs = 1 collapsed_revs = 1
rev.collapsed_revs = collapsed_revs rev.collapsed_revs = collapsed_revs
@ -138,6 +143,8 @@ 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. 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. If the pattern does not include a capture group, then only one output column will result.
""" """
class RegexPair(object): class RegexPair(object):
def __init__(self, pattern, label): def __init__(self, pattern, label):
self.pattern = re.compile(pattern) self.pattern = re.compile(pattern)
@ -145,10 +152,10 @@ class RegexPair(object):
self.has_groups = bool(self.pattern.groupindex) self.has_groups = bool(self.pattern.groupindex)
if self.has_groups: if self.has_groups:
self.capture_groups = list(self.pattern.groupindex.keys()) self.capture_groups = list(self.pattern.groupindex.keys())
def get_pyarrow_fields(self): def get_pyarrow_fields(self):
if self.has_groups: if self.has_groups:
fields = [pa.field(self._make_key(cap_group),pa.list_(pa.string())) fields = [pa.field(self._make_key(cap_group), pa.list_(pa.string()))
for cap_group in self.capture_groups] for cap_group in self.capture_groups]
else: else:
fields = [pa.field(self.label, pa.list_(pa.string()))] fields = [pa.field(self.label, pa.list_(pa.string()))]
@ -159,7 +166,7 @@ class RegexPair(object):
return ("{}_{}".format(self.label, cap_group)) return ("{}_{}".format(self.label, cap_group))
def matchmake(self, content, rev_data): def matchmake(self, content, rev_data):
temp_dict = {} temp_dict = {}
# if there are named capture groups in the regex # if there are named capture groups in the regex
if self.has_groups: if self.has_groups:
@ -178,7 +185,7 @@ class RegexPair(object):
temp_list.append(match.group(cap_group)) temp_list.append(match.group(cap_group))
# if temp_list of matches is empty just make that column None # if temp_list of matches is empty just make that column None
if len(temp_list)==0: if len(temp_list) == 0:
temp_dict[key] = None temp_dict[key] = None
# else we put in the list we made in the for-loop above # else we put in the list we made in the for-loop above
else: else:
@ -192,8 +199,8 @@ class RegexPair(object):
# there are no capture groups, we just search for all the matches of the regex # there are no capture groups, we just search for all the matches of the regex
else: else:
#given that there are matches to be made # given that there are matches to be made
if type(content) in(str, bytes): if type(content) in (str, bytes):
if self.pattern.search(content) is not None: if self.pattern.search(content) is not None:
m = self.pattern.findall(content) m = self.pattern.findall(content)
temp_dict[self.label] = ', '.join(m) temp_dict[self.label] = ', '.join(m)
@ -206,6 +213,7 @@ class RegexPair(object):
return rev_data return rev_data
""" """
We used to use a dictionary to collect fields for the output. We used to use a dictionary to collect fields for the output.
@ -222,9 +230,11 @@ It also needs to have the correct pyarrow schema so we can write parquet files.
The RevDataBase type has all the fields that will be output no matter how wikiq is invoked. The RevDataBase type has all the fields that will be output no matter how wikiq is invoked.
""" """
@dataclass() @dataclass()
class RevDataBase(): class RevDataBase():
revid: int revid: int
date_time: datetime date_time: datetime
articleid: int articleid: int
editorid: int editorid: int
@ -249,18 +259,18 @@ class RevDataBase():
pa_schema_fields = [ pa_schema_fields = [
pa.field("revid", pa.int64()), pa.field("revid", pa.int64()),
pa.field("date_time", pa.timestamp('ms')), pa.field("date_time", pa.timestamp('ms')),
pa.field("articleid",pa.int64()), pa.field("articleid", pa.int64()),
pa.field("editorid",pa.int64()), pa.field("editorid", pa.int64()),
pa.field("title",pa.string()), pa.field("title", pa.string()),
pa.field("namespace",pa.int32()), pa.field("namespace", pa.int32()),
pa.field("deleted",pa.bool_()), pa.field("deleted", pa.bool_()),
pa.field("test_chars",pa.int32()), pa.field("test_chars", pa.int32()),
pa.field("revert",pa.bool_()), pa.field("revert", pa.bool_()),
pa.field("reverteds",pa.list_(pa.int64())), pa.field("reverteds", pa.list_(pa.int64())),
pa.field("sha1",pa.string()), pa.field("sha1", pa.string()),
pa.field("minor",pa.bool_()), pa.field("minor", pa.bool_()),
pa.field("editor",pa.string()), pa.field("editor", pa.string()),
pa.field("anon",pa.bool_()) pa.field("anon", pa.bool_())
] ]
# pyarrow is a columnar format, so most of the work happens in the flush_parquet_buffer function # pyarrow is a columnar format, so most of the work happens in the flush_parquet_buffer function
@ -269,7 +279,7 @@ class RevDataBase():
# logic to convert each field into the wikiq tsv format goes here. # logic to convert each field into the wikiq tsv format goes here.
def to_tsv_row(self): def to_tsv_row(self):
row = [] row = []
for f in dc.fields(self): for f in dc.fields(self):
val = getattr(self, f.name) val = getattr(self, f.name)
@ -281,7 +291,7 @@ class RevDataBase():
elif f.type == datetime: elif f.type == datetime:
row.append(val.strftime('%Y-%m-%d %H:%M:%S')) row.append(val.strftime('%Y-%m-%d %H:%M:%S'))
elif f.name in {'editor','title'}: elif f.name in {'editor', 'title'}:
s = '"' + val + '"' s = '"' + val + '"'
if self.urlencode and f.name in TO_ENCODE: if self.urlencode and f.name in TO_ENCODE:
row.append(quote(str(s))) row.append(quote(str(s)))
@ -299,11 +309,12 @@ class RevDataBase():
else: else:
row.append(val) row.append(val)
return '\t'.join(map(str,row)) return '\t'.join(map(str, row))
def header_row(self): def header_row(self):
return '\t'.join(map(lambda f: f.name, dc.fields(self))) return '\t'.join(map(lambda f: f.name, dc.fields(self)))
""" """
If collapse=True we'll use a RevDataCollapse dataclass. If collapse=True we'll use a RevDataCollapse dataclass.
@ -312,43 +323,54 @@ This class inherits from RevDataBase. This means that it has all the same fields
It just adds a new field and updates the pyarrow schema. It just adds a new field and updates the pyarrow schema.
""" """
@dataclass() @dataclass()
class RevDataCollapse(RevDataBase): class RevDataCollapse(RevDataBase):
collapsed_revs:int = None collapsed_revs: int = None
pa_collapsed_revs_schema = pa.field('collapsed_revs',pa.int64()) pa_collapsed_revs_schema = pa.field('collapsed_revs', pa.int64())
pa_schema_fields = RevDataBase.pa_schema_fields + [pa_collapsed_revs_schema] pa_schema_fields = RevDataBase.pa_schema_fields + [pa_collapsed_revs_schema]
""" """
If persistence data is to be computed we'll need the fields added by RevDataPersistence. If persistence data is to be computed we'll need the fields added by RevDataPersistence.
""" """
@dataclass() @dataclass()
class RevDataPersistence(RevDataBase): class RevDataPersistence(RevDataBase):
token_revs:int = None token_revs: int = None
tokens_added:int = None tokens_added: int = None
tokens_removed:int = None tokens_removed: int = None
tokens_window:int = None tokens_window: int = None
pa_persistence_schema_fields = [ pa_persistence_schema_fields = [
pa.field("token_revs", pa.int64()), pa.field("token_revs", pa.int64()),
pa.field("tokens_added", pa.int64()), pa.field("tokens_added", pa.int64()),
pa.field("tokens_removed", pa.int64()), pa.field("tokens_removed", pa.int64()),
pa.field("tokens_window", pa.int64())] pa.field("tokens_window", pa.int64())]
pa_schema_fields = RevDataBase.pa_schema_fields + pa_persistence_schema_fields pa_schema_fields = RevDataBase.pa_schema_fields + pa_persistence_schema_fields
""" """
class RevDataCollapsePersistence uses multiple inheritence to make a class that has both persistence and collapse fields. class RevDataCollapsePersistence uses multiple inheritence to make a class that has both persistence and collapse fields.
""" """
@dataclass() @dataclass()
class RevDataCollapsePersistence(RevDataCollapse, RevDataPersistence): class RevDataCollapsePersistence(RevDataCollapse, RevDataPersistence):
pa_schema_fields = RevDataCollapse.pa_schema_fields + RevDataPersistence.pa_persistence_schema_fields pa_schema_fields = RevDataCollapse.pa_schema_fields + RevDataPersistence.pa_persistence_schema_fields
class WikiqParser():
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): class WikiqParser:
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):
""" """
Parameters: Parameters:
persist : what persistence method to use. Takes a PersistMethod value persist : what persistence method to use. Takes a PersistMethod value
@ -360,7 +382,7 @@ class WikiqParser():
self.namespaces = [] self.namespaces = []
self.urlencode = urlencode self.urlencode = urlencode
self.revert_radius = revert_radius self.revert_radius = revert_radius
if namespaces is not None: if namespaces is not None:
self.namespace_filter = set(namespaces) self.namespace_filter = set(namespaces)
else: else:
@ -370,9 +392,8 @@ class WikiqParser():
self.regex_revision_pairs = self.make_matchmake_pairs(regex_match_revision, regex_revision_label) self.regex_revision_pairs = self.make_matchmake_pairs(regex_match_revision, regex_revision_label)
self.regex_comment_pairs = self.make_matchmake_pairs(regex_match_comment, regex_comment_label) self.regex_comment_pairs = self.make_matchmake_pairs(regex_match_comment, regex_comment_label)
# This is where we set the type for revdata. # This is where we set the type for revdata.
if self.collapse_user is True: if self.collapse_user is True:
if self.persist == PersistMethod.none: if self.persist == PersistMethod.none:
revdata_type = RevDataCollapse revdata_type = RevDataCollapse
@ -391,10 +412,10 @@ class WikiqParser():
self.revdata_type = dc.make_dataclass('RevData_Parser', self.revdata_type = dc.make_dataclass('RevData_Parser',
fields=regex_fields, fields=regex_fields,
bases=(revdata_type,)) bases=(revdata_type,))
# we also need to make sure that we have the right pyarrow schema # we also need to make sure that we have the right pyarrow schema
self.revdata_type.pa_schema_fields = revdata_type.pa_schema_fields + self.regex_schemas self.revdata_type.pa_schema_fields = revdata_type.pa_schema_fields + self.regex_schemas
self.revdata_type.urlencode = self.urlencode self.revdata_type.urlencode = self.urlencode
self.schema = pa.schema(self.revdata_type.pa_schema_fields) self.schema = pa.schema(self.revdata_type.pa_schema_fields)
@ -409,15 +430,15 @@ class WikiqParser():
else: else:
self.print_header = True self.print_header = True
if output_file == sys.stdout: if output_file == sys.stdout:
self.output_file = output_file self.output_file = output_file
else: else:
self.output_file = open(output_file,'w') self.output_file = open(output_file, 'w')
self.output_parquet = False self.output_parquet = False
def make_matchmake_pairs(self, patterns, labels): def make_matchmake_pairs(self, patterns, labels):
if (patterns is not None and labels is not None) and \ if (patterns is not None and labels is not None) and \
(len(patterns) == len(labels)): (len(patterns) == len(labels)):
result = [] result = []
for pattern, label in zip(patterns, labels): for pattern, label in zip(patterns, labels):
rp = RegexPair(pattern, label) rp = RegexPair(pattern, label)
@ -435,7 +456,7 @@ class WikiqParser():
return rev_data return rev_data
def matchmake_text(self, text, rev_data): def matchmake_text(self, text, rev_data):
return self.matchmake_pairs(text, rev_data, self.regex_revision_pairs) return self.matchmake_pairs(text, rev_data, self.regex_revision_pairs)
def matchmake_comment(self, comment, rev_data): def matchmake_comment(self, comment, rev_data):
return self.matchmake_pairs(comment, rev_data, self.regex_comment_pairs) return self.matchmake_pairs(comment, rev_data, self.regex_comment_pairs)
@ -450,7 +471,7 @@ class WikiqParser():
for ns in self.namespaces: for ns in self.namespaces:
# skip if the namespace is not defined # skip if the namespace is not defined
if ns == None: if ns is None:
default_ns = self.namespaces[ns] default_ns = self.namespaces[ns]
continue continue
@ -460,7 +481,6 @@ class WikiqParser():
# if we've made it this far with no matches, we return the default namespace # if we've made it this far with no matches, we return the default namespace
return default_ns return default_ns
def process(self): def process(self):
# create a regex that creates the output filename # create a regex that creates the output filename
@ -472,12 +492,11 @@ class WikiqParser():
dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user) dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user)
# extract list of namspaces # extract list of namspaces
self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces} self.namespaces = {ns.name: ns.id for ns in dump.mwiterator.site_info.namespaces}
page_count = 0 page_count = 0
rev_count = 0 rev_count = 0
# Iterate through pages # Iterate through pages
for page in dump: for page in dump:
namespace = page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title) namespace = page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title)
@ -487,17 +506,17 @@ class WikiqParser():
if namespace not in self.namespace_filter: if namespace not in self.namespace_filter:
continue continue
rev_detector = mwreverts.Detector(radius = self.revert_radius) rev_detector = mwreverts.Detector(radius=self.revert_radius)
if self.persist != PersistMethod.none: if self.persist != PersistMethod.none:
window = deque(maxlen=PERSISTENCE_RADIUS) window = deque(maxlen=PERSISTENCE_RADIUS)
if self.persist == PersistMethod.sequence: if self.persist == PersistMethod.sequence:
state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split), state = mwpersistence.DiffState(SequenceMatcher(tokenizer=wikitext_split),
revert_radius=PERSISTENCE_RADIUS) revert_radius=PERSISTENCE_RADIUS)
elif self.persist == PersistMethod.segment: elif self.persist == PersistMethod.segment:
state = mwpersistence.DiffState(SegmentMatcher(tokenizer = wikitext_split), state = mwpersistence.DiffState(SegmentMatcher(tokenizer=wikitext_split),
revert_radius=PERSISTENCE_RADIUS) revert_radius=PERSISTENCE_RADIUS)
# self.persist == PersistMethod.legacy # self.persist == PersistMethod.legacy
@ -507,15 +526,15 @@ class WikiqParser():
# Iterate through a page's revisions # Iterate through a page's revisions
for rev in page: for rev in page:
# create a new data object instead of a dictionary. # create a new data object instead of a dictionary.
rev_data = self.revdata_type(revid = rev.id, rev_data = self.revdata_type(revid=rev.id,
date_time = datetime.fromtimestamp(rev.timestamp.unix(), tz=timezone.utc), date_time=datetime.fromtimestamp(rev.timestamp.unix(), tz=timezone.utc),
articleid = page.id, articleid=page.id,
editorid = "" if rev.deleted.user == True or rev.user.id is None else rev.user.id, editorid="" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
title = page.title, title=page.title,
deleted = rev.deleted.text, deleted=rev.deleted.text,
namespace = namespace namespace=namespace
) )
rev_data = self.matchmake_revision(rev, rev_data) rev_data = self.matchmake_revision(rev, rev_data)
@ -530,7 +549,7 @@ class WikiqParser():
text_sha1 = rev.sha1 text_sha1 = rev.sha1
else: else:
text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest() text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
rev_data.sha1 = text_sha1 rev_data.sha1 = text_sha1
# TODO rev.bytes doesn't work.. looks like a bug # TODO rev.bytes doesn't work.. looks like a bug
@ -538,7 +557,7 @@ class WikiqParser():
# generate revert data # generate revert data
revert = rev_detector.process(text_sha1, rev.id) revert = rev_detector.process(text_sha1, rev.id)
if revert: if revert:
rev_data.revert = True rev_data.revert = True
rev_data.reverteds = revert.reverteds rev_data.reverteds = revert.reverteds
@ -550,16 +569,16 @@ class WikiqParser():
if not rev.deleted.user: if not rev.deleted.user:
# wrap user-defined editors in quotes for fread # wrap user-defined editors in quotes for fread
rev_data.editor = rev.user.text rev_data.editor = rev.user.text
rev_data.anon = rev.user.id is None rev_data.anon = rev.user.id is None
#if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I): # if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
# redirect = True # redirect = True
#else: # else:
# redirect = False # redirect = False
#TODO missing: additions_size deletions_size # TODO missing: additions_size deletions_size
# if collapse user was on, lets run that # if collapse user was on, lets run that
if self.collapse_user: if self.collapse_user:
rev_data.collapsed_revs = rev.collapsed_revs rev_data.collapsed_revs = rev.collapsed_revs
@ -573,18 +592,18 @@ class WikiqParser():
else: else:
_, tokens_added, tokens_removed = state.process(rev.text, rev.id, text_sha1) _, tokens_added, tokens_removed = state.process(rev.text, rev.id, text_sha1)
window.append((rev.id, rev_data, tokens_added, tokens_removed)) window.append((rev.id, rev_data, tokens_added, tokens_removed))
if len(window) == PERSISTENCE_RADIUS: if len(window) == PERSISTENCE_RADIUS:
old_rev_id, old_rev_data, old_tokens_added, old_tokens_removed = window[0] old_rev_id, old_rev_data, old_tokens_added, old_tokens_removed = window[0]
num_token_revs, num_tokens = calculate_persistence(old_tokens_added) num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
old_rev_data.token_revs = num_token_revs old_rev_data.token_revs = num_token_revs
old_rev_data.tokens_added = num_tokens old_rev_data.tokens_added = num_tokens
old_rev_data.tokens_removed = len(old_tokens_removed) old_rev_data.tokens_removed = len(old_tokens_removed)
old_rev_data.tokens_window = PERSISTENCE_RADIUS-1 old_rev_data.tokens_window = PERSISTENCE_RADIUS - 1
self.print_rev_data(old_rev_data) self.print_rev_data(old_rev_data)
@ -606,7 +625,7 @@ class WikiqParser():
rev_data.token_revs = num_token_revs rev_data.token_revs = num_token_revs
rev_data.tokens_added = num_tokens rev_data.tokens_added = num_tokens
rev_data.tokens_removed = len(tokens_removed) rev_data.tokens_removed = len(tokens_removed)
rev_data.tokens_window = len(window)-(i+1) rev_data.tokens_window = len(window) - (i + 1)
self.print_rev_data(rev_data) self.print_rev_data(rev_data)
page_count += 1 page_count += 1
@ -622,11 +641,11 @@ class WikiqParser():
else: else:
self.output_file.close() self.output_file.close()
""" """
For performance reasons it's better to write parquet in batches instead of one row at a time. For performance reasons it's better to write parquet in batches instead of one row at a time.
So this function just puts the data on a buffer. If the buffer is full, then it gets flushed (written). So this function just puts the data on a buffer. If the buffer is full, then it gets flushed (written).
""" """
def write_parquet_row(self, rev_data): def write_parquet_row(self, rev_data):
padata = rev_data.to_pyarrow() padata = rev_data.to_pyarrow()
self.parquet_buffer.append(padata) self.parquet_buffer.append(padata)
@ -634,16 +653,17 @@ class WikiqParser():
if len(self.parquet_buffer) >= self.parquet_buffer_size: if len(self.parquet_buffer) >= self.parquet_buffer_size:
self.flush_parquet_buffer() self.flush_parquet_buffer()
""" """
Function that actually writes data to the parquet file. Function that actually writes data to the parquet file.
It needs to transpose the data from row-by-row to column-by-column It needs to transpose the data from row-by-row to column-by-column
""" """
def flush_parquet_buffer(self): def flush_parquet_buffer(self):
""" """
Returns the pyarrow table that we'll write Returns the pyarrow table that we'll write
""" """
def rows_to_table(rg, schema): def rows_to_table(rg, schema):
cols = [] cols = []
first = rg[0] first = rg[0]
@ -661,18 +681,18 @@ class WikiqParser():
outtable = rows_to_table(self.parquet_buffer, self.schema) outtable = rows_to_table(self.parquet_buffer, self.schema)
if self.pq_writer is None: if self.pq_writer is None:
self.pq_writer = pq.ParquetWriter(self.output_file, schema, flavor='spark') self.pq_writer = pq.ParquetWriter(self.output_file, self.schema, flavor='spark')
self.pq_writer.write_table(outtable) self.pq_writer.write_table(outtable)
self.parquet_buffer = [] self.parquet_buffer = []
# depending on if we are configured to write tsv or parquet, we'll call a different function. # depending on if we are configured to write tsv or parquet, we'll call a different function.
def print_rev_data(self, rev_data): def print_rev_data(self, rev_data):
if self.output_parquet is False: if self.output_parquet is False:
printfunc = self.write_tsv_row printfunc = self.write_tsv_row
else: else:
printfunc = self.write_parquet_row printfunc = self.write_parquet_row
printfunc(rev_data) printfunc(rev_data)
def write_tsv_row(self, rev_data): def write_tsv_row(self, rev_data):
@ -686,11 +706,11 @@ class WikiqParser():
def open_input_file(input_filename): def open_input_file(input_filename):
if re.match(r'.*\.7z$', input_filename): if re.match(r'.*\.7z$', input_filename):
cmd = ["7za", "x", "-so", input_filename, "*.xml"] cmd = ["7za", "x", "-so", input_filename, "*.xml"]
elif re.match(r'.*\.gz$', input_filename): elif re.match(r'.*\.gz$', input_filename):
cmd = ["zcat", input_filename] cmd = ["zcat", input_filename]
elif re.match(r'.*\.bz2$', input_filename): elif re.match(r'.*\.bz2$', input_filename):
cmd = ["bzcat", "-dk", input_filename] cmd = ["bzcat", "-dk", input_filename]
try: try:
input_file = Popen(cmd, stdout=PIPE).stdout input_file = Popen(cmd, stdout=PIPE).stdout
@ -699,7 +719,8 @@ def open_input_file(input_filename):
return input_file return input_file
def get_output_filename(input_filename, parquet = False):
def get_output_filename(input_filename, parquet=False):
output_filename = re.sub(r'\.(7z|gz|bz2)?$', '', input_filename) output_filename = re.sub(r'\.(7z|gz|bz2)?$', '', input_filename)
output_filename = re.sub(r'\.xml', '', output_filename) output_filename = re.sub(r'\.xml', '', output_filename)
if parquet is False: if parquet is False:
@ -708,16 +729,18 @@ def get_output_filename(input_filename, parquet = False):
output_filename = output_filename + ".parquet" output_filename = output_filename + ".parquet"
return output_filename return output_filename
def open_output_file(input_filename): def open_output_file(input_filename):
# create a regex that creates the output filename # create a regex that creates the output filename
output_filename = get_output_filename(input_filename, parquet = False) output_filename = get_output_filename(input_filename, parquet=False)
output_file = open(output_filename, "w") output_file = open(output_filename, "w")
return output_file return output_file
parser = argparse.ArgumentParser(description='Parse MediaWiki XML database dumps into tab delimitted data.') parser = argparse.ArgumentParser(description='Parse MediaWiki XML database dumps into tab delimitted data.')
# arguments for the input direction # arguments for the input direction
parser.add_argument('dumpfiles', metavar="DUMPFILE", nargs="*", type=str, 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.") 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-dir', metavar='DIR', dest='output_dir', type=str, nargs=1, parser.add_argument('-o', '--output-dir', metavar='DIR', dest='output_dir', type=str, nargs=1,
@ -729,7 +752,8 @@ parser.add_argument('-s', '--stdout', dest="stdout", action="store_true",
parser.add_argument('--collapse-user', dest="collapse_user", action="store_true", 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.") 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 = ['','segment','sequence','legacy'], nargs='?', parser.add_argument('-p', '--persistence', dest="persist", default=None, const='', type=str,
choices=['', '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 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.") 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.")
parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true", parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true",
@ -749,19 +773,19 @@ parser.add_argument('-rr',
parser.add_argument('-RP', '--revision-pattern', dest="regex_match_revision", default=None, type=str, action='append', 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.") 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', 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.") 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', 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.") 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', 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.") help="The label for the outputted column based on matching the regex in comments.")
args = parser.parse_args() args = parser.parse_args()
# set persistence method # set persistence method
if args.persist is None: if args.persist is None:
@ -798,7 +822,7 @@ if len(args.dumpfiles) > 0:
output_file = sys.stdout output_file = sys.stdout
else: else:
filename = os.path.join(output_dir, os.path.basename(filename)) filename = os.path.join(output_dir, os.path.basename(filename))
output_file = get_output_filename(filename, parquet = output_parquet) output_file = get_output_filename(filename, parquet=output_parquet)
wikiq = WikiqParser(input_file, wikiq = WikiqParser(input_file,
output_file, output_file,
@ -807,15 +831,15 @@ if len(args.dumpfiles) > 0:
urlencode=args.urlencode, urlencode=args.urlencode,
namespaces=namespaces, namespaces=namespaces,
revert_radius=args.revert_radius, revert_radius=args.revert_radius,
regex_match_revision = args.regex_match_revision, regex_match_revision=args.regex_match_revision,
regex_revision_label = args.regex_revision_label, regex_revision_label=args.regex_revision_label,
regex_match_comment = args.regex_match_comment, regex_match_comment=args.regex_match_comment,
regex_comment_label = args.regex_comment_label, regex_comment_label=args.regex_comment_label,
output_parquet=output_parquet) output_parquet=output_parquet)
wikiq.process() wikiq.process()
# close things # close things
input_file.close() input_file.close()
else: else:
@ -823,16 +847,16 @@ else:
sys.stdout, sys.stdout,
collapse_user=args.collapse_user, collapse_user=args.collapse_user,
persist=persist, persist=persist,
#persist_legacy=args.persist_legacy, # persist_legacy=args.persist_legacy,
urlencode=args.urlencode, urlencode=args.urlencode,
namespaces=namespaces, namespaces=namespaces,
revert_radius=args.revert_radius, revert_radius=args.revert_radius,
regex_match_revision = args.regex_match_revision, regex_match_revision=args.regex_match_revision,
regex_revision_label = args.regex_revision_label, regex_revision_label=args.regex_revision_label,
regex_match_comment = args.regex_match_comment, regex_match_comment=args.regex_match_comment,
regex_comment_label = args.regex_comment_label) regex_comment_label=args.regex_comment_label)
wikiq.process() 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 = "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(",") # stop_words = stop_words.split(",")