X-Git-Url: https://code.communitydata.science/mediawiki_dump_tools.git/blobdiff_plain/f784c77f60ae014aeda3c3b56ed8168692ddb13c..4089ebae9272bb74c8d6f5f4404969e125dfbd70:/wikiq diff --git a/wikiq b/wikiq index d7bc0bd..731f59a 100755 --- a/wikiq +++ b/wikiq @@ -3,7 +3,6 @@ # 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, os.path @@ -13,16 +12,25 @@ from subprocess import Popen, PIPE from collections import deque from hashlib import sha1 -from mwxml import Dump +from mwxml import Dump, Page from deltas.tokenizers import wikitext_split +from mwdiffs.utilities import dump2diffs import mwpersistence +from mwpersistence.state import DiffState + +from mwpersistence import Token +from mwpersistence.utilities import diffs2persistence import mwreverts from urllib.parse import quote -TO_ENCODE = ('title', 'editor') -PERSISTENCE_RADIUS=7 + from deltas import SequenceMatcher from deltas import SegmentMatcher +TO_ENCODE = ('title', 'editor') +PERSISTENCE_RADIUS=7 + +ws_lex = ['break','whitespace'] +punct_lex = ['period','qmark','epoint','comma','colon','scolon','paren_open','paren_close','brack_open','brack_close','dbrack_close','dbrack_open','tab_close','tab_open','dcurly_close','dcurly_open','equals','bar','etc','bold','italic','tag','comment_end','comment_start'] class PersistMethod: none = 0 @@ -30,101 +38,158 @@ class PersistMethod: segment = 2 legacy = 3 -def calculate_persistence(tokens_added): - return(sum([(len(x.revisions)-1) for x in tokens_added]), - len(tokens_added)) +def calculate_persistence(tokens_added, tokens_removed, exclude_ws = False, exclude_punct = False, legacy = False): -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 = 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) + if not legacy: + cond = lambda t: not (exclude_punct and (t.type in punct_lex)) \ + and not(exclude_ws and (t.type in ws_lex)) - def __iter__(self): - return self.__pages + tokens_added = [t for t in tokens_added if cond(t)] + tokens_removed = [t for t in tokens_removed if cond(t)] - def __next__(self): - return next(self._pages) + return(sum([(len(x.revisions)-1) for x in tokens_added]), + len(tokens_added), + len(tokens_removed) + ) + +class WikiqIterator(Dump): + + @classmethod + def from_file(cls, fh, collapse_user = False): + cls.fh = fh + cls.collapse_user = collapse_user + cls = super(WikiqIterator, cls).from_file(fh) + return cls + + @classmethod + def process_item(cls, item_element, namespace_map): + if not hasattr(cls,'inv_namespace_map'): + cls.inv_namespace_map = {ns.id:name for name, ns in namespace_map.items()} + + if item_element.tag == "page": + return WikiqPage.from_element(item_element, namespace_map, cls.inv_namespace_map, cls.collapse_user) + elif item_element.tag == "logitem": + return LogItem.from_element(item_element, namespace_map) + else: + raise MalformedXML("Expected to see or . " + + "Instead saw <{0}>".format(item_element.tag)) -class WikiqPage(): +class WikiqPage(Page): __slots__ = ('id', 'title', 'namespace', 'redirect', - 'restrictions', 'mwpage', '__revisions', - 'collapse_user') - - def __init__(self, page, namespace_map, collapse_user=False): - self.id = page.id - self.namespace = page.namespace + 'restrictions','collapse_user') + + @classmethod + def from_element(cls, item_element, namespace_map, inv_namespace_map, collapse_user = False): + cls.prev_rev = None + + cls = super(WikiqPage, cls).from_element(item_element, namespace_map) + # following mwxml, we assume namespace 0 in cases where # page.namespace is inconsistent with namespace_map - if page.namespace not in namespace_map: - self.title = page.title - page.namespace = 0 - if page.namespace != 0: - self.title = ':'.join([namespace_map[page.namespace], page.title]) + # this undoes the "correction" of the namespace in mwxml + + if cls.namespace not in inv_namespace_map: + cls.namespace = 0 + if cls.namespace != 0: + cls.title = ':'.join([inv_namespace_map[cls.namespace], cls.title]) + + cls.collapse_user = collapse_user + cls.revisions = cls._Page__revisions + return cls + + @staticmethod + def _correct_sha(rev_data): + + if rev_data.deleted.text: + rev_data.text = "" + rev_data.text_chars = 0 + rev_data.sha1 = "" + rev_data.revert = "" + rev_data.reverteds = "" + else: - self.title = page.title - self.restrictions = page.restrictions - self.collapse_user = collapse_user - self.mwpage = page - self.__revisions = self.rev_list() - - 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 - for i, rev in enumerate(self.mwpage): - # never yield the first time - if i == 0: - if self.collapse_user: - collapsed_revs = 1 - rev.collapsed_revs = collapsed_revs + if rev_data.text is None : + rev_data.text = "" + + rev_data.text_chars = len(rev_data.text) - else: - if self.collapse_user: - # yield if this is the last edit in a seq by a user and reset - # also yield if we do know who the user is - - if rev.deleted.user or prev_rev.deleted.user: - yield prev_rev - collapsed_revs = 1 - rev.collapsed_revs = collapsed_revs - - elif not rev.user.text == prev_rev.user.text: - yield prev_rev - collapsed_revs = 1 - rev.collapsed_revs = collapsed_revs - # otherwise, add one to the counter - else: - collapsed_revs += 1 - rev.collapsed_revs = collapsed_revs - # if collapse_user is false, we always yield - else: - yield prev_rev + if hasattr(rev_data,"sha1") and rev_data.sha1 is not None: + text_sha1 = rev_data.sha1 + + else: + text_sha1 = sha1(bytes(rev_data.text, "utf8")).hexdigest() + rev_data.sha1 = text_sha1 + + return rev_data + + # 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 + def __find_next_revision(self): + + if self.prev_rev is None: + prev_rev = WikiqPage._correct_sha(next(self.revisions)) + self.prev_rev = prev_rev + else: + prev_rev = self.prev_rev + + if self.collapse_user: + collapsed_revs = 1 + self.prev_rev.collapsed_revs = collapsed_revs + prev_rev = self.prev_rev + + for rev in self.revisions: + rev = WikiqPage._correct_sha(rev) + if self.collapse_user: + # yield if this is the last edit in a seq by a user and reset + # also yield if we do know who the user is + + if rev.deleted.user or prev_rev.deleted.user: + self.prev_rev = rev + if prev_rev is not None: + prev_rev.collapsed_revs = collapsed_revs + return prev_rev + + elif not rev.user.text == prev_rev.user.text: + self.prev_rev = rev + if prev_rev is not None: + prev_rev.collapsed_revs = collapsed_revs + return prev_rev + + # otherwise, add one to the counter + else: + collapsed_revs += 1 + rev.collapsed_revs = collapsed_revs + # if collapse_user is false, we always yield + else: + self.prev_rev = rev + if prev_rev is not None: + return prev_rev prev_rev = rev - # also yield the final time - yield prev_rev + self.prev_rev = None + + if self.collapse_user: + prev_rev.collapsed_revs = collapsed_revs + return prev_rev - def __iter__(self): - return self.__revisions def __next__(self): - return next(self.__revisions) + revision = self.__find_next_revision() + revision.page = self + return revision + + def __iter__(self): + while(True): + revision = self.__find_next_revision() + revision.page = self + yield revision class WikiqParser(): @@ -133,7 +198,6 @@ class WikiqParser(): Parameters: persist : what persistence method to use. Takes a PersistMethod value """ - self.input_file = input_file self.output_file = output_file self.collapse_user = collapse_user @@ -146,46 +210,54 @@ class WikiqParser(): else: self.namespace_filter = None - def __get_namespace_from_title(self, title): - default_ns = None - - for ns in self.namespaces: - # skip if the namespace is not defined - if ns == 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) + self.dump = WikiqIterator.from_file(self.input_file, self.collapse_user) + + self.diff_engine = None + + if self.persist == PersistMethod.sequence: + self.diff_engine = SequenceMatcher(tokenizer = wikitext_split) + + if self.persist == PersistMethod.segment: + self.diff_engine = SegmentMatcher(tokenizer = wikitext_split) + + # def __get_namespace_from_title(self, title): + # default_ns = None + + # for ns in self.namespaces: + # # skip if the namespace is not defined + # if ns == None: + # default_ns = self.namespaces[ns] + # continue + + # if title.startswith(ns + ":"): + # return self.namespaces[ns] - # extract list of namspaces - self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces} + # # if we've made it this far with no matches, we return the default namespace + # return default_ns + # def _set_namespace(self, rev_docs): + + # for rev_data in rev_docs: + # if 'namespace' not in rev_data['page']: + # namespace = self.__get_namespace_from_title(page['title']) + # rev_data['page']['namespace'] = namespace + # yield rev_data + + def process(self): page_count = 0 rev_count = 0 + for page in self.dump: - # Iterate through pages - for page in dump: - namespace = page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title) - - # skip namespaces not in the filter - if self.namespace_filter is not None: - if namespace not in self.namespace_filter: - continue + # skip pages not in the namespaces we want + if self.namespace_filter is not None and page.namespace not in self.namespace_filter: + continue rev_detector = mwreverts.Detector() @@ -193,27 +265,25 @@ class WikiqParser(): window = deque(maxlen=PERSISTENCE_RADIUS) if self.persist == PersistMethod.sequence: - state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split), + state = DiffState(SequenceMatcher(tokenizer = wikitext_split), revert_radius=PERSISTENCE_RADIUS) elif self.persist == PersistMethod.segment: - state = mwpersistence.DiffState(SegmentMatcher(tokenizer = wikitext_split), + state = DiffState(SegmentMatcher(tokenizer = wikitext_split), revert_radius=PERSISTENCE_RADIUS) - # self.persist == PersistMethod.legacy else: from mw.lib import persistence state = persistence.State() # Iterate through a page's revisions for rev in page: - rev_data = {'revid' : rev.id, 'date_time' : rev.timestamp.strftime('%Y-%m-%d %H:%M:%S'), 'articleid' : page.id, 'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id, 'title' : '"' + page.title + '"', - 'namespace' : namespace, + 'namespace' : page.namespace, 'deleted' : "TRUE" if rev.deleted.text else "FALSE" } # if revisions are deleted, /many/ things will be missing @@ -270,15 +340,15 @@ class WikiqParser(): #TODO missing: additions_size deletions_size # if collapse user was on, lets run that - if self.collapse_user: - rev_data['collapsed_revs'] = rev.collapsed_revs + # if self.collapse_user: + # rev_data.collapsed_revs = rev.collapsed_revs if self.persist != PersistMethod.none: if rev.deleted.text: for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]: old_rev_data[k] = None else: - + if self.persist != PersistMethod.legacy: _, tokens_added, tokens_removed = state.update(rev.text, rev.id) @@ -290,11 +360,11 @@ class WikiqParser(): if len(window) == PERSISTENCE_RADIUS: 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_added, num_tokens_removed = calculate_persistence(old_tokens_added, old_tokens_removed, legacy = self.persist == PersistMethod.legacy) old_rev_data["token_revs"] = num_token_revs - old_rev_data["tokens_added"] = num_tokens - old_rev_data["tokens_removed"] = len(old_tokens_removed) + old_rev_data["tokens_added"] = num_tokens_added + old_rev_data["tokens_removed"] = num_tokens_removed old_rev_data["tokens_window"] = PERSISTENCE_RADIUS-1 self.print_rev_data(old_rev_data) @@ -312,11 +382,12 @@ class WikiqParser(): continue rev_id, rev_data, tokens_added, tokens_removed = item - num_token_revs, num_tokens = calculate_persistence(tokens_added) + + num_token_revs, num_tokens_added, num_tokens_removed = calculate_persistence(tokens_added, tokens_removed, legacy = self.persist == PersistMethod.legacy) rev_data["token_revs"] = num_token_revs - rev_data["tokens_added"] = num_tokens - rev_data["tokens_removed"] = len(tokens_removed) + rev_data["tokens_added"] = num_tokens_added + rev_data["tokens_removed"] = num_tokens_removed rev_data["tokens_window"] = len(window)-(i+1) self.print_rev_data(rev_data) @@ -385,7 +456,7 @@ parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true", 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.") parser.add_argument('-n', '--namespace-include', dest="namespace_filter", type=int, action='append', - help="Id number of namspace to include.") + help="Id number of namspace to include. Can be specified more than once.")