X-Git-Url: https://code.communitydata.science/mediawiki_dump_tools.git/blobdiff_plain/d77b0a4965e59a3e5f84956b27dad58d98110c06..bf396ad366988d8dfa8afd00cbc49df8454cc611:/wikiq diff --git a/wikiq b/wikiq index 9260f35..bc6b06d 100755 --- a/wikiq +++ b/wikiq @@ -13,7 +13,7 @@ from subprocess import Popen, PIPE from collections import deque from hashlib import sha1 -from mw.xml_dump import Iterator +from mwxml import Dump from deltas.tokenizers import wikitext_split import mwpersistence @@ -32,12 +32,16 @@ class WikiqIterator(): def __init__(self, fh, collapse_user=False): self.fh = fh self.collapse_user = collapse_user - self.mwiterator = Iterator.from_file(self.fh) + 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, collapse_user=self.collapse_user) + yield WikiqPage(page, + namespace_map = self.namespace_map, + collapse_user=self.collapse_user) def __iter__(self): return self.__pages @@ -50,13 +54,14 @@ class WikiqPage(): 'restrictions', 'mwpage', '__revisions', 'collapse_user') - def __init__(self, page, collapse_user=False): + def __init__(self, page, namespace_map, collapse_user=False): self.id = page.id - self.title = page.title self.namespace = page.namespace - self.redirect = page.redirect + if page.namespace != 0: + self.title = ':'.join([namespace_map[page.namespace], page.title]) + else: + self.title = page.title self.restrictions = page.restrictions - self.collapse_user = collapse_user self.mwpage = page self.__revisions = self.rev_list() @@ -80,7 +85,14 @@ class WikiqPage(): else: if self.collapse_user: # yield if this is the last edit in a seq by a user and reset - if not rev.contributor.user_text == prev_rev.contributor.user_text: + # 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 @@ -93,6 +105,7 @@ class WikiqPage(): yield prev_rev prev_rev = rev + # also yield the final time yield prev_rev @@ -104,7 +117,6 @@ class WikiqPage(): class WikiqParser(): - def __init__(self, input_file, output_file, collapse_user=False, persist=False, urlencode=False, persist_legacy=False): self.input_file = input_file @@ -142,7 +154,7 @@ class WikiqParser(): dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user) # extract list of namspaces - self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.namespaces} + self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces} page_count = 0 rev_count = 0 @@ -169,23 +181,28 @@ class WikiqParser(): rev_data = {'revid' : rev.id, 'date_time' : rev.timestamp.strftime('%Y-%m-%d %H:%M:%S'), 'articleid' : page.id, - 'editor_id' : "" if rev.contributor.id == None else rev.contributor.id, + 'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id, 'title' : '"' + page.title + '"', - 'namespace' : page.namespace if page.namespace else self.__get_namespace_from_title(page.title), - 'deleted' : "TRUE" if rev.text.deleted else "FALSE" } + 'namespace' : page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title), + 'deleted' : "TRUE" if rev.deleted.text else "FALSE" } # if revisions are deleted, /many/ things will be missing - if rev.text.deleted: + if rev.deleted.text: rev_data['text_chars'] = "" rev_data['sha1'] = "" rev_data['revert'] = "" rev_data['reverteds'] = "" else: + # rev.text can be None if the page has no text + if not rev.text: + rev.text = "" # if text exists, we'll check for a sha1 and generate one otherwise + if rev.sha1: text_sha1 = rev.sha1 else: + text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest() rev_data['sha1'] = text_sha1 @@ -206,10 +223,10 @@ class WikiqParser(): # if the fact that the edit was minor can be hidden, this might be an issue rev_data['minor'] = "TRUE" if rev.minor else "FALSE" - if rev.contributor.user_text: + if not rev.deleted.user: # wrap user-defined editors in quotes for fread - rev_data['editor'] = '"' + rev.contributor.user_text + '"' - rev_data['anon'] = "TRUE" if rev.contributor.id == None else "FALSE" + rev_data['editor'] = '"' + rev.user.text + '"' + rev_data['anon'] = "TRUE" if rev.user.id == None else "FALSE" else: rev_data['anon'] = "" @@ -227,7 +244,8 @@ class WikiqParser(): rev_data['collapsed_revs'] = rev.collapsed_revs if self.persist or self.persist_legacy: - if rev.text.deleted: + if rev.deleted.text: + for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]: old_rev_data[k] = None else: @@ -236,7 +254,7 @@ class WikiqParser(): _, tokens_added, tokens_removed = state.update(rev.text, rev.id) 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)) @@ -284,7 +302,7 @@ class WikiqParser(): if self.urlencode: for field in TO_ENCODE: rev_data[field] = quote(str(rev_data[field])) - + if not self.printed_header: print("\t".join([str(k) for k in sorted(rev_data.keys())]), file=self.output_file) self.printed_header = True