def wiki2sentences(wiki, sent_detector,withTags=True):
# get rid of (nested) template calls
oldLen = 1E10
while len(wiki)<oldLen:
oldLen = len(wiki)
wiki = re.sub('{[^{}]*}',' ',wiki)
tree = parse_txt(wiki)
text = tree2string(tree)
lines = cleanup(text).split('\n')
sentences = []
tags = []
for line in lines:
if line.startswith('<s>'):
sentences.append(line[3:].strip())
tags.append('Section')
elif line.startswith('<i>'):
sentences.append(line[3:].strip())
tags.append('Item')
else:
newSentences = sent_detector(line.strip())
sentences += newSentences
tags += ['Sentence']*(len(newSentences)-1)
tags.append('LastSentence')
if withTags:
return sentences,tags
else:
return sentences
def _process_page(self, elem):
"""
Gather frequency distribution of a page,
category names, and linked page names.
"""
# Get the text we need.
#id = int(self._find(elem, 'id').text)
title = self._find(elem, 'title').text
#datetime = self._find(elem, 'revision', 'timestamp').text
text = self._find(elem, 'revision', 'text').text
redirect = self._find(elem, 'redirect')
# `title` should be the canonical title, i.e. the 'official'
# title of a page. If the page redirects to another (the canonical
# page), the <redirect> elem contains the canonical title to which
# the page redirects.
if redirect is not None:
title = redirect.attrib.get('title')
# Extract certain elements.
# pagelinks is a list of linked page titles.
# categories is a list of this page's category titles.
result = parse_txt(text)
pagelinks = [pagelink.full_target for pagelink in result.find(parser.ArticleLink)]
# Have to trim the beginning of Category links.
# So 'Category:Anarchism' becomes just 'Anarchism'.
cstart = len('Category:')
categories = [category.full_target[cstart:] for category in result.find(parser.CategoryLink)]
# Build the bag of words representation of the document.
clean_text = self._clean(text)
return clean_text
def simpleparse(
raw,
lang=None
): # !!! USE FOR DEBUGGING ONLY !!! does not use post processors
a = compat.parse_txt(raw, lang=lang)
core.show(a)
return a
def wiki2sentences(wiki, sent_detector, withTags=True):
# get rid of (nested) template calls
oldLen = 1E10
while len(wiki) < oldLen:
oldLen = len(wiki)
wiki = re.sub('{[^{}]*}', ' ', wiki)
tree = parse_txt(wiki)
text = tree2string(tree)
lines = cleanup(text).split('\n')
sentences = []
tags = []
for line in lines:
if line.startswith('<s>'):
sentences.append(line[3:].strip())
tags.append('Section')
elif line.startswith('<i>'):
sentences.append(line[3:].strip())
tags.append('Item')
else:
newSentences = sent_detector(line.strip())
sentences += newSentences
tags += ['Sentence'] * (len(newSentences) - 1)
tags.append('LastSentence')
if withTags:
return sentences, tags
else:
return sentences
def parseString(title=None, raw=None, wikidb=None, revision=None,
lang=None, magicwords=None, expandTemplates=True):
"""parse article with title from raw mediawiki text"""
uniquifier = None
siteinfo = None
assert title is not None, 'no title given'
if raw is None:
page = wikidb.normalize_and_get_page(title, 0)
if page:
raw = page.rawtext
else:
raw = None
assert raw is not None, "cannot get article %r" % (title,)
input_raw = raw
te = None
if wikidb:
if expandTemplates:
te = expander.Expander(raw, pagename=title, wikidb=wikidb)
input_raw = te.expandTemplates(True)
uniquifier = te.uniquifier
if hasattr(wikidb, 'get_siteinfo'):
siteinfo = wikidb.get_siteinfo()
src = None
if hasattr(wikidb, 'getSource'):
src = wikidb.getSource(title, revision=revision)
assert not isinstance(src, dict)
if not src:
src = metabook.source()
if lang is None:
lang = src.language
if magicwords is None:
if siteinfo is not None and 'magicwords' in siteinfo:
magicwords = siteinfo['magicwords']
else:
magicwords = src.get('magicwords')
if siteinfo is None:
nshandler = nshandling.get_nshandler_for_lang(lang)
else:
nshandler = nshandling.nshandler(siteinfo)
a = compat.parse_txt(input_raw, title=title, wikidb=wikidb, nshandler=nshandler, lang=lang, magicwords=magicwords, uniquifier=uniquifier, expander=te)
a.caption = title
if te and te.magic_displaytitle:
a.caption = te.magic_displaytitle
from mwlib.old_uparser import postprocessors
for x in postprocessors:
x(a, title=title, revision=revision, wikidb=wikidb, lang=lang)
return a
def get_media_for_article(self, topic):
media = []
try:
client = Client()
markup = self.get_markup_for_topic(topic)
article = compat.parse_txt(markup)
self.depth_find_media(article, topic, media)
#print "toplc: " + topic
#print media
except:
pass
return media
def get_article(self, topic):
global usejsoncache
global threaded
cacheDir = "./json_cache/{}".format(self.wiki_version)
try:
os.mkdir(cacheDir)
except:
pass
filename = "{}/{}".format(cacheDir, topic)
if usejsoncache is True and os.path.isfile(filename):
print "cache hit for " + topic
js = open(filename).read()
else:
markup = self.get_markup_for_topic(topic)
templParser = mwlib.templ.parser.Parser(markup)
templates = templParser.parse()
print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$"
markup = ""
for t in templates:
# print "==>{} {}<==".format(type(t), t)
if isinstance(t, unicode):
markup+= t
elif isinstance(t, mwlib.templ.nodes.Template):
print "==>{}<==".format(t[0])
if t[0] == "Wide image":
print " -->{}<--".format(t[1][0])
markup+= " [[File:{}]] ".format(t[1][0])
# print article
print "$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$"
markup = expandstr(markup)
# print markup
article = compat.parse_txt(markup)
self.reset()
if threaded:
self.build_related_media(article)
else:
print "Not performing threaded processing"
self.depth_first(article)
obj = {"content":self.content, "title":topic, "url":"https://{}.wikipedia.org/wiki/{}".format(self.wiki_version, topic), "url":"https://{}.wikipedia.org/wiki/{}".format(self.wiki_version, topic)}
js = json.dumps(obj, indent=2)
open(filename, "w").write(js)
return js
def wiki2sentences(wiki, sent_detector, withTags=True):
# save the dates (made obsolete by introduction of wikitemplate2text)
#wiki = re.sub('{{date\|([^{}]*?)}}', r'\1', wiki)
# repeat since everything can be nested
oldLen = 1E10
while len(wiki) < oldLen:
oldLen = len(wiki)
# eliminates html comments. e.g. <--This is a comment-->
wiki = re.sub('<!--.*?-->', '', wiki)
# eliminate wiki tables
# commented out because mwlib takes care of them
#wiki = re.sub('{\|[^{}]*?\|}', wikitable2text, wiki)
# eliminate wiki templates. e.g. {{date}}
wiki = re.sub('{{[^{}]*}}', wikitemplate2text, wiki)
# eliminate refrence tags. e.g. <ref>text</ref>
wiki = re.sub('<ref[^/>]*?>[^<]*?</ref>', '', wiki)
# eliminate html tags. e.g. <ref name="My Life"/>
wiki = re.sub('</?[A-Za-z][^>]*?>', '', wiki)
# remove angle brackets
# mwlib restores the changes, so do nothing.
#wiki = re.sub('<', '<', wiki)
#wiki = re.sub('>', '>', wiki)
#print wiki.encode('utf-8')
tree = parse_txt(wiki)
text = tree2string(tree)
lines = cleanup(text).split('\n')
sentences = []
tags = []
for line in lines:
if line.startswith('<s>'):
sentences.append(line[3:].strip())
tags.append('Section')
elif line.startswith('<i>'):
newSentences = sent_detector(line[3:].strip())
sentences += newSentences
tags += ['Item-Sentence']*(len(newSentences)-1)
tags.append('Item-LastSentence')
else:
newSentences = sent_detector(line.strip())
sentences += newSentences
tags += ['Sentence']*(len(newSentences)-1)
tags.append('LastSentence')
if withTags:
return sentences,tags
else:
return sentences
def extractPage(wiki_text_n, path, wiki_text, verbose=False):
try:
page = extractText(parse_txt(wiki_text))
music_info = music_re.search(page)
if music_info:
music_info_end = infobox_end(page, music_info.end())
if music_info_end > 0:
infobox = page[music_info.start() + 2:music_info_end]
if verbose:
print "Page:", wiki_text_n
print "Infobox:", infobox
infobox = "".join(infobox.split("\n"))
infobox = "".join(infobox.split("\r"))
infobox = stripmacros(infobox, " ")
if verbose:
print "Infobox 1:", infobox
infobox = infobox.split("|")
if verbose:
for ib in infobox:
print "Line: [%s]\n" % (ib)
attrs = "<Metadata MyRadioType=\"" + music_info.group(
2) + "\" "
ibattrs = {}
for info in infobox:
item = info.split("=", 1)
if len(item) == 2:
key = nopunct(item[0]).strip().replace(" ", "_")
value = item[1].strip()
if value.endswith("}}"):
value = value[:-2]
ibattrs[key] = saxutils.quoteattr(
striptags(saxutils.unescape(value), ","))
for (key, value) in ibattrs.items():
attrs += " %s=%s" % (key, value)
attrs += "/>\n"
if verbose:
print "Extracting:", attrs
page = page[:music_info.start()] + page[music_info_end:]
page = "<?xml version=\"1.0\" encoding=\"UTF-8\" ?>\n<Document>\n" + attrs + page + "</Document>\n"
page = stripmacros(page)
codecs.open(os.path.join(path, "%08d.xml" % (wiki_text_n)),
"w", "utf-8", "replace").write(page)
except Exception, e:
if verbose:
print "Error at %d was %s" % (wiki_text_n, str(e))
codecs.open(os.path.join(path, "%08d.raw" % (wiki_text_n)), "w",
"utf-8", "replace").write(wiki_text)
def _parse(txt):
"""parse text....and try to return a 'better' (some inner) node"""
from mwlib.refine.compat import parse_txt
from mwlib import parser
res = parse_txt(txt)
# res is an parser.Article.
if len(res.children) != 1:
res.__class__ = parser.Node
return res
res = res.children[0]
if res.__class__ == parser.Paragraph:
res.__class__ = parser.Node
return res
def _parse(txt):
"""parse text....and try to return a 'better' (some inner) node"""
from mwlib.refine.compat import parse_txt
from mwlib import parser
res = parse_txt(txt)
# res is an parser.Article.
if len(res.children) != 1:
res.__class__ = parser.Node
return res
res = res.children[0]
if res.__class__ == parser.Paragraph:
res.__class__ = parser.Node
return res
def extractPage( wiki_text_n, path, wiki_text, verbose = False ):
try:
page = extractText(parse_txt(wiki_text))
music_info = music_re.search(page)
if music_info:
music_info_end = infobox_end(page,music_info.end())
if music_info_end > 0:
infobox = page[music_info.start()+2:music_info_end]
if verbose:
print "Page:",wiki_text_n
print "Infobox:",infobox
infobox = "".join(infobox.split("\n"))
infobox = "".join(infobox.split("\r"))
infobox = stripmacros(infobox," ")
if verbose:
print "Infobox 1:",infobox
infobox = infobox.split("|")
if verbose:
for ib in infobox:
print "Line: [%s]\n"%(ib)
attrs = "<Metadata MyRadioType=\"" + music_info.group(2) + "\" "
ibattrs = {}
for info in infobox:
item = info.split("=",1)
if len(item) == 2:
key = nopunct(item[0]).strip().replace(" ","_")
value = item[1].strip()
if value.endswith("}}"):
value = value[:-2]
ibattrs[key] = saxutils.quoteattr(striptags(saxutils.unescape(value),","))
for (key,value) in ibattrs.items():
attrs += " %s=%s"%(key,value)
attrs += "/>\n"
if verbose:
print "Extracting:",attrs
page = page[:music_info.start()] + page[music_info_end:]
page = "<?xml version=\"1.0\" encoding=\"UTF-8\" ?>\n<Document>\n" + attrs + page + "</Document>\n"
page = stripmacros(page)
codecs.open(os.path.join(path,"%08d.xml"%(wiki_text_n)),"w","utf-8","replace").write(page)
except Exception, e:
if verbose:
print "Error at %d was %s"%(wiki_text_n,str(e))
codecs.open(os.path.join(path,"%08d.raw"%(wiki_text_n)),"w","utf-8","replace").write(wiki_text)
def _process_page(self, elem):
"""
Gather frequency distribution of a page,
category names, and linked page names.
"""
# Get the text we need.
#id = int(self._find(elem, 'id').text)
title = self._find(elem, 'title').text
#datetime = self._find(elem, 'revision', 'timestamp').text
text = self._find(elem, 'revision', 'text').text
redirect = self._find(elem, 'redirect')
# `title` should be the canonical title, i.e. the 'official'
# title of a page. If the page redirects to another (the canonical
# page), the <redirect> elem contains the canonical title to which
# the page redirects.
if redirect is not None:
title = redirect.attrib.get('title')
# Extract certain elements.
# pagelinks is a list of linked page titles.
# categories is a list of this page's category titles.
result = parse_txt(text)
pagelinks = [
pagelink.full_target
for pagelink in result.find(parser.ArticleLink)
]
# Have to trim the beginning of Category links.
# So 'Category:Anarchism' becomes just 'Anarchism'.
cstart = len('Category:')
categories = [
category.full_target[cstart:]
for category in result.find(parser.CategoryLink)
]
# Build the bag of words representation of the document.
clean_text = self._clean(text)
return clean_text
def get_random_image(self):
global random_image_updated
global random_image
now = time.time()
d = now - random_image_updated
if d > 60: # Refresh every minute.
i = 0
media = []
while len(media) == 0 and i<100:
url = "https://{}.wikipedia.org/wiki/Special:Random?action=raw".format(self.wiki_version)
response = requests.get(url, timeout=2.0)
markup = response.text
topic = urlparse.parse_qs(urlparse.urlparse(response.url).query)["title"]
# print markup
article = compat.parse_txt(markup)
self.depth_find_media(article, topic, media)
i+=1
if len(media) > 0:
random_image = media[0]
random_image_updated = now
return random_image
print "Fail!"
sys.exit(2)
raw = wc.read()
rdata = json.loads(raw)
wc.close()
page = rdata['query']['pages'].itervalues().next()
if not page:
print "NO page found"
sys.exit(3)
revision = page['revisions'][0]
if not revision:
print "NO revision found"
sys.exit(4)
content = revision[str(revision.keys()[0])]
parsed = compat.parse_txt(content)
table = parsed.find(parser.Table)[0]
if not table:
print "Table not found"
sys.exit(5)
for row in table.children:
cells = row.find(parser.Cell)
print cells[0].asText().replace("}}", "").replace("{{", "").strip() + \
" || " + cells[1].asText().strip() + " || " + cells[2].asText().strip() \
+ " || " + cells[3].asText().strip()
def parseString(title=None,
raw=None,
wikidb=None,
revision=None,
lang=None,
magicwords=None,
expandTemplates=True):
"""parse article with title from raw mediawiki text"""
uniquifier = None
siteinfo = None
assert title is not None, 'no title given'
if raw is None:
page = wikidb.normalize_and_get_page(title, 0)
if page:
raw = page.rawtext
else:
raw = None
assert raw is not None, "cannot get article %r" % (title, )
input = raw
te = None
if wikidb:
if expandTemplates:
te = expander.Expander(raw, pagename=title, wikidb=wikidb)
input = te.expandTemplates(True)
uniquifier = te.uniquifier
if hasattr(wikidb, 'get_siteinfo'):
siteinfo = wikidb.get_siteinfo()
src = None
if hasattr(wikidb, 'getSource'):
src = wikidb.getSource(title, revision=revision)
assert not isinstance(src, dict)
if not src:
src = metabook.source()
if lang is None:
lang = src.language
if magicwords is None:
if siteinfo is not None and 'magicwords' in siteinfo:
magicwords = siteinfo['magicwords']
else:
magicwords = src.get('magicwords')
if siteinfo is None:
nshandler = nshandling.get_nshandler_for_lang(lang)
else:
nshandler = nshandling.nshandler(siteinfo)
a = compat.parse_txt(input,
title=title,
wikidb=wikidb,
nshandler=nshandler,
lang=lang,
magicwords=magicwords,
uniquifier=uniquifier,
expander=te)
a.caption = title
if te and te.magic_displaytitle:
a.caption = te.magic_displaytitle
from mwlib.old_uparser import postprocessors
for x in postprocessors:
x(a, title=title, revision=revision, wikidb=wikidb, lang=lang)
return a
def simpleparse(raw,lang=None): # !!! USE FOR DEBUGGING ONLY !!! does not use post processors
a=compat.parse_txt(raw,lang=lang)
core.show(a)
return a
#! /usr/bin/env python # -*- coding: utf-8 -*- import sys import time s = unicode(open(sys.argv[1], "rb").read(), "utf-8") from mwlib import uparser, advtree, treecleaner from mwlib.refine import compat stime = time.time() r = compat.parse_txt(s) print "parse:", time.time() - stime stime = time.time() advtree.buildAdvancedTree(r) print "tree", time.time() - stime stime = time.time() tc = treecleaner.TreeCleaner(r) tc.cleanAll() print "clean:", time.time() - stime
