mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-11 03:45:01 -05:00
0e82c5f573
Adjusted code for custom RSS feed parsing now that feedparser does the parsing for us all we do is pull the data from the returned entries.
4014 lines
163 KiB
Python
4014 lines
163 KiB
Python
"""Universal feed parser
|
|
|
|
Handles RSS 0.9x, RSS 1.0, RSS 2.0, CDF, Atom 0.3, and Atom 1.0 feeds
|
|
|
|
Visit https://code.google.com/p/feedparser/ for the latest version
|
|
Visit http://packages.python.org/feedparser/ for the latest documentation
|
|
|
|
Required: Python 2.4 or later
|
|
Recommended: iconv_codec <http://cjkpython.i18n.org/>
|
|
"""
|
|
|
|
__version__ = "5.1.3"
|
|
__license__ = """
|
|
Copyright (c) 2010-2012 Kurt McKee <contactme@kurtmckee.org>
|
|
Copyright (c) 2002-2008 Mark Pilgrim
|
|
All rights reserved.
|
|
|
|
Redistribution and use in source and binary forms, with or without modification,
|
|
are permitted provided that the following conditions are met:
|
|
|
|
* Redistributions of source code must retain the above copyright notice,
|
|
this list of conditions and the following disclaimer.
|
|
* Redistributions in binary form must reproduce the above copyright notice,
|
|
this list of conditions and the following disclaimer in the documentation
|
|
and/or other materials provided with the distribution.
|
|
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 'AS IS'
|
|
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
|
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
POSSIBILITY OF SUCH DAMAGE."""
|
|
__author__ = "Mark Pilgrim <http://diveintomark.org/>"
|
|
__contributors__ = ["Jason Diamond <http://injektilo.org/>",
|
|
"John Beimler <http://john.beimler.org/>",
|
|
"Fazal Majid <http://www.majid.info/mylos/weblog/>",
|
|
"Aaron Swartz <http://aaronsw.com/>",
|
|
"Kevin Marks <http://epeus.blogspot.com/>",
|
|
"Sam Ruby <http://intertwingly.net/>",
|
|
"Ade Oshineye <http://blog.oshineye.com/>",
|
|
"Martin Pool <http://sourcefrog.net/>",
|
|
"Kurt McKee <http://kurtmckee.org/>",
|
|
"Bernd Schlapsi <https://github.com/brot>",]
|
|
|
|
# HTTP "User-Agent" header to send to servers when downloading feeds.
|
|
# If you are embedding feedparser in a larger application, you should
|
|
# change this to your application name and URL.
|
|
USER_AGENT = "UniversalFeedParser/%s +https://code.google.com/p/feedparser/" % __version__
|
|
|
|
# HTTP "Accept" header to send to servers when downloading feeds. If you don't
|
|
# want to send an Accept header, set this to None.
|
|
ACCEPT_HEADER = "application/atom+xml,application/rdf+xml,application/rss+xml,application/x-netcdf,application/xml;q=0.9,text/xml;q=0.2,*/*;q=0.1"
|
|
|
|
# List of preferred XML parsers, by SAX driver name. These will be tried first,
|
|
# but if they're not installed, Python will keep searching through its own list
|
|
# of pre-installed parsers until it finds one that supports everything we need.
|
|
PREFERRED_XML_PARSERS = ["drv_libxml2"]
|
|
|
|
# If you want feedparser to automatically run HTML markup through HTML Tidy, set
|
|
# this to 1. Requires mxTidy <http://www.egenix.com/files/python/mxTidy.html>
|
|
# or utidylib <http://utidylib.berlios.de/>.
|
|
TIDY_MARKUP = 0
|
|
|
|
# List of Python interfaces for HTML Tidy, in order of preference. Only useful
|
|
# if TIDY_MARKUP = 1
|
|
PREFERRED_TIDY_INTERFACES = ["uTidy", "mxTidy"]
|
|
|
|
# If you want feedparser to automatically resolve all relative URIs, set this
|
|
# to 1.
|
|
RESOLVE_RELATIVE_URIS = 1
|
|
|
|
# If you want feedparser to automatically sanitize all potentially unsafe
|
|
# HTML content, set this to 1.
|
|
SANITIZE_HTML = 1
|
|
|
|
# If you want feedparser to automatically parse microformat content embedded
|
|
# in entry contents, set this to 1
|
|
PARSE_MICROFORMATS = 1
|
|
|
|
# ---------- Python 3 modules (make it work if possible) ----------
|
|
try:
|
|
import rfc822
|
|
except ImportError:
|
|
from email import _parseaddr as rfc822
|
|
|
|
try:
|
|
# Python 3.1 introduces bytes.maketrans and simultaneously
|
|
# deprecates string.maketrans; use bytes.maketrans if possible
|
|
_maketrans = bytes.maketrans
|
|
except (NameError, AttributeError):
|
|
import string
|
|
_maketrans = string.maketrans
|
|
|
|
# base64 support for Atom feeds that contain embedded binary data
|
|
try:
|
|
import base64, binascii
|
|
except ImportError:
|
|
base64 = binascii = None
|
|
else:
|
|
# Python 3.1 deprecates decodestring in favor of decodebytes
|
|
_base64decode = getattr(base64, 'decodebytes', base64.decodestring)
|
|
|
|
# _s2bytes: convert a UTF-8 str to bytes if the interpreter is Python 3
|
|
# _l2bytes: convert a list of ints to bytes if the interpreter is Python 3
|
|
try:
|
|
if bytes is str:
|
|
# In Python 2.5 and below, bytes doesn't exist (NameError)
|
|
# In Python 2.6 and above, bytes and str are the same type
|
|
raise NameError
|
|
except NameError:
|
|
# Python 2
|
|
def _s2bytes(s):
|
|
return s
|
|
def _l2bytes(l):
|
|
return ''.join(map(chr, l))
|
|
else:
|
|
# Python 3
|
|
def _s2bytes(s):
|
|
return bytes(s, 'utf8')
|
|
def _l2bytes(l):
|
|
return bytes(l)
|
|
|
|
# If you want feedparser to allow all URL schemes, set this to ()
|
|
# List culled from Python's urlparse documentation at:
|
|
# http://docs.python.org/library/urlparse.html
|
|
# as well as from "URI scheme" at Wikipedia:
|
|
# https://secure.wikimedia.org/wikipedia/en/wiki/URI_scheme
|
|
# Many more will likely need to be added!
|
|
ACCEPTABLE_URI_SCHEMES = (
|
|
'file', 'ftp', 'gopher', 'h323', 'hdl', 'http', 'https', 'imap', 'magnet',
|
|
'mailto', 'mms', 'news', 'nntp', 'prospero', 'rsync', 'rtsp', 'rtspu',
|
|
'sftp', 'shttp', 'sip', 'sips', 'snews', 'svn', 'svn+ssh', 'telnet',
|
|
'wais',
|
|
# Additional common-but-unofficial schemes
|
|
'aim', 'callto', 'cvs', 'facetime', 'feed', 'git', 'gtalk', 'irc', 'ircs',
|
|
'irc6', 'itms', 'mms', 'msnim', 'skype', 'ssh', 'smb', 'svn', 'ymsg',
|
|
)
|
|
#ACCEPTABLE_URI_SCHEMES = ()
|
|
|
|
# ---------- required modules (should come with any Python distribution) ----------
|
|
import cgi
|
|
import codecs
|
|
import copy
|
|
import datetime
|
|
import re
|
|
import struct
|
|
import time
|
|
import types
|
|
import urllib
|
|
import urllib2
|
|
import urlparse
|
|
import warnings
|
|
|
|
from htmlentitydefs import name2codepoint, codepoint2name, entitydefs
|
|
|
|
try:
|
|
from io import BytesIO as _StringIO
|
|
except ImportError:
|
|
try:
|
|
from cStringIO import StringIO as _StringIO
|
|
except ImportError:
|
|
from StringIO import StringIO as _StringIO
|
|
|
|
# ---------- optional modules (feedparser will work without these, but with reduced functionality) ----------
|
|
|
|
# gzip is included with most Python distributions, but may not be available if you compiled your own
|
|
try:
|
|
import gzip
|
|
except ImportError:
|
|
gzip = None
|
|
try:
|
|
import zlib
|
|
except ImportError:
|
|
zlib = None
|
|
|
|
# If a real XML parser is available, feedparser will attempt to use it. feedparser has
|
|
# been tested with the built-in SAX parser and libxml2. On platforms where the
|
|
# Python distribution does not come with an XML parser (such as Mac OS X 10.2 and some
|
|
# versions of FreeBSD), feedparser will quietly fall back on regex-based parsing.
|
|
try:
|
|
import xml.sax
|
|
from xml.sax.saxutils import escape as _xmlescape
|
|
except ImportError:
|
|
_XML_AVAILABLE = 0
|
|
def _xmlescape(data,entities={}):
|
|
data = data.replace('&', '&')
|
|
data = data.replace('>', '>')
|
|
data = data.replace('<', '<')
|
|
for char, entity in entities:
|
|
data = data.replace(char, entity)
|
|
return data
|
|
else:
|
|
try:
|
|
xml.sax.make_parser(PREFERRED_XML_PARSERS) # test for valid parsers
|
|
except xml.sax.SAXReaderNotAvailable:
|
|
_XML_AVAILABLE = 0
|
|
else:
|
|
_XML_AVAILABLE = 1
|
|
|
|
# sgmllib is not available by default in Python 3; if the end user doesn't have
|
|
# it available then we'll lose illformed XML parsing, content santizing, and
|
|
# microformat support (at least while feedparser depends on BeautifulSoup).
|
|
try:
|
|
import sgmllib
|
|
except ImportError:
|
|
# This is probably Python 3, which doesn't include sgmllib anymore
|
|
_SGML_AVAILABLE = 0
|
|
|
|
# Mock sgmllib enough to allow subclassing later on
|
|
class sgmllib(object):
|
|
class SGMLParser(object):
|
|
def goahead(self, i):
|
|
pass
|
|
def parse_starttag(self, i):
|
|
pass
|
|
else:
|
|
_SGML_AVAILABLE = 1
|
|
|
|
# sgmllib defines a number of module-level regular expressions that are
|
|
# insufficient for the XML parsing feedparser needs. Rather than modify
|
|
# the variables directly in sgmllib, they're defined here using the same
|
|
# names, and the compiled code objects of several sgmllib.SGMLParser
|
|
# methods are copied into _BaseHTMLProcessor so that they execute in
|
|
# feedparser's scope instead of sgmllib's scope.
|
|
charref = re.compile('&#(\d+|[xX][0-9a-fA-F]+);')
|
|
tagfind = re.compile('[a-zA-Z][-_.:a-zA-Z0-9]*')
|
|
attrfind = re.compile(
|
|
r'\s*([a-zA-Z_][-:.a-zA-Z_0-9]*)[$]?(\s*=\s*'
|
|
r'(\'[^\']*\'|"[^"]*"|[][\-a-zA-Z0-9./,:;+*%?!&$\(\)_#=~\'"@]*))?'
|
|
)
|
|
|
|
# Unfortunately, these must be copied over to prevent NameError exceptions
|
|
entityref = sgmllib.entityref
|
|
incomplete = sgmllib.incomplete
|
|
interesting = sgmllib.interesting
|
|
shorttag = sgmllib.shorttag
|
|
shorttagopen = sgmllib.shorttagopen
|
|
starttagopen = sgmllib.starttagopen
|
|
|
|
class _EndBracketRegEx:
|
|
def __init__(self):
|
|
# Overriding the built-in sgmllib.endbracket regex allows the
|
|
# parser to find angle brackets embedded in element attributes.
|
|
self.endbracket = re.compile('''([^'"<>]|"[^"]*"(?=>|/|\s|\w+=)|'[^']*'(?=>|/|\s|\w+=))*(?=[<>])|.*?(?=[<>])''')
|
|
def search(self, target, index=0):
|
|
match = self.endbracket.match(target, index)
|
|
if match is not None:
|
|
# Returning a new object in the calling thread's context
|
|
# resolves a thread-safety.
|
|
return EndBracketMatch(match)
|
|
return None
|
|
class EndBracketMatch:
|
|
def __init__(self, match):
|
|
self.match = match
|
|
def start(self, n):
|
|
return self.match.end(n)
|
|
endbracket = _EndBracketRegEx()
|
|
|
|
|
|
# iconv_codec provides support for more character encodings.
|
|
# It's available from http://cjkpython.i18n.org/
|
|
try:
|
|
import iconv_codec
|
|
except ImportError:
|
|
pass
|
|
|
|
# chardet library auto-detects character encodings
|
|
# Download from http://chardet.feedparser.org/
|
|
try:
|
|
import chardet
|
|
except ImportError:
|
|
chardet = None
|
|
|
|
# BeautifulSoup is used to extract microformat content from HTML
|
|
# feedparser is tested using BeautifulSoup 3.2.0
|
|
# http://www.crummy.com/software/BeautifulSoup/
|
|
try:
|
|
import BeautifulSoup
|
|
except ImportError:
|
|
BeautifulSoup = None
|
|
PARSE_MICROFORMATS = False
|
|
|
|
# ---------- don't touch these ----------
|
|
class ThingsNobodyCaresAboutButMe(Exception): pass
|
|
class CharacterEncodingOverride(ThingsNobodyCaresAboutButMe): pass
|
|
class CharacterEncodingUnknown(ThingsNobodyCaresAboutButMe): pass
|
|
class NonXMLContentType(ThingsNobodyCaresAboutButMe): pass
|
|
class UndeclaredNamespace(Exception): pass
|
|
|
|
SUPPORTED_VERSIONS = {'': u'unknown',
|
|
'rss090': u'RSS 0.90',
|
|
'rss091n': u'RSS 0.91 (Netscape)',
|
|
'rss091u': u'RSS 0.91 (Userland)',
|
|
'rss092': u'RSS 0.92',
|
|
'rss093': u'RSS 0.93',
|
|
'rss094': u'RSS 0.94',
|
|
'rss20': u'RSS 2.0',
|
|
'rss10': u'RSS 1.0',
|
|
'rss': u'RSS (unknown version)',
|
|
'atom01': u'Atom 0.1',
|
|
'atom02': u'Atom 0.2',
|
|
'atom03': u'Atom 0.3',
|
|
'atom10': u'Atom 1.0',
|
|
'atom': u'Atom (unknown version)',
|
|
'cdf': u'CDF',
|
|
}
|
|
|
|
class FeedParserDict(dict):
|
|
keymap = {'channel': 'feed',
|
|
'items': 'entries',
|
|
'guid': 'id',
|
|
'date': 'updated',
|
|
'date_parsed': 'updated_parsed',
|
|
'description': ['summary', 'subtitle'],
|
|
'description_detail': ['summary_detail', 'subtitle_detail'],
|
|
'url': ['href'],
|
|
'modified': 'updated',
|
|
'modified_parsed': 'updated_parsed',
|
|
'issued': 'published',
|
|
'issued_parsed': 'published_parsed',
|
|
'copyright': 'rights',
|
|
'copyright_detail': 'rights_detail',
|
|
'tagline': 'subtitle',
|
|
'tagline_detail': 'subtitle_detail'}
|
|
def __getitem__(self, key):
|
|
if key == 'category':
|
|
try:
|
|
return dict.__getitem__(self, 'tags')[0]['term']
|
|
except IndexError:
|
|
raise KeyError, "object doesn't have key 'category'"
|
|
elif key == 'enclosures':
|
|
norel = lambda link: FeedParserDict([(name,value) for (name,value) in link.items() if name!='rel'])
|
|
return [norel(link) for link in dict.__getitem__(self, 'links') if link['rel']==u'enclosure']
|
|
elif key == 'license':
|
|
for link in dict.__getitem__(self, 'links'):
|
|
if link['rel']==u'license' and 'href' in link:
|
|
return link['href']
|
|
elif key == 'updated':
|
|
# Temporarily help developers out by keeping the old
|
|
# broken behavior that was reported in issue 310.
|
|
# This fix was proposed in issue 328.
|
|
if not dict.__contains__(self, 'updated') and \
|
|
dict.__contains__(self, 'published'):
|
|
warnings.warn("To avoid breaking existing software while "
|
|
"fixing issue 310, a temporary mapping has been created "
|
|
"from `updated` to `published` if `updated` doesn't "
|
|
"exist. This fallback will be removed in a future version "
|
|
"of feedparser.", DeprecationWarning)
|
|
return dict.__getitem__(self, 'published')
|
|
return dict.__getitem__(self, 'updated')
|
|
elif key == 'updated_parsed':
|
|
if not dict.__contains__(self, 'updated_parsed') and \
|
|
dict.__contains__(self, 'published_parsed'):
|
|
warnings.warn("To avoid breaking existing software while "
|
|
"fixing issue 310, a temporary mapping has been created "
|
|
"from `updated_parsed` to `published_parsed` if "
|
|
"`updated_parsed` doesn't exist. This fallback will be "
|
|
"removed in a future version of feedparser.",
|
|
DeprecationWarning)
|
|
return dict.__getitem__(self, 'published_parsed')
|
|
return dict.__getitem__(self, 'updated_parsed')
|
|
else:
|
|
realkey = self.keymap.get(key, key)
|
|
if isinstance(realkey, list):
|
|
for k in realkey:
|
|
if dict.__contains__(self, k):
|
|
return dict.__getitem__(self, k)
|
|
elif dict.__contains__(self, realkey):
|
|
return dict.__getitem__(self, realkey)
|
|
return dict.__getitem__(self, key)
|
|
|
|
def __contains__(self, key):
|
|
if key in ('updated', 'updated_parsed'):
|
|
# Temporarily help developers out by keeping the old
|
|
# broken behavior that was reported in issue 310.
|
|
# This fix was proposed in issue 328.
|
|
return dict.__contains__(self, key)
|
|
try:
|
|
self.__getitem__(key)
|
|
except KeyError:
|
|
return False
|
|
else:
|
|
return True
|
|
|
|
has_key = __contains__
|
|
|
|
def get(self, key, default=None):
|
|
try:
|
|
return self.__getitem__(key)
|
|
except KeyError:
|
|
return default
|
|
|
|
def __setitem__(self, key, value):
|
|
key = self.keymap.get(key, key)
|
|
if isinstance(key, list):
|
|
key = key[0]
|
|
return dict.__setitem__(self, key, value)
|
|
|
|
def setdefault(self, key, value):
|
|
if key not in self:
|
|
self[key] = value
|
|
return value
|
|
return self[key]
|
|
|
|
def __getattr__(self, key):
|
|
# __getattribute__() is called first; this will be called
|
|
# only if an attribute was not already found
|
|
try:
|
|
return self.__getitem__(key)
|
|
except KeyError:
|
|
raise AttributeError, "object has no attribute '%s'" % key
|
|
|
|
def __hash__(self):
|
|
return id(self)
|
|
|
|
_cp1252 = {
|
|
128: unichr(8364), # euro sign
|
|
130: unichr(8218), # single low-9 quotation mark
|
|
131: unichr( 402), # latin small letter f with hook
|
|
132: unichr(8222), # double low-9 quotation mark
|
|
133: unichr(8230), # horizontal ellipsis
|
|
134: unichr(8224), # dagger
|
|
135: unichr(8225), # double dagger
|
|
136: unichr( 710), # modifier letter circumflex accent
|
|
137: unichr(8240), # per mille sign
|
|
138: unichr( 352), # latin capital letter s with caron
|
|
139: unichr(8249), # single left-pointing angle quotation mark
|
|
140: unichr( 338), # latin capital ligature oe
|
|
142: unichr( 381), # latin capital letter z with caron
|
|
145: unichr(8216), # left single quotation mark
|
|
146: unichr(8217), # right single quotation mark
|
|
147: unichr(8220), # left double quotation mark
|
|
148: unichr(8221), # right double quotation mark
|
|
149: unichr(8226), # bullet
|
|
150: unichr(8211), # en dash
|
|
151: unichr(8212), # em dash
|
|
152: unichr( 732), # small tilde
|
|
153: unichr(8482), # trade mark sign
|
|
154: unichr( 353), # latin small letter s with caron
|
|
155: unichr(8250), # single right-pointing angle quotation mark
|
|
156: unichr( 339), # latin small ligature oe
|
|
158: unichr( 382), # latin small letter z with caron
|
|
159: unichr( 376), # latin capital letter y with diaeresis
|
|
}
|
|
|
|
_urifixer = re.compile('^([A-Za-z][A-Za-z0-9+-.]*://)(/*)(.*?)')
|
|
def _urljoin(base, uri):
|
|
uri = _urifixer.sub(r'\1\3', uri)
|
|
#try:
|
|
if not isinstance(uri, unicode):
|
|
uri = uri.decode('utf-8', 'ignore')
|
|
uri = urlparse.urljoin(base, uri)
|
|
if not isinstance(uri, unicode):
|
|
return uri.decode('utf-8', 'ignore')
|
|
return uri
|
|
#except:
|
|
# uri = urlparse.urlunparse([urllib.quote(part) for part in urlparse.urlparse(uri)])
|
|
# return urlparse.urljoin(base, uri)
|
|
|
|
class _FeedParserMixin:
|
|
namespaces = {
|
|
'': '',
|
|
'http://backend.userland.com/rss': '',
|
|
'http://blogs.law.harvard.edu/tech/rss': '',
|
|
'http://purl.org/rss/1.0/': '',
|
|
'http://my.netscape.com/rdf/simple/0.9/': '',
|
|
'http://example.com/newformat#': '',
|
|
'http://example.com/necho': '',
|
|
'http://purl.org/echo/': '',
|
|
'uri/of/echo/namespace#': '',
|
|
'http://purl.org/pie/': '',
|
|
'http://purl.org/atom/ns#': '',
|
|
'http://www.w3.org/2005/Atom': '',
|
|
'http://purl.org/rss/1.0/modules/rss091#': '',
|
|
|
|
'http://webns.net/mvcb/': 'admin',
|
|
'http://purl.org/rss/1.0/modules/aggregation/': 'ag',
|
|
'http://purl.org/rss/1.0/modules/annotate/': 'annotate',
|
|
'http://media.tangent.org/rss/1.0/': 'audio',
|
|
'http://backend.userland.com/blogChannelModule': 'blogChannel',
|
|
'http://web.resource.org/cc/': 'cc',
|
|
'http://backend.userland.com/creativeCommonsRssModule': 'creativeCommons',
|
|
'http://purl.org/rss/1.0/modules/company': 'co',
|
|
'http://purl.org/rss/1.0/modules/content/': 'content',
|
|
'http://my.theinfo.org/changed/1.0/rss/': 'cp',
|
|
'http://purl.org/dc/elements/1.1/': 'dc',
|
|
'http://purl.org/dc/terms/': 'dcterms',
|
|
'http://purl.org/rss/1.0/modules/email/': 'email',
|
|
'http://purl.org/rss/1.0/modules/event/': 'ev',
|
|
'http://rssnamespace.org/feedburner/ext/1.0': 'feedburner',
|
|
'http://freshmeat.net/rss/fm/': 'fm',
|
|
'http://xmlns.com/foaf/0.1/': 'foaf',
|
|
'http://www.w3.org/2003/01/geo/wgs84_pos#': 'geo',
|
|
'http://postneo.com/icbm/': 'icbm',
|
|
'http://purl.org/rss/1.0/modules/image/': 'image',
|
|
'http://www.itunes.com/DTDs/PodCast-1.0.dtd': 'itunes',
|
|
'http://example.com/DTDs/PodCast-1.0.dtd': 'itunes',
|
|
'http://purl.org/rss/1.0/modules/link/': 'l',
|
|
'http://search.yahoo.com/mrss': 'media',
|
|
# Version 1.1.2 of the Media RSS spec added the trailing slash on the namespace
|
|
'http://search.yahoo.com/mrss/': 'media',
|
|
'http://madskills.com/public/xml/rss/module/pingback/': 'pingback',
|
|
'http://prismstandard.org/namespaces/1.2/basic/': 'prism',
|
|
'http://www.w3.org/1999/02/22-rdf-syntax-ns#': 'rdf',
|
|
'http://www.w3.org/2000/01/rdf-schema#': 'rdfs',
|
|
'http://purl.org/rss/1.0/modules/reference/': 'ref',
|
|
'http://purl.org/rss/1.0/modules/richequiv/': 'reqv',
|
|
'http://purl.org/rss/1.0/modules/search/': 'search',
|
|
'http://purl.org/rss/1.0/modules/slash/': 'slash',
|
|
'http://schemas.xmlsoap.org/soap/envelope/': 'soap',
|
|
'http://purl.org/rss/1.0/modules/servicestatus/': 'ss',
|
|
'http://hacks.benhammersley.com/rss/streaming/': 'str',
|
|
'http://purl.org/rss/1.0/modules/subscription/': 'sub',
|
|
'http://purl.org/rss/1.0/modules/syndication/': 'sy',
|
|
'http://schemas.pocketsoap.com/rss/myDescModule/': 'szf',
|
|
'http://purl.org/rss/1.0/modules/taxonomy/': 'taxo',
|
|
'http://purl.org/rss/1.0/modules/threading/': 'thr',
|
|
'http://purl.org/rss/1.0/modules/textinput/': 'ti',
|
|
'http://madskills.com/public/xml/rss/module/trackback/': 'trackback',
|
|
'http://wellformedweb.org/commentAPI/': 'wfw',
|
|
'http://purl.org/rss/1.0/modules/wiki/': 'wiki',
|
|
'http://www.w3.org/1999/xhtml': 'xhtml',
|
|
'http://www.w3.org/1999/xlink': 'xlink',
|
|
'http://www.w3.org/XML/1998/namespace': 'xml',
|
|
}
|
|
_matchnamespaces = {}
|
|
|
|
can_be_relative_uri = set(['link', 'id', 'wfw_comment', 'wfw_commentrss', 'docs', 'url', 'href', 'comments', 'icon', 'logo'])
|
|
can_contain_relative_uris = set(['content', 'title', 'summary', 'info', 'tagline', 'subtitle', 'copyright', 'rights', 'description'])
|
|
can_contain_dangerous_markup = set(['content', 'title', 'summary', 'info', 'tagline', 'subtitle', 'copyright', 'rights', 'description'])
|
|
html_types = [u'text/html', u'application/xhtml+xml']
|
|
|
|
def __init__(self, baseuri=None, baselang=None, encoding=u'utf-8'):
|
|
if not self._matchnamespaces:
|
|
for k, v in self.namespaces.items():
|
|
self._matchnamespaces[k.lower()] = v
|
|
self.feeddata = FeedParserDict() # feed-level data
|
|
self.encoding = encoding # character encoding
|
|
self.entries = [] # list of entry-level data
|
|
self.version = u'' # feed type/version, see SUPPORTED_VERSIONS
|
|
self.namespacesInUse = {} # dictionary of namespaces defined by the feed
|
|
|
|
# the following are used internally to track state;
|
|
# this is really out of control and should be refactored
|
|
self.infeed = 0
|
|
self.inentry = 0
|
|
self.incontent = 0
|
|
self.intextinput = 0
|
|
self.inimage = 0
|
|
self.inauthor = 0
|
|
self.incontributor = 0
|
|
self.inpublisher = 0
|
|
self.insource = 0
|
|
self.sourcedata = FeedParserDict()
|
|
self.contentparams = FeedParserDict()
|
|
self._summaryKey = None
|
|
self.namespacemap = {}
|
|
self.elementstack = []
|
|
self.basestack = []
|
|
self.langstack = []
|
|
self.baseuri = baseuri or u''
|
|
self.lang = baselang or None
|
|
self.svgOK = 0
|
|
self.title_depth = -1
|
|
self.depth = 0
|
|
if baselang:
|
|
self.feeddata['language'] = baselang.replace('_','-')
|
|
|
|
# A map of the following form:
|
|
# {
|
|
# object_that_value_is_set_on: {
|
|
# property_name: depth_of_node_property_was_extracted_from,
|
|
# other_property: depth_of_node_property_was_extracted_from,
|
|
# },
|
|
# }
|
|
self.property_depth_map = {}
|
|
|
|
def _normalize_attributes(self, kv):
|
|
k = kv[0].lower()
|
|
v = k in ('rel', 'type') and kv[1].lower() or kv[1]
|
|
# the sgml parser doesn't handle entities in attributes, nor
|
|
# does it pass the attribute values through as unicode, while
|
|
# strict xml parsers do -- account for this difference
|
|
if isinstance(self, _LooseFeedParser):
|
|
v = v.replace('&', '&')
|
|
if not isinstance(v, unicode):
|
|
v = v.decode('utf-8')
|
|
return (k, v)
|
|
|
|
def unknown_starttag(self, tag, attrs):
|
|
# increment depth counter
|
|
self.depth += 1
|
|
|
|
# normalize attrs
|
|
attrs = map(self._normalize_attributes, attrs)
|
|
|
|
# track xml:base and xml:lang
|
|
attrsD = dict(attrs)
|
|
baseuri = attrsD.get('xml:base', attrsD.get('base')) or self.baseuri
|
|
if not isinstance(baseuri, unicode):
|
|
baseuri = baseuri.decode(self.encoding, 'ignore')
|
|
# ensure that self.baseuri is always an absolute URI that
|
|
# uses a whitelisted URI scheme (e.g. not `javscript:`)
|
|
if self.baseuri:
|
|
self.baseuri = _makeSafeAbsoluteURI(self.baseuri, baseuri) or self.baseuri
|
|
else:
|
|
self.baseuri = _urljoin(self.baseuri, baseuri)
|
|
lang = attrsD.get('xml:lang', attrsD.get('lang'))
|
|
if lang == '':
|
|
# xml:lang could be explicitly set to '', we need to capture that
|
|
lang = None
|
|
elif lang is None:
|
|
# if no xml:lang is specified, use parent lang
|
|
lang = self.lang
|
|
if lang:
|
|
if tag in ('feed', 'rss', 'rdf:RDF'):
|
|
self.feeddata['language'] = lang.replace('_','-')
|
|
self.lang = lang
|
|
self.basestack.append(self.baseuri)
|
|
self.langstack.append(lang)
|
|
|
|
# track namespaces
|
|
for prefix, uri in attrs:
|
|
if prefix.startswith('xmlns:'):
|
|
self.trackNamespace(prefix[6:], uri)
|
|
elif prefix == 'xmlns':
|
|
self.trackNamespace(None, uri)
|
|
|
|
# track inline content
|
|
if self.incontent and not self.contentparams.get('type', u'xml').endswith(u'xml'):
|
|
if tag in ('xhtml:div', 'div'):
|
|
return # typepad does this 10/2007
|
|
# element declared itself as escaped markup, but it isn't really
|
|
self.contentparams['type'] = u'application/xhtml+xml'
|
|
if self.incontent and self.contentparams.get('type') == u'application/xhtml+xml':
|
|
if tag.find(':') <> -1:
|
|
prefix, tag = tag.split(':', 1)
|
|
namespace = self.namespacesInUse.get(prefix, '')
|
|
if tag=='math' and namespace=='http://www.w3.org/1998/Math/MathML':
|
|
attrs.append(('xmlns',namespace))
|
|
if tag=='svg' and namespace=='http://www.w3.org/2000/svg':
|
|
attrs.append(('xmlns',namespace))
|
|
if tag == 'svg':
|
|
self.svgOK += 1
|
|
return self.handle_data('<%s%s>' % (tag, self.strattrs(attrs)), escape=0)
|
|
|
|
# match namespaces
|
|
if tag.find(':') <> -1:
|
|
prefix, suffix = tag.split(':', 1)
|
|
else:
|
|
prefix, suffix = '', tag
|
|
prefix = self.namespacemap.get(prefix, prefix)
|
|
if prefix:
|
|
prefix = prefix + '_'
|
|
|
|
# special hack for better tracking of empty textinput/image elements in illformed feeds
|
|
if (not prefix) and tag not in ('title', 'link', 'description', 'name'):
|
|
self.intextinput = 0
|
|
if (not prefix) and tag not in ('title', 'link', 'description', 'url', 'href', 'width', 'height'):
|
|
self.inimage = 0
|
|
|
|
# call special handler (if defined) or default handler
|
|
methodname = '_start_' + prefix + suffix
|
|
try:
|
|
method = getattr(self, methodname)
|
|
return method(attrsD)
|
|
except AttributeError:
|
|
# Since there's no handler or something has gone wrong we explicitly add the element and its attributes
|
|
unknown_tag = prefix + suffix
|
|
if len(attrsD) == 0:
|
|
# No attributes so merge it into the encosing dictionary
|
|
return self.push(unknown_tag, 1)
|
|
else:
|
|
# Has attributes so create it in its own dictionary
|
|
context = self._getContext()
|
|
context[unknown_tag] = attrsD
|
|
|
|
def unknown_endtag(self, tag):
|
|
# match namespaces
|
|
if tag.find(':') <> -1:
|
|
prefix, suffix = tag.split(':', 1)
|
|
else:
|
|
prefix, suffix = '', tag
|
|
prefix = self.namespacemap.get(prefix, prefix)
|
|
if prefix:
|
|
prefix = prefix + '_'
|
|
if suffix == 'svg' and self.svgOK:
|
|
self.svgOK -= 1
|
|
|
|
# call special handler (if defined) or default handler
|
|
methodname = '_end_' + prefix + suffix
|
|
try:
|
|
if self.svgOK:
|
|
raise AttributeError()
|
|
method = getattr(self, methodname)
|
|
method()
|
|
except AttributeError:
|
|
self.pop(prefix + suffix)
|
|
|
|
# track inline content
|
|
if self.incontent and not self.contentparams.get('type', u'xml').endswith(u'xml'):
|
|
# element declared itself as escaped markup, but it isn't really
|
|
if tag in ('xhtml:div', 'div'):
|
|
return # typepad does this 10/2007
|
|
self.contentparams['type'] = u'application/xhtml+xml'
|
|
if self.incontent and self.contentparams.get('type') == u'application/xhtml+xml':
|
|
tag = tag.split(':')[-1]
|
|
self.handle_data('</%s>' % tag, escape=0)
|
|
|
|
# track xml:base and xml:lang going out of scope
|
|
if self.basestack:
|
|
self.basestack.pop()
|
|
if self.basestack and self.basestack[-1]:
|
|
self.baseuri = self.basestack[-1]
|
|
if self.langstack:
|
|
self.langstack.pop()
|
|
if self.langstack: # and (self.langstack[-1] is not None):
|
|
self.lang = self.langstack[-1]
|
|
|
|
self.depth -= 1
|
|
|
|
def handle_charref(self, ref):
|
|
# called for each character reference, e.g. for ' ', ref will be '160'
|
|
if not self.elementstack:
|
|
return
|
|
ref = ref.lower()
|
|
if ref in ('34', '38', '39', '60', '62', 'x22', 'x26', 'x27', 'x3c', 'x3e'):
|
|
text = '&#%s;' % ref
|
|
else:
|
|
if ref[0] == 'x':
|
|
c = int(ref[1:], 16)
|
|
else:
|
|
c = int(ref)
|
|
text = unichr(c).encode('utf-8')
|
|
self.elementstack[-1][2].append(text)
|
|
|
|
def handle_entityref(self, ref):
|
|
# called for each entity reference, e.g. for '©', ref will be 'copy'
|
|
if not self.elementstack:
|
|
return
|
|
if ref in ('lt', 'gt', 'quot', 'amp', 'apos'):
|
|
text = '&%s;' % ref
|
|
elif ref in self.entities:
|
|
text = self.entities[ref]
|
|
if text.startswith('&#') and text.endswith(';'):
|
|
return self.handle_entityref(text)
|
|
else:
|
|
try:
|
|
name2codepoint[ref]
|
|
except KeyError:
|
|
text = '&%s;' % ref
|
|
else:
|
|
text = unichr(name2codepoint[ref]).encode('utf-8')
|
|
self.elementstack[-1][2].append(text)
|
|
|
|
def handle_data(self, text, escape=1):
|
|
# called for each block of plain text, i.e. outside of any tag and
|
|
# not containing any character or entity references
|
|
if not self.elementstack:
|
|
return
|
|
if escape and self.contentparams.get('type') == u'application/xhtml+xml':
|
|
text = _xmlescape(text)
|
|
self.elementstack[-1][2].append(text)
|
|
|
|
def handle_comment(self, text):
|
|
# called for each comment, e.g. <!-- insert message here -->
|
|
pass
|
|
|
|
def handle_pi(self, text):
|
|
# called for each processing instruction, e.g. <?instruction>
|
|
pass
|
|
|
|
def handle_decl(self, text):
|
|
pass
|
|
|
|
def parse_declaration(self, i):
|
|
# override internal declaration handler to handle CDATA blocks
|
|
if self.rawdata[i:i+9] == '<![CDATA[':
|
|
k = self.rawdata.find(']]>', i)
|
|
if k == -1:
|
|
# CDATA block began but didn't finish
|
|
k = len(self.rawdata)
|
|
return k
|
|
self.handle_data(_xmlescape(self.rawdata[i+9:k]), 0)
|
|
return k+3
|
|
else:
|
|
k = self.rawdata.find('>', i)
|
|
if k >= 0:
|
|
return k+1
|
|
else:
|
|
# We have an incomplete CDATA block.
|
|
return k
|
|
|
|
def mapContentType(self, contentType):
|
|
contentType = contentType.lower()
|
|
if contentType == 'text' or contentType == 'plain':
|
|
contentType = u'text/plain'
|
|
elif contentType == 'html':
|
|
contentType = u'text/html'
|
|
elif contentType == 'xhtml':
|
|
contentType = u'application/xhtml+xml'
|
|
return contentType
|
|
|
|
def trackNamespace(self, prefix, uri):
|
|
loweruri = uri.lower()
|
|
if not self.version:
|
|
if (prefix, loweruri) == (None, 'http://my.netscape.com/rdf/simple/0.9/'):
|
|
self.version = u'rss090'
|
|
elif loweruri == 'http://purl.org/rss/1.0/':
|
|
self.version = u'rss10'
|
|
elif loweruri == 'http://www.w3.org/2005/atom':
|
|
self.version = u'atom10'
|
|
if loweruri.find(u'backend.userland.com/rss') <> -1:
|
|
# match any backend.userland.com namespace
|
|
uri = u'http://backend.userland.com/rss'
|
|
loweruri = uri
|
|
if loweruri in self._matchnamespaces:
|
|
self.namespacemap[prefix] = self._matchnamespaces[loweruri]
|
|
self.namespacesInUse[self._matchnamespaces[loweruri]] = uri
|
|
else:
|
|
self.namespacesInUse[prefix or ''] = uri
|
|
|
|
def resolveURI(self, uri):
|
|
return _urljoin(self.baseuri or u'', uri)
|
|
|
|
def decodeEntities(self, element, data):
|
|
return data
|
|
|
|
def strattrs(self, attrs):
|
|
return ''.join([' %s="%s"' % (t[0],_xmlescape(t[1],{'"':'"'})) for t in attrs])
|
|
|
|
def push(self, element, expectingText):
|
|
self.elementstack.append([element, expectingText, []])
|
|
|
|
def pop(self, element, stripWhitespace=1):
|
|
if not self.elementstack:
|
|
return
|
|
if self.elementstack[-1][0] != element:
|
|
return
|
|
|
|
element, expectingText, pieces = self.elementstack.pop()
|
|
|
|
if self.version == u'atom10' and self.contentparams.get('type', u'text') == u'application/xhtml+xml':
|
|
# remove enclosing child element, but only if it is a <div> and
|
|
# only if all the remaining content is nested underneath it.
|
|
# This means that the divs would be retained in the following:
|
|
# <div>foo</div><div>bar</div>
|
|
while pieces and len(pieces)>1 and not pieces[-1].strip():
|
|
del pieces[-1]
|
|
while pieces and len(pieces)>1 and not pieces[0].strip():
|
|
del pieces[0]
|
|
if pieces and (pieces[0] == '<div>' or pieces[0].startswith('<div ')) and pieces[-1]=='</div>':
|
|
depth = 0
|
|
for piece in pieces[:-1]:
|
|
if piece.startswith('</'):
|
|
depth -= 1
|
|
if depth == 0:
|
|
break
|
|
elif piece.startswith('<') and not piece.endswith('/>'):
|
|
depth += 1
|
|
else:
|
|
pieces = pieces[1:-1]
|
|
|
|
# Ensure each piece is a str for Python 3
|
|
for (i, v) in enumerate(pieces):
|
|
if not isinstance(v, unicode):
|
|
pieces[i] = v.decode('utf-8')
|
|
|
|
output = u''.join(pieces)
|
|
if stripWhitespace:
|
|
output = output.strip()
|
|
if not expectingText:
|
|
return output
|
|
|
|
# decode base64 content
|
|
if base64 and self.contentparams.get('base64', 0):
|
|
try:
|
|
output = _base64decode(output)
|
|
except binascii.Error:
|
|
pass
|
|
except binascii.Incomplete:
|
|
pass
|
|
except TypeError:
|
|
# In Python 3, base64 takes and outputs bytes, not str
|
|
# This may not be the most correct way to accomplish this
|
|
output = _base64decode(output.encode('utf-8')).decode('utf-8')
|
|
|
|
# resolve relative URIs
|
|
if (element in self.can_be_relative_uri) and output:
|
|
output = self.resolveURI(output)
|
|
|
|
# decode entities within embedded markup
|
|
if not self.contentparams.get('base64', 0):
|
|
output = self.decodeEntities(element, output)
|
|
|
|
# some feed formats require consumers to guess
|
|
# whether the content is html or plain text
|
|
if not self.version.startswith(u'atom') and self.contentparams.get('type') == u'text/plain':
|
|
if self.lookslikehtml(output):
|
|
self.contentparams['type'] = u'text/html'
|
|
|
|
# remove temporary cruft from contentparams
|
|
try:
|
|
del self.contentparams['mode']
|
|
except KeyError:
|
|
pass
|
|
try:
|
|
del self.contentparams['base64']
|
|
except KeyError:
|
|
pass
|
|
|
|
is_htmlish = self.mapContentType(self.contentparams.get('type', u'text/html')) in self.html_types
|
|
# resolve relative URIs within embedded markup
|
|
if is_htmlish and RESOLVE_RELATIVE_URIS:
|
|
if element in self.can_contain_relative_uris:
|
|
output = _resolveRelativeURIs(output, self.baseuri, self.encoding, self.contentparams.get('type', u'text/html'))
|
|
|
|
# parse microformats
|
|
# (must do this before sanitizing because some microformats
|
|
# rely on elements that we sanitize)
|
|
if PARSE_MICROFORMATS and is_htmlish and element in ['content', 'description', 'summary']:
|
|
mfresults = _parseMicroformats(output, self.baseuri, self.encoding)
|
|
if mfresults:
|
|
for tag in mfresults.get('tags', []):
|
|
self._addTag(tag['term'], tag['scheme'], tag['label'])
|
|
for enclosure in mfresults.get('enclosures', []):
|
|
self._start_enclosure(enclosure)
|
|
for xfn in mfresults.get('xfn', []):
|
|
self._addXFN(xfn['relationships'], xfn['href'], xfn['name'])
|
|
vcard = mfresults.get('vcard')
|
|
if vcard:
|
|
self._getContext()['vcard'] = vcard
|
|
|
|
# sanitize embedded markup
|
|
if is_htmlish and SANITIZE_HTML:
|
|
if element in self.can_contain_dangerous_markup:
|
|
output = _sanitizeHTML(output, self.encoding, self.contentparams.get('type', u'text/html'))
|
|
|
|
if self.encoding and not isinstance(output, unicode):
|
|
output = output.decode(self.encoding, 'ignore')
|
|
|
|
# address common error where people take data that is already
|
|
# utf-8, presume that it is iso-8859-1, and re-encode it.
|
|
if self.encoding in (u'utf-8', u'utf-8_INVALID_PYTHON_3') and isinstance(output, unicode):
|
|
try:
|
|
output = output.encode('iso-8859-1').decode('utf-8')
|
|
except (UnicodeEncodeError, UnicodeDecodeError):
|
|
pass
|
|
|
|
# map win-1252 extensions to the proper code points
|
|
if isinstance(output, unicode):
|
|
output = output.translate(_cp1252)
|
|
|
|
# categories/tags/keywords/whatever are handled in _end_category
|
|
if element == 'category':
|
|
return output
|
|
|
|
if element == 'title' and -1 < self.title_depth <= self.depth:
|
|
return output
|
|
|
|
# store output in appropriate place(s)
|
|
if self.inentry and not self.insource:
|
|
if element == 'content':
|
|
self.entries[-1].setdefault(element, [])
|
|
contentparams = copy.deepcopy(self.contentparams)
|
|
contentparams['value'] = output
|
|
self.entries[-1][element].append(contentparams)
|
|
elif element == 'link':
|
|
if not self.inimage:
|
|
# query variables in urls in link elements are improperly
|
|
# converted from `?a=1&b=2` to `?a=1&b;=2` as if they're
|
|
# unhandled character references. fix this special case.
|
|
output = re.sub("&([A-Za-z0-9_]+);", "&\g<1>", output)
|
|
self.entries[-1][element] = output
|
|
if output:
|
|
self.entries[-1]['links'][-1]['href'] = output
|
|
else:
|
|
if element == 'description':
|
|
element = 'summary'
|
|
old_value_depth = self.property_depth_map.setdefault(self.entries[-1], {}).get(element)
|
|
if old_value_depth is None or self.depth <= old_value_depth:
|
|
self.property_depth_map[self.entries[-1]][element] = self.depth
|
|
self.entries[-1][element] = output
|
|
if self.incontent:
|
|
contentparams = copy.deepcopy(self.contentparams)
|
|
contentparams['value'] = output
|
|
self.entries[-1][element + '_detail'] = contentparams
|
|
elif (self.infeed or self.insource):# and (not self.intextinput) and (not self.inimage):
|
|
context = self._getContext()
|
|
if element == 'description':
|
|
element = 'subtitle'
|
|
context[element] = output
|
|
if element == 'link':
|
|
# fix query variables; see above for the explanation
|
|
output = re.sub("&([A-Za-z0-9_]+);", "&\g<1>", output)
|
|
context[element] = output
|
|
context['links'][-1]['href'] = output
|
|
elif self.incontent:
|
|
contentparams = copy.deepcopy(self.contentparams)
|
|
contentparams['value'] = output
|
|
context[element + '_detail'] = contentparams
|
|
return output
|
|
|
|
def pushContent(self, tag, attrsD, defaultContentType, expectingText):
|
|
self.incontent += 1
|
|
if self.lang:
|
|
self.lang=self.lang.replace('_','-')
|
|
self.contentparams = FeedParserDict({
|
|
'type': self.mapContentType(attrsD.get('type', defaultContentType)),
|
|
'language': self.lang,
|
|
'base': self.baseuri})
|
|
self.contentparams['base64'] = self._isBase64(attrsD, self.contentparams)
|
|
self.push(tag, expectingText)
|
|
|
|
def popContent(self, tag):
|
|
value = self.pop(tag)
|
|
self.incontent -= 1
|
|
self.contentparams.clear()
|
|
return value
|
|
|
|
# a number of elements in a number of RSS variants are nominally plain
|
|
# text, but this is routinely ignored. This is an attempt to detect
|
|
# the most common cases. As false positives often result in silent
|
|
# data loss, this function errs on the conservative side.
|
|
@staticmethod
|
|
def lookslikehtml(s):
|
|
# must have a close tag or an entity reference to qualify
|
|
if not (re.search(r'</(\w+)>',s) or re.search("&#?\w+;",s)):
|
|
return
|
|
|
|
# all tags must be in a restricted subset of valid HTML tags
|
|
if filter(lambda t: t.lower() not in _HTMLSanitizer.acceptable_elements,
|
|
re.findall(r'</?(\w+)',s)):
|
|
return
|
|
|
|
# all entities must have been defined as valid HTML entities
|
|
if filter(lambda e: e not in entitydefs.keys(), re.findall(r'&(\w+);', s)):
|
|
return
|
|
|
|
return 1
|
|
|
|
def _mapToStandardPrefix(self, name):
|
|
colonpos = name.find(':')
|
|
if colonpos <> -1:
|
|
prefix = name[:colonpos]
|
|
suffix = name[colonpos+1:]
|
|
prefix = self.namespacemap.get(prefix, prefix)
|
|
name = prefix + ':' + suffix
|
|
return name
|
|
|
|
def _getAttribute(self, attrsD, name):
|
|
return attrsD.get(self._mapToStandardPrefix(name))
|
|
|
|
def _isBase64(self, attrsD, contentparams):
|
|
if attrsD.get('mode', '') == 'base64':
|
|
return 1
|
|
if self.contentparams['type'].startswith(u'text/'):
|
|
return 0
|
|
if self.contentparams['type'].endswith(u'+xml'):
|
|
return 0
|
|
if self.contentparams['type'].endswith(u'/xml'):
|
|
return 0
|
|
return 1
|
|
|
|
def _itsAnHrefDamnIt(self, attrsD):
|
|
href = attrsD.get('url', attrsD.get('uri', attrsD.get('href', None)))
|
|
if href:
|
|
try:
|
|
del attrsD['url']
|
|
except KeyError:
|
|
pass
|
|
try:
|
|
del attrsD['uri']
|
|
except KeyError:
|
|
pass
|
|
attrsD['href'] = href
|
|
return attrsD
|
|
|
|
def _save(self, key, value, overwrite=False):
|
|
context = self._getContext()
|
|
if overwrite:
|
|
context[key] = value
|
|
else:
|
|
context.setdefault(key, value)
|
|
|
|
def _start_rss(self, attrsD):
|
|
versionmap = {'0.91': u'rss091u',
|
|
'0.92': u'rss092',
|
|
'0.93': u'rss093',
|
|
'0.94': u'rss094'}
|
|
#If we're here then this is an RSS feed.
|
|
#If we don't have a version or have a version that starts with something
|
|
#other than RSS then there's been a mistake. Correct it.
|
|
if not self.version or not self.version.startswith(u'rss'):
|
|
attr_version = attrsD.get('version', '')
|
|
version = versionmap.get(attr_version)
|
|
if version:
|
|
self.version = version
|
|
elif attr_version.startswith('2.'):
|
|
self.version = u'rss20'
|
|
else:
|
|
self.version = u'rss'
|
|
|
|
def _start_channel(self, attrsD):
|
|
self.infeed = 1
|
|
self._cdf_common(attrsD)
|
|
|
|
def _cdf_common(self, attrsD):
|
|
if 'lastmod' in attrsD:
|
|
self._start_modified({})
|
|
self.elementstack[-1][-1] = attrsD['lastmod']
|
|
self._end_modified()
|
|
if 'href' in attrsD:
|
|
self._start_link({})
|
|
self.elementstack[-1][-1] = attrsD['href']
|
|
self._end_link()
|
|
|
|
def _start_feed(self, attrsD):
|
|
self.infeed = 1
|
|
versionmap = {'0.1': u'atom01',
|
|
'0.2': u'atom02',
|
|
'0.3': u'atom03'}
|
|
if not self.version:
|
|
attr_version = attrsD.get('version')
|
|
version = versionmap.get(attr_version)
|
|
if version:
|
|
self.version = version
|
|
else:
|
|
self.version = u'atom'
|
|
|
|
def _end_channel(self):
|
|
self.infeed = 0
|
|
_end_feed = _end_channel
|
|
|
|
def _start_image(self, attrsD):
|
|
context = self._getContext()
|
|
if not self.inentry:
|
|
context.setdefault('image', FeedParserDict())
|
|
self.inimage = 1
|
|
self.title_depth = -1
|
|
self.push('image', 0)
|
|
|
|
def _end_image(self):
|
|
self.pop('image')
|
|
self.inimage = 0
|
|
|
|
def _start_textinput(self, attrsD):
|
|
context = self._getContext()
|
|
context.setdefault('textinput', FeedParserDict())
|
|
self.intextinput = 1
|
|
self.title_depth = -1
|
|
self.push('textinput', 0)
|
|
_start_textInput = _start_textinput
|
|
|
|
def _end_textinput(self):
|
|
self.pop('textinput')
|
|
self.intextinput = 0
|
|
_end_textInput = _end_textinput
|
|
|
|
def _start_author(self, attrsD):
|
|
self.inauthor = 1
|
|
self.push('author', 1)
|
|
# Append a new FeedParserDict when expecting an author
|
|
context = self._getContext()
|
|
context.setdefault('authors', [])
|
|
context['authors'].append(FeedParserDict())
|
|
_start_managingeditor = _start_author
|
|
_start_dc_author = _start_author
|
|
_start_dc_creator = _start_author
|
|
_start_itunes_author = _start_author
|
|
|
|
def _end_author(self):
|
|
self.pop('author')
|
|
self.inauthor = 0
|
|
self._sync_author_detail()
|
|
_end_managingeditor = _end_author
|
|
_end_dc_author = _end_author
|
|
_end_dc_creator = _end_author
|
|
_end_itunes_author = _end_author
|
|
|
|
def _start_itunes_owner(self, attrsD):
|
|
self.inpublisher = 1
|
|
self.push('publisher', 0)
|
|
|
|
def _end_itunes_owner(self):
|
|
self.pop('publisher')
|
|
self.inpublisher = 0
|
|
self._sync_author_detail('publisher')
|
|
|
|
def _start_contributor(self, attrsD):
|
|
self.incontributor = 1
|
|
context = self._getContext()
|
|
context.setdefault('contributors', [])
|
|
context['contributors'].append(FeedParserDict())
|
|
self.push('contributor', 0)
|
|
|
|
def _end_contributor(self):
|
|
self.pop('contributor')
|
|
self.incontributor = 0
|
|
|
|
def _start_dc_contributor(self, attrsD):
|
|
self.incontributor = 1
|
|
context = self._getContext()
|
|
context.setdefault('contributors', [])
|
|
context['contributors'].append(FeedParserDict())
|
|
self.push('name', 0)
|
|
|
|
def _end_dc_contributor(self):
|
|
self._end_name()
|
|
self.incontributor = 0
|
|
|
|
def _start_name(self, attrsD):
|
|
self.push('name', 0)
|
|
_start_itunes_name = _start_name
|
|
|
|
def _end_name(self):
|
|
value = self.pop('name')
|
|
if self.inpublisher:
|
|
self._save_author('name', value, 'publisher')
|
|
elif self.inauthor:
|
|
self._save_author('name', value)
|
|
elif self.incontributor:
|
|
self._save_contributor('name', value)
|
|
elif self.intextinput:
|
|
context = self._getContext()
|
|
context['name'] = value
|
|
_end_itunes_name = _end_name
|
|
|
|
def _start_width(self, attrsD):
|
|
self.push('width', 0)
|
|
|
|
def _end_width(self):
|
|
value = self.pop('width')
|
|
try:
|
|
value = int(value)
|
|
except ValueError:
|
|
value = 0
|
|
if self.inimage:
|
|
context = self._getContext()
|
|
context['width'] = value
|
|
|
|
def _start_height(self, attrsD):
|
|
self.push('height', 0)
|
|
|
|
def _end_height(self):
|
|
value = self.pop('height')
|
|
try:
|
|
value = int(value)
|
|
except ValueError:
|
|
value = 0
|
|
if self.inimage:
|
|
context = self._getContext()
|
|
context['height'] = value
|
|
|
|
def _start_url(self, attrsD):
|
|
self.push('href', 1)
|
|
_start_homepage = _start_url
|
|
_start_uri = _start_url
|
|
|
|
def _end_url(self):
|
|
value = self.pop('href')
|
|
if self.inauthor:
|
|
self._save_author('href', value)
|
|
elif self.incontributor:
|
|
self._save_contributor('href', value)
|
|
_end_homepage = _end_url
|
|
_end_uri = _end_url
|
|
|
|
def _start_email(self, attrsD):
|
|
self.push('email', 0)
|
|
_start_itunes_email = _start_email
|
|
|
|
def _end_email(self):
|
|
value = self.pop('email')
|
|
if self.inpublisher:
|
|
self._save_author('email', value, 'publisher')
|
|
elif self.inauthor:
|
|
self._save_author('email', value)
|
|
elif self.incontributor:
|
|
self._save_contributor('email', value)
|
|
_end_itunes_email = _end_email
|
|
|
|
def _getContext(self):
|
|
if self.insource:
|
|
context = self.sourcedata
|
|
elif self.inimage and 'image' in self.feeddata:
|
|
context = self.feeddata['image']
|
|
elif self.intextinput:
|
|
context = self.feeddata['textinput']
|
|
elif self.inentry:
|
|
context = self.entries[-1]
|
|
else:
|
|
context = self.feeddata
|
|
return context
|
|
|
|
def _save_author(self, key, value, prefix='author'):
|
|
context = self._getContext()
|
|
context.setdefault(prefix + '_detail', FeedParserDict())
|
|
context[prefix + '_detail'][key] = value
|
|
self._sync_author_detail()
|
|
context.setdefault('authors', [FeedParserDict()])
|
|
context['authors'][-1][key] = value
|
|
|
|
def _save_contributor(self, key, value):
|
|
context = self._getContext()
|
|
context.setdefault('contributors', [FeedParserDict()])
|
|
context['contributors'][-1][key] = value
|
|
|
|
def _sync_author_detail(self, key='author'):
|
|
context = self._getContext()
|
|
detail = context.get('%s_detail' % key)
|
|
if detail:
|
|
name = detail.get('name')
|
|
email = detail.get('email')
|
|
if name and email:
|
|
context[key] = u'%s (%s)' % (name, email)
|
|
elif name:
|
|
context[key] = name
|
|
elif email:
|
|
context[key] = email
|
|
else:
|
|
author, email = context.get(key), None
|
|
if not author:
|
|
return
|
|
emailmatch = re.search(ur'''(([a-zA-Z0-9\_\-\.\+]+)@((\[[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.)|(([a-zA-Z0-9\-]+\.)+))([a-zA-Z]{2,4}|[0-9]{1,3})(\]?))(\?subject=\S+)?''', author)
|
|
if emailmatch:
|
|
email = emailmatch.group(0)
|
|
# probably a better way to do the following, but it passes all the tests
|
|
author = author.replace(email, u'')
|
|
author = author.replace(u'()', u'')
|
|
author = author.replace(u'<>', u'')
|
|
author = author.replace(u'<>', u'')
|
|
author = author.strip()
|
|
if author and (author[0] == u'('):
|
|
author = author[1:]
|
|
if author and (author[-1] == u')'):
|
|
author = author[:-1]
|
|
author = author.strip()
|
|
if author or email:
|
|
context.setdefault('%s_detail' % key, FeedParserDict())
|
|
if author:
|
|
context['%s_detail' % key]['name'] = author
|
|
if email:
|
|
context['%s_detail' % key]['email'] = email
|
|
|
|
def _start_subtitle(self, attrsD):
|
|
self.pushContent('subtitle', attrsD, u'text/plain', 1)
|
|
_start_tagline = _start_subtitle
|
|
_start_itunes_subtitle = _start_subtitle
|
|
|
|
def _end_subtitle(self):
|
|
self.popContent('subtitle')
|
|
_end_tagline = _end_subtitle
|
|
_end_itunes_subtitle = _end_subtitle
|
|
|
|
def _start_rights(self, attrsD):
|
|
self.pushContent('rights', attrsD, u'text/plain', 1)
|
|
_start_dc_rights = _start_rights
|
|
_start_copyright = _start_rights
|
|
|
|
def _end_rights(self):
|
|
self.popContent('rights')
|
|
_end_dc_rights = _end_rights
|
|
_end_copyright = _end_rights
|
|
|
|
def _start_item(self, attrsD):
|
|
self.entries.append(FeedParserDict())
|
|
self.push('item', 0)
|
|
self.inentry = 1
|
|
self.guidislink = 0
|
|
self.title_depth = -1
|
|
id = self._getAttribute(attrsD, 'rdf:about')
|
|
if id:
|
|
context = self._getContext()
|
|
context['id'] = id
|
|
self._cdf_common(attrsD)
|
|
_start_entry = _start_item
|
|
|
|
def _end_item(self):
|
|
self.pop('item')
|
|
self.inentry = 0
|
|
_end_entry = _end_item
|
|
|
|
def _start_dc_language(self, attrsD):
|
|
self.push('language', 1)
|
|
_start_language = _start_dc_language
|
|
|
|
def _end_dc_language(self):
|
|
self.lang = self.pop('language')
|
|
_end_language = _end_dc_language
|
|
|
|
def _start_dc_publisher(self, attrsD):
|
|
self.push('publisher', 1)
|
|
_start_webmaster = _start_dc_publisher
|
|
|
|
def _end_dc_publisher(self):
|
|
self.pop('publisher')
|
|
self._sync_author_detail('publisher')
|
|
_end_webmaster = _end_dc_publisher
|
|
|
|
def _start_published(self, attrsD):
|
|
self.push('published', 1)
|
|
_start_dcterms_issued = _start_published
|
|
_start_issued = _start_published
|
|
_start_pubdate = _start_published
|
|
|
|
def _end_published(self):
|
|
value = self.pop('published')
|
|
self._save('published_parsed', _parse_date(value), overwrite=True)
|
|
_end_dcterms_issued = _end_published
|
|
_end_issued = _end_published
|
|
_end_pubdate = _end_published
|
|
|
|
def _start_updated(self, attrsD):
|
|
self.push('updated', 1)
|
|
_start_modified = _start_updated
|
|
_start_dcterms_modified = _start_updated
|
|
_start_dc_date = _start_updated
|
|
_start_lastbuilddate = _start_updated
|
|
|
|
def _end_updated(self):
|
|
value = self.pop('updated')
|
|
parsed_value = _parse_date(value)
|
|
self._save('updated_parsed', parsed_value, overwrite=True)
|
|
_end_modified = _end_updated
|
|
_end_dcterms_modified = _end_updated
|
|
_end_dc_date = _end_updated
|
|
_end_lastbuilddate = _end_updated
|
|
|
|
def _start_created(self, attrsD):
|
|
self.push('created', 1)
|
|
_start_dcterms_created = _start_created
|
|
|
|
def _end_created(self):
|
|
value = self.pop('created')
|
|
self._save('created_parsed', _parse_date(value), overwrite=True)
|
|
_end_dcterms_created = _end_created
|
|
|
|
def _start_expirationdate(self, attrsD):
|
|
self.push('expired', 1)
|
|
|
|
def _end_expirationdate(self):
|
|
self._save('expired_parsed', _parse_date(self.pop('expired')), overwrite=True)
|
|
|
|
def _start_cc_license(self, attrsD):
|
|
context = self._getContext()
|
|
value = self._getAttribute(attrsD, 'rdf:resource')
|
|
attrsD = FeedParserDict()
|
|
attrsD['rel'] = u'license'
|
|
if value:
|
|
attrsD['href']=value
|
|
context.setdefault('links', []).append(attrsD)
|
|
|
|
def _start_creativecommons_license(self, attrsD):
|
|
self.push('license', 1)
|
|
_start_creativeCommons_license = _start_creativecommons_license
|
|
|
|
def _end_creativecommons_license(self):
|
|
value = self.pop('license')
|
|
context = self._getContext()
|
|
attrsD = FeedParserDict()
|
|
attrsD['rel'] = u'license'
|
|
if value:
|
|
attrsD['href'] = value
|
|
context.setdefault('links', []).append(attrsD)
|
|
del context['license']
|
|
_end_creativeCommons_license = _end_creativecommons_license
|
|
|
|
def _addXFN(self, relationships, href, name):
|
|
context = self._getContext()
|
|
xfn = context.setdefault('xfn', [])
|
|
value = FeedParserDict({'relationships': relationships, 'href': href, 'name': name})
|
|
if value not in xfn:
|
|
xfn.append(value)
|
|
|
|
def _addTag(self, term, scheme, label):
|
|
context = self._getContext()
|
|
tags = context.setdefault('tags', [])
|
|
if (not term) and (not scheme) and (not label):
|
|
return
|
|
value = FeedParserDict({'term': term, 'scheme': scheme, 'label': label})
|
|
if value not in tags:
|
|
tags.append(value)
|
|
|
|
def _start_category(self, attrsD):
|
|
term = attrsD.get('term')
|
|
scheme = attrsD.get('scheme', attrsD.get('domain'))
|
|
label = attrsD.get('label')
|
|
self._addTag(term, scheme, label)
|
|
self.push('category', 1)
|
|
_start_dc_subject = _start_category
|
|
_start_keywords = _start_category
|
|
|
|
def _start_media_category(self, attrsD):
|
|
attrsD.setdefault('scheme', u'http://search.yahoo.com/mrss/category_schema')
|
|
self._start_category(attrsD)
|
|
|
|
def _end_itunes_keywords(self):
|
|
for term in self.pop('itunes_keywords').split(','):
|
|
if term.strip():
|
|
self._addTag(term.strip(), u'http://www.itunes.com/', None)
|
|
|
|
def _start_itunes_category(self, attrsD):
|
|
self._addTag(attrsD.get('text'), u'http://www.itunes.com/', None)
|
|
self.push('category', 1)
|
|
|
|
def _end_category(self):
|
|
value = self.pop('category')
|
|
if not value:
|
|
return
|
|
context = self._getContext()
|
|
tags = context['tags']
|
|
if value and len(tags) and not tags[-1]['term']:
|
|
tags[-1]['term'] = value
|
|
else:
|
|
self._addTag(value, None, None)
|
|
_end_dc_subject = _end_category
|
|
_end_keywords = _end_category
|
|
_end_itunes_category = _end_category
|
|
_end_media_category = _end_category
|
|
|
|
def _start_cloud(self, attrsD):
|
|
self._getContext()['cloud'] = FeedParserDict(attrsD)
|
|
|
|
def _start_link(self, attrsD):
|
|
attrsD.setdefault('rel', u'alternate')
|
|
if attrsD['rel'] == u'self':
|
|
attrsD.setdefault('type', u'application/atom+xml')
|
|
else:
|
|
attrsD.setdefault('type', u'text/html')
|
|
context = self._getContext()
|
|
attrsD = self._itsAnHrefDamnIt(attrsD)
|
|
if 'href' in attrsD:
|
|
attrsD['href'] = self.resolveURI(attrsD['href'])
|
|
expectingText = self.infeed or self.inentry or self.insource
|
|
context.setdefault('links', [])
|
|
if not (self.inentry and self.inimage):
|
|
context['links'].append(FeedParserDict(attrsD))
|
|
if 'href' in attrsD:
|
|
expectingText = 0
|
|
if (attrsD.get('rel') == u'alternate') and (self.mapContentType(attrsD.get('type')) in self.html_types):
|
|
context['link'] = attrsD['href']
|
|
else:
|
|
self.push('link', expectingText)
|
|
|
|
def _end_link(self):
|
|
value = self.pop('link')
|
|
|
|
def _start_guid(self, attrsD):
|
|
self.guidislink = (attrsD.get('ispermalink', 'true') == 'true')
|
|
self.push('id', 1)
|
|
_start_id = _start_guid
|
|
|
|
def _end_guid(self):
|
|
value = self.pop('id')
|
|
self._save('guidislink', self.guidislink and 'link' not in self._getContext())
|
|
if self.guidislink:
|
|
# guid acts as link, but only if 'ispermalink' is not present or is 'true',
|
|
# and only if the item doesn't already have a link element
|
|
self._save('link', value)
|
|
_end_id = _end_guid
|
|
|
|
def _start_title(self, attrsD):
|
|
if self.svgOK:
|
|
return self.unknown_starttag('title', attrsD.items())
|
|
self.pushContent('title', attrsD, u'text/plain', self.infeed or self.inentry or self.insource)
|
|
_start_dc_title = _start_title
|
|
_start_media_title = _start_title
|
|
|
|
def _end_title(self):
|
|
if self.svgOK:
|
|
return
|
|
value = self.popContent('title')
|
|
if not value:
|
|
return
|
|
self.title_depth = self.depth
|
|
_end_dc_title = _end_title
|
|
|
|
def _end_media_title(self):
|
|
title_depth = self.title_depth
|
|
self._end_title()
|
|
self.title_depth = title_depth
|
|
|
|
def _start_description(self, attrsD):
|
|
context = self._getContext()
|
|
if 'summary' in context:
|
|
self._summaryKey = 'content'
|
|
self._start_content(attrsD)
|
|
else:
|
|
self.pushContent('description', attrsD, u'text/html', self.infeed or self.inentry or self.insource)
|
|
_start_dc_description = _start_description
|
|
|
|
def _start_abstract(self, attrsD):
|
|
self.pushContent('description', attrsD, u'text/plain', self.infeed or self.inentry or self.insource)
|
|
|
|
def _end_description(self):
|
|
if self._summaryKey == 'content':
|
|
self._end_content()
|
|
else:
|
|
value = self.popContent('description')
|
|
self._summaryKey = None
|
|
_end_abstract = _end_description
|
|
_end_dc_description = _end_description
|
|
|
|
def _start_info(self, attrsD):
|
|
self.pushContent('info', attrsD, u'text/plain', 1)
|
|
_start_feedburner_browserfriendly = _start_info
|
|
|
|
def _end_info(self):
|
|
self.popContent('info')
|
|
_end_feedburner_browserfriendly = _end_info
|
|
|
|
def _start_generator(self, attrsD):
|
|
if attrsD:
|
|
attrsD = self._itsAnHrefDamnIt(attrsD)
|
|
if 'href' in attrsD:
|
|
attrsD['href'] = self.resolveURI(attrsD['href'])
|
|
self._getContext()['generator_detail'] = FeedParserDict(attrsD)
|
|
self.push('generator', 1)
|
|
|
|
def _end_generator(self):
|
|
value = self.pop('generator')
|
|
context = self._getContext()
|
|
if 'generator_detail' in context:
|
|
context['generator_detail']['name'] = value
|
|
|
|
def _start_admin_generatoragent(self, attrsD):
|
|
self.push('generator', 1)
|
|
value = self._getAttribute(attrsD, 'rdf:resource')
|
|
if value:
|
|
self.elementstack[-1][2].append(value)
|
|
self.pop('generator')
|
|
self._getContext()['generator_detail'] = FeedParserDict({'href': value})
|
|
|
|
def _start_admin_errorreportsto(self, attrsD):
|
|
self.push('errorreportsto', 1)
|
|
value = self._getAttribute(attrsD, 'rdf:resource')
|
|
if value:
|
|
self.elementstack[-1][2].append(value)
|
|
self.pop('errorreportsto')
|
|
|
|
def _start_summary(self, attrsD):
|
|
context = self._getContext()
|
|
if 'summary' in context:
|
|
self._summaryKey = 'content'
|
|
self._start_content(attrsD)
|
|
else:
|
|
self._summaryKey = 'summary'
|
|
self.pushContent(self._summaryKey, attrsD, u'text/plain', 1)
|
|
_start_itunes_summary = _start_summary
|
|
|
|
def _end_summary(self):
|
|
if self._summaryKey == 'content':
|
|
self._end_content()
|
|
else:
|
|
self.popContent(self._summaryKey or 'summary')
|
|
self._summaryKey = None
|
|
_end_itunes_summary = _end_summary
|
|
|
|
def _start_enclosure(self, attrsD):
|
|
attrsD = self._itsAnHrefDamnIt(attrsD)
|
|
context = self._getContext()
|
|
attrsD['rel'] = u'enclosure'
|
|
context.setdefault('links', []).append(FeedParserDict(attrsD))
|
|
|
|
def _start_source(self, attrsD):
|
|
if 'url' in attrsD:
|
|
# This means that we're processing a source element from an RSS 2.0 feed
|
|
self.sourcedata['href'] = attrsD[u'url']
|
|
self.push('source', 1)
|
|
self.insource = 1
|
|
self.title_depth = -1
|
|
|
|
def _end_source(self):
|
|
self.insource = 0
|
|
value = self.pop('source')
|
|
if value:
|
|
self.sourcedata['title'] = value
|
|
self._getContext()['source'] = copy.deepcopy(self.sourcedata)
|
|
self.sourcedata.clear()
|
|
|
|
def _start_content(self, attrsD):
|
|
self.pushContent('content', attrsD, u'text/plain', 1)
|
|
src = attrsD.get('src')
|
|
if src:
|
|
self.contentparams['src'] = src
|
|
self.push('content', 1)
|
|
|
|
def _start_body(self, attrsD):
|
|
self.pushContent('content', attrsD, u'application/xhtml+xml', 1)
|
|
_start_xhtml_body = _start_body
|
|
|
|
def _start_content_encoded(self, attrsD):
|
|
self.pushContent('content', attrsD, u'text/html', 1)
|
|
_start_fullitem = _start_content_encoded
|
|
|
|
def _end_content(self):
|
|
copyToSummary = self.mapContentType(self.contentparams.get('type')) in ([u'text/plain'] + self.html_types)
|
|
value = self.popContent('content')
|
|
if copyToSummary:
|
|
self._save('summary', value)
|
|
|
|
_end_body = _end_content
|
|
_end_xhtml_body = _end_content
|
|
_end_content_encoded = _end_content
|
|
_end_fullitem = _end_content
|
|
|
|
def _start_itunes_image(self, attrsD):
|
|
self.push('itunes_image', 0)
|
|
if attrsD.get('href'):
|
|
self._getContext()['image'] = FeedParserDict({'href': attrsD.get('href')})
|
|
elif attrsD.get('url'):
|
|
self._getContext()['image'] = FeedParserDict({'href': attrsD.get('url')})
|
|
_start_itunes_link = _start_itunes_image
|
|
|
|
def _end_itunes_block(self):
|
|
value = self.pop('itunes_block', 0)
|
|
self._getContext()['itunes_block'] = (value == 'yes') and 1 or 0
|
|
|
|
def _end_itunes_explicit(self):
|
|
value = self.pop('itunes_explicit', 0)
|
|
# Convert 'yes' -> True, 'clean' to False, and any other value to None
|
|
# False and None both evaluate as False, so the difference can be ignored
|
|
# by applications that only need to know if the content is explicit.
|
|
self._getContext()['itunes_explicit'] = (None, False, True)[(value == 'yes' and 2) or value == 'clean' or 0]
|
|
|
|
def _start_media_content(self, attrsD):
|
|
context = self._getContext()
|
|
context.setdefault('media_content', [])
|
|
context['media_content'].append(attrsD)
|
|
|
|
def _start_media_thumbnail(self, attrsD):
|
|
context = self._getContext()
|
|
context.setdefault('media_thumbnail', [])
|
|
self.push('url', 1) # new
|
|
context['media_thumbnail'].append(attrsD)
|
|
|
|
def _end_media_thumbnail(self):
|
|
url = self.pop('url')
|
|
context = self._getContext()
|
|
if url != None and len(url.strip()) != 0:
|
|
if 'url' not in context['media_thumbnail'][-1]:
|
|
context['media_thumbnail'][-1]['url'] = url
|
|
|
|
def _start_media_player(self, attrsD):
|
|
self.push('media_player', 0)
|
|
self._getContext()['media_player'] = FeedParserDict(attrsD)
|
|
|
|
def _end_media_player(self):
|
|
value = self.pop('media_player')
|
|
context = self._getContext()
|
|
context['media_player']['content'] = value
|
|
|
|
def _start_newlocation(self, attrsD):
|
|
self.push('newlocation', 1)
|
|
|
|
def _end_newlocation(self):
|
|
url = self.pop('newlocation')
|
|
context = self._getContext()
|
|
# don't set newlocation if the context isn't right
|
|
if context is not self.feeddata:
|
|
return
|
|
context['newlocation'] = _makeSafeAbsoluteURI(self.baseuri, url.strip())
|
|
|
|
if _XML_AVAILABLE:
|
|
class _StrictFeedParser(_FeedParserMixin, xml.sax.handler.ContentHandler):
|
|
def __init__(self, baseuri, baselang, encoding):
|
|
xml.sax.handler.ContentHandler.__init__(self)
|
|
_FeedParserMixin.__init__(self, baseuri, baselang, encoding)
|
|
self.bozo = 0
|
|
self.exc = None
|
|
self.decls = {}
|
|
|
|
def startPrefixMapping(self, prefix, uri):
|
|
if not uri:
|
|
return
|
|
# Jython uses '' instead of None; standardize on None
|
|
prefix = prefix or None
|
|
self.trackNamespace(prefix, uri)
|
|
if prefix and uri == 'http://www.w3.org/1999/xlink':
|
|
self.decls['xmlns:' + prefix] = uri
|
|
|
|
def startElementNS(self, name, qname, attrs):
|
|
namespace, localname = name
|
|
lowernamespace = str(namespace or '').lower()
|
|
if lowernamespace.find(u'backend.userland.com/rss') <> -1:
|
|
# match any backend.userland.com namespace
|
|
namespace = u'http://backend.userland.com/rss'
|
|
lowernamespace = namespace
|
|
if qname and qname.find(':') > 0:
|
|
givenprefix = qname.split(':')[0]
|
|
else:
|
|
givenprefix = None
|
|
prefix = self._matchnamespaces.get(lowernamespace, givenprefix)
|
|
if givenprefix and (prefix == None or (prefix == '' and lowernamespace == '')) and givenprefix not in self.namespacesInUse:
|
|
raise UndeclaredNamespace, "'%s' is not associated with a namespace" % givenprefix
|
|
localname = str(localname).lower()
|
|
|
|
# qname implementation is horribly broken in Python 2.1 (it
|
|
# doesn't report any), and slightly broken in Python 2.2 (it
|
|
# doesn't report the xml: namespace). So we match up namespaces
|
|
# with a known list first, and then possibly override them with
|
|
# the qnames the SAX parser gives us (if indeed it gives us any
|
|
# at all). Thanks to MatejC for helping me test this and
|
|
# tirelessly telling me that it didn't work yet.
|
|
attrsD, self.decls = self.decls, {}
|
|
if localname=='math' and namespace=='http://www.w3.org/1998/Math/MathML':
|
|
attrsD['xmlns']=namespace
|
|
if localname=='svg' and namespace=='http://www.w3.org/2000/svg':
|
|
attrsD['xmlns']=namespace
|
|
|
|
if prefix:
|
|
localname = prefix.lower() + ':' + localname
|
|
elif namespace and not qname: #Expat
|
|
for name,value in self.namespacesInUse.items():
|
|
if name and value == namespace:
|
|
localname = name + ':' + localname
|
|
break
|
|
|
|
for (namespace, attrlocalname), attrvalue in attrs.items():
|
|
lowernamespace = (namespace or '').lower()
|
|
prefix = self._matchnamespaces.get(lowernamespace, '')
|
|
if prefix:
|
|
attrlocalname = prefix + ':' + attrlocalname
|
|
attrsD[str(attrlocalname).lower()] = attrvalue
|
|
for qname in attrs.getQNames():
|
|
attrsD[str(qname).lower()] = attrs.getValueByQName(qname)
|
|
self.unknown_starttag(localname, attrsD.items())
|
|
|
|
def characters(self, text):
|
|
self.handle_data(text)
|
|
|
|
def endElementNS(self, name, qname):
|
|
namespace, localname = name
|
|
lowernamespace = str(namespace or '').lower()
|
|
if qname and qname.find(':') > 0:
|
|
givenprefix = qname.split(':')[0]
|
|
else:
|
|
givenprefix = ''
|
|
prefix = self._matchnamespaces.get(lowernamespace, givenprefix)
|
|
if prefix:
|
|
localname = prefix + ':' + localname
|
|
elif namespace and not qname: #Expat
|
|
for name,value in self.namespacesInUse.items():
|
|
if name and value == namespace:
|
|
localname = name + ':' + localname
|
|
break
|
|
localname = str(localname).lower()
|
|
self.unknown_endtag(localname)
|
|
|
|
def error(self, exc):
|
|
self.bozo = 1
|
|
self.exc = exc
|
|
|
|
# drv_libxml2 calls warning() in some cases
|
|
warning = error
|
|
|
|
def fatalError(self, exc):
|
|
self.error(exc)
|
|
raise exc
|
|
|
|
class _BaseHTMLProcessor(sgmllib.SGMLParser):
|
|
special = re.compile('''[<>'"]''')
|
|
bare_ampersand = re.compile("&(?!#\d+;|#x[0-9a-fA-F]+;|\w+;)")
|
|
elements_no_end_tag = set([
|
|
'area', 'base', 'basefont', 'br', 'col', 'command', 'embed', 'frame',
|
|
'hr', 'img', 'input', 'isindex', 'keygen', 'link', 'meta', 'param',
|
|
'source', 'track', 'wbr'
|
|
])
|
|
|
|
def __init__(self, encoding, _type):
|
|
self.encoding = encoding
|
|
self._type = _type
|
|
sgmllib.SGMLParser.__init__(self)
|
|
|
|
def reset(self):
|
|
self.pieces = []
|
|
sgmllib.SGMLParser.reset(self)
|
|
|
|
def _shorttag_replace(self, match):
|
|
tag = match.group(1)
|
|
if tag in self.elements_no_end_tag:
|
|
return '<' + tag + ' />'
|
|
else:
|
|
return '<' + tag + '></' + tag + '>'
|
|
|
|
# By declaring these methods and overriding their compiled code
|
|
# with the code from sgmllib, the original code will execute in
|
|
# feedparser's scope instead of sgmllib's. This means that the
|
|
# `tagfind` and `charref` regular expressions will be found as
|
|
# they're declared above, not as they're declared in sgmllib.
|
|
def goahead(self, i):
|
|
pass
|
|
goahead.func_code = sgmllib.SGMLParser.goahead.func_code
|
|
|
|
def __parse_starttag(self, i):
|
|
pass
|
|
__parse_starttag.func_code = sgmllib.SGMLParser.parse_starttag.func_code
|
|
|
|
def parse_starttag(self,i):
|
|
j = self.__parse_starttag(i)
|
|
if self._type == 'application/xhtml+xml':
|
|
if j>2 and self.rawdata[j-2:j]=='/>':
|
|
self.unknown_endtag(self.lasttag)
|
|
return j
|
|
|
|
def feed(self, data):
|
|
data = re.compile(r'<!((?!DOCTYPE|--|\[))', re.IGNORECASE).sub(r'<!\1', data)
|
|
data = re.sub(r'<([^<>\s]+?)\s*/>', self._shorttag_replace, data)
|
|
data = data.replace(''', "'")
|
|
data = data.replace('"', '"')
|
|
try:
|
|
bytes
|
|
if bytes is str:
|
|
raise NameError
|
|
self.encoding = self.encoding + u'_INVALID_PYTHON_3'
|
|
except NameError:
|
|
if self.encoding and isinstance(data, unicode):
|
|
data = data.encode(self.encoding)
|
|
sgmllib.SGMLParser.feed(self, data)
|
|
sgmllib.SGMLParser.close(self)
|
|
|
|
def normalize_attrs(self, attrs):
|
|
if not attrs:
|
|
return attrs
|
|
# utility method to be called by descendants
|
|
attrs = dict([(k.lower(), v) for k, v in attrs]).items()
|
|
attrs = [(k, k in ('rel', 'type') and v.lower() or v) for k, v in attrs]
|
|
attrs.sort()
|
|
return attrs
|
|
|
|
def unknown_starttag(self, tag, attrs):
|
|
# called for each start tag
|
|
# attrs is a list of (attr, value) tuples
|
|
# e.g. for <pre class='screen'>, tag='pre', attrs=[('class', 'screen')]
|
|
uattrs = []
|
|
strattrs=''
|
|
if attrs:
|
|
for key, value in attrs:
|
|
value=value.replace('>','>').replace('<','<').replace('"','"')
|
|
value = self.bare_ampersand.sub("&", value)
|
|
# thanks to Kevin Marks for this breathtaking hack to deal with (valid) high-bit attribute values in UTF-8 feeds
|
|
if not isinstance(value, unicode):
|
|
value = value.decode(self.encoding, 'ignore')
|
|
try:
|
|
# Currently, in Python 3 the key is already a str, and cannot be decoded again
|
|
uattrs.append((unicode(key, self.encoding), value))
|
|
except TypeError:
|
|
uattrs.append((key, value))
|
|
strattrs = u''.join([u' %s="%s"' % (key, value) for key, value in uattrs])
|
|
if self.encoding:
|
|
try:
|
|
strattrs = strattrs.encode(self.encoding)
|
|
except (UnicodeEncodeError, LookupError):
|
|
pass
|
|
if tag in self.elements_no_end_tag:
|
|
self.pieces.append('<%s%s />' % (tag, strattrs))
|
|
else:
|
|
self.pieces.append('<%s%s>' % (tag, strattrs))
|
|
|
|
def unknown_endtag(self, tag):
|
|
# called for each end tag, e.g. for </pre>, tag will be 'pre'
|
|
# Reconstruct the original end tag.
|
|
if tag not in self.elements_no_end_tag:
|
|
self.pieces.append("</%s>" % tag)
|
|
|
|
def handle_charref(self, ref):
|
|
# called for each character reference, e.g. for ' ', ref will be '160'
|
|
# Reconstruct the original character reference.
|
|
ref = ref.lower()
|
|
if ref.startswith('x'):
|
|
value = int(ref[1:], 16)
|
|
else:
|
|
value = int(ref)
|
|
|
|
if value in _cp1252:
|
|
self.pieces.append('&#%s;' % hex(ord(_cp1252[value]))[1:])
|
|
else:
|
|
self.pieces.append('&#%s;' % ref)
|
|
|
|
def handle_entityref(self, ref):
|
|
# called for each entity reference, e.g. for '©', ref will be 'copy'
|
|
# Reconstruct the original entity reference.
|
|
if ref in name2codepoint or ref == 'apos':
|
|
self.pieces.append('&%s;' % ref)
|
|
else:
|
|
self.pieces.append('&%s' % ref)
|
|
|
|
def handle_data(self, text):
|
|
# called for each block of plain text, i.e. outside of any tag and
|
|
# not containing any character or entity references
|
|
# Store the original text verbatim.
|
|
self.pieces.append(text)
|
|
|
|
def handle_comment(self, text):
|
|
# called for each HTML comment, e.g. <!-- insert Javascript code here -->
|
|
# Reconstruct the original comment.
|
|
self.pieces.append('<!--%s-->' % text)
|
|
|
|
def handle_pi(self, text):
|
|
# called for each processing instruction, e.g. <?instruction>
|
|
# Reconstruct original processing instruction.
|
|
self.pieces.append('<?%s>' % text)
|
|
|
|
def handle_decl(self, text):
|
|
# called for the DOCTYPE, if present, e.g.
|
|
# <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
|
|
# "http://www.w3.org/TR/html4/loose.dtd">
|
|
# Reconstruct original DOCTYPE
|
|
self.pieces.append('<!%s>' % text)
|
|
|
|
_new_declname_match = re.compile(r'[a-zA-Z][-_.a-zA-Z0-9:]*\s*').match
|
|
def _scan_name(self, i, declstartpos):
|
|
rawdata = self.rawdata
|
|
n = len(rawdata)
|
|
if i == n:
|
|
return None, -1
|
|
m = self._new_declname_match(rawdata, i)
|
|
if m:
|
|
s = m.group()
|
|
name = s.strip()
|
|
if (i + len(s)) == n:
|
|
return None, -1 # end of buffer
|
|
return name.lower(), m.end()
|
|
else:
|
|
self.handle_data(rawdata)
|
|
# self.updatepos(declstartpos, i)
|
|
return None, -1
|
|
|
|
def convert_charref(self, name):
|
|
return '&#%s;' % name
|
|
|
|
def convert_entityref(self, name):
|
|
return '&%s;' % name
|
|
|
|
def output(self):
|
|
'''Return processed HTML as a single string'''
|
|
return ''.join([str(p) for p in self.pieces])
|
|
|
|
def parse_declaration(self, i):
|
|
try:
|
|
return sgmllib.SGMLParser.parse_declaration(self, i)
|
|
except sgmllib.SGMLParseError:
|
|
# escape the doctype declaration and continue parsing
|
|
self.handle_data('<')
|
|
return i+1
|
|
|
|
class _LooseFeedParser(_FeedParserMixin, _BaseHTMLProcessor):
|
|
def __init__(self, baseuri, baselang, encoding, entities):
|
|
sgmllib.SGMLParser.__init__(self)
|
|
_FeedParserMixin.__init__(self, baseuri, baselang, encoding)
|
|
_BaseHTMLProcessor.__init__(self, encoding, 'application/xhtml+xml')
|
|
self.entities=entities
|
|
|
|
def decodeEntities(self, element, data):
|
|
data = data.replace('<', '<')
|
|
data = data.replace('<', '<')
|
|
data = data.replace('<', '<')
|
|
data = data.replace('>', '>')
|
|
data = data.replace('>', '>')
|
|
data = data.replace('>', '>')
|
|
data = data.replace('&', '&')
|
|
data = data.replace('&', '&')
|
|
data = data.replace('"', '"')
|
|
data = data.replace('"', '"')
|
|
data = data.replace(''', ''')
|
|
data = data.replace(''', ''')
|
|
if not self.contentparams.get('type', u'xml').endswith(u'xml'):
|
|
data = data.replace('<', '<')
|
|
data = data.replace('>', '>')
|
|
data = data.replace('&', '&')
|
|
data = data.replace('"', '"')
|
|
data = data.replace(''', "'")
|
|
return data
|
|
|
|
def strattrs(self, attrs):
|
|
return ''.join([' %s="%s"' % (n,v.replace('"','"')) for n,v in attrs])
|
|
|
|
class _MicroformatsParser:
|
|
STRING = 1
|
|
DATE = 2
|
|
URI = 3
|
|
NODE = 4
|
|
EMAIL = 5
|
|
|
|
known_xfn_relationships = set(['contact', 'acquaintance', 'friend', 'met', 'co-worker', 'coworker', 'colleague', 'co-resident', 'coresident', 'neighbor', 'child', 'parent', 'sibling', 'brother', 'sister', 'spouse', 'wife', 'husband', 'kin', 'relative', 'muse', 'crush', 'date', 'sweetheart', 'me'])
|
|
known_binary_extensions = set(['zip','rar','exe','gz','tar','tgz','tbz2','bz2','z','7z','dmg','img','sit','sitx','hqx','deb','rpm','bz2','jar','rar','iso','bin','msi','mp2','mp3','ogg','ogm','mp4','m4v','m4a','avi','wma','wmv'])
|
|
|
|
def __init__(self, data, baseuri, encoding):
|
|
self.document = BeautifulSoup.BeautifulSoup(data)
|
|
self.baseuri = baseuri
|
|
self.encoding = encoding
|
|
if isinstance(data, unicode):
|
|
data = data.encode(encoding)
|
|
self.tags = []
|
|
self.enclosures = []
|
|
self.xfn = []
|
|
self.vcard = None
|
|
|
|
def vcardEscape(self, s):
|
|
if isinstance(s, basestring):
|
|
s = s.replace(',', '\\,').replace(';', '\\;').replace('\n', '\\n')
|
|
return s
|
|
|
|
def vcardFold(self, s):
|
|
s = re.sub(';+$', '', s)
|
|
sFolded = ''
|
|
iMax = 75
|
|
sPrefix = ''
|
|
while len(s) > iMax:
|
|
sFolded += sPrefix + s[:iMax] + '\n'
|
|
s = s[iMax:]
|
|
sPrefix = ' '
|
|
iMax = 74
|
|
sFolded += sPrefix + s
|
|
return sFolded
|
|
|
|
def normalize(self, s):
|
|
return re.sub(r'\s+', ' ', s).strip()
|
|
|
|
def unique(self, aList):
|
|
results = []
|
|
for element in aList:
|
|
if element not in results:
|
|
results.append(element)
|
|
return results
|
|
|
|
def toISO8601(self, dt):
|
|
return time.strftime('%Y-%m-%dT%H:%M:%SZ', dt)
|
|
|
|
def getPropertyValue(self, elmRoot, sProperty, iPropertyType=4, bAllowMultiple=0, bAutoEscape=0):
|
|
all = lambda x: 1
|
|
sProperty = sProperty.lower()
|
|
bFound = 0
|
|
bNormalize = 1
|
|
propertyMatch = {'class': re.compile(r'\b%s\b' % sProperty)}
|
|
if bAllowMultiple and (iPropertyType != self.NODE):
|
|
snapResults = []
|
|
containers = elmRoot(['ul', 'ol'], propertyMatch)
|
|
for container in containers:
|
|
snapResults.extend(container('li'))
|
|
bFound = (len(snapResults) != 0)
|
|
if not bFound:
|
|
snapResults = elmRoot(all, propertyMatch)
|
|
bFound = (len(snapResults) != 0)
|
|
if (not bFound) and (sProperty == 'value'):
|
|
snapResults = elmRoot('pre')
|
|
bFound = (len(snapResults) != 0)
|
|
bNormalize = not bFound
|
|
if not bFound:
|
|
snapResults = [elmRoot]
|
|
bFound = (len(snapResults) != 0)
|
|
arFilter = []
|
|
if sProperty == 'vcard':
|
|
snapFilter = elmRoot(all, propertyMatch)
|
|
for node in snapFilter:
|
|
if node.findParent(all, propertyMatch):
|
|
arFilter.append(node)
|
|
arResults = []
|
|
for node in snapResults:
|
|
if node not in arFilter:
|
|
arResults.append(node)
|
|
bFound = (len(arResults) != 0)
|
|
if not bFound:
|
|
if bAllowMultiple:
|
|
return []
|
|
elif iPropertyType == self.STRING:
|
|
return ''
|
|
elif iPropertyType == self.DATE:
|
|
return None
|
|
elif iPropertyType == self.URI:
|
|
return ''
|
|
elif iPropertyType == self.NODE:
|
|
return None
|
|
else:
|
|
return None
|
|
arValues = []
|
|
for elmResult in arResults:
|
|
sValue = None
|
|
if iPropertyType == self.NODE:
|
|
if bAllowMultiple:
|
|
arValues.append(elmResult)
|
|
continue
|
|
else:
|
|
return elmResult
|
|
sNodeName = elmResult.name.lower()
|
|
if (iPropertyType == self.EMAIL) and (sNodeName == 'a'):
|
|
sValue = (elmResult.get('href') or '').split('mailto:').pop().split('?')[0]
|
|
if sValue:
|
|
sValue = bNormalize and self.normalize(sValue) or sValue.strip()
|
|
if (not sValue) and (sNodeName == 'abbr'):
|
|
sValue = elmResult.get('title')
|
|
if sValue:
|
|
sValue = bNormalize and self.normalize(sValue) or sValue.strip()
|
|
if (not sValue) and (iPropertyType == self.URI):
|
|
if sNodeName == 'a':
|
|
sValue = elmResult.get('href')
|
|
elif sNodeName == 'img':
|
|
sValue = elmResult.get('src')
|
|
elif sNodeName == 'object':
|
|
sValue = elmResult.get('data')
|
|
if sValue:
|
|
sValue = bNormalize and self.normalize(sValue) or sValue.strip()
|
|
if (not sValue) and (sNodeName == 'img'):
|
|
sValue = elmResult.get('alt')
|
|
if sValue:
|
|
sValue = bNormalize and self.normalize(sValue) or sValue.strip()
|
|
if not sValue:
|
|
sValue = elmResult.renderContents()
|
|
sValue = re.sub(r'<\S[^>]*>', '', sValue)
|
|
sValue = sValue.replace('\r\n', '\n')
|
|
sValue = sValue.replace('\r', '\n')
|
|
if sValue:
|
|
sValue = bNormalize and self.normalize(sValue) or sValue.strip()
|
|
if not sValue:
|
|
continue
|
|
if iPropertyType == self.DATE:
|
|
sValue = _parse_date_iso8601(sValue)
|
|
if bAllowMultiple:
|
|
arValues.append(bAutoEscape and self.vcardEscape(sValue) or sValue)
|
|
else:
|
|
return bAutoEscape and self.vcardEscape(sValue) or sValue
|
|
return arValues
|
|
|
|
def findVCards(self, elmRoot, bAgentParsing=0):
|
|
sVCards = ''
|
|
|
|
if not bAgentParsing:
|
|
arCards = self.getPropertyValue(elmRoot, 'vcard', bAllowMultiple=1)
|
|
else:
|
|
arCards = [elmRoot]
|
|
|
|
for elmCard in arCards:
|
|
arLines = []
|
|
|
|
def processSingleString(sProperty):
|
|
sValue = self.getPropertyValue(elmCard, sProperty, self.STRING, bAutoEscape=1).decode(self.encoding)
|
|
if sValue:
|
|
arLines.append(self.vcardFold(sProperty.upper() + ':' + sValue))
|
|
return sValue or u''
|
|
|
|
def processSingleURI(sProperty):
|
|
sValue = self.getPropertyValue(elmCard, sProperty, self.URI)
|
|
if sValue:
|
|
sContentType = ''
|
|
sEncoding = ''
|
|
sValueKey = ''
|
|
if sValue.startswith('data:'):
|
|
sEncoding = ';ENCODING=b'
|
|
sContentType = sValue.split(';')[0].split('/').pop()
|
|
sValue = sValue.split(',', 1).pop()
|
|
else:
|
|
elmValue = self.getPropertyValue(elmCard, sProperty)
|
|
if elmValue:
|
|
if sProperty != 'url':
|
|
sValueKey = ';VALUE=uri'
|
|
sContentType = elmValue.get('type', '').strip().split('/').pop().strip()
|
|
sContentType = sContentType.upper()
|
|
if sContentType == 'OCTET-STREAM':
|
|
sContentType = ''
|
|
if sContentType:
|
|
sContentType = ';TYPE=' + sContentType.upper()
|
|
arLines.append(self.vcardFold(sProperty.upper() + sEncoding + sContentType + sValueKey + ':' + sValue))
|
|
|
|
def processTypeValue(sProperty, arDefaultType, arForceType=None):
|
|
arResults = self.getPropertyValue(elmCard, sProperty, bAllowMultiple=1)
|
|
for elmResult in arResults:
|
|
arType = self.getPropertyValue(elmResult, 'type', self.STRING, 1, 1)
|
|
if arForceType:
|
|
arType = self.unique(arForceType + arType)
|
|
if not arType:
|
|
arType = arDefaultType
|
|
sValue = self.getPropertyValue(elmResult, 'value', self.EMAIL, 0)
|
|
if sValue:
|
|
arLines.append(self.vcardFold(sProperty.upper() + ';TYPE=' + ','.join(arType) + ':' + sValue))
|
|
|
|
# AGENT
|
|
# must do this before all other properties because it is destructive
|
|
# (removes nested class="vcard" nodes so they don't interfere with
|
|
# this vcard's other properties)
|
|
arAgent = self.getPropertyValue(elmCard, 'agent', bAllowMultiple=1)
|
|
for elmAgent in arAgent:
|
|
if re.compile(r'\bvcard\b').search(elmAgent.get('class')):
|
|
sAgentValue = self.findVCards(elmAgent, 1) + '\n'
|
|
sAgentValue = sAgentValue.replace('\n', '\\n')
|
|
sAgentValue = sAgentValue.replace(';', '\\;')
|
|
if sAgentValue:
|
|
arLines.append(self.vcardFold('AGENT:' + sAgentValue))
|
|
# Completely remove the agent element from the parse tree
|
|
elmAgent.extract()
|
|
else:
|
|
sAgentValue = self.getPropertyValue(elmAgent, 'value', self.URI, bAutoEscape=1);
|
|
if sAgentValue:
|
|
arLines.append(self.vcardFold('AGENT;VALUE=uri:' + sAgentValue))
|
|
|
|
# FN (full name)
|
|
sFN = processSingleString('fn')
|
|
|
|
# N (name)
|
|
elmName = self.getPropertyValue(elmCard, 'n')
|
|
if elmName:
|
|
sFamilyName = self.getPropertyValue(elmName, 'family-name', self.STRING, bAutoEscape=1)
|
|
sGivenName = self.getPropertyValue(elmName, 'given-name', self.STRING, bAutoEscape=1)
|
|
arAdditionalNames = self.getPropertyValue(elmName, 'additional-name', self.STRING, 1, 1) + self.getPropertyValue(elmName, 'additional-names', self.STRING, 1, 1)
|
|
arHonorificPrefixes = self.getPropertyValue(elmName, 'honorific-prefix', self.STRING, 1, 1) + self.getPropertyValue(elmName, 'honorific-prefixes', self.STRING, 1, 1)
|
|
arHonorificSuffixes = self.getPropertyValue(elmName, 'honorific-suffix', self.STRING, 1, 1) + self.getPropertyValue(elmName, 'honorific-suffixes', self.STRING, 1, 1)
|
|
arLines.append(self.vcardFold('N:' + sFamilyName + ';' +
|
|
sGivenName + ';' +
|
|
','.join(arAdditionalNames) + ';' +
|
|
','.join(arHonorificPrefixes) + ';' +
|
|
','.join(arHonorificSuffixes)))
|
|
elif sFN:
|
|
# implied "N" optimization
|
|
# http://microformats.org/wiki/hcard#Implied_.22N.22_Optimization
|
|
arNames = self.normalize(sFN).split()
|
|
if len(arNames) == 2:
|
|
bFamilyNameFirst = (arNames[0].endswith(',') or
|
|
len(arNames[1]) == 1 or
|
|
((len(arNames[1]) == 2) and (arNames[1].endswith('.'))))
|
|
if bFamilyNameFirst:
|
|
arLines.append(self.vcardFold('N:' + arNames[0] + ';' + arNames[1]))
|
|
else:
|
|
arLines.append(self.vcardFold('N:' + arNames[1] + ';' + arNames[0]))
|
|
|
|
# SORT-STRING
|
|
sSortString = self.getPropertyValue(elmCard, 'sort-string', self.STRING, bAutoEscape=1)
|
|
if sSortString:
|
|
arLines.append(self.vcardFold('SORT-STRING:' + sSortString))
|
|
|
|
# NICKNAME
|
|
arNickname = self.getPropertyValue(elmCard, 'nickname', self.STRING, 1, 1)
|
|
if arNickname:
|
|
arLines.append(self.vcardFold('NICKNAME:' + ','.join(arNickname)))
|
|
|
|
# PHOTO
|
|
processSingleURI('photo')
|
|
|
|
# BDAY
|
|
dtBday = self.getPropertyValue(elmCard, 'bday', self.DATE)
|
|
if dtBday:
|
|
arLines.append(self.vcardFold('BDAY:' + self.toISO8601(dtBday)))
|
|
|
|
# ADR (address)
|
|
arAdr = self.getPropertyValue(elmCard, 'adr', bAllowMultiple=1)
|
|
for elmAdr in arAdr:
|
|
arType = self.getPropertyValue(elmAdr, 'type', self.STRING, 1, 1)
|
|
if not arType:
|
|
arType = ['intl','postal','parcel','work'] # default adr types, see RFC 2426 section 3.2.1
|
|
sPostOfficeBox = self.getPropertyValue(elmAdr, 'post-office-box', self.STRING, 0, 1)
|
|
sExtendedAddress = self.getPropertyValue(elmAdr, 'extended-address', self.STRING, 0, 1)
|
|
sStreetAddress = self.getPropertyValue(elmAdr, 'street-address', self.STRING, 0, 1)
|
|
sLocality = self.getPropertyValue(elmAdr, 'locality', self.STRING, 0, 1)
|
|
sRegion = self.getPropertyValue(elmAdr, 'region', self.STRING, 0, 1)
|
|
sPostalCode = self.getPropertyValue(elmAdr, 'postal-code', self.STRING, 0, 1)
|
|
sCountryName = self.getPropertyValue(elmAdr, 'country-name', self.STRING, 0, 1)
|
|
arLines.append(self.vcardFold('ADR;TYPE=' + ','.join(arType) + ':' +
|
|
sPostOfficeBox + ';' +
|
|
sExtendedAddress + ';' +
|
|
sStreetAddress + ';' +
|
|
sLocality + ';' +
|
|
sRegion + ';' +
|
|
sPostalCode + ';' +
|
|
sCountryName))
|
|
|
|
# LABEL
|
|
processTypeValue('label', ['intl','postal','parcel','work'])
|
|
|
|
# TEL (phone number)
|
|
processTypeValue('tel', ['voice'])
|
|
|
|
# EMAIL
|
|
processTypeValue('email', ['internet'], ['internet'])
|
|
|
|
# MAILER
|
|
processSingleString('mailer')
|
|
|
|
# TZ (timezone)
|
|
processSingleString('tz')
|
|
|
|
# GEO (geographical information)
|
|
elmGeo = self.getPropertyValue(elmCard, 'geo')
|
|
if elmGeo:
|
|
sLatitude = self.getPropertyValue(elmGeo, 'latitude', self.STRING, 0, 1)
|
|
sLongitude = self.getPropertyValue(elmGeo, 'longitude', self.STRING, 0, 1)
|
|
arLines.append(self.vcardFold('GEO:' + sLatitude + ';' + sLongitude))
|
|
|
|
# TITLE
|
|
processSingleString('title')
|
|
|
|
# ROLE
|
|
processSingleString('role')
|
|
|
|
# LOGO
|
|
processSingleURI('logo')
|
|
|
|
# ORG (organization)
|
|
elmOrg = self.getPropertyValue(elmCard, 'org')
|
|
if elmOrg:
|
|
sOrganizationName = self.getPropertyValue(elmOrg, 'organization-name', self.STRING, 0, 1)
|
|
if not sOrganizationName:
|
|
# implied "organization-name" optimization
|
|
# http://microformats.org/wiki/hcard#Implied_.22organization-name.22_Optimization
|
|
sOrganizationName = self.getPropertyValue(elmCard, 'org', self.STRING, 0, 1)
|
|
if sOrganizationName:
|
|
arLines.append(self.vcardFold('ORG:' + sOrganizationName))
|
|
else:
|
|
arOrganizationUnit = self.getPropertyValue(elmOrg, 'organization-unit', self.STRING, 1, 1)
|
|
arLines.append(self.vcardFold('ORG:' + sOrganizationName + ';' + ';'.join(arOrganizationUnit)))
|
|
|
|
# CATEGORY
|
|
arCategory = self.getPropertyValue(elmCard, 'category', self.STRING, 1, 1) + self.getPropertyValue(elmCard, 'categories', self.STRING, 1, 1)
|
|
if arCategory:
|
|
arLines.append(self.vcardFold('CATEGORIES:' + ','.join(arCategory)))
|
|
|
|
# NOTE
|
|
processSingleString('note')
|
|
|
|
# REV
|
|
processSingleString('rev')
|
|
|
|
# SOUND
|
|
processSingleURI('sound')
|
|
|
|
# UID
|
|
processSingleString('uid')
|
|
|
|
# URL
|
|
processSingleURI('url')
|
|
|
|
# CLASS
|
|
processSingleString('class')
|
|
|
|
# KEY
|
|
processSingleURI('key')
|
|
|
|
if arLines:
|
|
arLines = [u'BEGIN:vCard',u'VERSION:3.0'] + arLines + [u'END:vCard']
|
|
# XXX - this is super ugly; properly fix this with issue 148
|
|
for i, s in enumerate(arLines):
|
|
if not isinstance(s, unicode):
|
|
arLines[i] = s.decode('utf-8', 'ignore')
|
|
sVCards += u'\n'.join(arLines) + u'\n'
|
|
|
|
return sVCards.strip()
|
|
|
|
def isProbablyDownloadable(self, elm):
|
|
attrsD = elm.attrMap
|
|
if 'href' not in attrsD:
|
|
return 0
|
|
linktype = attrsD.get('type', '').strip()
|
|
if linktype.startswith('audio/') or \
|
|
linktype.startswith('video/') or \
|
|
(linktype.startswith('application/') and not linktype.endswith('xml')):
|
|
return 1
|
|
try:
|
|
path = urlparse.urlparse(attrsD['href'])[2]
|
|
except ValueError:
|
|
return 0
|
|
if path.find('.') == -1:
|
|
return 0
|
|
fileext = path.split('.').pop().lower()
|
|
return fileext in self.known_binary_extensions
|
|
|
|
def findTags(self):
|
|
all = lambda x: 1
|
|
for elm in self.document(all, {'rel': re.compile(r'\btag\b')}):
|
|
href = elm.get('href')
|
|
if not href:
|
|
continue
|
|
urlscheme, domain, path, params, query, fragment = \
|
|
urlparse.urlparse(_urljoin(self.baseuri, href))
|
|
segments = path.split('/')
|
|
tag = segments.pop()
|
|
if not tag:
|
|
if segments:
|
|
tag = segments.pop()
|
|
else:
|
|
# there are no tags
|
|
continue
|
|
tagscheme = urlparse.urlunparse((urlscheme, domain, '/'.join(segments), '', '', ''))
|
|
if not tagscheme.endswith('/'):
|
|
tagscheme += '/'
|
|
self.tags.append(FeedParserDict({"term": tag, "scheme": tagscheme, "label": elm.string or ''}))
|
|
|
|
def findEnclosures(self):
|
|
all = lambda x: 1
|
|
enclosure_match = re.compile(r'\benclosure\b')
|
|
for elm in self.document(all, {'href': re.compile(r'.+')}):
|
|
if not enclosure_match.search(elm.get('rel', u'')) and not self.isProbablyDownloadable(elm):
|
|
continue
|
|
if elm.attrMap not in self.enclosures:
|
|
self.enclosures.append(elm.attrMap)
|
|
if elm.string and not elm.get('title'):
|
|
self.enclosures[-1]['title'] = elm.string
|
|
|
|
def findXFN(self):
|
|
all = lambda x: 1
|
|
for elm in self.document(all, {'rel': re.compile('.+'), 'href': re.compile('.+')}):
|
|
rels = elm.get('rel', u'').split()
|
|
xfn_rels = [r for r in rels if r in self.known_xfn_relationships]
|
|
if xfn_rels:
|
|
self.xfn.append({"relationships": xfn_rels, "href": elm.get('href', ''), "name": elm.string})
|
|
|
|
def _parseMicroformats(htmlSource, baseURI, encoding):
|
|
if not BeautifulSoup:
|
|
return
|
|
try:
|
|
p = _MicroformatsParser(htmlSource, baseURI, encoding)
|
|
except UnicodeEncodeError:
|
|
# sgmllib throws this exception when performing lookups of tags
|
|
# with non-ASCII characters in them.
|
|
return
|
|
p.vcard = p.findVCards(p.document)
|
|
p.findTags()
|
|
p.findEnclosures()
|
|
p.findXFN()
|
|
return {"tags": p.tags, "enclosures": p.enclosures, "xfn": p.xfn, "vcard": p.vcard}
|
|
|
|
class _RelativeURIResolver(_BaseHTMLProcessor):
|
|
relative_uris = set([('a', 'href'),
|
|
('applet', 'codebase'),
|
|
('area', 'href'),
|
|
('blockquote', 'cite'),
|
|
('body', 'background'),
|
|
('del', 'cite'),
|
|
('form', 'action'),
|
|
('frame', 'longdesc'),
|
|
('frame', 'src'),
|
|
('iframe', 'longdesc'),
|
|
('iframe', 'src'),
|
|
('head', 'profile'),
|
|
('img', 'longdesc'),
|
|
('img', 'src'),
|
|
('img', 'usemap'),
|
|
('input', 'src'),
|
|
('input', 'usemap'),
|
|
('ins', 'cite'),
|
|
('link', 'href'),
|
|
('object', 'classid'),
|
|
('object', 'codebase'),
|
|
('object', 'data'),
|
|
('object', 'usemap'),
|
|
('q', 'cite'),
|
|
('script', 'src'),
|
|
('video', 'poster')])
|
|
|
|
def __init__(self, baseuri, encoding, _type):
|
|
_BaseHTMLProcessor.__init__(self, encoding, _type)
|
|
self.baseuri = baseuri
|
|
|
|
def resolveURI(self, uri):
|
|
return _makeSafeAbsoluteURI(self.baseuri, uri.strip())
|
|
|
|
def unknown_starttag(self, tag, attrs):
|
|
attrs = self.normalize_attrs(attrs)
|
|
attrs = [(key, ((tag, key) in self.relative_uris) and self.resolveURI(value) or value) for key, value in attrs]
|
|
_BaseHTMLProcessor.unknown_starttag(self, tag, attrs)
|
|
|
|
def _resolveRelativeURIs(htmlSource, baseURI, encoding, _type):
|
|
if not _SGML_AVAILABLE:
|
|
return htmlSource
|
|
|
|
p = _RelativeURIResolver(baseURI, encoding, _type)
|
|
p.feed(htmlSource)
|
|
return p.output()
|
|
|
|
def _makeSafeAbsoluteURI(base, rel=None):
|
|
# bail if ACCEPTABLE_URI_SCHEMES is empty
|
|
if not ACCEPTABLE_URI_SCHEMES:
|
|
try:
|
|
return _urljoin(base, rel or u'')
|
|
except ValueError:
|
|
return u''
|
|
if not base:
|
|
return rel or u''
|
|
if not rel:
|
|
try:
|
|
scheme = urlparse.urlparse(base)[0]
|
|
except ValueError:
|
|
return u''
|
|
if not scheme or scheme in ACCEPTABLE_URI_SCHEMES:
|
|
return base
|
|
return u''
|
|
try:
|
|
uri = _urljoin(base, rel)
|
|
except ValueError:
|
|
return u''
|
|
if uri.strip().split(':', 1)[0] not in ACCEPTABLE_URI_SCHEMES:
|
|
return u''
|
|
return uri
|
|
|
|
class _HTMLSanitizer(_BaseHTMLProcessor):
|
|
acceptable_elements = set(['a', 'abbr', 'acronym', 'address', 'area',
|
|
'article', 'aside', 'audio', 'b', 'big', 'blockquote', 'br', 'button',
|
|
'canvas', 'caption', 'center', 'cite', 'code', 'col', 'colgroup',
|
|
'command', 'datagrid', 'datalist', 'dd', 'del', 'details', 'dfn',
|
|
'dialog', 'dir', 'div', 'dl', 'dt', 'em', 'event-source', 'fieldset',
|
|
'figcaption', 'figure', 'footer', 'font', 'form', 'header', 'h1',
|
|
'h2', 'h3', 'h4', 'h5', 'h6', 'hr', 'i', 'img', 'input', 'ins',
|
|
'keygen', 'kbd', 'label', 'legend', 'li', 'm', 'map', 'menu', 'meter',
|
|
'multicol', 'nav', 'nextid', 'ol', 'output', 'optgroup', 'option',
|
|
'p', 'pre', 'progress', 'q', 's', 'samp', 'section', 'select',
|
|
'small', 'sound', 'source', 'spacer', 'span', 'strike', 'strong',
|
|
'sub', 'sup', 'table', 'tbody', 'td', 'textarea', 'time', 'tfoot',
|
|
'th', 'thead', 'tr', 'tt', 'u', 'ul', 'var', 'video', 'noscript'])
|
|
|
|
acceptable_attributes = set(['abbr', 'accept', 'accept-charset', 'accesskey',
|
|
'action', 'align', 'alt', 'autocomplete', 'autofocus', 'axis',
|
|
'background', 'balance', 'bgcolor', 'bgproperties', 'border',
|
|
'bordercolor', 'bordercolordark', 'bordercolorlight', 'bottompadding',
|
|
'cellpadding', 'cellspacing', 'ch', 'challenge', 'char', 'charoff',
|
|
'choff', 'charset', 'checked', 'cite', 'class', 'clear', 'color', 'cols',
|
|
'colspan', 'compact', 'contenteditable', 'controls', 'coords', 'data',
|
|
'datafld', 'datapagesize', 'datasrc', 'datetime', 'default', 'delay',
|
|
'dir', 'disabled', 'draggable', 'dynsrc', 'enctype', 'end', 'face', 'for',
|
|
'form', 'frame', 'galleryimg', 'gutter', 'headers', 'height', 'hidefocus',
|
|
'hidden', 'high', 'href', 'hreflang', 'hspace', 'icon', 'id', 'inputmode',
|
|
'ismap', 'keytype', 'label', 'leftspacing', 'lang', 'list', 'longdesc',
|
|
'loop', 'loopcount', 'loopend', 'loopstart', 'low', 'lowsrc', 'max',
|
|
'maxlength', 'media', 'method', 'min', 'multiple', 'name', 'nohref',
|
|
'noshade', 'nowrap', 'open', 'optimum', 'pattern', 'ping', 'point-size',
|
|
'poster', 'pqg', 'preload', 'prompt', 'radiogroup', 'readonly', 'rel',
|
|
'repeat-max', 'repeat-min', 'replace', 'required', 'rev', 'rightspacing',
|
|
'rows', 'rowspan', 'rules', 'scope', 'selected', 'shape', 'size', 'span',
|
|
'src', 'start', 'step', 'summary', 'suppress', 'tabindex', 'target',
|
|
'template', 'title', 'toppadding', 'type', 'unselectable', 'usemap',
|
|
'urn', 'valign', 'value', 'variable', 'volume', 'vspace', 'vrml',
|
|
'width', 'wrap', 'xml:lang'])
|
|
|
|
unacceptable_elements_with_end_tag = set(['script', 'applet', 'style'])
|
|
|
|
acceptable_css_properties = set(['azimuth', 'background-color',
|
|
'border-bottom-color', 'border-collapse', 'border-color',
|
|
'border-left-color', 'border-right-color', 'border-top-color', 'clear',
|
|
'color', 'cursor', 'direction', 'display', 'elevation', 'float', 'font',
|
|
'font-family', 'font-size', 'font-style', 'font-variant', 'font-weight',
|
|
'height', 'letter-spacing', 'line-height', 'overflow', 'pause',
|
|
'pause-after', 'pause-before', 'pitch', 'pitch-range', 'richness',
|
|
'speak', 'speak-header', 'speak-numeral', 'speak-punctuation',
|
|
'speech-rate', 'stress', 'text-align', 'text-decoration', 'text-indent',
|
|
'unicode-bidi', 'vertical-align', 'voice-family', 'volume',
|
|
'white-space', 'width'])
|
|
|
|
# survey of common keywords found in feeds
|
|
acceptable_css_keywords = set(['auto', 'aqua', 'black', 'block', 'blue',
|
|
'bold', 'both', 'bottom', 'brown', 'center', 'collapse', 'dashed',
|
|
'dotted', 'fuchsia', 'gray', 'green', '!important', 'italic', 'left',
|
|
'lime', 'maroon', 'medium', 'none', 'navy', 'normal', 'nowrap', 'olive',
|
|
'pointer', 'purple', 'red', 'right', 'solid', 'silver', 'teal', 'top',
|
|
'transparent', 'underline', 'white', 'yellow'])
|
|
|
|
valid_css_values = re.compile('^(#[0-9a-f]+|rgb\(\d+%?,\d*%?,?\d*%?\)?|' +
|
|
'\d{0,2}\.?\d{0,2}(cm|em|ex|in|mm|pc|pt|px|%|,|\))?)$')
|
|
|
|
mathml_elements = set(['annotation', 'annotation-xml', 'maction', 'math',
|
|
'merror', 'mfenced', 'mfrac', 'mi', 'mmultiscripts', 'mn', 'mo', 'mover', 'mpadded',
|
|
'mphantom', 'mprescripts', 'mroot', 'mrow', 'mspace', 'msqrt', 'mstyle',
|
|
'msub', 'msubsup', 'msup', 'mtable', 'mtd', 'mtext', 'mtr', 'munder',
|
|
'munderover', 'none', 'semantics'])
|
|
|
|
mathml_attributes = set(['actiontype', 'align', 'columnalign', 'columnalign',
|
|
'columnalign', 'close', 'columnlines', 'columnspacing', 'columnspan', 'depth',
|
|
'display', 'displaystyle', 'encoding', 'equalcolumns', 'equalrows',
|
|
'fence', 'fontstyle', 'fontweight', 'frame', 'height', 'linethickness',
|
|
'lspace', 'mathbackground', 'mathcolor', 'mathvariant', 'mathvariant',
|
|
'maxsize', 'minsize', 'open', 'other', 'rowalign', 'rowalign', 'rowalign',
|
|
'rowlines', 'rowspacing', 'rowspan', 'rspace', 'scriptlevel', 'selection',
|
|
'separator', 'separators', 'stretchy', 'width', 'width', 'xlink:href',
|
|
'xlink:show', 'xlink:type', 'xmlns', 'xmlns:xlink'])
|
|
|
|
# svgtiny - foreignObject + linearGradient + radialGradient + stop
|
|
svg_elements = set(['a', 'animate', 'animateColor', 'animateMotion',
|
|
'animateTransform', 'circle', 'defs', 'desc', 'ellipse', 'foreignObject',
|
|
'font-face', 'font-face-name', 'font-face-src', 'g', 'glyph', 'hkern',
|
|
'linearGradient', 'line', 'marker', 'metadata', 'missing-glyph', 'mpath',
|
|
'path', 'polygon', 'polyline', 'radialGradient', 'rect', 'set', 'stop',
|
|
'svg', 'switch', 'text', 'title', 'tspan', 'use'])
|
|
|
|
# svgtiny + class + opacity + offset + xmlns + xmlns:xlink
|
|
svg_attributes = set(['accent-height', 'accumulate', 'additive', 'alphabetic',
|
|
'arabic-form', 'ascent', 'attributeName', 'attributeType',
|
|
'baseProfile', 'bbox', 'begin', 'by', 'calcMode', 'cap-height',
|
|
'class', 'color', 'color-rendering', 'content', 'cx', 'cy', 'd', 'dx',
|
|
'dy', 'descent', 'display', 'dur', 'end', 'fill', 'fill-opacity',
|
|
'fill-rule', 'font-family', 'font-size', 'font-stretch', 'font-style',
|
|
'font-variant', 'font-weight', 'from', 'fx', 'fy', 'g1', 'g2',
|
|
'glyph-name', 'gradientUnits', 'hanging', 'height', 'horiz-adv-x',
|
|
'horiz-origin-x', 'id', 'ideographic', 'k', 'keyPoints', 'keySplines',
|
|
'keyTimes', 'lang', 'mathematical', 'marker-end', 'marker-mid',
|
|
'marker-start', 'markerHeight', 'markerUnits', 'markerWidth', 'max',
|
|
'min', 'name', 'offset', 'opacity', 'orient', 'origin',
|
|
'overline-position', 'overline-thickness', 'panose-1', 'path',
|
|
'pathLength', 'points', 'preserveAspectRatio', 'r', 'refX', 'refY',
|
|
'repeatCount', 'repeatDur', 'requiredExtensions', 'requiredFeatures',
|
|
'restart', 'rotate', 'rx', 'ry', 'slope', 'stemh', 'stemv',
|
|
'stop-color', 'stop-opacity', 'strikethrough-position',
|
|
'strikethrough-thickness', 'stroke', 'stroke-dasharray',
|
|
'stroke-dashoffset', 'stroke-linecap', 'stroke-linejoin',
|
|
'stroke-miterlimit', 'stroke-opacity', 'stroke-width', 'systemLanguage',
|
|
'target', 'text-anchor', 'to', 'transform', 'type', 'u1', 'u2',
|
|
'underline-position', 'underline-thickness', 'unicode', 'unicode-range',
|
|
'units-per-em', 'values', 'version', 'viewBox', 'visibility', 'width',
|
|
'widths', 'x', 'x-height', 'x1', 'x2', 'xlink:actuate', 'xlink:arcrole',
|
|
'xlink:href', 'xlink:role', 'xlink:show', 'xlink:title', 'xlink:type',
|
|
'xml:base', 'xml:lang', 'xml:space', 'xmlns', 'xmlns:xlink', 'y', 'y1',
|
|
'y2', 'zoomAndPan'])
|
|
|
|
svg_attr_map = None
|
|
svg_elem_map = None
|
|
|
|
acceptable_svg_properties = set([ 'fill', 'fill-opacity', 'fill-rule',
|
|
'stroke', 'stroke-width', 'stroke-linecap', 'stroke-linejoin',
|
|
'stroke-opacity'])
|
|
|
|
def reset(self):
|
|
_BaseHTMLProcessor.reset(self)
|
|
self.unacceptablestack = 0
|
|
self.mathmlOK = 0
|
|
self.svgOK = 0
|
|
|
|
def unknown_starttag(self, tag, attrs):
|
|
acceptable_attributes = self.acceptable_attributes
|
|
keymap = {}
|
|
if not tag in self.acceptable_elements or self.svgOK:
|
|
if tag in self.unacceptable_elements_with_end_tag:
|
|
self.unacceptablestack += 1
|
|
|
|
# add implicit namespaces to html5 inline svg/mathml
|
|
if self._type.endswith('html'):
|
|
if not dict(attrs).get('xmlns'):
|
|
if tag=='svg':
|
|
attrs.append( ('xmlns','http://www.w3.org/2000/svg') )
|
|
if tag=='math':
|
|
attrs.append( ('xmlns','http://www.w3.org/1998/Math/MathML') )
|
|
|
|
# not otherwise acceptable, perhaps it is MathML or SVG?
|
|
if tag=='math' and ('xmlns','http://www.w3.org/1998/Math/MathML') in attrs:
|
|
self.mathmlOK += 1
|
|
if tag=='svg' and ('xmlns','http://www.w3.org/2000/svg') in attrs:
|
|
self.svgOK += 1
|
|
|
|
# chose acceptable attributes based on tag class, else bail
|
|
if self.mathmlOK and tag in self.mathml_elements:
|
|
acceptable_attributes = self.mathml_attributes
|
|
elif self.svgOK and tag in self.svg_elements:
|
|
# for most vocabularies, lowercasing is a good idea. Many
|
|
# svg elements, however, are camel case
|
|
if not self.svg_attr_map:
|
|
lower=[attr.lower() for attr in self.svg_attributes]
|
|
mix=[a for a in self.svg_attributes if a not in lower]
|
|
self.svg_attributes = lower
|
|
self.svg_attr_map = dict([(a.lower(),a) for a in mix])
|
|
|
|
lower=[attr.lower() for attr in self.svg_elements]
|
|
mix=[a for a in self.svg_elements if a not in lower]
|
|
self.svg_elements = lower
|
|
self.svg_elem_map = dict([(a.lower(),a) for a in mix])
|
|
acceptable_attributes = self.svg_attributes
|
|
tag = self.svg_elem_map.get(tag,tag)
|
|
keymap = self.svg_attr_map
|
|
elif not tag in self.acceptable_elements:
|
|
return
|
|
|
|
# declare xlink namespace, if needed
|
|
if self.mathmlOK or self.svgOK:
|
|
if filter(lambda (n,v): n.startswith('xlink:'),attrs):
|
|
if not ('xmlns:xlink','http://www.w3.org/1999/xlink') in attrs:
|
|
attrs.append(('xmlns:xlink','http://www.w3.org/1999/xlink'))
|
|
|
|
clean_attrs = []
|
|
for key, value in self.normalize_attrs(attrs):
|
|
if key in acceptable_attributes:
|
|
key=keymap.get(key,key)
|
|
# make sure the uri uses an acceptable uri scheme
|
|
if key == u'href':
|
|
value = _makeSafeAbsoluteURI(value)
|
|
clean_attrs.append((key,value))
|
|
elif key=='style':
|
|
clean_value = self.sanitize_style(value)
|
|
if clean_value:
|
|
clean_attrs.append((key,clean_value))
|
|
_BaseHTMLProcessor.unknown_starttag(self, tag, clean_attrs)
|
|
|
|
def unknown_endtag(self, tag):
|
|
if not tag in self.acceptable_elements:
|
|
if tag in self.unacceptable_elements_with_end_tag:
|
|
self.unacceptablestack -= 1
|
|
if self.mathmlOK and tag in self.mathml_elements:
|
|
if tag == 'math' and self.mathmlOK:
|
|
self.mathmlOK -= 1
|
|
elif self.svgOK and tag in self.svg_elements:
|
|
tag = self.svg_elem_map.get(tag,tag)
|
|
if tag == 'svg' and self.svgOK:
|
|
self.svgOK -= 1
|
|
else:
|
|
return
|
|
_BaseHTMLProcessor.unknown_endtag(self, tag)
|
|
|
|
def handle_pi(self, text):
|
|
pass
|
|
|
|
def handle_decl(self, text):
|
|
pass
|
|
|
|
def handle_data(self, text):
|
|
if not self.unacceptablestack:
|
|
_BaseHTMLProcessor.handle_data(self, text)
|
|
|
|
def sanitize_style(self, style):
|
|
# disallow urls
|
|
style=re.compile('url\s*\(\s*[^\s)]+?\s*\)\s*').sub(' ',style)
|
|
|
|
# gauntlet
|
|
if not re.match("""^([:,;#%.\sa-zA-Z0-9!]|\w-\w|'[\s\w]+'|"[\s\w]+"|\([\d,\s]+\))*$""", style):
|
|
return ''
|
|
# This replaced a regexp that used re.match and was prone to pathological back-tracking.
|
|
if re.sub("\s*[-\w]+\s*:\s*[^:;]*;?", '', style).strip():
|
|
return ''
|
|
|
|
clean = []
|
|
for prop,value in re.findall("([-\w]+)\s*:\s*([^:;]*)",style):
|
|
if not value:
|
|
continue
|
|
if prop.lower() in self.acceptable_css_properties:
|
|
clean.append(prop + ': ' + value + ';')
|
|
elif prop.split('-')[0].lower() in ['background','border','margin','padding']:
|
|
for keyword in value.split():
|
|
if not keyword in self.acceptable_css_keywords and \
|
|
not self.valid_css_values.match(keyword):
|
|
break
|
|
else:
|
|
clean.append(prop + ': ' + value + ';')
|
|
elif self.svgOK and prop.lower() in self.acceptable_svg_properties:
|
|
clean.append(prop + ': ' + value + ';')
|
|
|
|
return ' '.join(clean)
|
|
|
|
def parse_comment(self, i, report=1):
|
|
ret = _BaseHTMLProcessor.parse_comment(self, i, report)
|
|
if ret >= 0:
|
|
return ret
|
|
# if ret == -1, this may be a malicious attempt to circumvent
|
|
# sanitization, or a page-destroying unclosed comment
|
|
match = re.compile(r'--[^>]*>').search(self.rawdata, i+4)
|
|
if match:
|
|
return match.end()
|
|
# unclosed comment; deliberately fail to handle_data()
|
|
return len(self.rawdata)
|
|
|
|
|
|
def _sanitizeHTML(htmlSource, encoding, _type):
|
|
if not _SGML_AVAILABLE:
|
|
return htmlSource
|
|
p = _HTMLSanitizer(encoding, _type)
|
|
htmlSource = htmlSource.replace('<![CDATA[', '<![CDATA[')
|
|
p.feed(htmlSource)
|
|
data = p.output()
|
|
if TIDY_MARKUP:
|
|
# loop through list of preferred Tidy interfaces looking for one that's installed,
|
|
# then set up a common _tidy function to wrap the interface-specific API.
|
|
_tidy = None
|
|
for tidy_interface in PREFERRED_TIDY_INTERFACES:
|
|
try:
|
|
if tidy_interface == "uTidy":
|
|
from tidy import parseString as _utidy
|
|
def _tidy(data, **kwargs):
|
|
return str(_utidy(data, **kwargs))
|
|
break
|
|
elif tidy_interface == "mxTidy":
|
|
from mx.Tidy import Tidy as _mxtidy
|
|
def _tidy(data, **kwargs):
|
|
nerrors, nwarnings, data, errordata = _mxtidy.tidy(data, **kwargs)
|
|
return data
|
|
break
|
|
except:
|
|
pass
|
|
if _tidy:
|
|
utf8 = isinstance(data, unicode)
|
|
if utf8:
|
|
data = data.encode('utf-8')
|
|
data = _tidy(data, output_xhtml=1, numeric_entities=1, wrap=0, char_encoding="utf8")
|
|
if utf8:
|
|
data = unicode(data, 'utf-8')
|
|
if data.count('<body'):
|
|
data = data.split('<body', 1)[1]
|
|
if data.count('>'):
|
|
data = data.split('>', 1)[1]
|
|
if data.count('</body'):
|
|
data = data.split('</body', 1)[0]
|
|
data = data.strip().replace('\r\n', '\n')
|
|
return data
|
|
|
|
class _FeedURLHandler(urllib2.HTTPDigestAuthHandler, urllib2.HTTPRedirectHandler, urllib2.HTTPDefaultErrorHandler):
|
|
def http_error_default(self, req, fp, code, msg, headers):
|
|
# The default implementation just raises HTTPError.
|
|
# Forget that.
|
|
fp.status = code
|
|
return fp
|
|
|
|
def http_error_301(self, req, fp, code, msg, hdrs):
|
|
result = urllib2.HTTPRedirectHandler.http_error_301(self, req, fp,
|
|
code, msg, hdrs)
|
|
result.status = code
|
|
result.newurl = result.geturl()
|
|
return result
|
|
# The default implementations in urllib2.HTTPRedirectHandler
|
|
# are identical, so hardcoding a http_error_301 call above
|
|
# won't affect anything
|
|
http_error_300 = http_error_301
|
|
http_error_302 = http_error_301
|
|
http_error_303 = http_error_301
|
|
http_error_307 = http_error_301
|
|
|
|
def http_error_401(self, req, fp, code, msg, headers):
|
|
# Check if
|
|
# - server requires digest auth, AND
|
|
# - we tried (unsuccessfully) with basic auth, AND
|
|
# If all conditions hold, parse authentication information
|
|
# out of the Authorization header we sent the first time
|
|
# (for the username and password) and the WWW-Authenticate
|
|
# header the server sent back (for the realm) and retry
|
|
# the request with the appropriate digest auth headers instead.
|
|
# This evil genius hack has been brought to you by Aaron Swartz.
|
|
host = urlparse.urlparse(req.get_full_url())[1]
|
|
if base64 is None or 'Authorization' not in req.headers \
|
|
or 'WWW-Authenticate' not in headers:
|
|
return self.http_error_default(req, fp, code, msg, headers)
|
|
auth = _base64decode(req.headers['Authorization'].split(' ')[1])
|
|
user, passw = auth.split(':')
|
|
realm = re.findall('realm="([^"]*)"', headers['WWW-Authenticate'])[0]
|
|
self.add_password(realm, host, user, passw)
|
|
retry = self.http_error_auth_reqed('www-authenticate', host, req, headers)
|
|
self.reset_retry_count()
|
|
return retry
|
|
|
|
def _open_resource(url_file_stream_or_string, etag, modified, agent, referrer, handlers, request_headers):
|
|
"""URL, filename, or string --> stream
|
|
|
|
This function lets you define parsers that take any input source
|
|
(URL, pathname to local or network file, or actual data as a string)
|
|
and deal with it in a uniform manner. Returned object is guaranteed
|
|
to have all the basic stdio read methods (read, readline, readlines).
|
|
Just .close() the object when you're done with it.
|
|
|
|
If the etag argument is supplied, it will be used as the value of an
|
|
If-None-Match request header.
|
|
|
|
If the modified argument is supplied, it can be a tuple of 9 integers
|
|
(as returned by gmtime() in the standard Python time module) or a date
|
|
string in any format supported by feedparser. Regardless, it MUST
|
|
be in GMT (Greenwich Mean Time). It will be reformatted into an
|
|
RFC 1123-compliant date and used as the value of an If-Modified-Since
|
|
request header.
|
|
|
|
If the agent argument is supplied, it will be used as the value of a
|
|
User-Agent request header.
|
|
|
|
If the referrer argument is supplied, it will be used as the value of a
|
|
Referer[sic] request header.
|
|
|
|
If handlers is supplied, it is a list of handlers used to build a
|
|
urllib2 opener.
|
|
|
|
if request_headers is supplied it is a dictionary of HTTP request headers
|
|
that will override the values generated by FeedParser.
|
|
"""
|
|
|
|
if hasattr(url_file_stream_or_string, 'read'):
|
|
return url_file_stream_or_string
|
|
|
|
if isinstance(url_file_stream_or_string, basestring) \
|
|
and urlparse.urlparse(url_file_stream_or_string)[0] in ('http', 'https', 'ftp', 'file', 'feed'):
|
|
# Deal with the feed URI scheme
|
|
if url_file_stream_or_string.startswith('feed:http'):
|
|
url_file_stream_or_string = url_file_stream_or_string[5:]
|
|
elif url_file_stream_or_string.startswith('feed:'):
|
|
url_file_stream_or_string = 'http:' + url_file_stream_or_string[5:]
|
|
if not agent:
|
|
agent = USER_AGENT
|
|
# Test for inline user:password credentials for HTTP basic auth
|
|
auth = None
|
|
if base64 and not url_file_stream_or_string.startswith('ftp:'):
|
|
urltype, rest = urllib.splittype(url_file_stream_or_string)
|
|
realhost, rest = urllib.splithost(rest)
|
|
if realhost:
|
|
user_passwd, realhost = urllib.splituser(realhost)
|
|
if user_passwd:
|
|
url_file_stream_or_string = '%s://%s%s' % (urltype, realhost, rest)
|
|
auth = base64.standard_b64encode(user_passwd).strip()
|
|
|
|
# iri support
|
|
if isinstance(url_file_stream_or_string, unicode):
|
|
url_file_stream_or_string = _convert_to_idn(url_file_stream_or_string)
|
|
|
|
# try to open with urllib2 (to use optional headers)
|
|
request = _build_urllib2_request(url_file_stream_or_string, agent, etag, modified, referrer, auth, request_headers)
|
|
opener = urllib2.build_opener(*tuple(handlers + [_FeedURLHandler()]))
|
|
opener.addheaders = [] # RMK - must clear so we only send our custom User-Agent
|
|
try:
|
|
return opener.open(request)
|
|
finally:
|
|
opener.close() # JohnD
|
|
|
|
# try to open with native open function (if url_file_stream_or_string is a filename)
|
|
try:
|
|
return open(url_file_stream_or_string, 'rb')
|
|
except (IOError, UnicodeEncodeError, TypeError):
|
|
# if url_file_stream_or_string is a unicode object that
|
|
# cannot be converted to the encoding returned by
|
|
# sys.getfilesystemencoding(), a UnicodeEncodeError
|
|
# will be thrown
|
|
# If url_file_stream_or_string is a string that contains NULL
|
|
# (such as an XML document encoded in UTF-32), TypeError will
|
|
# be thrown.
|
|
pass
|
|
|
|
# treat url_file_stream_or_string as string
|
|
if isinstance(url_file_stream_or_string, unicode):
|
|
return _StringIO(url_file_stream_or_string.encode('utf-8'))
|
|
return _StringIO(url_file_stream_or_string)
|
|
|
|
def _convert_to_idn(url):
|
|
"""Convert a URL to IDN notation"""
|
|
# this function should only be called with a unicode string
|
|
# strategy: if the host cannot be encoded in ascii, then
|
|
# it'll be necessary to encode it in idn form
|
|
parts = list(urlparse.urlsplit(url))
|
|
try:
|
|
parts[1].encode('ascii')
|
|
except UnicodeEncodeError:
|
|
# the url needs to be converted to idn notation
|
|
host = parts[1].rsplit(':', 1)
|
|
newhost = []
|
|
port = u''
|
|
if len(host) == 2:
|
|
port = host.pop()
|
|
for h in host[0].split('.'):
|
|
newhost.append(h.encode('idna').decode('utf-8'))
|
|
parts[1] = '.'.join(newhost)
|
|
if port:
|
|
parts[1] += ':' + port
|
|
return urlparse.urlunsplit(parts)
|
|
else:
|
|
return url
|
|
|
|
def _build_urllib2_request(url, agent, etag, modified, referrer, auth, request_headers):
|
|
request = urllib2.Request(url)
|
|
request.add_header('User-Agent', agent)
|
|
if etag:
|
|
request.add_header('If-None-Match', etag)
|
|
if isinstance(modified, basestring):
|
|
modified = _parse_date(modified)
|
|
elif isinstance(modified, datetime.datetime):
|
|
modified = modified.utctimetuple()
|
|
if modified:
|
|
# format into an RFC 1123-compliant timestamp. We can't use
|
|
# time.strftime() since the %a and %b directives can be affected
|
|
# by the current locale, but RFC 2616 states that dates must be
|
|
# in English.
|
|
short_weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
|
|
months = ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec']
|
|
request.add_header('If-Modified-Since', '%s, %02d %s %04d %02d:%02d:%02d GMT' % (short_weekdays[modified[6]], modified[2], months[modified[1] - 1], modified[0], modified[3], modified[4], modified[5]))
|
|
if referrer:
|
|
request.add_header('Referer', referrer)
|
|
if gzip and zlib:
|
|
request.add_header('Accept-encoding', 'gzip, deflate')
|
|
elif gzip:
|
|
request.add_header('Accept-encoding', 'gzip')
|
|
elif zlib:
|
|
request.add_header('Accept-encoding', 'deflate')
|
|
else:
|
|
request.add_header('Accept-encoding', '')
|
|
if auth:
|
|
request.add_header('Authorization', 'Basic %s' % auth)
|
|
if ACCEPT_HEADER:
|
|
request.add_header('Accept', ACCEPT_HEADER)
|
|
# use this for whatever -- cookies, special headers, etc
|
|
# [('Cookie','Something'),('x-special-header','Another Value')]
|
|
for header_name, header_value in request_headers.items():
|
|
request.add_header(header_name, header_value)
|
|
request.add_header('A-IM', 'feed') # RFC 3229 support
|
|
return request
|
|
|
|
_date_handlers = []
|
|
def registerDateHandler(func):
|
|
'''Register a date handler function (takes string, returns 9-tuple date in GMT)'''
|
|
_date_handlers.insert(0, func)
|
|
|
|
# ISO-8601 date parsing routines written by Fazal Majid.
|
|
# The ISO 8601 standard is very convoluted and irregular - a full ISO 8601
|
|
# parser is beyond the scope of feedparser and would be a worthwhile addition
|
|
# to the Python library.
|
|
# A single regular expression cannot parse ISO 8601 date formats into groups
|
|
# as the standard is highly irregular (for instance is 030104 2003-01-04 or
|
|
# 0301-04-01), so we use templates instead.
|
|
# Please note the order in templates is significant because we need a
|
|
# greedy match.
|
|
_iso8601_tmpl = ['YYYY-?MM-?DD', 'YYYY-0MM?-?DD', 'YYYY-MM', 'YYYY-?OOO',
|
|
'YY-?MM-?DD', 'YY-?OOO', 'YYYY',
|
|
'-YY-?MM', '-OOO', '-YY',
|
|
'--MM-?DD', '--MM',
|
|
'---DD',
|
|
'CC', '']
|
|
_iso8601_re = [
|
|
tmpl.replace(
|
|
'YYYY', r'(?P<year>\d{4})').replace(
|
|
'YY', r'(?P<year>\d\d)').replace(
|
|
'MM', r'(?P<month>[01]\d)').replace(
|
|
'DD', r'(?P<day>[0123]\d)').replace(
|
|
'OOO', r'(?P<ordinal>[0123]\d\d)').replace(
|
|
'CC', r'(?P<century>\d\d$)')
|
|
+ r'(T?(?P<hour>\d{2}):(?P<minute>\d{2})'
|
|
+ r'(:(?P<second>\d{2}))?'
|
|
+ r'(\.(?P<fracsecond>\d+))?'
|
|
+ r'(?P<tz>[+-](?P<tzhour>\d{2})(:(?P<tzmin>\d{2}))?|Z)?)?'
|
|
for tmpl in _iso8601_tmpl]
|
|
try:
|
|
del tmpl
|
|
except NameError:
|
|
pass
|
|
_iso8601_matches = [re.compile(regex).match for regex in _iso8601_re]
|
|
try:
|
|
del regex
|
|
except NameError:
|
|
pass
|
|
def _parse_date_iso8601(dateString):
|
|
'''Parse a variety of ISO-8601-compatible formats like 20040105'''
|
|
m = None
|
|
for _iso8601_match in _iso8601_matches:
|
|
m = _iso8601_match(dateString)
|
|
if m:
|
|
break
|
|
if not m:
|
|
return
|
|
if m.span() == (0, 0):
|
|
return
|
|
params = m.groupdict()
|
|
ordinal = params.get('ordinal', 0)
|
|
if ordinal:
|
|
ordinal = int(ordinal)
|
|
else:
|
|
ordinal = 0
|
|
year = params.get('year', '--')
|
|
if not year or year == '--':
|
|
year = time.gmtime()[0]
|
|
elif len(year) == 2:
|
|
# ISO 8601 assumes current century, i.e. 93 -> 2093, NOT 1993
|
|
year = 100 * int(time.gmtime()[0] / 100) + int(year)
|
|
else:
|
|
year = int(year)
|
|
month = params.get('month', '-')
|
|
if not month or month == '-':
|
|
# ordinals are NOT normalized by mktime, we simulate them
|
|
# by setting month=1, day=ordinal
|
|
if ordinal:
|
|
month = 1
|
|
else:
|
|
month = time.gmtime()[1]
|
|
month = int(month)
|
|
day = params.get('day', 0)
|
|
if not day:
|
|
# see above
|
|
if ordinal:
|
|
day = ordinal
|
|
elif params.get('century', 0) or \
|
|
params.get('year', 0) or params.get('month', 0):
|
|
day = 1
|
|
else:
|
|
day = time.gmtime()[2]
|
|
else:
|
|
day = int(day)
|
|
# special case of the century - is the first year of the 21st century
|
|
# 2000 or 2001 ? The debate goes on...
|
|
if 'century' in params:
|
|
year = (int(params['century']) - 1) * 100 + 1
|
|
# in ISO 8601 most fields are optional
|
|
for field in ['hour', 'minute', 'second', 'tzhour', 'tzmin']:
|
|
if not params.get(field, None):
|
|
params[field] = 0
|
|
hour = int(params.get('hour', 0))
|
|
minute = int(params.get('minute', 0))
|
|
second = int(float(params.get('second', 0)))
|
|
# weekday is normalized by mktime(), we can ignore it
|
|
weekday = 0
|
|
daylight_savings_flag = -1
|
|
tm = [year, month, day, hour, minute, second, weekday,
|
|
ordinal, daylight_savings_flag]
|
|
# ISO 8601 time zone adjustments
|
|
tz = params.get('tz')
|
|
if tz and tz != 'Z':
|
|
if tz[0] == '-':
|
|
tm[3] += int(params.get('tzhour', 0))
|
|
tm[4] += int(params.get('tzmin', 0))
|
|
elif tz[0] == '+':
|
|
tm[3] -= int(params.get('tzhour', 0))
|
|
tm[4] -= int(params.get('tzmin', 0))
|
|
else:
|
|
return None
|
|
# Python's time.mktime() is a wrapper around the ANSI C mktime(3c)
|
|
# which is guaranteed to normalize d/m/y/h/m/s.
|
|
# Many implementations have bugs, but we'll pretend they don't.
|
|
return time.localtime(time.mktime(tuple(tm)))
|
|
registerDateHandler(_parse_date_iso8601)
|
|
|
|
# 8-bit date handling routines written by ytrewq1.
|
|
_korean_year = u'\ub144' # b3e2 in euc-kr
|
|
_korean_month = u'\uc6d4' # bff9 in euc-kr
|
|
_korean_day = u'\uc77c' # c0cf in euc-kr
|
|
_korean_am = u'\uc624\uc804' # bfc0 c0fc in euc-kr
|
|
_korean_pm = u'\uc624\ud6c4' # bfc0 c8c4 in euc-kr
|
|
|
|
_korean_onblog_date_re = \
|
|
re.compile('(\d{4})%s\s+(\d{2})%s\s+(\d{2})%s\s+(\d{2}):(\d{2}):(\d{2})' % \
|
|
(_korean_year, _korean_month, _korean_day))
|
|
_korean_nate_date_re = \
|
|
re.compile(u'(\d{4})-(\d{2})-(\d{2})\s+(%s|%s)\s+(\d{,2}):(\d{,2}):(\d{,2})' % \
|
|
(_korean_am, _korean_pm))
|
|
def _parse_date_onblog(dateString):
|
|
'''Parse a string according to the OnBlog 8-bit date format'''
|
|
m = _korean_onblog_date_re.match(dateString)
|
|
if not m:
|
|
return
|
|
w3dtfdate = '%(year)s-%(month)s-%(day)sT%(hour)s:%(minute)s:%(second)s%(zonediff)s' % \
|
|
{'year': m.group(1), 'month': m.group(2), 'day': m.group(3),\
|
|
'hour': m.group(4), 'minute': m.group(5), 'second': m.group(6),\
|
|
'zonediff': '+09:00'}
|
|
return _parse_date_w3dtf(w3dtfdate)
|
|
registerDateHandler(_parse_date_onblog)
|
|
|
|
def _parse_date_nate(dateString):
|
|
'''Parse a string according to the Nate 8-bit date format'''
|
|
m = _korean_nate_date_re.match(dateString)
|
|
if not m:
|
|
return
|
|
hour = int(m.group(5))
|
|
ampm = m.group(4)
|
|
if (ampm == _korean_pm):
|
|
hour += 12
|
|
hour = str(hour)
|
|
if len(hour) == 1:
|
|
hour = '0' + hour
|
|
w3dtfdate = '%(year)s-%(month)s-%(day)sT%(hour)s:%(minute)s:%(second)s%(zonediff)s' % \
|
|
{'year': m.group(1), 'month': m.group(2), 'day': m.group(3),\
|
|
'hour': hour, 'minute': m.group(6), 'second': m.group(7),\
|
|
'zonediff': '+09:00'}
|
|
return _parse_date_w3dtf(w3dtfdate)
|
|
registerDateHandler(_parse_date_nate)
|
|
|
|
# Unicode strings for Greek date strings
|
|
_greek_months = \
|
|
{ \
|
|
u'\u0399\u03b1\u03bd': u'Jan', # c9e1ed in iso-8859-7
|
|
u'\u03a6\u03b5\u03b2': u'Feb', # d6e5e2 in iso-8859-7
|
|
u'\u039c\u03ac\u03ce': u'Mar', # ccdcfe in iso-8859-7
|
|
u'\u039c\u03b1\u03ce': u'Mar', # cce1fe in iso-8859-7
|
|
u'\u0391\u03c0\u03c1': u'Apr', # c1f0f1 in iso-8859-7
|
|
u'\u039c\u03ac\u03b9': u'May', # ccdce9 in iso-8859-7
|
|
u'\u039c\u03b1\u03ca': u'May', # cce1fa in iso-8859-7
|
|
u'\u039c\u03b1\u03b9': u'May', # cce1e9 in iso-8859-7
|
|
u'\u0399\u03bf\u03cd\u03bd': u'Jun', # c9effded in iso-8859-7
|
|
u'\u0399\u03bf\u03bd': u'Jun', # c9efed in iso-8859-7
|
|
u'\u0399\u03bf\u03cd\u03bb': u'Jul', # c9effdeb in iso-8859-7
|
|
u'\u0399\u03bf\u03bb': u'Jul', # c9f9eb in iso-8859-7
|
|
u'\u0391\u03cd\u03b3': u'Aug', # c1fde3 in iso-8859-7
|
|
u'\u0391\u03c5\u03b3': u'Aug', # c1f5e3 in iso-8859-7
|
|
u'\u03a3\u03b5\u03c0': u'Sep', # d3e5f0 in iso-8859-7
|
|
u'\u039f\u03ba\u03c4': u'Oct', # cfeaf4 in iso-8859-7
|
|
u'\u039d\u03bf\u03ad': u'Nov', # cdefdd in iso-8859-7
|
|
u'\u039d\u03bf\u03b5': u'Nov', # cdefe5 in iso-8859-7
|
|
u'\u0394\u03b5\u03ba': u'Dec', # c4e5ea in iso-8859-7
|
|
}
|
|
|
|
_greek_wdays = \
|
|
{ \
|
|
u'\u039a\u03c5\u03c1': u'Sun', # caf5f1 in iso-8859-7
|
|
u'\u0394\u03b5\u03c5': u'Mon', # c4e5f5 in iso-8859-7
|
|
u'\u03a4\u03c1\u03b9': u'Tue', # d4f1e9 in iso-8859-7
|
|
u'\u03a4\u03b5\u03c4': u'Wed', # d4e5f4 in iso-8859-7
|
|
u'\u03a0\u03b5\u03bc': u'Thu', # d0e5ec in iso-8859-7
|
|
u'\u03a0\u03b1\u03c1': u'Fri', # d0e1f1 in iso-8859-7
|
|
u'\u03a3\u03b1\u03b2': u'Sat', # d3e1e2 in iso-8859-7
|
|
}
|
|
|
|
_greek_date_format_re = \
|
|
re.compile(u'([^,]+),\s+(\d{2})\s+([^\s]+)\s+(\d{4})\s+(\d{2}):(\d{2}):(\d{2})\s+([^\s]+)')
|
|
|
|
def _parse_date_greek(dateString):
|
|
'''Parse a string according to a Greek 8-bit date format.'''
|
|
m = _greek_date_format_re.match(dateString)
|
|
if not m:
|
|
return
|
|
wday = _greek_wdays[m.group(1)]
|
|
month = _greek_months[m.group(3)]
|
|
rfc822date = '%(wday)s, %(day)s %(month)s %(year)s %(hour)s:%(minute)s:%(second)s %(zonediff)s' % \
|
|
{'wday': wday, 'day': m.group(2), 'month': month, 'year': m.group(4),\
|
|
'hour': m.group(5), 'minute': m.group(6), 'second': m.group(7),\
|
|
'zonediff': m.group(8)}
|
|
return _parse_date_rfc822(rfc822date)
|
|
registerDateHandler(_parse_date_greek)
|
|
|
|
# Unicode strings for Hungarian date strings
|
|
_hungarian_months = \
|
|
{ \
|
|
u'janu\u00e1r': u'01', # e1 in iso-8859-2
|
|
u'febru\u00e1ri': u'02', # e1 in iso-8859-2
|
|
u'm\u00e1rcius': u'03', # e1 in iso-8859-2
|
|
u'\u00e1prilis': u'04', # e1 in iso-8859-2
|
|
u'm\u00e1ujus': u'05', # e1 in iso-8859-2
|
|
u'j\u00fanius': u'06', # fa in iso-8859-2
|
|
u'j\u00falius': u'07', # fa in iso-8859-2
|
|
u'augusztus': u'08',
|
|
u'szeptember': u'09',
|
|
u'okt\u00f3ber': u'10', # f3 in iso-8859-2
|
|
u'november': u'11',
|
|
u'december': u'12',
|
|
}
|
|
|
|
_hungarian_date_format_re = \
|
|
re.compile(u'(\d{4})-([^-]+)-(\d{,2})T(\d{,2}):(\d{2})((\+|-)(\d{,2}:\d{2}))')
|
|
|
|
def _parse_date_hungarian(dateString):
|
|
'''Parse a string according to a Hungarian 8-bit date format.'''
|
|
m = _hungarian_date_format_re.match(dateString)
|
|
if not m or m.group(2) not in _hungarian_months:
|
|
return None
|
|
month = _hungarian_months[m.group(2)]
|
|
day = m.group(3)
|
|
if len(day) == 1:
|
|
day = '0' + day
|
|
hour = m.group(4)
|
|
if len(hour) == 1:
|
|
hour = '0' + hour
|
|
w3dtfdate = '%(year)s-%(month)s-%(day)sT%(hour)s:%(minute)s%(zonediff)s' % \
|
|
{'year': m.group(1), 'month': month, 'day': day,\
|
|
'hour': hour, 'minute': m.group(5),\
|
|
'zonediff': m.group(6)}
|
|
return _parse_date_w3dtf(w3dtfdate)
|
|
registerDateHandler(_parse_date_hungarian)
|
|
|
|
# W3DTF-style date parsing adapted from PyXML xml.utils.iso8601, written by
|
|
# Drake and licensed under the Python license. Removed all range checking
|
|
# for month, day, hour, minute, and second, since mktime will normalize
|
|
# these later
|
|
# Modified to also support MSSQL-style datetimes as defined at:
|
|
# http://msdn.microsoft.com/en-us/library/ms186724.aspx
|
|
# (which basically means allowing a space as a date/time/timezone separator)
|
|
def _parse_date_w3dtf(dateString):
|
|
def __extract_date(m):
|
|
year = int(m.group('year'))
|
|
if year < 100:
|
|
year = 100 * int(time.gmtime()[0] / 100) + int(year)
|
|
if year < 1000:
|
|
return 0, 0, 0
|
|
julian = m.group('julian')
|
|
if julian:
|
|
julian = int(julian)
|
|
month = julian / 30 + 1
|
|
day = julian % 30 + 1
|
|
jday = None
|
|
while jday != julian:
|
|
t = time.mktime((year, month, day, 0, 0, 0, 0, 0, 0))
|
|
jday = time.gmtime(t)[-2]
|
|
diff = abs(jday - julian)
|
|
if jday > julian:
|
|
if diff < day:
|
|
day = day - diff
|
|
else:
|
|
month = month - 1
|
|
day = 31
|
|
elif jday < julian:
|
|
if day + diff < 28:
|
|
day = day + diff
|
|
else:
|
|
month = month + 1
|
|
return year, month, day
|
|
month = m.group('month')
|
|
day = 1
|
|
if month is None:
|
|
month = 1
|
|
else:
|
|
month = int(month)
|
|
day = m.group('day')
|
|
if day:
|
|
day = int(day)
|
|
else:
|
|
day = 1
|
|
return year, month, day
|
|
|
|
def __extract_time(m):
|
|
if not m:
|
|
return 0, 0, 0
|
|
hours = m.group('hours')
|
|
if not hours:
|
|
return 0, 0, 0
|
|
hours = int(hours)
|
|
minutes = int(m.group('minutes'))
|
|
seconds = m.group('seconds')
|
|
if seconds:
|
|
seconds = int(seconds)
|
|
else:
|
|
seconds = 0
|
|
return hours, minutes, seconds
|
|
|
|
def __extract_tzd(m):
|
|
'''Return the Time Zone Designator as an offset in seconds from UTC.'''
|
|
if not m:
|
|
return 0
|
|
tzd = m.group('tzd')
|
|
if not tzd:
|
|
return 0
|
|
if tzd == 'Z':
|
|
return 0
|
|
hours = int(m.group('tzdhours'))
|
|
minutes = m.group('tzdminutes')
|
|
if minutes:
|
|
minutes = int(minutes)
|
|
else:
|
|
minutes = 0
|
|
offset = (hours*60 + minutes) * 60
|
|
if tzd[0] == '+':
|
|
return -offset
|
|
return offset
|
|
|
|
__date_re = ('(?P<year>\d\d\d\d)'
|
|
'(?:(?P<dsep>-|)'
|
|
'(?:(?P<month>\d\d)(?:(?P=dsep)(?P<day>\d\d))?'
|
|
'|(?P<julian>\d\d\d)))?')
|
|
__tzd_re = ' ?(?P<tzd>[-+](?P<tzdhours>\d\d)(?::?(?P<tzdminutes>\d\d))|Z)?'
|
|
__time_re = ('(?P<hours>\d\d)(?P<tsep>:|)(?P<minutes>\d\d)'
|
|
'(?:(?P=tsep)(?P<seconds>\d\d)(?:[.,]\d+)?)?'
|
|
+ __tzd_re)
|
|
__datetime_re = '%s(?:[T ]%s)?' % (__date_re, __time_re)
|
|
__datetime_rx = re.compile(__datetime_re)
|
|
m = __datetime_rx.match(dateString)
|
|
if (m is None) or (m.group() != dateString):
|
|
return
|
|
gmt = __extract_date(m) + __extract_time(m) + (0, 0, 0)
|
|
if gmt[0] == 0:
|
|
return
|
|
return time.gmtime(time.mktime(gmt) + __extract_tzd(m) - time.timezone)
|
|
registerDateHandler(_parse_date_w3dtf)
|
|
|
|
# Define the strings used by the RFC822 datetime parser
|
|
_rfc822_months = ['jan', 'feb', 'mar', 'apr', 'may', 'jun',
|
|
'jul', 'aug', 'sep', 'oct', 'nov', 'dec']
|
|
_rfc822_daynames = ['mon', 'tue', 'wed', 'thu', 'fri', 'sat', 'sun']
|
|
|
|
# Only the first three letters of the month name matter
|
|
_rfc822_month = "(?P<month>%s)(?:[a-z]*,?)" % ('|'.join(_rfc822_months))
|
|
# The year may be 2 or 4 digits; capture the century if it exists
|
|
_rfc822_year = "(?P<year>(?:\d{2})?\d{2})"
|
|
_rfc822_day = "(?P<day> *\d{1,2})"
|
|
_rfc822_date = "%s %s %s" % (_rfc822_day, _rfc822_month, _rfc822_year)
|
|
|
|
_rfc822_hour = "(?P<hour>\d{2}):(?P<minute>\d{2})(?::(?P<second>\d{2}))?"
|
|
_rfc822_tz = "(?P<tz>ut|gmt(?:[+-]\d{2}:\d{2})?|[aecmp][sd]?t|[zamny]|[+-]\d{4})"
|
|
_rfc822_tznames = {
|
|
'ut': 0, 'gmt': 0, 'z': 0,
|
|
'adt': -3, 'ast': -4, 'at': -4,
|
|
'edt': -4, 'est': -5, 'et': -5,
|
|
'cdt': -5, 'cst': -6, 'ct': -6,
|
|
'mdt': -6, 'mst': -7, 'mt': -7,
|
|
'pdt': -7, 'pst': -8, 'pt': -8,
|
|
'a': -1, 'n': 1,
|
|
'm': -12, 'y': 12,
|
|
}
|
|
# The timezone may be prefixed by 'Etc/'
|
|
_rfc822_time = "%s (?:etc/)?%s" % (_rfc822_hour, _rfc822_tz)
|
|
|
|
_rfc822_dayname = "(?P<dayname>%s)" % ('|'.join(_rfc822_daynames))
|
|
_rfc822_match = re.compile(
|
|
"(?:%s, )?%s(?: %s)?" % (_rfc822_dayname, _rfc822_date, _rfc822_time)
|
|
).match
|
|
|
|
def _parse_date_group_rfc822(m):
|
|
# Calculate a date and timestamp
|
|
for k in ('year', 'day', 'hour', 'minute', 'second'):
|
|
m[k] = int(m[k])
|
|
m['month'] = _rfc822_months.index(m['month']) + 1
|
|
# If the year is 2 digits, assume everything in the 90's is the 1990's
|
|
if m['year'] < 100:
|
|
m['year'] += (1900, 2000)[m['year'] < 90]
|
|
stamp = datetime.datetime(*[m[i] for i in
|
|
('year', 'month', 'day', 'hour', 'minute', 'second')])
|
|
|
|
# Use the timezone information to calculate the difference between
|
|
# the given date and timestamp and Universal Coordinated Time
|
|
tzhour = 0
|
|
tzmin = 0
|
|
if m['tz'] and m['tz'].startswith('gmt'):
|
|
# Handle GMT and GMT+hh:mm timezone syntax (the trailing
|
|
# timezone info will be handled by the next `if` block)
|
|
m['tz'] = ''.join(m['tz'][3:].split(':')) or 'gmt'
|
|
if not m['tz']:
|
|
pass
|
|
elif m['tz'].startswith('+'):
|
|
tzhour = int(m['tz'][1:3])
|
|
tzmin = int(m['tz'][3:])
|
|
elif m['tz'].startswith('-'):
|
|
tzhour = int(m['tz'][1:3]) * -1
|
|
tzmin = int(m['tz'][3:]) * -1
|
|
else:
|
|
tzhour = _rfc822_tznames[m['tz']]
|
|
delta = datetime.timedelta(0, 0, 0, 0, tzmin, tzhour)
|
|
|
|
# Return the date and timestamp in UTC
|
|
return (stamp - delta).utctimetuple()
|
|
|
|
def _parse_date_rfc822(dt):
|
|
"""Parse RFC 822 dates and times, with one minor
|
|
difference: years may be 4DIGIT or 2DIGIT.
|
|
http://tools.ietf.org/html/rfc822#section-5"""
|
|
try:
|
|
m = _rfc822_match(dt.lower()).groupdict(0)
|
|
except AttributeError:
|
|
return None
|
|
|
|
return _parse_date_group_rfc822(m)
|
|
registerDateHandler(_parse_date_rfc822)
|
|
|
|
def _parse_date_rfc822_grubby(dt):
|
|
"""Parse date format similar to RFC 822, but
|
|
the comma after the dayname is optional and
|
|
month/day are inverted"""
|
|
_rfc822_date_grubby = "%s %s %s" % (_rfc822_month, _rfc822_day, _rfc822_year)
|
|
_rfc822_match_grubby = re.compile(
|
|
"(?:%s[,]? )?%s(?: %s)?" % (_rfc822_dayname, _rfc822_date_grubby, _rfc822_time)
|
|
).match
|
|
|
|
try:
|
|
m = _rfc822_match_grubby(dt.lower()).groupdict(0)
|
|
except AttributeError:
|
|
return None
|
|
|
|
return _parse_date_group_rfc822(m)
|
|
registerDateHandler(_parse_date_rfc822_grubby)
|
|
|
|
def _parse_date_asctime(dt):
|
|
"""Parse asctime-style dates"""
|
|
dayname, month, day, remainder = dt.split(None, 3)
|
|
# Convert month and day into zero-padded integers
|
|
month = '%02i ' % (_rfc822_months.index(month.lower()) + 1)
|
|
day = '%02i ' % (int(day),)
|
|
dt = month + day + remainder
|
|
return time.strptime(dt, '%m %d %H:%M:%S %Y')[:-1] + (0, )
|
|
registerDateHandler(_parse_date_asctime)
|
|
|
|
def _parse_date_perforce(aDateString):
|
|
"""parse a date in yyyy/mm/dd hh:mm:ss TTT format"""
|
|
# Fri, 2006/09/15 08:19:53 EDT
|
|
_my_date_pattern = re.compile( \
|
|
r'(\w{,3}), (\d{,4})/(\d{,2})/(\d{2}) (\d{,2}):(\d{2}):(\d{2}) (\w{,3})')
|
|
|
|
m = _my_date_pattern.search(aDateString)
|
|
if m is None:
|
|
return None
|
|
dow, year, month, day, hour, minute, second, tz = m.groups()
|
|
months = ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec']
|
|
dateString = "%s, %s %s %s %s:%s:%s %s" % (dow, day, months[int(month) - 1], year, hour, minute, second, tz)
|
|
tm = rfc822.parsedate_tz(dateString)
|
|
if tm:
|
|
return time.gmtime(rfc822.mktime_tz(tm))
|
|
registerDateHandler(_parse_date_perforce)
|
|
|
|
def _parse_date(dateString):
|
|
'''Parses a variety of date formats into a 9-tuple in GMT'''
|
|
if not dateString:
|
|
return None
|
|
for handler in _date_handlers:
|
|
try:
|
|
date9tuple = handler(dateString)
|
|
except (KeyError, OverflowError, ValueError):
|
|
continue
|
|
if not date9tuple:
|
|
continue
|
|
if len(date9tuple) != 9:
|
|
continue
|
|
return date9tuple
|
|
return None
|
|
|
|
# Each marker represents some of the characters of the opening XML
|
|
# processing instruction ('<?xm') in the specified encoding.
|
|
EBCDIC_MARKER = _l2bytes([0x4C, 0x6F, 0xA7, 0x94])
|
|
UTF16BE_MARKER = _l2bytes([0x00, 0x3C, 0x00, 0x3F])
|
|
UTF16LE_MARKER = _l2bytes([0x3C, 0x00, 0x3F, 0x00])
|
|
UTF32BE_MARKER = _l2bytes([0x00, 0x00, 0x00, 0x3C])
|
|
UTF32LE_MARKER = _l2bytes([0x3C, 0x00, 0x00, 0x00])
|
|
|
|
ZERO_BYTES = _l2bytes([0x00, 0x00])
|
|
|
|
# Match the opening XML declaration.
|
|
# Example: <?xml version="1.0" encoding="utf-8"?>
|
|
RE_XML_DECLARATION = re.compile('^<\?xml[^>]*?>')
|
|
|
|
# Capture the value of the XML processing instruction's encoding attribute.
|
|
# Example: <?xml version="1.0" encoding="utf-8"?>
|
|
RE_XML_PI_ENCODING = re.compile(_s2bytes('^<\?.*encoding=[\'"](.*?)[\'"].*\?>'))
|
|
|
|
def convert_to_utf8(http_headers, data):
|
|
'''Detect and convert the character encoding to UTF-8.
|
|
|
|
http_headers is a dictionary
|
|
data is a raw string (not Unicode)'''
|
|
|
|
# This is so much trickier than it sounds, it's not even funny.
|
|
# According to RFC 3023 ('XML Media Types'), if the HTTP Content-Type
|
|
# is application/xml, application/*+xml,
|
|
# application/xml-external-parsed-entity, or application/xml-dtd,
|
|
# the encoding given in the charset parameter of the HTTP Content-Type
|
|
# takes precedence over the encoding given in the XML prefix within the
|
|
# document, and defaults to 'utf-8' if neither are specified. But, if
|
|
# the HTTP Content-Type is text/xml, text/*+xml, or
|
|
# text/xml-external-parsed-entity, the encoding given in the XML prefix
|
|
# within the document is ALWAYS IGNORED and only the encoding given in
|
|
# the charset parameter of the HTTP Content-Type header should be
|
|
# respected, and it defaults to 'us-ascii' if not specified.
|
|
|
|
# Furthermore, discussion on the atom-syntax mailing list with the
|
|
# author of RFC 3023 leads me to the conclusion that any document
|
|
# served with a Content-Type of text/* and no charset parameter
|
|
# must be treated as us-ascii. (We now do this.) And also that it
|
|
# must always be flagged as non-well-formed. (We now do this too.)
|
|
|
|
# If Content-Type is unspecified (input was local file or non-HTTP source)
|
|
# or unrecognized (server just got it totally wrong), then go by the
|
|
# encoding given in the XML prefix of the document and default to
|
|
# 'iso-8859-1' as per the HTTP specification (RFC 2616).
|
|
|
|
# Then, assuming we didn't find a character encoding in the HTTP headers
|
|
# (and the HTTP Content-type allowed us to look in the body), we need
|
|
# to sniff the first few bytes of the XML data and try to determine
|
|
# whether the encoding is ASCII-compatible. Section F of the XML
|
|
# specification shows the way here:
|
|
# http://www.w3.org/TR/REC-xml/#sec-guessing-no-ext-info
|
|
|
|
# If the sniffed encoding is not ASCII-compatible, we need to make it
|
|
# ASCII compatible so that we can sniff further into the XML declaration
|
|
# to find the encoding attribute, which will tell us the true encoding.
|
|
|
|
# Of course, none of this guarantees that we will be able to parse the
|
|
# feed in the declared character encoding (assuming it was declared
|
|
# correctly, which many are not). iconv_codec can help a lot;
|
|
# you should definitely install it if you can.
|
|
# http://cjkpython.i18n.org/
|
|
|
|
bom_encoding = u''
|
|
xml_encoding = u''
|
|
rfc3023_encoding = u''
|
|
|
|
# Look at the first few bytes of the document to guess what
|
|
# its encoding may be. We only need to decode enough of the
|
|
# document that we can use an ASCII-compatible regular
|
|
# expression to search for an XML encoding declaration.
|
|
# The heuristic follows the XML specification, section F:
|
|
# http://www.w3.org/TR/REC-xml/#sec-guessing-no-ext-info
|
|
# Check for BOMs first.
|
|
if data[:4] == codecs.BOM_UTF32_BE:
|
|
bom_encoding = u'utf-32be'
|
|
data = data[4:]
|
|
elif data[:4] == codecs.BOM_UTF32_LE:
|
|
bom_encoding = u'utf-32le'
|
|
data = data[4:]
|
|
elif data[:2] == codecs.BOM_UTF16_BE and data[2:4] != ZERO_BYTES:
|
|
bom_encoding = u'utf-16be'
|
|
data = data[2:]
|
|
elif data[:2] == codecs.BOM_UTF16_LE and data[2:4] != ZERO_BYTES:
|
|
bom_encoding = u'utf-16le'
|
|
data = data[2:]
|
|
elif data[:3] == codecs.BOM_UTF8:
|
|
bom_encoding = u'utf-8'
|
|
data = data[3:]
|
|
# Check for the characters '<?xm' in several encodings.
|
|
elif data[:4] == EBCDIC_MARKER:
|
|
bom_encoding = u'cp037'
|
|
elif data[:4] == UTF16BE_MARKER:
|
|
bom_encoding = u'utf-16be'
|
|
elif data[:4] == UTF16LE_MARKER:
|
|
bom_encoding = u'utf-16le'
|
|
elif data[:4] == UTF32BE_MARKER:
|
|
bom_encoding = u'utf-32be'
|
|
elif data[:4] == UTF32LE_MARKER:
|
|
bom_encoding = u'utf-32le'
|
|
|
|
tempdata = data
|
|
try:
|
|
if bom_encoding:
|
|
tempdata = data.decode(bom_encoding).encode('utf-8')
|
|
except (UnicodeDecodeError, LookupError):
|
|
# feedparser recognizes UTF-32 encodings that aren't
|
|
# available in Python 2.4 and 2.5, so it's possible to
|
|
# encounter a LookupError during decoding.
|
|
xml_encoding_match = None
|
|
else:
|
|
xml_encoding_match = RE_XML_PI_ENCODING.match(tempdata)
|
|
|
|
if xml_encoding_match:
|
|
xml_encoding = xml_encoding_match.groups()[0].decode('utf-8').lower()
|
|
# Normalize the xml_encoding if necessary.
|
|
if bom_encoding and (xml_encoding in (
|
|
u'u16', u'utf-16', u'utf16', u'utf_16',
|
|
u'u32', u'utf-32', u'utf32', u'utf_32',
|
|
u'iso-10646-ucs-2', u'iso-10646-ucs-4',
|
|
u'csucs4', u'csunicode', u'ucs-2', u'ucs-4'
|
|
)):
|
|
xml_encoding = bom_encoding
|
|
|
|
# Find the HTTP Content-Type and, hopefully, a character
|
|
# encoding provided by the server. The Content-Type is used
|
|
# to choose the "correct" encoding among the BOM encoding,
|
|
# XML declaration encoding, and HTTP encoding, following the
|
|
# heuristic defined in RFC 3023.
|
|
http_content_type = http_headers.get('content-type') or ''
|
|
http_content_type, params = cgi.parse_header(http_content_type)
|
|
http_encoding = params.get('charset', '').replace("'", "")
|
|
if not isinstance(http_encoding, unicode):
|
|
http_encoding = http_encoding.decode('utf-8', 'ignore')
|
|
|
|
acceptable_content_type = 0
|
|
application_content_types = (u'application/xml', u'application/xml-dtd',
|
|
u'application/xml-external-parsed-entity')
|
|
text_content_types = (u'text/xml', u'text/xml-external-parsed-entity')
|
|
if (http_content_type in application_content_types) or \
|
|
(http_content_type.startswith(u'application/') and
|
|
http_content_type.endswith(u'+xml')):
|
|
acceptable_content_type = 1
|
|
rfc3023_encoding = http_encoding or xml_encoding or u'utf-8'
|
|
elif (http_content_type in text_content_types) or \
|
|
(http_content_type.startswith(u'text/') and
|
|
http_content_type.endswith(u'+xml')):
|
|
acceptable_content_type = 1
|
|
rfc3023_encoding = http_encoding or u'us-ascii'
|
|
elif http_content_type.startswith(u'text/'):
|
|
rfc3023_encoding = http_encoding or u'us-ascii'
|
|
elif http_headers and 'content-type' not in http_headers:
|
|
rfc3023_encoding = xml_encoding or u'iso-8859-1'
|
|
else:
|
|
rfc3023_encoding = xml_encoding or u'utf-8'
|
|
# gb18030 is a superset of gb2312, so always replace gb2312
|
|
# with gb18030 for greater compatibility.
|
|
if rfc3023_encoding.lower() == u'gb2312':
|
|
rfc3023_encoding = u'gb18030'
|
|
if xml_encoding.lower() == u'gb2312':
|
|
xml_encoding = u'gb18030'
|
|
|
|
# there are four encodings to keep track of:
|
|
# - http_encoding is the encoding declared in the Content-Type HTTP header
|
|
# - xml_encoding is the encoding declared in the <?xml declaration
|
|
# - bom_encoding is the encoding sniffed from the first 4 bytes of the XML data
|
|
# - rfc3023_encoding is the actual encoding, as per RFC 3023 and a variety of other conflicting specifications
|
|
error = None
|
|
|
|
if http_headers and (not acceptable_content_type):
|
|
if 'content-type' in http_headers:
|
|
msg = '%s is not an XML media type' % http_headers['content-type']
|
|
else:
|
|
msg = 'no Content-type specified'
|
|
error = NonXMLContentType(msg)
|
|
|
|
# determine character encoding
|
|
known_encoding = 0
|
|
chardet_encoding = None
|
|
tried_encodings = []
|
|
if chardet:
|
|
chardet_encoding = unicode(chardet.detect(data)['encoding'] or '', 'ascii', 'ignore')
|
|
# try: HTTP encoding, declared XML encoding, encoding sniffed from BOM
|
|
for proposed_encoding in (rfc3023_encoding, xml_encoding, bom_encoding,
|
|
chardet_encoding, u'utf-8', u'windows-1252', u'iso-8859-2'):
|
|
if not proposed_encoding:
|
|
continue
|
|
if proposed_encoding in tried_encodings:
|
|
continue
|
|
tried_encodings.append(proposed_encoding)
|
|
try:
|
|
data = data.decode(proposed_encoding)
|
|
except (UnicodeDecodeError, LookupError):
|
|
pass
|
|
else:
|
|
known_encoding = 1
|
|
# Update the encoding in the opening XML processing instruction.
|
|
new_declaration = '''<?xml version='1.0' encoding='utf-8'?>'''
|
|
if RE_XML_DECLARATION.search(data):
|
|
data = RE_XML_DECLARATION.sub(new_declaration, data)
|
|
else:
|
|
data = new_declaration + u'\n' + data
|
|
data = data.encode('utf-8')
|
|
break
|
|
# if still no luck, give up
|
|
if not known_encoding:
|
|
error = CharacterEncodingUnknown(
|
|
'document encoding unknown, I tried ' +
|
|
'%s, %s, utf-8, windows-1252, and iso-8859-2 but nothing worked' %
|
|
(rfc3023_encoding, xml_encoding))
|
|
rfc3023_encoding = u''
|
|
elif proposed_encoding != rfc3023_encoding:
|
|
error = CharacterEncodingOverride(
|
|
'document declared as %s, but parsed as %s' %
|
|
(rfc3023_encoding, proposed_encoding))
|
|
rfc3023_encoding = proposed_encoding
|
|
|
|
return data, rfc3023_encoding, error
|
|
|
|
# Match XML entity declarations.
|
|
# Example: <!ENTITY copyright "(C)">
|
|
RE_ENTITY_PATTERN = re.compile(_s2bytes(r'^\s*<!ENTITY([^>]*?)>'), re.MULTILINE)
|
|
|
|
# Match XML DOCTYPE declarations.
|
|
# Example: <!DOCTYPE feed [ ]>
|
|
RE_DOCTYPE_PATTERN = re.compile(_s2bytes(r'^\s*<!DOCTYPE([^>]*?)>'), re.MULTILINE)
|
|
|
|
# Match safe entity declarations.
|
|
# This will allow hexadecimal character references through,
|
|
# as well as text, but not arbitrary nested entities.
|
|
# Example: cubed "³"
|
|
# Example: copyright "(C)"
|
|
# Forbidden: explode1 "&explode2;&explode2;"
|
|
RE_SAFE_ENTITY_PATTERN = re.compile(_s2bytes('\s+(\w+)\s+"(&#\w+;|[^&"]*)"'))
|
|
|
|
def replace_doctype(data):
|
|
'''Strips and replaces the DOCTYPE, returns (rss_version, stripped_data)
|
|
|
|
rss_version may be 'rss091n' or None
|
|
stripped_data is the same XML document with a replaced DOCTYPE
|
|
'''
|
|
|
|
# Divide the document into two groups by finding the location
|
|
# of the first element that doesn't begin with '<?' or '<!'.
|
|
start = re.search(_s2bytes('<\w'), data)
|
|
start = start and start.start() or -1
|
|
head, data = data[:start+1], data[start+1:]
|
|
|
|
# Save and then remove all of the ENTITY declarations.
|
|
entity_results = RE_ENTITY_PATTERN.findall(head)
|
|
head = RE_ENTITY_PATTERN.sub(_s2bytes(''), head)
|
|
|
|
# Find the DOCTYPE declaration and check the feed type.
|
|
doctype_results = RE_DOCTYPE_PATTERN.findall(head)
|
|
doctype = doctype_results and doctype_results[0] or _s2bytes('')
|
|
if _s2bytes('netscape') in doctype.lower():
|
|
version = u'rss091n'
|
|
else:
|
|
version = None
|
|
|
|
# Re-insert the safe ENTITY declarations if a DOCTYPE was found.
|
|
replacement = _s2bytes('')
|
|
if len(doctype_results) == 1 and entity_results:
|
|
match_safe_entities = lambda e: RE_SAFE_ENTITY_PATTERN.match(e)
|
|
safe_entities = filter(match_safe_entities, entity_results)
|
|
if safe_entities:
|
|
replacement = _s2bytes('<!DOCTYPE feed [\n<!ENTITY') \
|
|
+ _s2bytes('>\n<!ENTITY ').join(safe_entities) \
|
|
+ _s2bytes('>\n]>')
|
|
data = RE_DOCTYPE_PATTERN.sub(replacement, head) + data
|
|
|
|
# Precompute the safe entities for the loose parser.
|
|
safe_entities = dict((k.decode('utf-8'), v.decode('utf-8'))
|
|
for k, v in RE_SAFE_ENTITY_PATTERN.findall(replacement))
|
|
return version, data, safe_entities
|
|
|
|
def parse(url_file_stream_or_string, etag=None, modified=None, agent=None, referrer=None, handlers=None, request_headers=None, response_headers=None):
|
|
'''Parse a feed from a URL, file, stream, or string.
|
|
|
|
request_headers, if given, is a dict from http header name to value to add
|
|
to the request; this overrides internally generated values.
|
|
'''
|
|
|
|
if handlers is None:
|
|
handlers = []
|
|
if request_headers is None:
|
|
request_headers = {}
|
|
if response_headers is None:
|
|
response_headers = {}
|
|
|
|
result = FeedParserDict()
|
|
result['feed'] = FeedParserDict()
|
|
result['entries'] = []
|
|
result['bozo'] = 0
|
|
if not isinstance(handlers, list):
|
|
handlers = [handlers]
|
|
try:
|
|
f = _open_resource(url_file_stream_or_string, etag, modified, agent, referrer, handlers, request_headers)
|
|
data = f.read()
|
|
except Exception, e:
|
|
result['bozo'] = 1
|
|
result['bozo_exception'] = e
|
|
data = None
|
|
f = None
|
|
|
|
if hasattr(f, 'headers'):
|
|
result['headers'] = dict(f.headers)
|
|
# overwrite existing headers using response_headers
|
|
if 'headers' in result:
|
|
result['headers'].update(response_headers)
|
|
elif response_headers:
|
|
result['headers'] = copy.deepcopy(response_headers)
|
|
|
|
# lowercase all of the HTTP headers for comparisons per RFC 2616
|
|
if 'headers' in result:
|
|
http_headers = dict((k.lower(), v) for k, v in result['headers'].items())
|
|
else:
|
|
http_headers = {}
|
|
|
|
# if feed is gzip-compressed, decompress it
|
|
if f and data and http_headers:
|
|
if gzip and 'gzip' in http_headers.get('content-encoding', ''):
|
|
try:
|
|
data = gzip.GzipFile(fileobj=_StringIO(data)).read()
|
|
except (IOError, struct.error), e:
|
|
# IOError can occur if the gzip header is bad.
|
|
# struct.error can occur if the data is damaged.
|
|
result['bozo'] = 1
|
|
result['bozo_exception'] = e
|
|
if isinstance(e, struct.error):
|
|
# A gzip header was found but the data is corrupt.
|
|
# Ideally, we should re-request the feed without the
|
|
# 'Accept-encoding: gzip' header, but we don't.
|
|
data = None
|
|
elif zlib and 'deflate' in http_headers.get('content-encoding', ''):
|
|
try:
|
|
data = zlib.decompress(data)
|
|
except zlib.error, e:
|
|
try:
|
|
# The data may have no headers and no checksum.
|
|
data = zlib.decompress(data, -15)
|
|
except zlib.error, e:
|
|
result['bozo'] = 1
|
|
result['bozo_exception'] = e
|
|
|
|
# save HTTP headers
|
|
if http_headers:
|
|
if 'etag' in http_headers:
|
|
etag = http_headers.get('etag', u'')
|
|
if not isinstance(etag, unicode):
|
|
etag = etag.decode('utf-8', 'ignore')
|
|
if etag:
|
|
result['etag'] = etag
|
|
if 'last-modified' in http_headers:
|
|
modified = http_headers.get('last-modified', u'')
|
|
if modified:
|
|
result['modified'] = modified
|
|
result['modified_parsed'] = _parse_date(modified)
|
|
if hasattr(f, 'url'):
|
|
if not isinstance(f.url, unicode):
|
|
result['href'] = f.url.decode('utf-8', 'ignore')
|
|
else:
|
|
result['href'] = f.url
|
|
result['status'] = 200
|
|
if hasattr(f, 'status'):
|
|
result['status'] = f.status
|
|
if hasattr(f, 'close'):
|
|
f.close()
|
|
|
|
if data is None:
|
|
return result
|
|
|
|
# Stop processing if the server sent HTTP 304 Not Modified.
|
|
if getattr(f, 'code', 0) == 304:
|
|
result['version'] = u''
|
|
result['debug_message'] = 'The feed has not changed since you last checked, ' + \
|
|
'so the server sent no data. This is a feature, not a bug!'
|
|
return result
|
|
|
|
data, result['encoding'], error = convert_to_utf8(http_headers, data)
|
|
use_strict_parser = result['encoding'] and True or False
|
|
if error is not None:
|
|
result['bozo'] = 1
|
|
result['bozo_exception'] = error
|
|
|
|
result['version'], data, entities = replace_doctype(data)
|
|
|
|
# Ensure that baseuri is an absolute URI using an acceptable URI scheme.
|
|
contentloc = http_headers.get('content-location', u'')
|
|
href = result.get('href', u'')
|
|
baseuri = _makeSafeAbsoluteURI(href, contentloc) or _makeSafeAbsoluteURI(contentloc) or href
|
|
|
|
baselang = http_headers.get('content-language', None)
|
|
if not isinstance(baselang, unicode) and baselang is not None:
|
|
baselang = baselang.decode('utf-8', 'ignore')
|
|
|
|
if not _XML_AVAILABLE:
|
|
use_strict_parser = 0
|
|
if use_strict_parser:
|
|
# initialize the SAX parser
|
|
feedparser = _StrictFeedParser(baseuri, baselang, 'utf-8')
|
|
saxparser = xml.sax.make_parser(PREFERRED_XML_PARSERS)
|
|
saxparser.setFeature(xml.sax.handler.feature_namespaces, 1)
|
|
try:
|
|
# disable downloading external doctype references, if possible
|
|
saxparser.setFeature(xml.sax.handler.feature_external_ges, 0)
|
|
except xml.sax.SAXNotSupportedException:
|
|
pass
|
|
saxparser.setContentHandler(feedparser)
|
|
saxparser.setErrorHandler(feedparser)
|
|
source = xml.sax.xmlreader.InputSource()
|
|
source.setByteStream(_StringIO(data))
|
|
try:
|
|
saxparser.parse(source)
|
|
except xml.sax.SAXException, e:
|
|
result['bozo'] = 1
|
|
result['bozo_exception'] = feedparser.exc or e
|
|
use_strict_parser = 0
|
|
if not use_strict_parser and _SGML_AVAILABLE:
|
|
feedparser = _LooseFeedParser(baseuri, baselang, 'utf-8', entities)
|
|
feedparser.feed(data.decode('utf-8', 'replace'))
|
|
result['feed'] = feedparser.feeddata
|
|
result['entries'] = feedparser.entries
|
|
result['version'] = result['version'] or feedparser.version
|
|
result['namespaces'] = feedparser.namespacesInUse
|
|
return result
|