mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-17 14:55:07 -05:00
0d9fbc1ad7
This version of SickBeard uses both TVDB and TVRage to search and gather it's series data from allowing you to now have access to and download shows that you couldn't before because of being locked into only what TheTVDB had to offer. Also this edition is based off the code we used in our XEM editon so it does come with scene numbering support as well as all the other features our XEM edition has to offer. Please before using this with your existing database (sickbeard.db) please make a backup copy of it and delete any other database files such as cache.db and failed.db if present, we HIGHLY recommend starting out with no database files at all to make this a fresh start but the choice is at your own risk! Enjoy!
877 lines
34 KiB
Python
877 lines
34 KiB
Python
"""
|
|
parser.http.utils module (imdb package).
|
|
|
|
This module provides miscellaneous utilities used by
|
|
the imdb.parser.http classes.
|
|
|
|
Copyright 2004-2012 Davide Alberani <da@erlug.linux.it>
|
|
2008 H. Turgut Uyar <uyar@tekir.org>
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 2 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program; if not, write to the Free Software
|
|
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
"""
|
|
|
|
import re
|
|
import logging
|
|
import warnings
|
|
|
|
from imdb._exceptions import IMDbError
|
|
|
|
from imdb.utils import flatten, _Container
|
|
from imdb.Movie import Movie
|
|
from imdb.Person import Person
|
|
from imdb.Character import Character
|
|
|
|
|
|
# Year, imdbIndex and kind.
|
|
re_yearKind_index = re.compile(r'(\([0-9\?]{4}(?:/[IVXLCDM]+)?\)(?: \(mini\)| \(TV\)| \(V\)| \(VG\))?)')
|
|
|
|
# Match imdb ids in href tags
|
|
re_imdbid = re.compile(r'(title/tt|name/nm|character/ch|company/co)([0-9]+)')
|
|
|
|
def analyze_imdbid(href):
|
|
"""Return an imdbID from an URL."""
|
|
if not href:
|
|
return None
|
|
match = re_imdbid.search(href)
|
|
if not match:
|
|
return None
|
|
return str(match.group(2))
|
|
|
|
|
|
_modify_keys = list(Movie.keys_tomodify_list) + list(Person.keys_tomodify_list)
|
|
def _putRefs(d, re_titles, re_names, re_characters, lastKey=None):
|
|
"""Iterate over the strings inside list items or dictionary values,
|
|
substitutes movie titles and person names with the (qv) references."""
|
|
if isinstance(d, list):
|
|
for i in xrange(len(d)):
|
|
if isinstance(d[i], (unicode, str)):
|
|
if lastKey in _modify_keys:
|
|
if re_names:
|
|
d[i] = re_names.sub(ur"'\1' (qv)", d[i])
|
|
if re_titles:
|
|
d[i] = re_titles.sub(ur'_\1_ (qv)', d[i])
|
|
if re_characters:
|
|
d[i] = re_characters.sub(ur'#\1# (qv)', d[i])
|
|
elif isinstance(d[i], (list, dict)):
|
|
_putRefs(d[i], re_titles, re_names, re_characters,
|
|
lastKey=lastKey)
|
|
elif isinstance(d, dict):
|
|
for k, v in d.items():
|
|
lastKey = k
|
|
if isinstance(v, (unicode, str)):
|
|
if lastKey in _modify_keys:
|
|
if re_names:
|
|
d[k] = re_names.sub(ur"'\1' (qv)", v)
|
|
if re_titles:
|
|
d[k] = re_titles.sub(ur'_\1_ (qv)', v)
|
|
if re_characters:
|
|
d[k] = re_characters.sub(ur'#\1# (qv)', v)
|
|
elif isinstance(v, (list, dict)):
|
|
_putRefs(d[k], re_titles, re_names, re_characters,
|
|
lastKey=lastKey)
|
|
|
|
|
|
# Handle HTML/XML/SGML entities.
|
|
from htmlentitydefs import entitydefs
|
|
entitydefs = entitydefs.copy()
|
|
entitydefsget = entitydefs.get
|
|
entitydefs['nbsp'] = ' '
|
|
|
|
sgmlentity = {'lt': '<', 'gt': '>', 'amp': '&', 'quot': '"', 'apos': '\'', 'ndash': '-'}
|
|
sgmlentityget = sgmlentity.get
|
|
_sgmlentkeys = sgmlentity.keys()
|
|
|
|
entcharrefs = {}
|
|
entcharrefsget = entcharrefs.get
|
|
for _k, _v in entitydefs.items():
|
|
if _k in _sgmlentkeys: continue
|
|
if _v[0:2] == '&#':
|
|
dec_code = _v[1:-1]
|
|
_v = unichr(int(_v[2:-1]))
|
|
entcharrefs[dec_code] = _v
|
|
else:
|
|
dec_code = '#' + str(ord(_v))
|
|
_v = unicode(_v, 'latin_1', 'replace')
|
|
entcharrefs[dec_code] = _v
|
|
entcharrefs[_k] = _v
|
|
del _sgmlentkeys, _k, _v
|
|
entcharrefs['#160'] = u' '
|
|
entcharrefs['#xA0'] = u' '
|
|
entcharrefs['#xa0'] = u' '
|
|
entcharrefs['#XA0'] = u' '
|
|
entcharrefs['#x22'] = u'"'
|
|
entcharrefs['#X22'] = u'"'
|
|
# convert &x26; to &, to make BeautifulSoup happy; beware that this
|
|
# leaves lone '&' in the html broken, but I assume this is better than
|
|
# the contrary...
|
|
entcharrefs['#38'] = u'&'
|
|
entcharrefs['#x26'] = u'&'
|
|
entcharrefs['#x26'] = u'&'
|
|
|
|
re_entcharrefs = re.compile('&(%s|\#160|\#\d{1,5}|\#x[0-9a-f]{1,4});' %
|
|
'|'.join(map(re.escape, entcharrefs)), re.I)
|
|
re_entcharrefssub = re_entcharrefs.sub
|
|
|
|
sgmlentity.update(dict([('#34', u'"'), ('#38', u'&'),
|
|
('#60', u'<'), ('#62', u'>'), ('#39', u"'")]))
|
|
re_sgmlref = re.compile('&(%s);' % '|'.join(map(re.escape, sgmlentity)))
|
|
re_sgmlrefsub = re_sgmlref.sub
|
|
|
|
# Matches XML-only single tags, like <br/> ; they are invalid in HTML,
|
|
# but widely used by IMDb web site. :-/
|
|
re_xmltags = re.compile('<([a-zA-Z]+)/>')
|
|
|
|
|
|
def _replXMLRef(match):
|
|
"""Replace the matched XML/HTML entities and references;
|
|
replace everything except sgml entities like <, >, ..."""
|
|
ref = match.group(1)
|
|
value = entcharrefsget(ref)
|
|
if value is None:
|
|
if ref[0] == '#':
|
|
ref_code = ref[1:]
|
|
if ref_code in ('34', '38', '60', '62', '39'):
|
|
return match.group(0)
|
|
elif ref_code[0].lower() == 'x':
|
|
#if ref[2:] == '26':
|
|
# # Don't convert &x26; to &, to make BeautifulSoup happy.
|
|
# return '&'
|
|
return unichr(int(ref[2:], 16))
|
|
else:
|
|
return unichr(int(ref[1:]))
|
|
else:
|
|
return ref
|
|
return value
|
|
|
|
def subXMLRefs(s):
|
|
"""Return the given html string with entity and char references
|
|
replaced."""
|
|
return re_entcharrefssub(_replXMLRef, s)
|
|
|
|
# XXX: no more used here; move it to mobile (they are imported by helpers, too)?
|
|
def _replSGMLRefs(match):
|
|
"""Replace the matched SGML entity."""
|
|
ref = match.group(1)
|
|
return sgmlentityget(ref, ref)
|
|
|
|
def subSGMLRefs(s):
|
|
"""Return the given html string with sgml entity and char references
|
|
replaced."""
|
|
return re_sgmlrefsub(_replSGMLRefs, s)
|
|
|
|
|
|
_b_p_logger = logging.getLogger('imdbpy.parser.http.build_person')
|
|
def build_person(txt, personID=None, billingPos=None,
|
|
roleID=None, accessSystem='http', modFunct=None):
|
|
"""Return a Person instance from the tipical <tr>...</tr> strings
|
|
found in the IMDb's web site."""
|
|
#if personID is None
|
|
# _b_p_logger.debug('empty name or personID for "%s"', txt)
|
|
notes = u''
|
|
role = u''
|
|
# Search the (optional) separator between name and role/notes.
|
|
if txt.find('....') != -1:
|
|
sep = '....'
|
|
elif txt.find('...') != -1:
|
|
sep = '...'
|
|
else:
|
|
sep = '...'
|
|
# Replace the first parenthesis, assuming there are only
|
|
# notes, after.
|
|
# Rationale: no imdbIndex is (ever?) showed on the web site.
|
|
txt = txt.replace('(', '...(', 1)
|
|
txt_split = txt.split(sep, 1)
|
|
name = txt_split[0].strip()
|
|
if len(txt_split) == 2:
|
|
role_comment = txt_split[1].strip()
|
|
# Strip common endings.
|
|
if role_comment[-4:] == ' and':
|
|
role_comment = role_comment[:-4].rstrip()
|
|
elif role_comment[-2:] == ' &':
|
|
role_comment = role_comment[:-2].rstrip()
|
|
elif role_comment[-6:] == '& ....':
|
|
role_comment = role_comment[:-6].rstrip()
|
|
# Get the notes.
|
|
if roleID is not None:
|
|
if not isinstance(roleID, list):
|
|
cmt_idx = role_comment.find('(')
|
|
if cmt_idx != -1:
|
|
role = role_comment[:cmt_idx].rstrip()
|
|
notes = role_comment[cmt_idx:]
|
|
else:
|
|
# Just a role, without notes.
|
|
role = role_comment
|
|
else:
|
|
role = role_comment
|
|
else:
|
|
# We're managing something that doesn't have a 'role', so
|
|
# everything are notes.
|
|
notes = role_comment
|
|
if role == '....': role = u''
|
|
roleNotes = []
|
|
# Manages multiple roleIDs.
|
|
if isinstance(roleID, list):
|
|
rolesplit = role.split('/')
|
|
role = []
|
|
for r in rolesplit:
|
|
nidx = r.find('(')
|
|
if nidx != -1:
|
|
role.append(r[:nidx].rstrip())
|
|
roleNotes.append(r[nidx:])
|
|
else:
|
|
role.append(r)
|
|
roleNotes.append(None)
|
|
lr = len(role)
|
|
lrid = len(roleID)
|
|
if lr > lrid:
|
|
roleID += [None] * (lrid - lr)
|
|
elif lr < lrid:
|
|
roleID = roleID[:lr]
|
|
for i, rid in enumerate(roleID):
|
|
if rid is not None:
|
|
roleID[i] = str(rid)
|
|
if lr == 1:
|
|
role = role[0]
|
|
roleID = roleID[0]
|
|
notes = roleNotes[0] or u''
|
|
elif roleID is not None:
|
|
roleID = str(roleID)
|
|
if personID is not None:
|
|
personID = str(personID)
|
|
if (not name) or (personID is None):
|
|
# Set to 'debug', since build_person is expected to receive some crap.
|
|
_b_p_logger.debug('empty name or personID for "%s"', txt)
|
|
# XXX: return None if something strange is detected?
|
|
person = Person(name=name, personID=personID, currentRole=role,
|
|
roleID=roleID, notes=notes, billingPos=billingPos,
|
|
modFunct=modFunct, accessSystem=accessSystem)
|
|
if roleNotes and len(roleNotes) == len(roleID):
|
|
for idx, role in enumerate(person.currentRole):
|
|
if roleNotes[idx]:
|
|
role.notes = roleNotes[idx]
|
|
return person
|
|
|
|
|
|
_re_chrIDs = re.compile('[0-9]{7}')
|
|
|
|
_b_m_logger = logging.getLogger('imdbpy.parser.http.build_movie')
|
|
# To shrink spaces.
|
|
re_spaces = re.compile(r'\s+')
|
|
def build_movie(txt, movieID=None, roleID=None, status=None,
|
|
accessSystem='http', modFunct=None, _parsingCharacter=False,
|
|
_parsingCompany=False, year=None, chrRoles=None,
|
|
rolesNoChar=None, additionalNotes=None):
|
|
"""Given a string as normally seen on the "categorized" page of
|
|
a person on the IMDb's web site, returns a Movie instance."""
|
|
# FIXME: Oook, lets face it: build_movie and build_person are now
|
|
# two horrible sets of patches to support the new IMDb design. They
|
|
# must be rewritten from scratch.
|
|
if _parsingCharacter:
|
|
_defSep = ' Played by '
|
|
elif _parsingCompany:
|
|
_defSep = ' ... '
|
|
else:
|
|
_defSep = ' .... '
|
|
title = re_spaces.sub(' ', txt).strip()
|
|
# Split the role/notes from the movie title.
|
|
tsplit = title.split(_defSep, 1)
|
|
role = u''
|
|
notes = u''
|
|
roleNotes = []
|
|
if len(tsplit) == 2:
|
|
title = tsplit[0].rstrip()
|
|
role = tsplit[1].lstrip()
|
|
if title[-9:] == 'TV Series':
|
|
title = title[:-9].rstrip()
|
|
#elif title[-7:] == '(short)':
|
|
# title = title[:-7].rstrip()
|
|
#elif title[-11:] == '(TV series)':
|
|
# title = title[:-11].rstrip()
|
|
#elif title[-10:] == '(TV movie)':
|
|
# title = title[:-10].rstrip()
|
|
elif title[-14:] == 'TV mini-series':
|
|
title = title[:-14] + ' (mini)'
|
|
if title and title.endswith(_defSep.rstrip()):
|
|
title = title[:-len(_defSep)+1]
|
|
# Try to understand where the movie title ends.
|
|
while True:
|
|
if year:
|
|
break
|
|
if title[-1:] != ')':
|
|
# Ignore the silly "TV Series" notice.
|
|
if title[-9:] == 'TV Series':
|
|
title = title[:-9].rstrip()
|
|
continue
|
|
else:
|
|
# Just a title: stop here.
|
|
break
|
|
# Try to match paired parentheses; yes: sometimes there are
|
|
# parentheses inside comments...
|
|
nidx = title.rfind('(')
|
|
while (nidx != -1 and \
|
|
title[nidx:].count('(') != title[nidx:].count(')')):
|
|
nidx = title[:nidx].rfind('(')
|
|
# Unbalanced parentheses: stop here.
|
|
if nidx == -1: break
|
|
# The last item in parentheses seems to be a year: stop here.
|
|
first4 = title[nidx+1:nidx+5]
|
|
if (first4.isdigit() or first4 == '????') and \
|
|
title[nidx+5:nidx+6] in (')', '/'): break
|
|
# The last item in parentheses is a known kind: stop here.
|
|
if title[nidx+1:-1] in ('TV', 'V', 'mini', 'VG', 'TV movie',
|
|
'TV series', 'short'): break
|
|
# Else, in parentheses there are some notes.
|
|
# XXX: should the notes in the role half be kept separated
|
|
# from the notes in the movie title half?
|
|
if notes: notes = '%s %s' % (title[nidx:], notes)
|
|
else: notes = title[nidx:]
|
|
title = title[:nidx].rstrip()
|
|
if year:
|
|
year = year.strip()
|
|
if title[-1] == ')':
|
|
fpIdx = title.rfind('(')
|
|
if fpIdx != -1:
|
|
if notes: notes = '%s %s' % (title[fpIdx:], notes)
|
|
else: notes = title[fpIdx:]
|
|
title = title[:fpIdx].rstrip()
|
|
title = u'%s (%s)' % (title, year)
|
|
if _parsingCharacter and roleID and not role:
|
|
roleID = None
|
|
if not roleID:
|
|
roleID = None
|
|
elif len(roleID) == 1:
|
|
roleID = roleID[0]
|
|
if not role and chrRoles and isinstance(roleID, (str, unicode)):
|
|
roleID = _re_chrIDs.findall(roleID)
|
|
role = ' / '.join(filter(None, chrRoles.split('@@')))
|
|
# Manages multiple roleIDs.
|
|
if isinstance(roleID, list):
|
|
tmprole = role.split('/')
|
|
role = []
|
|
for r in tmprole:
|
|
nidx = r.find('(')
|
|
if nidx != -1:
|
|
role.append(r[:nidx].rstrip())
|
|
roleNotes.append(r[nidx:])
|
|
else:
|
|
role.append(r)
|
|
roleNotes.append(None)
|
|
lr = len(role)
|
|
lrid = len(roleID)
|
|
if lr > lrid:
|
|
roleID += [None] * (lrid - lr)
|
|
elif lr < lrid:
|
|
roleID = roleID[:lr]
|
|
for i, rid in enumerate(roleID):
|
|
if rid is not None:
|
|
roleID[i] = str(rid)
|
|
if lr == 1:
|
|
role = role[0]
|
|
roleID = roleID[0]
|
|
elif roleID is not None:
|
|
roleID = str(roleID)
|
|
if movieID is not None:
|
|
movieID = str(movieID)
|
|
if (not title) or (movieID is None):
|
|
_b_m_logger.error('empty title or movieID for "%s"', txt)
|
|
if rolesNoChar:
|
|
rolesNoChar = filter(None, [x.strip() for x in rolesNoChar.split('/')])
|
|
if not role:
|
|
role = []
|
|
elif not isinstance(role, list):
|
|
role = [role]
|
|
role += rolesNoChar
|
|
notes = notes.strip()
|
|
if additionalNotes:
|
|
additionalNotes = re_spaces.sub(' ', additionalNotes).strip()
|
|
if notes:
|
|
notes += u' '
|
|
notes += additionalNotes
|
|
if role and isinstance(role, list) and notes.endswith(role[-1].replace('\n', ' ')):
|
|
role = role[:-1]
|
|
m = Movie(title=title, movieID=movieID, notes=notes, currentRole=role,
|
|
roleID=roleID, roleIsPerson=_parsingCharacter,
|
|
modFunct=modFunct, accessSystem=accessSystem)
|
|
if roleNotes and len(roleNotes) == len(roleID):
|
|
for idx, role in enumerate(m.currentRole):
|
|
try:
|
|
if roleNotes[idx]:
|
|
role.notes = roleNotes[idx]
|
|
except IndexError:
|
|
break
|
|
# Status can't be checked here, and must be detected by the parser.
|
|
if status:
|
|
m['status'] = status
|
|
return m
|
|
|
|
|
|
class DOMParserBase(object):
|
|
"""Base parser to handle HTML data from the IMDb's web server."""
|
|
_defGetRefs = False
|
|
_containsObjects = False
|
|
|
|
preprocessors = []
|
|
extractors = []
|
|
usingModule = None
|
|
|
|
_logger = logging.getLogger('imdbpy.parser.http.domparser')
|
|
|
|
def __init__(self, useModule=None):
|
|
"""Initialize the parser. useModule can be used to force it
|
|
to use 'BeautifulSoup' or 'lxml'; by default, it's auto-detected,
|
|
using 'lxml' if available and falling back to 'BeautifulSoup'
|
|
otherwise."""
|
|
# Module to use.
|
|
if useModule is None:
|
|
useModule = ('lxml', 'BeautifulSoup')
|
|
if not isinstance(useModule, (tuple, list)):
|
|
useModule = [useModule]
|
|
self._useModule = useModule
|
|
nrMods = len(useModule)
|
|
_gotError = False
|
|
for idx, mod in enumerate(useModule):
|
|
mod = mod.strip().lower()
|
|
try:
|
|
if mod == 'lxml':
|
|
from lxml.html import fromstring
|
|
from lxml.etree import tostring
|
|
self._is_xml_unicode = False
|
|
self.usingModule = 'lxml'
|
|
elif mod == 'beautifulsoup':
|
|
from bsouplxml.html import fromstring
|
|
from bsouplxml.etree import tostring
|
|
self._is_xml_unicode = True
|
|
self.usingModule = 'beautifulsoup'
|
|
else:
|
|
self._logger.warn('unknown module "%s"' % mod)
|
|
continue
|
|
self.fromstring = fromstring
|
|
self._tostring = tostring
|
|
if _gotError:
|
|
warnings.warn('falling back to "%s"' % mod)
|
|
break
|
|
except ImportError, e:
|
|
if idx+1 >= nrMods:
|
|
# Raise the exception, if we don't have any more
|
|
# options to try.
|
|
raise IMDbError('unable to use any parser in %s: %s' % \
|
|
(str(useModule), str(e)))
|
|
else:
|
|
warnings.warn('unable to use "%s": %s' % (mod, str(e)))
|
|
_gotError = True
|
|
continue
|
|
else:
|
|
raise IMDbError('unable to use parsers in %s' % str(useModule))
|
|
# Fall-back defaults.
|
|
self._modFunct = None
|
|
self._as = 'http'
|
|
self._cname = self.__class__.__name__
|
|
self._init()
|
|
self.reset()
|
|
|
|
def reset(self):
|
|
"""Reset the parser."""
|
|
# Names and titles references.
|
|
self._namesRefs = {}
|
|
self._titlesRefs = {}
|
|
self._charactersRefs = {}
|
|
self._reset()
|
|
|
|
def _init(self):
|
|
"""Subclasses can override this method, if needed."""
|
|
pass
|
|
|
|
def _reset(self):
|
|
"""Subclasses can override this method, if needed."""
|
|
pass
|
|
|
|
def parse(self, html_string, getRefs=None, **kwds):
|
|
"""Return the dictionary generated from the given html string;
|
|
getRefs can be used to force the gathering of movies/persons/characters
|
|
references."""
|
|
self.reset()
|
|
if getRefs is not None:
|
|
self.getRefs = getRefs
|
|
else:
|
|
self.getRefs = self._defGetRefs
|
|
# Useful only for the testsuite.
|
|
if not isinstance(html_string, unicode):
|
|
html_string = unicode(html_string, 'latin_1', 'replace')
|
|
html_string = subXMLRefs(html_string)
|
|
# Temporary fix: self.parse_dom must work even for empty strings.
|
|
html_string = self.preprocess_string(html_string)
|
|
html_string = html_string.strip()
|
|
if self.usingModule == 'beautifulsoup':
|
|
# tag attributes like title=""Family Guy"" will be
|
|
# converted to title=""Family Guy"" and this confuses BeautifulSoup.
|
|
html_string = html_string.replace('""', '"')
|
|
# Browser-specific escapes create problems to BeautifulSoup.
|
|
html_string = html_string.replace('<!--[if IE]>', '"')
|
|
html_string = html_string.replace('<![endif]-->', '"')
|
|
#print html_string.encode('utf8')
|
|
if html_string:
|
|
dom = self.get_dom(html_string)
|
|
#print self.tostring(dom).encode('utf8')
|
|
try:
|
|
dom = self.preprocess_dom(dom)
|
|
except Exception, e:
|
|
self._logger.error('%s: caught exception preprocessing DOM',
|
|
self._cname, exc_info=True)
|
|
if self.getRefs:
|
|
try:
|
|
self.gather_refs(dom)
|
|
except Exception, e:
|
|
self._logger.warn('%s: unable to gather refs: %s',
|
|
self._cname, exc_info=True)
|
|
data = self.parse_dom(dom)
|
|
else:
|
|
data = {}
|
|
try:
|
|
data = self.postprocess_data(data)
|
|
except Exception, e:
|
|
self._logger.error('%s: caught exception postprocessing data',
|
|
self._cname, exc_info=True)
|
|
if self._containsObjects:
|
|
self.set_objects_params(data)
|
|
data = self.add_refs(data)
|
|
return data
|
|
|
|
def _build_empty_dom(self):
|
|
from bsouplxml import _bsoup
|
|
return _bsoup.BeautifulSoup('')
|
|
|
|
def get_dom(self, html_string):
|
|
"""Return a dom object, from the given string."""
|
|
try:
|
|
dom = self.fromstring(html_string)
|
|
if dom is None:
|
|
dom = self._build_empty_dom()
|
|
self._logger.error('%s: using a fake empty DOM', self._cname)
|
|
return dom
|
|
except Exception, e:
|
|
self._logger.error('%s: caught exception parsing DOM',
|
|
self._cname, exc_info=True)
|
|
return self._build_empty_dom()
|
|
|
|
def xpath(self, element, path):
|
|
"""Return elements matching the given XPath."""
|
|
try:
|
|
xpath_result = element.xpath(path)
|
|
if self._is_xml_unicode:
|
|
return xpath_result
|
|
result = []
|
|
for item in xpath_result:
|
|
if isinstance(item, str):
|
|
item = unicode(item)
|
|
result.append(item)
|
|
return result
|
|
except Exception, e:
|
|
self._logger.error('%s: caught exception extracting XPath "%s"',
|
|
self._cname, path, exc_info=True)
|
|
return []
|
|
|
|
def tostring(self, element):
|
|
"""Convert the element to a string."""
|
|
if isinstance(element, (unicode, str)):
|
|
return unicode(element)
|
|
else:
|
|
try:
|
|
return self._tostring(element, encoding=unicode)
|
|
except Exception, e:
|
|
self._logger.error('%s: unable to convert to string',
|
|
self._cname, exc_info=True)
|
|
return u''
|
|
|
|
def clone(self, element):
|
|
"""Clone an element."""
|
|
return self.fromstring(self.tostring(element))
|
|
|
|
def preprocess_string(self, html_string):
|
|
"""Here we can modify the text, before it's parsed."""
|
|
if not html_string:
|
|
return html_string
|
|
# Remove silly » and – chars.
|
|
html_string = html_string.replace(u' \xbb', u'')
|
|
html_string = html_string.replace(u'–', u'-')
|
|
try:
|
|
preprocessors = self.preprocessors
|
|
except AttributeError:
|
|
return html_string
|
|
for src, sub in preprocessors:
|
|
# re._pattern_type is present only since Python 2.5.
|
|
if callable(getattr(src, 'sub', None)):
|
|
html_string = src.sub(sub, html_string)
|
|
elif isinstance(src, str):
|
|
html_string = html_string.replace(src, sub)
|
|
elif callable(src):
|
|
try:
|
|
html_string = src(html_string)
|
|
except Exception, e:
|
|
_msg = '%s: caught exception preprocessing html'
|
|
self._logger.error(_msg, self._cname, exc_info=True)
|
|
continue
|
|
##print html_string.encode('utf8')
|
|
return html_string
|
|
|
|
def gather_refs(self, dom):
|
|
"""Collect references."""
|
|
grParser = GatherRefs(useModule=self._useModule)
|
|
grParser._as = self._as
|
|
grParser._modFunct = self._modFunct
|
|
refs = grParser.parse_dom(dom)
|
|
refs = grParser.postprocess_data(refs)
|
|
self._namesRefs = refs['names refs']
|
|
self._titlesRefs = refs['titles refs']
|
|
self._charactersRefs = refs['characters refs']
|
|
|
|
def preprocess_dom(self, dom):
|
|
"""Last chance to modify the dom, before the rules in self.extractors
|
|
are applied by the parse_dom method."""
|
|
return dom
|
|
|
|
def parse_dom(self, dom):
|
|
"""Parse the given dom according to the rules specified
|
|
in self.extractors."""
|
|
result = {}
|
|
for extractor in self.extractors:
|
|
##print extractor.label
|
|
if extractor.group is None:
|
|
elements = [(extractor.label, element)
|
|
for element in self.xpath(dom, extractor.path)]
|
|
else:
|
|
groups = self.xpath(dom, extractor.group)
|
|
elements = []
|
|
for group in groups:
|
|
group_key = self.xpath(group, extractor.group_key)
|
|
if not group_key: continue
|
|
group_key = group_key[0]
|
|
# XXX: always tries the conversion to unicode:
|
|
# BeautifulSoup.NavigableString is a subclass
|
|
# of unicode, and so it's never converted.
|
|
group_key = self.tostring(group_key)
|
|
normalizer = extractor.group_key_normalize
|
|
if normalizer is not None:
|
|
if callable(normalizer):
|
|
try:
|
|
group_key = normalizer(group_key)
|
|
except Exception, e:
|
|
_m = '%s: unable to apply group_key normalizer'
|
|
self._logger.error(_m, self._cname,
|
|
exc_info=True)
|
|
group_elements = self.xpath(group, extractor.path)
|
|
elements.extend([(group_key, element)
|
|
for element in group_elements])
|
|
for group_key, element in elements:
|
|
for attr in extractor.attrs:
|
|
if isinstance(attr.path, dict):
|
|
data = {}
|
|
for field in attr.path.keys():
|
|
path = attr.path[field]
|
|
value = self.xpath(element, path)
|
|
if not value:
|
|
data[field] = None
|
|
else:
|
|
# XXX: use u'' , to join?
|
|
data[field] = ''.join(value)
|
|
else:
|
|
data = self.xpath(element, attr.path)
|
|
if not data:
|
|
data = None
|
|
else:
|
|
data = attr.joiner.join(data)
|
|
if not data:
|
|
continue
|
|
attr_postprocess = attr.postprocess
|
|
if callable(attr_postprocess):
|
|
try:
|
|
data = attr_postprocess(data)
|
|
except Exception, e:
|
|
_m = '%s: unable to apply attr postprocess'
|
|
self._logger.error(_m, self._cname, exc_info=True)
|
|
key = attr.key
|
|
if key is None:
|
|
key = group_key
|
|
elif key.startswith('.'):
|
|
# assuming this is an xpath
|
|
try:
|
|
key = self.xpath(element, key)[0]
|
|
except IndexError:
|
|
self._logger.error('%s: XPath returned no items',
|
|
self._cname, exc_info=True)
|
|
elif key.startswith('self.'):
|
|
key = getattr(self, key[5:])
|
|
if attr.multi:
|
|
if key not in result:
|
|
result[key] = []
|
|
result[key].append(data)
|
|
else:
|
|
if isinstance(data, dict):
|
|
result.update(data)
|
|
else:
|
|
result[key] = data
|
|
return result
|
|
|
|
def postprocess_data(self, data):
|
|
"""Here we can modify the data."""
|
|
return data
|
|
|
|
def set_objects_params(self, data):
|
|
"""Set parameters of Movie/Person/... instances, since they are
|
|
not always set in the parser's code."""
|
|
for obj in flatten(data, yieldDictKeys=True, scalar=_Container):
|
|
obj.accessSystem = self._as
|
|
obj.modFunct = self._modFunct
|
|
|
|
def add_refs(self, data):
|
|
"""Modify data according to the expected output."""
|
|
if self.getRefs:
|
|
titl_re = ur'(%s)' % '|'.join([re.escape(x) for x
|
|
in self._titlesRefs.keys()])
|
|
if titl_re != ur'()': re_titles = re.compile(titl_re, re.U)
|
|
else: re_titles = None
|
|
nam_re = ur'(%s)' % '|'.join([re.escape(x) for x
|
|
in self._namesRefs.keys()])
|
|
if nam_re != ur'()': re_names = re.compile(nam_re, re.U)
|
|
else: re_names = None
|
|
chr_re = ur'(%s)' % '|'.join([re.escape(x) for x
|
|
in self._charactersRefs.keys()])
|
|
if chr_re != ur'()': re_characters = re.compile(chr_re, re.U)
|
|
else: re_characters = None
|
|
_putRefs(data, re_titles, re_names, re_characters)
|
|
return {'data': data, 'titlesRefs': self._titlesRefs,
|
|
'namesRefs': self._namesRefs,
|
|
'charactersRefs': self._charactersRefs}
|
|
|
|
|
|
class Extractor(object):
|
|
"""Instruct the DOM parser about how to parse a document."""
|
|
def __init__(self, label, path, attrs, group=None, group_key=None,
|
|
group_key_normalize=None):
|
|
"""Initialize an Extractor object, used to instruct the DOM parser
|
|
about how to parse a document."""
|
|
# rarely (never?) used, mostly for debugging purposes.
|
|
self.label = label
|
|
self.group = group
|
|
if group_key is None:
|
|
self.group_key = ".//text()"
|
|
else:
|
|
self.group_key = group_key
|
|
self.group_key_normalize = group_key_normalize
|
|
self.path = path
|
|
# A list of attributes to fetch.
|
|
if isinstance(attrs, Attribute):
|
|
attrs = [attrs]
|
|
self.attrs = attrs
|
|
|
|
def __repr__(self):
|
|
"""String representation of an Extractor object."""
|
|
r = '<Extractor id:%s (label=%s, path=%s, attrs=%s, group=%s, ' \
|
|
'group_key=%s group_key_normalize=%s)>' % (id(self),
|
|
self.label, self.path, repr(self.attrs), self.group,
|
|
self.group_key, self.group_key_normalize)
|
|
return r
|
|
|
|
|
|
class Attribute(object):
|
|
"""The attribute to consider, for a given node."""
|
|
def __init__(self, key, multi=False, path=None, joiner=None,
|
|
postprocess=None):
|
|
"""Initialize an Attribute object, used to specify the
|
|
attribute to consider, for a given node."""
|
|
# The key under which information will be saved; can be a string or an
|
|
# XPath. If None, the label of the containing extractor will be used.
|
|
self.key = key
|
|
self.multi = multi
|
|
self.path = path
|
|
if joiner is None:
|
|
joiner = ''
|
|
self.joiner = joiner
|
|
# Post-process this set of information.
|
|
self.postprocess = postprocess
|
|
|
|
def __repr__(self):
|
|
"""String representation of an Attribute object."""
|
|
r = '<Attribute id:%s (key=%s, multi=%s, path=%s, joiner=%s, ' \
|
|
'postprocess=%s)>' % (id(self), self.key,
|
|
self.multi, repr(self.path),
|
|
self.joiner, repr(self.postprocess))
|
|
return r
|
|
|
|
|
|
def _parse_ref(text, link, info):
|
|
"""Manage links to references."""
|
|
if link.find('/title/tt') != -1:
|
|
yearK = re_yearKind_index.match(info)
|
|
if yearK and yearK.start() == 0:
|
|
text += ' %s' % info[:yearK.end()]
|
|
return (text.replace('\n', ' '), link)
|
|
|
|
|
|
class GatherRefs(DOMParserBase):
|
|
"""Parser used to gather references to movies, persons and characters."""
|
|
_attrs = [Attribute(key=None, multi=True,
|
|
path={
|
|
'text': './text()',
|
|
'link': './@href',
|
|
'info': './following::text()[1]'
|
|
},
|
|
postprocess=lambda x: _parse_ref(x.get('text') or u'', x.get('link') or '',
|
|
(x.get('info') or u'').strip()))]
|
|
extractors = [
|
|
Extractor(label='names refs',
|
|
path="//a[starts-with(@href, '/name/nm')][string-length(@href)=16]",
|
|
attrs=_attrs),
|
|
|
|
Extractor(label='titles refs',
|
|
path="//a[starts-with(@href, '/title/tt')]" \
|
|
"[string-length(@href)=17]",
|
|
attrs=_attrs),
|
|
|
|
Extractor(label='characters refs',
|
|
path="//a[starts-with(@href, '/character/ch')]" \
|
|
"[string-length(@href)=21]",
|
|
attrs=_attrs),
|
|
]
|
|
|
|
def postprocess_data(self, data):
|
|
result = {}
|
|
for item in ('names refs', 'titles refs', 'characters refs'):
|
|
result[item] = {}
|
|
for k, v in data.get(item, []):
|
|
k = k.strip()
|
|
v = v.strip()
|
|
if not (k and v):
|
|
continue
|
|
if not v.endswith('/'): continue
|
|
imdbID = analyze_imdbid(v)
|
|
if item == 'names refs':
|
|
obj = Person(personID=imdbID, name=k,
|
|
accessSystem=self._as, modFunct=self._modFunct)
|
|
elif item == 'titles refs':
|
|
obj = Movie(movieID=imdbID, title=k,
|
|
accessSystem=self._as, modFunct=self._modFunct)
|
|
else:
|
|
obj = Character(characterID=imdbID, name=k,
|
|
accessSystem=self._as, modFunct=self._modFunct)
|
|
# XXX: companies aren't handled: are they ever found in text,
|
|
# as links to their page?
|
|
result[item][k] = obj
|
|
return result
|
|
|
|
def add_refs(self, data):
|
|
return data
|
|
|
|
|