Fix CSS parsing for Patreon

Some of the CSS classes end in " double", so this commit refines the
HTML parsing to account for both kinds of classes, and also adds an
additional test case.
This commit is contained in:
Erik Johnson 2014-08-05 00:26:23 -05:00
parent c3f0b12b0f
commit 6994e70651
1 changed files with 25 additions and 5 deletions

View File

@ -16,15 +16,25 @@ class PatreonHTMLParser(compat_html_parser.HTMLParser):
_PREFIX = 'http://www.patreon.com' _PREFIX = 'http://www.patreon.com'
_ATTACH_TAGS = 5 * ['div'] _ATTACH_TAGS = 5 * ['div']
_ATTACH_CLASSES = [ _ATTACH_CLASSES = [
'fancyboxhidden', 'box photo double', 'boxwrapper double', 'fancyboxhidden', 'box photo', 'boxwrapper',
'hiddendisplay shareinfo', 'attach' 'hiddendisplay shareinfo', 'attach'
] ]
_INFO_TAGS = 4 * ['div'] _INFO_TAGS = 4 * ['div']
_INFO_CLASSES = [ _INFO_CLASSES = [
'fancyboxhidden', 'box photo double', 'boxwrapper double', 'fancyboxhidden', 'box photo', 'boxwrapper',
'hiddendisplay shareinfo' 'hiddendisplay shareinfo'
] ]
def _match(self, attrs_classes, desired):
if attrs_classes == desired:
return True
elif len(attrs_classes) == len(desired):
return all(
x.startswith(y)
for x, y in zip(attrs_classes, desired)
)
return False
def get_creation_info(self, html_data): def get_creation_info(self, html_data):
self.tag_stack = [] self.tag_stack = []
self.attrs_stack = [] self.attrs_stack = []
@ -45,7 +55,7 @@ class PatreonHTMLParser(compat_html_parser.HTMLParser):
attrs_classes = [ attrs_classes = [
x.get('class', '').lower() for x in self.attrs_stack[-6:-1] x.get('class', '').lower() for x in self.attrs_stack[-6:-1]
] ]
if attrs_classes == self._ATTACH_CLASSES: if self._match(attrs_classes, self._ATTACH_CLASSES):
if self.tag_stack[-1] == 'a': if self.tag_stack[-1] == 'a':
url = self._PREFIX + self.attrs_stack[-1].get('href') url = self._PREFIX + self.attrs_stack[-1].get('href')
self.creation_info['url'] = url self.creation_info['url'] = url
@ -56,7 +66,7 @@ class PatreonHTMLParser(compat_html_parser.HTMLParser):
attrs_classes = [ attrs_classes = [
x.get('class', '').lower() for x in self.attrs_stack[-5:-1] x.get('class', '').lower() for x in self.attrs_stack[-5:-1]
] ]
if attrs_classes == self._INFO_CLASSES: if self._match(attrs_classes, self._INFO_CLASSES):
if self.attrs_stack[-1].get('class') == 'utitle': if self.attrs_stack[-1].get('class') == 'utitle':
self.creation_info['title'] = data.strip() self.creation_info['title'] = data.strip()
@ -65,10 +75,10 @@ class PatreonIE(InfoExtractor):
IE_NAME = 'patreon' IE_NAME = 'patreon'
_VALID_URL = r'https?://(?:www\.)?patreon\.com/creation\?hid=(.+)' _VALID_URL = r'https?://(?:www\.)?patreon\.com/creation\?hid=(.+)'
_TESTS = [ _TESTS = [
# CSS names with "double" in the name, i.e. "boxwrapper double"
{ {
'url': 'http://www.patreon.com/creation?hid=743933', 'url': 'http://www.patreon.com/creation?hid=743933',
'md5': 'e25505eec1053a6e6813b8ed369875cc', 'md5': 'e25505eec1053a6e6813b8ed369875cc',
'name': 'Patreon',
'info_dict': { 'info_dict': {
'id': '743933', 'id': '743933',
'ext': 'mp3', 'ext': 'mp3',
@ -76,6 +86,16 @@ class PatreonIE(InfoExtractor):
'uploader': 'Cognitive Dissonance Podcast', 'uploader': 'Cognitive Dissonance Podcast',
}, },
}, },
{
'url': 'http://www.patreon.com/creation?hid=754133',
'md5': '3eb09345bf44bf60451b8b0b81759d0a',
'info_dict': {
'id': '754133',
'ext': 'mp3',
'title': 'CD 167 Extra',
'uploader': 'Cognitive Dissonance Podcast',
},
},
] ]
# Currently Patreon exposes download URL via hidden CSS, so login is not # Currently Patreon exposes download URL via hidden CSS, so login is not