fix assorted redirect, unicode and screenscraping bugs

This commit is contained in:
Matthew Hodgson 2016-04-01 02:17:48 +01:00
parent 683e564815
commit c60b751694

View File

@ -74,84 +74,93 @@ class PreviewUrlResource(BaseMediaResource):
elif self._is_html(media_info['media_type']): elif self._is_html(media_info['media_type']):
# TODO: somehow stop a big HTML tree from exploding synapse's RAM # TODO: somehow stop a big HTML tree from exploding synapse's RAM
# XXX: can't work out how to make lxml ignore UTF8 decoding errors def _calc_og():
# so slurp as a string at this point. # suck it up into lxml and define our OG response.
file = open(media_info['filename']) # if we see any URLs in the OG response, then spider them
body = file.read() # (although the client could choose to do this by asking for previews of those URLs to avoid DoSing the server)
file.close()
# FIXME: we shouldn't be forcing utf-8 if the page isn't actually utf-8...
tree = html.fromstring(body.decode('utf-8','ignore'))
# tree = html.parse(media_info['filename'])
# suck it up into lxml and define our OG response. # "og:type" : "article"
# if we see any URLs in the OG response, then spider them # "og:url" : "https://twitter.com/matrixdotorg/status/684074366691356672"
# (although the client could choose to do this by asking for previews of those URLs to avoid DoSing the server) # "og:title" : "Matrix on Twitter"
# "og:image" : "https://pbs.twimg.com/profile_images/500400952029888512/yI0qtFi7_400x400.png"
# "og:description" : "Synapse 0.12 is out! Lots of polishing, performance & bugfixes: /sync API, /r0 prefix, fulltext search, 3PID invites https://t.co/5alhXLLEGP"
# "og:site_name" : "Twitter"
# or:
# "og:type" : "article" # "og:type" : "video",
# "og:url" : "https://twitter.com/matrixdotorg/status/684074366691356672" # "og:url" : "https://www.youtube.com/watch?v=LXDBoHyjmtw",
# "og:title" : "Matrix on Twitter" # "og:site_name" : "YouTube",
# "og:image" : "https://pbs.twimg.com/profile_images/500400952029888512/yI0qtFi7_400x400.png" # "og:video:type" : "application/x-shockwave-flash",
# "og:description" : "Synapse 0.12 is out! Lots of polishing, performance & bugfixes: /sync API, /r0 prefix, fulltext search, 3PID invites https://t.co/5alhXLLEGP" # "og:description" : " ",
# "og:site_name" : "Twitter" # "og:title" : "RemoteJam - Matrix team hack for Disrupt Europe Hackathon",
# "og:image" : "https://i.ytimg.com/vi/LXDBoHyjmtw/maxresdefault.jpg",
# or: # "og:video:url" : "http://www.youtube.com/v/LXDBoHyjmtw?version=3&autohide=1",
# "og:video:width" : "1280"
# "og:video:height" : "720",
# "og:video:secure_url": "https://www.youtube.com/v/LXDBoHyjmtw?version=3&autohide=1",
# "og:type" : "video", og = {}
# "og:url" : "https://www.youtube.com/watch?v=LXDBoHyjmtw", for tag in tree.xpath("//*/meta[starts-with(@property, 'og:')]"):
# "og:site_name" : "YouTube", og[tag.attrib['property']] = tag.attrib['content']
# "og:video:type" : "application/x-shockwave-flash",
# "og:description" : " ",
# "og:title" : "RemoteJam - Matrix team hack for Disrupt Europe Hackathon",
# "og:image" : "https://i.ytimg.com/vi/LXDBoHyjmtw/maxresdefault.jpg",
# "og:video:url" : "http://www.youtube.com/v/LXDBoHyjmtw?version=3&autohide=1",
# "og:video:width" : "1280"
# "og:video:height" : "720",
# "og:video:secure_url": "https://www.youtube.com/v/LXDBoHyjmtw?version=3&autohide=1",
og = {} if 'og:title' not in og:
for tag in tree.xpath("//*/meta[starts-with(@property, 'og:')]"): # do some basic spidering of the HTML
og[tag.attrib['property']] = tag.attrib['content'] title = tree.xpath("(//title)[1] | (//h1)[1] | (//h2)[1] | (//h3)[1]")
og['og:title'] = title[0].text if title else None
if not og:
# do some basic spidering of the HTML
title = tree.xpath("(//title)[1] | (//h1)[1] | (//h2)[1] | (//h3)[1]")
og['og:title'] = title[0].text if title else None
images = [ i for i in tree.xpath("//img") if 'src' in i.attrib ] if 'og:image' not in og:
big_images = [ i for i in images if ( meta_image = tree.xpath("//*/meta[@itemprop='image']/@content");
'width' in i.attrib and 'height' in i.attrib and if meta_image:
i.attrib['width'] > 64 and i.attrib['height'] > 64 og['og:image'] = self._rebase_url(meta_image[0], media_info['uri'])
)]
big_images = big_images.sort(key=lambda i: (-1 * int(i.attrib['width']) * int(i.attrib['height'])))
images = big_images if big_images else images
if images:
base = list(urlparse(media_info['uri']))
src = list(urlparse(images[0].attrib['src']))
if not src[0] and not src[1]:
src[0] = base[0]
src[1] = base[1]
if not src[2].startswith('/'):
src[2] = re.sub(r'/[^/]+$', '/', base[2]) + src[2]
og['og:image'] = urlunparse(src)
text_nodes = tree.xpath("//h1/text() | //h2/text() | //h3/text() | //p/text() | //div/text() | //span/text() | //a/text()")
# text_nodes = tree.xpath("//h1/text() | //h2/text() | //h3/text() | //p/text() | //div/text()")
text = ''
for text_node in text_nodes:
if len(text) < 1024:
text += text_node + ' '
else: else:
break images = [ i for i in tree.xpath("//img") if 'src' in i.attrib ]
text = re.sub(r'[\t ]+', ' ', text) big_images = [ i for i in images if (
text = re.sub(r'[\t \r\n]*[\r\n]+', '\n', text) 'width' in i.attrib and 'height' in i.attrib and
text = text.strip()[:1024] i.attrib['width'] > 64 and i.attrib['height'] > 64
og['og:description'] = text if text else None )]
big_images = big_images.sort(key=lambda i: (-1 * int(i.attrib['width']) * int(i.attrib['height'])))
images = big_images if big_images else images
if images:
og['og:image'] = self._rebase_url(images[0].attrib['src'], media_info['uri'])
if 'og:description' not in og:
meta_description = tree.xpath("//*/meta[@name='description']/@content");
if meta_description:
og['og:description'] = meta_description[0]
else:
text_nodes = tree.xpath("//h1/text() | //h2/text() | //h3/text() | //p/text() | //div/text() | //span/text() | //a/text()")
# text_nodes = tree.xpath("//h1/text() | //h2/text() | //h3/text() | //p/text() | //div/text()")
text = ''
for text_node in text_nodes:
if len(text) < 500:
text += text_node + ' '
else:
break
text = re.sub(r'[\t ]+', ' ', text)
text = re.sub(r'[\t \r\n]*[\r\n]+', '\n', text)
text = text.strip()[:500]
og['og:description'] = text if text else None
# TODO: extract a favicon?
# TODO: turn any OG media URLs into mxc URLs to capture and thumbnail them too
# TODO: store our OG details in a cache (and expire them when stale)
# TODO: delete the content to stop diskfilling, as we only ever cared about its OG
return og
try:
tree = html.parse(media_info['filename'])
og = _calc_og()
except UnicodeDecodeError:
# XXX: evil evil bodge
file = open(media_info['filename'])
body = file.read()
file.close()
tree = html.fromstring(body.decode('utf-8','ignore'))
og = _calc_og()
# TODO: extract a favicon?
# TODO: turn any OG media URLs into mxc URLs to capture and thumbnail them too
# TODO: store our OG details in a cache (and expire them when stale)
# TODO: delete the content to stop diskfilling, as we only ever cared about its OG
else: else:
logger.warn("Failed to find any OG data in %s", url) logger.warn("Failed to find any OG data in %s", url)
og = {} og = {}
@ -173,6 +182,15 @@ class PreviewUrlResource(BaseMediaResource):
) )
raise raise
def _rebase_url(self, url, base):
base = list(urlparse(base))
url = list(urlparse(url))
if not url[0] and not url[1]:
url[0] = base[0]
url[1] = base[1]
if not url[2].startswith('/'):
url[2] = re.sub(r'/[^/]+$', '/', base[2]) + url[2]
return urlunparse(url)
@defer.inlineCallbacks @defer.inlineCallbacks
def _download_url(self, url, user): def _download_url(self, url, user):
@ -223,7 +241,7 @@ class PreviewUrlResource(BaseMediaResource):
download_name = None download_name = None
yield self.store.store_local_media( yield self.store.store_local_media(
media_id=fname, media_id=file_id,
media_type=media_type, media_type=media_type,
time_now_ms=self.clock.time_msec(), time_now_ms=self.clock.time_msec(),
upload_name=download_name, upload_name=download_name,