Use <meta> tags to discover the per-page encoding of html previews (#4183)

This commit is contained in:
Amber Brown 2018-11-15 11:05:08 -06:00 committed by GitHub
parent a51288e5d6
commit df758e155d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 100 additions and 9 deletions

1
changelog.d/4183.bugfix Normal file
View File

@ -0,0 +1 @@
URL previews now correctly decode non-UTF-8 text if the header contains a `<meta http-equiv="Content-Type"` header.

View File

@ -53,6 +53,9 @@ from ._base import FileInfo
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
_charset_match = re.compile(br"<\s*meta[^>]*charset\s*=\s*([a-z0-9-]+)", flags=re.I)
_content_type_match = re.compile(r'.*; *charset="?(.*?)"?(;|$)', flags=re.I)
class PreviewUrlResource(Resource): class PreviewUrlResource(Resource):
isLeaf = True isLeaf = True
@ -223,15 +226,25 @@ class PreviewUrlResource(Resource):
with open(media_info['filename'], 'rb') as file: with open(media_info['filename'], 'rb') as file:
body = file.read() body = file.read()
# clobber the encoding from the content-type, or default to utf-8 encoding = None
# XXX: this overrides any <meta/> or XML charset headers in the body
# which may pose problems, but so far seems to work okay. # Let's try and figure out if it has an encoding set in a meta tag.
match = re.match( # Limit it to the first 1kb, since it ought to be in the meta tags
r'.*; *charset="?(.*?)"?(;|$)', # at the top.
media_info['media_type'], match = _charset_match.search(body[:1000])
re.I
) # If we find a match, it should take precedence over the
encoding = match.group(1) if match else "utf-8" # Content-Type header, so set it here.
if match:
encoding = match.group(1).decode('ascii')
# If we don't find a match, we'll look at the HTTP Content-Type, and
# if that doesn't exist, we'll fall back to UTF-8.
if not encoding:
match = _content_type_match.match(
media_info['media_type']
)
encoding = match.group(1) if match else "utf-8"
og = decode_and_calc_og(body, media_info['uri'], encoding) og = decode_and_calc_og(body, media_info['uri'], encoding)

View File

@ -162,3 +162,80 @@ class URLPreviewTests(unittest.HomeserverTestCase):
self.assertEqual( self.assertEqual(
channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} channel.json_body, {"og:title": "~matrix~", "og:description": "hi"}
) )
def test_non_ascii_preview_httpequiv(self):
request, channel = self.make_request(
"GET", "url_preview?url=matrix.org", shorthand=False
)
request.render(self.preview_url)
self.pump()
# We've made one fetch
self.assertEqual(len(self.fetches), 1)
end_content = (
b'<html><head>'
b'<meta http-equiv="Content-Type" content="text/html; charset=windows-1251"/>'
b'<meta property="og:title" content="\xe4\xea\xe0" />'
b'<meta property="og:description" content="hi" />'
b'</head></html>'
)
self.fetches[0][0].callback(
(
end_content,
(
len(end_content),
{
b"Content-Length": [b"%d" % (len(end_content))],
# This charset=utf-8 should be ignored, because the
# document has a meta tag overriding it.
b"Content-Type": [b'text/html; charset="utf8"'],
},
"https://example.com",
200,
),
)
)
self.pump()
self.assertEqual(channel.code, 200)
self.assertEqual(channel.json_body["og:title"], u"\u0434\u043a\u0430")
def test_non_ascii_preview_content_type(self):
request, channel = self.make_request(
"GET", "url_preview?url=matrix.org", shorthand=False
)
request.render(self.preview_url)
self.pump()
# We've made one fetch
self.assertEqual(len(self.fetches), 1)
end_content = (
b'<html><head>'
b'<meta property="og:title" content="\xe4\xea\xe0" />'
b'<meta property="og:description" content="hi" />'
b'</head></html>'
)
self.fetches[0][0].callback(
(
end_content,
(
len(end_content),
{
b"Content-Length": [b"%d" % (len(end_content))],
b"Content-Type": [b'text/html; charset="windows-1251"'],
},
"https://example.com",
200,
),
)
)
self.pump()
self.assertEqual(channel.code, 200)
self.assertEqual(channel.json_body["og:title"], u"\u0434\u043a\u0430")