diff --git a/synapse/rest/media/v1/preview_html.py b/synapse/rest/media/v1/preview_html.py
index 872a9e72e8..4cc9c66fbe 100644
--- a/synapse/rest/media/v1/preview_html.py
+++ b/synapse/rest/media/v1/preview_html.py
@@ -16,7 +16,6 @@ import itertools
import logging
import re
from typing import TYPE_CHECKING, Dict, Generator, Iterable, Optional, Set, Union
-from urllib import parse as urlparse
if TYPE_CHECKING:
from lxml import etree
@@ -144,9 +143,7 @@ def decode_body(
return etree.fromstring(body, parser)
-def parse_html_to_open_graph(
- tree: "etree.Element", media_uri: str
-) -> Dict[str, Optional[str]]:
+def parse_html_to_open_graph(tree: "etree.Element") -> Dict[str, Optional[str]]:
"""
Parse the HTML document into an Open Graph response.
@@ -155,7 +152,6 @@ def parse_html_to_open_graph(
Args:
tree: The parsed HTML document.
- media_url: The URI used to download the body.
Returns:
The Open Graph response as a dictionary.
@@ -209,7 +205,7 @@ def parse_html_to_open_graph(
"//*/meta[translate(@itemprop, 'IMAGE', 'image')='image']/@content"
)
if meta_image:
- og["og:image"] = rebase_url(meta_image[0], media_uri)
+ og["og:image"] = meta_image[0]
else:
# TODO: consider inlined CSS styles as well as width & height attribs
images = tree.xpath("//img[@src][number(@width)>10][number(@height)>10]")
@@ -320,37 +316,6 @@ def _iterate_over_text(
)
-def rebase_url(url: str, base: str) -> str:
- """
- Resolves a potentially relative `url` against an absolute `base` URL.
-
- For example:
-
- >>> rebase_url("subpage", "https://example.com/foo/")
- 'https://example.com/foo/subpage'
- >>> rebase_url("sibling", "https://example.com/foo")
- 'https://example.com/sibling'
- >>> rebase_url("/bar", "https://example.com/foo/")
- 'https://example.com/bar'
- >>> rebase_url("https://alice.com/a/", "https://example.com/foo/")
- 'https://alice.com/a'
- """
- base_parts = urlparse.urlparse(base)
- # Convert the parsed URL to a list for (potential) modification.
- url_parts = list(urlparse.urlparse(url))
- # Add a scheme, if one does not exist.
- if not url_parts[0]:
- url_parts[0] = base_parts.scheme or "http"
- # Fix up the hostname, if this is not a data URL.
- if url_parts[0] != "data" and not url_parts[1]:
- url_parts[1] = base_parts.netloc
- # If the path does not start with a /, nest it under the base path's last
- # directory.
- if not url_parts[2].startswith("/"):
- url_parts[2] = re.sub(r"/[^/]+$", "/", base_parts.path) + url_parts[2]
- return urlparse.urlunparse(url_parts)
-
-
def summarize_paragraphs(
text_nodes: Iterable[str], min_size: int = 200, max_size: int = 500
) -> Optional[str]:
diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py
index 14ea88b240..d47af8ead6 100644
--- a/synapse/rest/media/v1/preview_url_resource.py
+++ b/synapse/rest/media/v1/preview_url_resource.py
@@ -22,7 +22,7 @@ import shutil
import sys
import traceback
from typing import TYPE_CHECKING, BinaryIO, Iterable, Optional, Tuple
-from urllib import parse as urlparse
+from urllib.parse import urljoin, urlparse, urlsplit
from urllib.request import urlopen
import attr
@@ -44,11 +44,7 @@ from synapse.metrics.background_process_metrics import run_as_background_process
from synapse.rest.media.v1._base import get_filename_from_headers
from synapse.rest.media.v1.media_storage import MediaStorage
from synapse.rest.media.v1.oembed import OEmbedProvider
-from synapse.rest.media.v1.preview_html import (
- decode_body,
- parse_html_to_open_graph,
- rebase_url,
-)
+from synapse.rest.media.v1.preview_html import decode_body, parse_html_to_open_graph
from synapse.types import JsonDict, UserID
from synapse.util import json_encoder
from synapse.util.async_helpers import ObservableDeferred
@@ -187,7 +183,7 @@ class PreviewUrlResource(DirectServeJsonResource):
ts = self.clock.time_msec()
# XXX: we could move this into _do_preview if we wanted.
- url_tuple = urlparse.urlsplit(url)
+ url_tuple = urlsplit(url)
for entry in self.url_preview_url_blacklist:
match = True
for attrib in entry:
@@ -322,7 +318,7 @@ class PreviewUrlResource(DirectServeJsonResource):
# Parse Open Graph information from the HTML in case the oEmbed
# response failed or is incomplete.
- og_from_html = parse_html_to_open_graph(tree, media_info.uri)
+ og_from_html = parse_html_to_open_graph(tree)
# Compile the Open Graph response by using the scraped
# information from the HTML and overlaying any information
@@ -588,12 +584,17 @@ class PreviewUrlResource(DirectServeJsonResource):
if "og:image" not in og or not og["og:image"]:
return
+ # The image URL from the HTML might be relative to the previewed page,
+ # convert it to an URL which can be requested directly.
+ image_url = og["og:image"]
+ url_parts = urlparse(image_url)
+ if url_parts.scheme != "data":
+ image_url = urljoin(media_info.uri, image_url)
+
# FIXME: it might be cleaner to use the same flow as the main /preview_url
# request itself and benefit from the same caching etc. But for now we
# just rely on the caching on the master request to speed things up.
- image_info = await self._handle_url(
- rebase_url(og["og:image"], media_info.uri), user, allow_data_urls=True
- )
+ image_info = await self._handle_url(image_url, user, allow_data_urls=True)
if _is_media(image_info.media_type):
# TODO: make sure we don't choke on white-on-transparent images
|
