mirror of
https://git.anonymousland.org/anonymousland/synapse.git
synced 2025-08-09 06:12:12 -04:00
Use attrs internally for the URL preview code & add documentation. (#10753)
This commit is contained in:
Patrick Cloke
GitHub
parent
a23f3abb9b
commit
89ba834818
5 changed files with 132 additions and 119 deletions
|
|
@ -22,9 +22,11 @@ import re
|
|||
import shutil
|
||||
import sys
|
||||
import traceback
|
||||
from typing import TYPE_CHECKING, Any, Dict, Generator, Iterable, Optional, Union
|
||||
from typing import TYPE_CHECKING, Dict, Generator, Iterable, Optional, Union
|
||||
from urllib import parse as urlparse
|
||||
|
||||
import attr
|
||||
|
||||
from twisted.internet.error import DNSLookupError
|
||||
from twisted.web.server import Request
|
||||
|
||||
|
|
@ -42,6 +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 OEmbedError, OEmbedProvider
|
||||
from synapse.types import JsonDict
|
||||
from synapse.util import json_encoder
|
||||
from synapse.util.async_helpers import ObservableDeferred
|
||||
from synapse.util.caches.expiringcache import ExpiringCache
|
||||
|
|
@ -71,7 +74,43 @@ OG_TAG_VALUE_MAXLEN = 1000
|
|||
ONE_HOUR = 60 * 60 * 1000
|
||||
|
||||
|
||||
@attr.s(slots=True, frozen=True, auto_attribs=True)
|
||||
class MediaInfo:
|
||||
"""
|
||||
Information parsed from downloading media being previewed.
|
||||
"""
|
||||
|
||||
# The Content-Type header of the response.
|
||||
media_type: str
|
||||
# The length (in bytes) of the downloaded media.
|
||||
media_length: int
|
||||
# The media filename, according to the server. This is parsed from the
|
||||
# returned headers, if possible.
|
||||
download_name: Optional[str]
|
||||
# The time of the preview.
|
||||
created_ts_ms: int
|
||||
# Information from the media storage provider about where the file is stored
|
||||
# on disk.
|
||||
filesystem_id: str
|
||||
filename: str
|
||||
# The URI being previewed.
|
||||
uri: str
|
||||
# The HTTP response code.
|
||||
response_code: int
|
||||
# The timestamp (in milliseconds) of when this preview expires.
|
||||
expires: int
|
||||
# The ETag header of the response.
|
||||
etag: Optional[str]
|
||||
|
||||
|
||||
class PreviewUrlResource(DirectServeJsonResource):
|
||||
"""
|
||||
Generating URL previews is a complicated task which many potential pitfalls.
|
||||
|
||||
See docs/development/url_previews.md for discussion of the design and
|
||||
algorithm followed in this module.
|
||||
"""
|
||||
|
||||
isLeaf = True
|
||||
|
||||
def __init__(
|
||||
|
|
@ -219,18 +258,17 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
|
||||
logger.debug("got media_info of '%s'", media_info)
|
||||
|
||||
if _is_media(media_info["media_type"]):
|
||||
file_id = media_info["filesystem_id"]
|
||||
if _is_media(media_info.media_type):
|
||||
file_id = media_info.filesystem_id
|
||||
dims = await self.media_repo._generate_thumbnails(
|
||||
None, file_id, file_id, media_info["media_type"], url_cache=True
|
||||
None, file_id, file_id, media_info.media_type, url_cache=True
|
||||
)
|
||||
|
||||
og = {
|
||||
"og:description": media_info["download_name"],
|
||||
"og:image": "mxc://%s/%s"
|
||||
% (self.server_name, media_info["filesystem_id"]),
|
||||
"og:image:type": media_info["media_type"],
|
||||
"matrix:image:size": media_info["media_length"],
|
||||
"og:description": media_info.download_name,
|
||||
"og:image": f"mxc://{self.server_name}/{media_info.filesystem_id}",
|
||||
"og:image:type": media_info.media_type,
|
||||
"matrix:image:size": media_info.media_length,
|
||||
}
|
||||
|
||||
if dims:
|
||||
|
|
@ -240,14 +278,14 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
logger.warning("Couldn't get dims for %s" % url)
|
||||
|
||||
# define our OG response for this media
|
||||
elif _is_html(media_info["media_type"]):
|
||||
elif _is_html(media_info.media_type):
|
||||
# TODO: somehow stop a big HTML tree from exploding synapse's RAM
|
||||
|
||||
with open(media_info["filename"], "rb") as file:
|
||||
with open(media_info.filename, "rb") as file:
|
||||
body = file.read()
|
||||
|
||||
encoding = get_html_media_encoding(body, media_info["media_type"])
|
||||
og = decode_and_calc_og(body, media_info["uri"], encoding)
|
||||
encoding = get_html_media_encoding(body, media_info.media_type)
|
||||
og = decode_and_calc_og(body, media_info.uri, encoding)
|
||||
|
||||
# pre-cache the image for posterity
|
||||
# FIXME: it might be cleaner to use the same flow as the main /preview_url
|
||||
|
|
@ -255,14 +293,14 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
# just rely on the caching on the master request to speed things up.
|
||||
if "og:image" in og and og["og:image"]:
|
||||
image_info = await self._download_url(
|
||||
_rebase_url(og["og:image"], media_info["uri"]), user
|
||||
_rebase_url(og["og:image"], media_info.uri), user
|
||||
)
|
||||
|
||||
if _is_media(image_info["media_type"]):
|
||||
if _is_media(image_info.media_type):
|
||||
# TODO: make sure we don't choke on white-on-transparent images
|
||||
file_id = image_info["filesystem_id"]
|
||||
file_id = image_info.filesystem_id
|
||||
dims = await self.media_repo._generate_thumbnails(
|
||||
None, file_id, file_id, image_info["media_type"], url_cache=True
|
||||
None, file_id, file_id, image_info.media_type, url_cache=True
|
||||
)
|
||||
if dims:
|
||||
og["og:image:width"] = dims["width"]
|
||||
|
|
@ -270,12 +308,11 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
else:
|
||||
logger.warning("Couldn't get dims for %s", og["og:image"])
|
||||
|
||||
og["og:image"] = "mxc://%s/%s" % (
|
||||
self.server_name,
|
||||
image_info["filesystem_id"],
|
||||
)
|
||||
og["og:image:type"] = image_info["media_type"]
|
||||
og["matrix:image:size"] = image_info["media_length"]
|
||||
og[
|
||||
"og:image"
|
||||
] = f"mxc://{self.server_name}/{image_info.filesystem_id}"
|
||||
og["og:image:type"] = image_info.media_type
|
||||
og["matrix:image:size"] = image_info.media_length
|
||||
else:
|
||||
del og["og:image"]
|
||||
else:
|
||||
|
|
@ -301,17 +338,17 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
# store OG in history-aware DB cache
|
||||
await self.store.store_url_cache(
|
||||
url,
|
||||
media_info["response_code"],
|
||||
media_info["etag"],
|
||||
media_info["expires"] + media_info["created_ts"],
|
||||
media_info.response_code,
|
||||
media_info.etag,
|
||||
media_info.expires + media_info.created_ts_ms,
|
||||
jsonog,
|
||||
media_info["filesystem_id"],
|
||||
media_info["created_ts"],
|
||||
media_info.filesystem_id,
|
||||
media_info.created_ts_ms,
|
||||
)
|
||||
|
||||
return jsonog.encode("utf8")
|
||||
|
||||
async def _download_url(self, url: str, user: str) -> Dict[str, Any]:
|
||||
async def _download_url(self, url: str, user: str) -> MediaInfo:
|
||||
# TODO: we should probably honour robots.txt... except in practice
|
||||
# we're most likely being explicitly triggered by a human rather than a
|
||||
# bot, so are we really a robot?
|
||||
|
|
@ -423,18 +460,18 @@ class PreviewUrlResource(DirectServeJsonResource):
|
|||
# therefore not expire it.
|
||||
raise
|
||||
|
||||
return {
|
||||
"media_type": media_type,
|
||||
"media_length": length,
|
||||
"download_name": download_name,
|
||||
"created_ts": time_now_ms,
|
||||
"filesystem_id": file_id,
|
||||
"filename": fname,
|
||||
"uri": uri,
|
||||
"response_code": code,
|
||||
"expires": expires,
|
||||
"etag": etag,
|
||||
}
|
||||
return MediaInfo(
|
||||
media_type=media_type,
|
||||
media_length=length,
|
||||
download_name=download_name,
|
||||
created_ts_ms=time_now_ms,
|
||||
filesystem_id=file_id,
|
||||
filename=fname,
|
||||
uri=uri,
|
||||
response_code=code,
|
||||
expires=expires,
|
||||
etag=etag,
|
||||
)
|
||||
|
||||
def _start_expire_url_cache_data(self):
|
||||
return run_as_background_process(
|
||||
|
|
@ -580,7 +617,7 @@ def get_html_media_encoding(body: bytes, content_type: str) -> str:
|
|||
|
||||
def decode_and_calc_og(
|
||||
body: bytes, media_uri: str, request_encoding: Optional[str] = None
|
||||
) -> Dict[str, Optional[str]]:
|
||||
) -> JsonDict:
|
||||
"""
|
||||
Calculate metadata for an HTML document.
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue