Merge branch 'todos-and-issues' into 'master'

Fixes addressing various manually tracked content issues.

See merge request jaywink/federation!183

CHANGELOG.md

@@ -1,5 +1,41 @@
 # Changelog
 
+## Unreleased
+
+### Changed
+
+* This is actually both a change and a fix. AP Image objects do not define properties matching the 
+  HTML img tag alt and title properties. Image.name is used to render both alt and title, which IMHO is
+  wrong. With this change, markdown images defining the title property will be recognized instead of
+  being thrown away (the fix) and the title property, if defined, will have precedence over the
+  alt property as the Image.name value (the change). Before this change, the client app would properly
+  render the img tag from the markdown source (with distinct alt and title properties), but the Image
+  object would not federate and hence not be displayed on other platforms (namely Mastodon).
+
+### Fixed
+
+* Note._find_and_mark_mentions: When an AP Mention object href can't be found in the rendered content,
+  try the name property.
+
+* Ignore media objects that don't define a media type.
+
+* Prevent rendered content image duplication when an image is both in the AP payload rendered content
+  and defined as an attachment that doesn't set the inlineImage property.
+
+* Instead of discarding the whole AP payload out when encountering an undefined or unlisted AP object,
+  log a warning and keep going. Ensure None is returned when a nested field only contains an undefined
+  object.
+
+* Accept the application/ld+json type for webfinger AP links.
+
+* Mark an AP mention only if profile.finger is defined.
+
+* Handle escape sequences for inbound markdown mentions.
+
+* Extend the Unicode character range allowed in markdown mentions.
+
+* Discard illegal characters from tag text. Previously, this was done only on tag links.
+
 ## [0.25.1] - 2024-02-18
 
 ### Fixed

federation/entities/activitypub/models.py

@@ -209,11 +209,16 @@ class MixedField(fields.Nested):
         ret = []
         for item in value:
             if item.get('@type'):
-                res = super()._deserialize(item, attr, data, **kwargs)
+                try:
+                    res = super()._deserialize(item, attr, data, **kwargs)
+                except KeyError as ex:
+                    logger.warning("nested field: undefined JSON-LD type %s", ex)
+                    continue
                 ret.append(res if not isinstance(res, list) else res[0])
             else:
                 ret.append(self.iri._deserialize(item, attr, data, **kwargs))
 
+        if not ret: ret.append(None)
         return ret if len(ret) > 1 or self.many else ret[0]
         
 
@@ -247,7 +252,7 @@ class Object(BaseEntity, metaclass=JsonLDAnnotation):
     icon = MixedField(as2.icon, nested='ImageSchema')
     image = MixedField(as2.image, nested='ImageSchema')
     tag_objects = MixedField(as2.tag, nested=['NoteSchema', 'HashtagSchema','MentionSchema','PropertyValueSchema','EmojiSchema'], many=True)
-    attachment = fields.Nested(as2.attachment, nested=['LinkSchema', 'NoteSchema', 'ImageSchema', 'AudioSchema', 'DocumentSchema','PropertyValueSchema','IdentityProofSchema'],
+    attachment = MixedField(as2.attachment, nested=['LinkSchema', 'NoteSchema', 'ImageSchema', 'AudioSchema', 'DocumentSchema','PropertyValueSchema','IdentityProofSchema'],
                                many=True, default=[])
     content_map = LanguageMap(as2.content)  # language maps are not implemented in calamus
     context = fields.RawJsonLD(as2.context)
@@ -421,6 +426,8 @@ class Document(Object):
     url = MixedField(as2.url, nested='LinkSchema')
 
     def to_base(self):
+        if self.media_type is missing:
+            return self
         self.__dict__.update({'schema': True})
         if self.media_type.startswith('image'):
             return Image(**get_base_attributes(self))
@@ -866,7 +873,7 @@ class Note(Object, RawContentMixin):
             normalized_url = f'{parsed.scheme}://{parsed.netloc}{normalized_path.decode()}'
             links = {link['href'].lower(), unquote(link['href']).lower(), url, normalized_url}
             if links.intersection(hrefs):
-                tag = re.match(r'^#?([\w\-]+$)', link.text)
+                tag = re.match(r'^#?([\w\-]+)', link.text)
                 if tag:
                     link['data-hashtag'] = tag.group(1).lower()
 
@@ -879,17 +886,28 @@ class Note(Object, RawContentMixin):
         for mention in mentions:
             hrefs = []
             profile = get_profile_or_entity(fid=mention.href, remote_url=mention.href)
-            if profile and not profile.url:
-                # This should be removed when we are confident that the remote_url property
-                # has been populated for most profiles on the client app side.
+            if profile and not (profile.url and profile.finger):
+                # This should be removed when we are confident that the remote_url and
+                # finger properties have been populated for most profiles on the client app side.
                 profile = retrieve_and_parse_profile(profile.id)
-            if profile:
+            if profile and profile.finger:
                 hrefs.extend([profile.id, profile.url])
+            else:
+                continue
             for href in hrefs:
                 links = self._soup.find_all(href=href)
                 for link in links:
                     link['data-mention'] = profile.finger
                     self._mentions.add(profile.finger)
+            if profile.finger not in self._mentions:
+                # can't find some mentions using their href property value
+                # try with the name property
+                matches = self._soup.find_all(string=mention.name)
+                for match in matches:
+                    link = match.find_parent('a')
+                    if link:
+                        link['data-mention'] = profile.finger
+                        self._mentions.add(profile.finger)
 
     def extract_mentions(self):
         """
@@ -953,7 +971,7 @@ class Note(Object, RawContentMixin):
                     if hasattr(child, 'to_base'):
                         child = child.to_base()
                     if isinstance(child, Image):
-                        if child.inline or (child.image and child.image in self.raw_content):
+                        if child.inline or self._soup.find('img', src=child.url):
                             continue
                     children.append(child)
             self._cached_children = children

federation/entities/mixins.py

@@ -6,6 +6,7 @@ from typing import List, Set, Union, Dict, Tuple
 
 from bs4 import BeautifulSoup
 from commonmark import commonmark
+from markdownify import markdownify
 from marshmallow import missing
 
 from federation.entities.activitypub.enums import ActivityType
@@ -224,13 +225,17 @@ class RawContentMixin(BaseEntity):
         Returns a Tuple of (url, filename).
         """
         images = []
-        if self._media_type != "text/markdown" or self.raw_content is None:
-            return images
-        regex = r"!\[([\w\s\-\']*)\]\((https?://[\w\d\-\./]+\.[\w]*((?<=jpg)|(?<=gif)|(?<=png)|(?<=jpeg)))\)"
-        matches = re.finditer(regex, self.raw_content, re.MULTILINE | re.IGNORECASE)
-        for match in matches:
-            groups = match.groups()
-            images.append((groups[1], groups[0] or ""))
+        if hasattr(self, '_soup'):
+            for img in self._soup.find_all('img', src=re.compile(r'^http')):
+                images.append((img['src'], img.get('title', '') or img.get('alt', '')))
+        else:
+            if self._media_type != "text/markdown" or self.raw_content is None:
+                return images
+            regex = r"!\[([\w\s\-\']*)\]\((https?://[\w\d\-\./]+\.[\w]*((?<=jpg)|(?<=gif)|(?<=png)|(?<=jpeg)))\)"
+            matches = re.finditer(regex, self.raw_content, re.MULTILINE | re.IGNORECASE)
+            for match in matches:
+                groups = match.groups()
+                images.append((groups[1], groups[0] or ""))
         return images
 
     # Legacy. Keep this until tests are reworked
@@ -258,6 +263,9 @@ class RawContentMixin(BaseEntity):
             if handle:
                 self._mentions.add(handle)
                 self.raw_content = self.raw_content.replace(mention, '@' + handle)
+                # mardownify the extracted mention in case some characters are escaped in
+                # raw_content
+                self.raw_content = self.raw_content.replace(markdownify(mention), '@' + handle)
 
 
 class OptionalRawContentMixin(RawContentMixin):

federation/tests/fixtures/entities.py

@@ -1,5 +1,6 @@
 import pytest
 # noinspection PyPackageRequirements
+from commonmark import commonmark
 from freezegun import freeze_time
 from unittest.mock import patch
 
@@ -152,8 +153,7 @@ def activitypubpost_tags():
 @pytest.fixture
 def activitypubpost_embedded_images():
     with freeze_time("2019-04-27"):
-        obj = models.Post(
-            raw_content="""
+        raw_content="""
 #Cycling #lauttasaari #sea #sun
 
 
@@ -166,7 +166,10 @@ def activitypubpost_embedded_images():
 [foo](https://jasonrobinson.me/media/uploads/2019/07/16/daa24d89-cedf-4fc7-bad8-74a9025414710.jpg)
 #only a link, not embedded
 https://jasonrobinson.me/media/uploads/2019/07/16/daa24d89-cedf-4fc7-bad8-74a9025414711.jpg
-""",
+"""
+        obj = models.Post(
+            raw_content=raw_content,
+            rendered_content=commonmark(raw_content, ignore_html_blocks=True),
             public=True,
             provider_display_name="Socialhome",
             id=f"http://127.0.0.1:8000/post/123456/",

federation/utils/activitypub.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import re
 from typing import Optional, Any
 from urllib.parse import urlparse
 
@@ -16,6 +17,7 @@ except Exception as exc:
     federation_user = None
     logger.warning("django is required for get requests signing: %s", exc)
 
+type_path = re.compile(r'^application/(activity|ld)\+json')
 
 def get_profile_id_from_webfinger(handle: str) -> Optional[str]:
     """
@@ -30,7 +32,7 @@ def get_profile_id_from_webfinger(handle: str) -> Optional[str]:
     except json.JSONDecodeError:
         return
     for link in doc.get("links", []):
-        if link.get("rel") == "self" and link.get("type") == "application/activity+json":
+        if link.get("rel") == "self" and type_path.match(link.get("type")):
             return link["href"]
     logger.debug("get_profile_id_from_webfinger: found webfinger but it has no as2 self href")
 

federation/utils/text.py

@@ -8,8 +8,8 @@ from commonmark import commonmark
 
 ILLEGAL_TAG_CHARS = "!#$%^&*+.,@£/()=?`'\\{[]}~;:\"’”—\xa0"
 TAG_PATTERN = re.compile(r'(#[\w\-]+)([)\]_!?*%/.,;\s]+\s*|\Z)', re.UNICODE)
-# This will match non matching braces. I don't think it's an issue.
-MENTION_PATTERN = re.compile(r'(@\{?(?:[\w\-. \u263a-\U0001f645]*; *)?[\w]+@[\w\-.]+\.[\w]+}?)', re.UNICODE)
+# This will match non-matching braces. I don't think it's an issue.
+MENTION_PATTERN = re.compile(r'(@\{?(?:[\w\-. \u0250-\U0001f64f]*; *)?[\w]+@[\w\-.]+\.[\w]+}?)', re.UNICODE)
 # based on https://stackoverflow.com/a/6041965
 URL_PATTERN = re.compile(r'((?:(?:https?|ftp)://|^|(?<=[("<\s]))+(?:[\w\-]+(?:(?:\.[\w\-]+)+))'
                          r'[\w.,;:@?!$()*^=%&/~+\-#]*(?<![:;,).>"]))',

setup.py

@@ -37,6 +37,7 @@ setup(
         "lxml>=3.4.0",
         "iteration_utilities",
         "jsonschema>=2.0.0",
+        "markdownify",
         "pycryptodome>=3.4.10",
         "python-dateutil>=2.4.0",
         "python-httpsig-socialhome",