Add language support to posts

activities/migrations/0017_post_language.py

@@ -0,0 +1,18 @@
+# Generated by Django 4.2.1 on 2023-05-15 09:26
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ("activities", "0016_index_together_migration"),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name="post",
+            name="language",
+            field=models.CharField(max_length=2, null=True),
+        ),
+    ]

activities/models/post.py

@@ -31,6 +31,7 @@ from core.html import ContentRenderer, FediverseHtmlParser
 from core.ld import (
     canonicalise,
     format_ld_date,
+    get_language,
     get_list,
     get_value_or_map,
     parse_ld_date,
@@ -252,6 +253,9 @@ class Post(StatorModel):
     # The main (HTML) content
     content = models.TextField()
 
+    # The language of the content
+    language = models.CharField(max_length=2, null=True)
+
     type = models.CharField(
         max_length=20,
         choices=Types.choices,
@@ -474,6 +478,7 @@ class Post(StatorModel):
         reply_to: Optional["Post"] = None,
         attachments: list | None = None,
         question: dict | None = None,
+        language: str | None = None,
     ) -> "Post":
         with transaction.atomic():
             # Find mentions in this post
@@ -492,6 +497,9 @@ class Post(StatorModel):
                 sorted([tag[: Hashtag.MAXIMUM_LENGTH] for tag in parser.hashtags])
                 or None
             )
+            if language is None:
+                language = author.config_identity.preferred_posting_language
+
             # Make the Post object
             post = cls.objects.create(
                 author=author,
@@ -502,6 +510,7 @@ class Post(StatorModel):
                 visibility=visibility,
                 hashtags=hashtags,
                 in_reply_to=reply_to.object_uri if reply_to else None,
+                language=language,
             )
             post.object_uri = post.urls.object_uri
             post.url = post.absolute_object_uri()
@@ -526,6 +535,7 @@ class Post(StatorModel):
         visibility: int = Visibilities.public,
         attachments: list | None = None,
         attachment_attributes: list | None = None,
+        language: str | None = None,
     ):
         with transaction.atomic():
             # Strip all HTML and apply linebreaks filter
@@ -538,6 +548,9 @@ class Post(StatorModel):
             self.summary = summary or None
             self.sensitive = bool(summary) if sensitive is None else sensitive
             self.visibility = visibility
+            if language is None:
+                language = self.author.config_identity.preferred_posting_language
+            self.language = language
             self.edited = timezone.now()
             self.mentions.set(self.mentions_from_content(content, self.author))
             self.emojis.set(Emoji.emojis_from_content(content, None))
@@ -649,6 +662,10 @@ class Post(StatorModel):
             "tag": [],
             "attachment": [],
         }
+        if self.language is not None:
+            value["contentMap"] = {
+                self.language: value["content"],
+            }
         if self.type == Post.Types.question and self.type_data:
             value[self.type_data.mode] = [
                 {
@@ -872,6 +889,7 @@ class Post(StatorModel):
             post.published = parse_ld_date(data.get("published"))
             post.edited = parse_ld_date(data.get("updated"))
             post.in_reply_to = data.get("inReplyTo")
+            post.language = get_language(data)
             # Mentions and hashtags
             post.hashtags = []
             for tag in get_list(data, "tag"):
@@ -1112,6 +1130,7 @@ class Post(StatorModel):
             "created_at": format_ld_date(self.published),
             "account": self.author.to_mastodon_json(include_counts=False),
             "content": self.safe_content_remote(),
+            "language": self.language,
             "visibility": visibility_mapping[self.visibility],
             "sensitive": self.sensitive,
             "spoiler_text": self.summary or "",
@@ -1152,7 +1171,6 @@ class Post(StatorModel):
             if isinstance(self.type_data, QuestionData)
             else None,
             "card": None,
-            "language": None,
             "text": self.safe_content_remote(),
             "edited_at": format_ld_date(self.edited) if self.edited else None,
         }

api/schemas.py

@@ -151,7 +151,7 @@ class Status(Schema):
     reblog: Optional["Status"] = Field(...)
     poll: Poll | None = Field(...)
     card: None = Field(...)
-    language: None = Field(...)
+    language: str | None = Field(...)
     text: str | None = Field(...)
     edited_at: str | None
     favourited: bool = False

api/views/statuses.py

@@ -110,6 +110,7 @@ def post_status(request, details: PostStatusSchema) -> schemas.Status:
         reply_to=reply_post,
         attachments=attachments,
         question=details.poll.dict() if details.poll else None,
+        language=details.language,
     )
     # Add their own timeline event for immediate visibility
     TimelineEvent.add_post(request.identity, post)
@@ -141,6 +142,7 @@ def edit_status(request, id: str, details: EditStatusSchema) -> schemas.Status:
         sensitive=details.sensitive,
         attachments=attachments,
         attachment_attributes=details.media_attributes,
+        language=details.language,
     )
     return schemas.Status.from_post(post)
 

core/ld.py

@@ -1,5 +1,6 @@
 import datetime
 import os
+import re
 import urllib.parse as urllib_parse
 
 from dateutil import parser
@@ -692,3 +693,24 @@ def media_type_from_filename(filename):
         return "image/webp"
     else:
         return "application/octet-stream"
+
+
+def get_language(data) -> str | None:
+    """Detects and returns a document's language"""
+    map_ = None
+    if "contentMap" in data:
+        map_ = data["contentMap"]
+    elif "nameMap" in data:
+        map_ = data["nameMap"]
+    elif "summaryMap" in data:
+        map_ = data["summaryMap"]
+
+    if not map_:
+        return None
+
+    lang = list(map_.keys())[0]
+    if not lang or lang == "und":
+        return None
+
+    lang = re.split("-|_", lang)[0]
+    return lang.lower()

tests/activities/models/test_post.py

@@ -259,6 +259,7 @@ def test_content_map(remote_identity):
         create=True,
     )
     assert post.content == "Hi World"
+    assert post.language is None
 
     post2 = Post.by_ap(
         data={
@@ -271,6 +272,7 @@ def test_content_map(remote_identity):
         create=True,
     )
     assert post2.content == "Hey World"
+    assert post2.language is None
 
     post3 = Post.by_ap(
         data={
@@ -283,6 +285,7 @@ def test_content_map(remote_identity):
         create=True,
     )
     assert post3.content == "Hello World"
+    assert post3.language == "en"
 
 
 @pytest.mark.django_db

tests/core/test_ld.py

@@ -2,7 +2,7 @@ import datetime
 
 from dateutil.tz import tzutc
 
-from core.ld import parse_ld_date
+from core.ld import get_language, parse_ld_date
 
 
 def test_parse_ld_date():
@@ -41,3 +41,41 @@ def test_parse_ld_date():
         tzinfo=tzutc(),
     )
     assert difference.total_seconds() == 0
+
+
+def test_get_language():
+    assert (
+        get_language(
+            {
+                "contentMap": {
+                    "en": "<p>Hello</p>",
+                    "es": "<p>hola</p>",
+                },
+                "nameMap": {"de": "Hallo"},
+                "summaryMap": {"fr": "Bonjour"},
+            }
+        )
+        == "en"
+    )
+    assert (
+        get_language(
+            {
+                "nameMap": {"de": "Hallo"},
+                "summaryMap": {"fr": "Bonjour"},
+            }
+        )
+        == "de"
+    )
+    assert (
+        get_language(
+            {
+                "summaryMap": {"fr": "Bonjour"},
+            }
+        )
+        == "fr"
+    )
+    assert get_language({"contentMap": {"en-gb": "<p>Hello</p>"}}) == "en"
+    assert get_language({"contentMap": {"en_GB": "<p>Hello</p>"}}) == "en"
+    assert get_language({"contentMap": {"EN": "<p>Hello</p>"}}) == "en"
+    assert get_language({"contentMap": {"und": "<p>Hello</p>"}}) is None
+    assert get_language({}) is None