fix(downloads): Facebook files get their upload date, non-Latin names their letters

Since 2026-10-28 the signed-in Facebook read gives no upload date, so
every Facebook file since was named 0000-00-00 in Season_0 (31 on prod,
none before). The signed-out read that already titles these downloads
carries the date, so it now lends that too when the signed-in read has
none (backfill.public_title_info is now public_page_info).

On-disk names dropped every letter outside the Latin script, so a Hebrew,
Cyrillic, Greek or CJK uploader went to one shared "untitled" folder
(prod asset 503). anyascii now writes them in Latin letters
(Сергей Иванов -> Sergei_Ivanov); Latin and accented names are unchanged.

backend/tests/test_facebook_titles.py

@@ -22,6 +22,9 @@ _PUBLIC = {
     "description": "sorry not helping\n\n#workhumor #officelife",
     "uploader": "Work dot exe",
 }
+# A signed-out read carries the upload date too (measured 2026-11-14 on watch?v=1140126121708919:
+# "20261113"), which the signed-in read no longer gives.
+_PUBLIC_DATED = {**_PUBLIC, "upload_date": "20261113"}
 
 
 def _fake_ydl(monkeypatch, outcome):
@@ -54,7 +57,7 @@ def test_a_placeholder_with_nothing_behind_it_takes_the_signed_out_pages_text(mo
     seen = _fake_ydl(monkeypatch, _PUBLIC)
     fb = adapters.FacebookAdapter()
 
-    info = backfill.public_title_info({"title": "Facebook", "description": None}, _URL, fb, 7)
+    info = backfill.public_page_info({"title": "Facebook", "description": None}, _URL, fb, 7)
 
     assert fb.title(info) == "sorry not helping"
     # One metadata-only read, and signed out: no cookie goes with it.
@@ -64,11 +67,36 @@ def test_a_placeholder_with_nothing_behind_it_takes_the_signed_out_pages_text(mo
     assert seen[0][egress.GUARDED] is True and seen[0]["proxy"] == egress.proxy_url()
 
 
+def test_the_signed_out_read_also_lends_the_upload_date(monkeypatch):
+    # Measured on prod: since 2026-10-28 no Facebook download has an upload date (37 before had one,
+    # 31 since none) — the signed-in read lost it with the title. The signed-out read still has it.
+    _fake_ydl(monkeypatch, _PUBLIC_DATED)
+
+    info = backfill.public_page_info(
+        {"title": "Facebook", "description": None}, _URL, adapters.FacebookAdapter(), 7
+    )
+
+    assert info["upload_date"] == "20261113"
+
+
+def test_a_date_the_signed_in_read_gave_stays(monkeypatch):
+    _fake_ydl(monkeypatch, _PUBLIC_DATED)
+
+    info = backfill.public_page_info(
+        {"title": "Facebook", "description": None, "upload_date": "20261001"},
+        _URL,
+        adapters.FacebookAdapter(),
+        7,
+    )
+
+    assert info["upload_date"] == "20261001"
+
+
 def test_a_real_title_reads_nothing_more(monkeypatch):
     seen = _fake_ydl(monkeypatch, _PUBLIC)
     info = {"title": "Funny Dark Humor Dad Joke!", "description": None}
 
-    assert backfill.public_title_info(info, _URL, adapters.FacebookAdapter(), 7) is info
+    assert backfill.public_page_info(info, _URL, adapters.FacebookAdapter(), 7) is info
     assert seen == []
 
 
@@ -76,7 +104,7 @@ def test_a_placeholder_with_the_posts_text_behind_it_reads_nothing_more(monkeypa
     seen = _fake_ydl(monkeypatch, _PUBLIC)
     info = {"title": "Video", "description": "Office jokes, part two"}
 
-    assert backfill.public_title_info(info, _URL, adapters.FacebookAdapter(), 7) is info
+    assert backfill.public_page_info(info, _URL, adapters.FacebookAdapter(), 7) is info
     assert seen == []
 
 
@@ -84,7 +112,7 @@ def test_a_placeholder_the_signed_out_page_cannot_better_stays(monkeypatch):
     _fake_ydl(monkeypatch, {"title": "Facebook", "description": None})
     fb = adapters.FacebookAdapter()
 
-    info = backfill.public_title_info({"title": "Facebook", "description": None}, _URL, fb, 7)
+    info = backfill.public_page_info({"title": "Facebook", "description": None}, _URL, fb, 7)
 
     assert fb.title(info) == "Facebook"
 
@@ -93,7 +121,7 @@ def test_another_source_is_never_read_again(monkeypatch):
     seen = _fake_ydl(monkeypatch, _PUBLIC)
     info = {"title": "Video", "description": None}
 
-    assert backfill.public_title_info(info, _URL, adapters.SourceAdapter(), 7) is info
+    assert backfill.public_page_info(info, _URL, adapters.SourceAdapter(), 7) is info
     assert seen == []
 
 
@@ -180,7 +208,7 @@ def test_a_name_given_while_the_backfill_runs_stays(db, make_user, monkeypatch):
     db.add(job)
     db.commit()
 
-    read_page = backfill.public_title_info
+    read_page = backfill.public_page_info
 
     def renamed_while_reading(*args):
         with SessionLocal() as other:
@@ -192,7 +220,7 @@ def test_a_name_given_while_the_backfill_runs_stays(db, make_user, monkeypatch):
             other.commit()
         return read_page(*args)
 
-    monkeypatch.setattr(backfill, "public_title_info", renamed_while_reading)
+    monkeypatch.setattr(backfill, "public_page_info", renamed_while_reading)
     monkeypatch.setattr(backfill.editmod, "ensure_poster", lambda a, root: None)
 
     assert backfill.run()["retitled"] == 1
@@ -252,8 +280,8 @@ def test_a_download_titled_facebook_is_named_from_the_signed_out_page(
     db, make_user, monkeypatch, tmp_path
 ):
     # The whole download path with yt-dlp and the media tools faked: the signed-in read gives only
-    # the placeholder, and the file and the download's name get the post's text.
-    seen = _fake_ydl(monkeypatch, _PUBLIC)
+    # the placeholder, and the file and the download's name get the post's text — and its date.
+    seen = _fake_ydl(monkeypatch, _PUBLIC_DATED)
     monkeypatch.setattr(settings, "files_root", str(tmp_path))
     monkeypatch.setattr(pipeline.urlguard, "check_url", lambda url: None)
     monkeypatch.setattr(pipeline, "_ensure_phone_safe", lambda _j, produced, *_a: produced)
@@ -301,3 +329,7 @@ def test_a_download_titled_facebook_is_named_from_the_signed_out_page(
     assert db.get(DownloadJob, job.id).status == "done"
     assert db.get(MediaAsset, asset.id).title == "sorry not helping"
     assert db.get(DownloadJob, job.id).display_name == "sorry not helping"
+    done = db.get(MediaAsset, asset.id)
+    assert done.upload_date is not None and done.upload_date.isoformat() == "2026-11-13"
+    # The date names the file and its season, not 0000-00-00 / Season_0.
+    assert done.rel_path.startswith("Work_dot_exe/Season_2026/Work_dot_exe_-_2026-11-13_-_")

backend/tests/test_naming.py

@@ -26,9 +26,21 @@ def test_sanitize_output_is_pure_ascii_without_spaces():
         assert _is_ascii_slug(out), out
 
 
-def test_sanitize_non_latin_falls_back_not_crashes():
-    # A title with no ASCII base collapses to the untitled fallback rather than an empty name.
-    assert storage.sanitize("日本語") == "untitled"
+def test_sanitize_writes_a_non_latin_name_in_latin_letters():
+    # Prod asset 503: a Hebrew uploader name was dropped whole, so the file went to "untitled/", the
+    # folder every Hebrew, Cyrillic, Greek or CJK name would share — one show in Plex for all of them.
+    names = ("מַלכָּה אָדוֹם", "Сергей Иванов", "Ελληνικά Νέα", "日本語", "محمد علي")
+    slugs = [storage.sanitize(n) for n in names]
+    for name, slug in zip(names, slugs, strict=True):
+        assert _is_ascii_slug(slug) and slug != "untitled", (name, slug)
+    assert len(set(slugs)) == len(slugs), slugs  # two names, two folders
+    # The measured spellings (anyascii 0.3.3, the locked one).
+    assert slugs == ["mlkhh_dvm", "Sergei_Ivanov", "Ellinika_Nea", "RiBenYu", "mhmd_ly"]
+
+
+def test_sanitize_falls_back_when_nothing_is_left():
+    # Emoji and symbols are dropped, and a name of nothing else still gets a folder.
+    assert storage.sanitize("🎉🎉") == "untitled"
 
 
 def test_display_filename_uses_underscores_and_ascii():