Fix: change tantivy token limit to title field max length (#13046)

src/documents/search/_tokenizer.py

@@ -48,6 +48,9 @@ _LANGUAGE_MAP: dict[str, str] = {
 }
 
 SUPPORTED_LANGUAGES: frozenset[str] = frozenset(_LANGUAGE_MAP)
+# Document.title is max_length=128, so use 129 as the limit for
+# Tantivy's remove_long filter
+_TOKEN_REMOVE_LONG_LIMIT: Final[int] = 129
 
 
 def register_tokenizers(index: tantivy.Index, language: str | None) -> None:
@@ -77,10 +80,10 @@ def register_tokenizers(index: tantivy.Index, language: str | None) -> None:
 
 
 def _paperless_text(language: str | None) -> tantivy.TextAnalyzer:
-    """Main full-text tokenizer for content, title, etc: simple -> remove_long(65) -> lowercase -> ascii_fold [-> stemmer]"""
+    """Main full-text tokenizer for content, title, etc: simple -> remove_long(129) -> lowercase -> ascii_fold [-> stemmer]"""
     builder = (
         tantivy.TextAnalyzerBuilder(tantivy.Tokenizer.simple())
-        .filter(tantivy.Filter.remove_long(65))
+        .filter(tantivy.Filter.remove_long(_TOKEN_REMOVE_LONG_LIMIT))
         .filter(tantivy.Filter.lowercase())
         .filter(tantivy.Filter.ascii_fold())
     )
@@ -119,12 +122,12 @@ def _bigram_analyzer() -> tantivy.TextAnalyzer:
 
 
 def _simple_search_analyzer() -> tantivy.TextAnalyzer:
-    """Tokenizer for simple substring search fields: non-whitespace chunks -> remove_long(65) -> lowercase -> ascii_fold."""
+    """Tokenizer for simple substring search fields: non-whitespace chunks -> remove_long(129) -> lowercase -> ascii_fold."""
     return (
         tantivy.TextAnalyzerBuilder(
             tantivy.Tokenizer.regex(r"\S+"),
         )
-        .filter(tantivy.Filter.remove_long(65))
+        .filter(tantivy.Filter.remove_long(_TOKEN_REMOVE_LONG_LIMIT))
         .filter(tantivy.Filter.lowercase())
         .filter(tantivy.Filter.ascii_fold())
         .build()

src/documents/tests/search/test_backend.py

@@ -261,6 +261,36 @@ class TestSearch:
             == 1
         )
 
+    @pytest.mark.parametrize(
+        ("search_mode", "query"),
+        [
+            pytest.param(SearchMode.TITLE, "12345", id="title_search"),
+            pytest.param(SearchMode.TEXT, "12345", id="text_search"),
+            pytest.param(SearchMode.QUERY, None, id="query_title_exact"),
+        ],
+    )
+    def test_search_modes_match_model_limit_title_tokens(
+        self,
+        backend: TantivyBackend,
+        search_mode: SearchMode,
+        query: str | None,
+    ) -> None:
+        """Search must keep filename-like title tokens up to the model limit."""
+        long_title = "1234567890" * 12 + "12345678"
+        doc = Document.objects.create(
+            title=long_title,
+            content="ordinary content",
+            checksum="TXT12",
+            pk=18,
+        )
+        backend.add_or_update(doc)
+
+        assert backend.search_ids(
+            query or f"title:{long_title}",
+            user=None,
+            search_mode=search_mode,
+        ) == [doc.pk]
+
     @pytest.mark.parametrize(
         ("mode", "title", "content", "hits", "misses"),
         [

src/documents/tests/search/test_tokenizer.py

@@ -99,6 +99,25 @@ class TestTokenizers:
         )
         assert simple_search_index.searcher().search(q, limit=5).count == 1
 
+    def test_simple_search_analyzer_supports_model_limit_token_substrings(
+        self,
+        simple_search_index: tantivy.Index,
+    ) -> None:
+        """Simple substring search keeps tokens up to Document.title's model limit."""
+        long_token = "abcdefghij" * 12 + "abcdefgh"
+        writer = simple_search_index.writer()
+        doc = tantivy.Document()
+        doc.add_text("simple_content", long_token)
+        writer.add_document(doc)
+        writer.commit()
+        simple_search_index.reload()
+        q = tantivy.Query.regex_query(
+            simple_search_index.schema,
+            "simple_content",
+            ".*cdefg.*",
+        )
+        assert simple_search_index.searcher().search(q, limit=5).count == 1
+
     def test_unsupported_language_logs_warning(self, caplog: LogCaptureFixture) -> None:
         """Unsupported language codes should log a warning and disable stemming gracefully."""
         sb = tantivy.SchemaBuilder()