Adds notes for where we can improve, if either fixes, features or a new release drop in from Tantivy

src/documents/search/_backend.py

@@ -224,10 +224,13 @@ class WriteBatch:
         """
         Remove a document from the batch by its primary key.
 
-        Uses range query instead of term query to work around unsigned integer
-        type detection bug in tantivy-py 0.25.
+        Uses range_query instead of term_query to work around a tantivy-py bug
+        where Python integers are inferred as i64, producing Terms that never
+        match u64 fields.
+
+        TODO: Replace with term_query("id", doc_id) once
+        https://github.com/quickwit-oss/tantivy-py/pull/642 lands.
         """
-        # Use range query to work around u64 deletion bug
         self._writer.delete_documents_by_query(
             tantivy.Query.range_query(
                 self._backend._schema,
@@ -518,10 +521,14 @@ class TantivyBackend:
         search_ids + ORM filtering) and just need highlight data.
 
         Note: Each doc_id requires an individual index lookup because tantivy-py
-        does not expose a batch doc-address-by-ID API. This is acceptable for
+        does not yet expose a batch fast-field read API. This is acceptable for
         page-sized batches (typically 25 docs) but should not be called with
         thousands of IDs.
 
+        TODO: When https://github.com/quickwit-oss/tantivy-py/pull/641 lands,
+        the per-doc range_query lookups here can be replaced with a single
+        collect_u64_fast_field("id", doc_addresses) call.
+
         Args:
             query: The search query (used for snippet generation)
             doc_ids: Ordered list of document IDs to generate hits for
@@ -665,6 +672,9 @@ class TantivyBackend:
             if threshold is not None:
                 all_hits = [hit for hit in all_hits if hit[1] >= threshold]
 
+        # TODO: Replace with searcher.collect_u64_fast_field("id", addrs) once
+        # https://github.com/quickwit-oss/tantivy-py/pull/641 lands — eliminates
+        # one stored-doc fetch per result (~80% reduction in search_ids latency).
         return [searcher.doc(doc_addr).to_dict()["id"][0] for doc_addr, *_ in all_hits]
 
     def autocomplete(
@@ -812,6 +822,8 @@ class TantivyBackend:
         # Fetch one extra to account for excluding the original document
         results = searcher.search(final_query, limit=effective_limit + 1)
 
+        # TODO: Replace with collect_u64_fast_field("id", addrs) once
+        # https://github.com/quickwit-oss/tantivy-py/pull/641 lands.
         ids = []
         for _score, doc_address in results.hits:
             result_doc_id = searcher.doc(doc_address).to_dict()["id"][0]

src/documents/search/_query.py

@@ -396,10 +396,17 @@ def build_permission_filter(
         Tantivy query that filters results to visible documents
 
     Implementation Notes:
-        - Uses range_query instead of term_query to work around unsigned integer
-          type detection bug in tantivy-py 0.25
-        - Uses boolean_query for "no owner" check since exists_query is not
-          available in tantivy-py 0.25.1 (available in master)
+        - Uses range_query instead of term_query for owner_id/viewer_id to work
+          around a tantivy-py bug where Python ints are inferred as i64, causing
+          term_query to return no hits on u64 fields.
+          TODO: Replace with term_query once
+          https://github.com/quickwit-oss/tantivy-py/pull/642 lands.
+
+        - Uses range_query(owner_id, 1, MAX_U64) as an "owner exists" check
+          because exists_query is not yet available in tantivy-py 0.25.
+          TODO: Replace with exists_query("owner_id") once that is exposed in
+          a tantivy-py release.
+
         - Uses disjunction_max_query to combine permission clauses with OR logic
     """
     owner_any = tantivy.Query.range_query(