Query.to_search_structure raises RecursionError on a long chain of ORs #100

What happens

Building an OR of many queries one at a time, the way a loop naturally does, gives a search structure nested one level per OR:

q = Query("base.id", "exact_string", ids[0])
for i in ids[1:]:
    q = q | Query("base.id", "exact_string", i)
db.search(q)   # RecursionError with ~1,000 ids

SQLiteDB.search calls query_obj.to_search_structure(). That calls Query._resolve_search_structure → _resolve_single, and the or branch of _resolve_single recurses into param1/param2. With about 1,000 ids that exceeds Python's default recursion limit (1000), so the search fails before it runs.

Why it matters

ORing a batch of ids built in a loop is a normal pattern. NDI, for example, looks up several documents by id at once.

The search itself already handles this depth. Since #98, _base_id_lookup_values walks the OR tree with a stack, and ORed base.id lookups run as one SQL query per 1,000 ids. The recursion in to_search_structure is the only thing in the way.

Found by

Writing tests/test_search_indexes.py (#98). A test for a left-nested OR of 1,200 ids was dropped there because of this.

Possible fix

Make _resolve_search_structure / _resolve_single iterative: walk the tree with an explicit stack and rebuild the resolved copy bottom-up.

Or rebalance a long left-nested OR chain into a balanced tree when Query.__or__ builds it. That keeps depth at log₂(n), but it changes the shape of search_structure, which callers might inspect.

Then restore the dropped test: an OR of 1,200 ids built in a loop returns all of them.

Generated by Claude Code