test(python): clarify prefilter regression attribution

python/python/tests/test_hybrid_query.py

@@ -365,6 +365,20 @@ def test_hybrid_prefilter_explain_plan(table_with_id: Table):
     )
 
 
+def test_hybrid_prefilter_returns_filtered_row(table_with_id: Table):
+    """Regression test for the hybrid prefilter inversion fixed in #3096."""
+    results = (
+        table_with_id.search(query_type="hybrid")
+        .vector([0.1, 0.1])
+        .text("dog")
+        .where("id = 3", prefilter=True)
+        .limit(1)
+        .to_arrow()
+    )
+
+    assert results["id"].to_pylist() == [3]
+
+
 def test_normalize_scores():
     cases = [
         (pa.array([0.1, 0.4]), pa.array([0.0, 1.0])),

python/python/tests/test_table.py

@@ -17,7 +17,7 @@ from unittest.mock import patch
 
 import lancedb
 from lancedb.dependencies import _PANDAS_AVAILABLE
-from lancedb.index import BTree, FTS, HnswFlat, HnswPq, HnswSq, IvfFlat, IvfPq
+from lancedb.index import BTree, FTS, HnswFlat, HnswPq, HnswSq, IvfPq
 import numpy as np
 import polars as pl
 import pyarrow as pa
@@ -2848,52 +2848,6 @@ def test_create_f16_table(mem_db: DBConnection):
     assert "s-2" in expected["text"].to_pylist()
 
 
-@pytest.mark.parametrize(
-    "index_config",
-    [
-        IvfPq(distance_type="cosine", num_partitions=2, num_sub_vectors=2),
-        IvfFlat(distance_type="cosine", num_partitions=2),
-    ],
-    ids=["ivf-pq", "ivf-flat"],
-)
-def test_f16_index_search_with_open_batch_reader(mem_db: DBConnection, index_config):
-    """Regression test for https://github.com/lancedb/lancedb/issues/2611."""
-    pytest.importorskip("pandas")
-    dimension = 32
-    num_rows = 512
-    rng = np.random.default_rng(42)
-    text_vectors = rng.standard_normal((num_rows, dimension)).astype(np.float16)
-    image_vectors = rng.standard_normal((num_rows, dimension)).astype(np.float16)
-    data = pa.table(
-        {
-            "id": np.arange(num_rows),
-            "text_embedding": pa.FixedSizeListArray.from_arrays(
-                pa.array(text_vectors.reshape(-1)), dimension
-            ),
-            "image_embedding": pa.FixedSizeListArray.from_arrays(
-                pa.array(image_vectors.reshape(-1)), dimension
-            ),
-        }
-    )
-    table = mem_db.create_table("f16_index_with_open_reader", data=data)
-    table.create_index("image_embedding", config=index_config)
-
-    reader = table.search().select(["id", "text_embedding"]).to_batches()
-    for batch in reader:
-        for _, _row in batch.to_pandas().iterrows():
-            result = (
-                table.search(image_vectors[2], vector_column_name="image_embedding")
-                .select(["id", "_distance"])
-                .distance_type("cosine")
-                .limit(10)
-                .to_pandas()
-            )
-            assert result.iloc[0]["id"] == 2
-            return
-
-    pytest.fail("expected the outer query to return a batch")
-
-
 def test_add_with_embedding_function(mem_db: DBConnection):
     emb = EmbeddingFunctionRegistry.get_instance().get("test").create()