Qian/make vector col optional (#950)

remote SDK tests were completed through lancedb_integtest

python/tests/test_embeddings_slow.py

@@ -69,10 +69,14 @@ def test_basic_text_embeddings(alias, tmp_path):
     )
 
     query = "greetings"
-    actual = table.search(query).limit(1).to_pydantic(Words)[0]
+    actual = (
+        table.search(query, vector_column_name="vector").limit(1).to_pydantic(Words)[0]
+    )
 
     vec = func.compute_query_embeddings(query)[0]
-    expected = table.search(vec).limit(1).to_pydantic(Words)[0]
+    expected = (
+        table.search(vec, vector_column_name="vector").limit(1).to_pydantic(Words)[0]
+    )
     assert actual.text == expected.text
     assert actual.text == "hello world"
     assert not np.allclose(actual.vector, actual.vector2)
@@ -116,7 +120,11 @@ def test_openclip(tmp_path):
     )
 
     # text search
-    actual = table.search("man's best friend").limit(1).to_pydantic(Images)[0]
+    actual = (
+        table.search("man's best friend", vector_column_name="vector")
+        .limit(1)
+        .to_pydantic(Images)[0]
+    )
     assert actual.label == "dog"
     frombytes = (
         table.search("man's best friend", vector_column_name="vec_from_bytes")
@@ -130,7 +138,11 @@ def test_openclip(tmp_path):
     query_image_uri = "http://farm1.staticflickr.com/200/467715466_ed4a31801f_z.jpg"
     image_bytes = requests.get(query_image_uri).content
     query_image = Image.open(io.BytesIO(image_bytes))
-    actual = table.search(query_image).limit(1).to_pydantic(Images)[0]
+    actual = (
+        table.search(query_image, vector_column_name="vector")
+        .limit(1)
+        .to_pydantic(Images)[0]
+    )
     assert actual.label == "dog"
     other = (
         table.search(query_image, vector_column_name="vec_from_bytes")

python/tests/test_remote_db.py

@@ -38,4 +38,5 @@ def test_remote_db():
     setattr(conn, "_client", FakeLanceDBClient())
 
     table = conn["test"]
+    table.schema = pa.schema([pa.field("vector", pa.list_(pa.float32(), 2))])
     table.search([1.0, 2.0]).to_pandas()

python/tests/test_table.py

@@ -710,6 +710,59 @@ def test_empty_query(db):
     assert len(df) == 100
 
 
+def test_search_with_schema_inf_single_vector(db):
+    class MyTable(LanceModel):
+        text: str
+        vector_col: Vector(10)
+
+    table = LanceTable.create(
+        db,
+        "my_table",
+        schema=MyTable,
+    )
+
+    v1 = np.random.randn(10)
+    v2 = np.random.randn(10)
+    data = [
+        {"vector_col": v1, "text": "foo"},
+        {"vector_col": v2, "text": "bar"},
+    ]
+    df = pd.DataFrame(data)
+    table.add(df)
+
+    q = np.random.randn(10)
+    result1 = table.search(q, vector_column_name="vector_col").limit(1).to_pandas()
+    result2 = table.search(q).limit(1).to_pandas()
+
+    assert result1["text"].iloc[0] == result2["text"].iloc[0]
+
+
+def test_search_with_schema_inf_multiple_vector(db):
+    class MyTable(LanceModel):
+        text: str
+        vector1: Vector(10)
+        vector2: Vector(10)
+
+    table = LanceTable.create(
+        db,
+        "my_table",
+        schema=MyTable,
+    )
+
+    v1 = np.random.randn(10)
+    v2 = np.random.randn(10)
+    data = [
+        {"vector1": v1, "vector2": v2, "text": "foo"},
+        {"vector1": v2, "vector2": v1, "text": "bar"},
+    ]
+    df = pd.DataFrame(data)
+    table.add(df)
+
+    q = np.random.randn(10)
+    with pytest.raises(ValueError):
+        table.search(q).limit(1).to_pandas()
+
+
 def test_compact_cleanup(db):
     table = LanceTable.create(
         db,