feat: support remote empty query (#1828)

Support sending empty query types to remote lancedb. also include offset
and limit, where were previously omitted.

python/python/lancedb/query.py

@@ -943,12 +943,16 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
 
 class LanceEmptyQueryBuilder(LanceQueryBuilder):
     def to_arrow(self) -> pa.Table:
-        ds = self._table.to_lance()
-        return ds.to_table(
+        query = Query(
             columns=self._columns,
             filter=self._where,
-            limit=self._limit,
+            k=self._limit or 10,
+            with_row_id=self._with_row_id,
+            vector=[],
+            # not actually respected in remote query
+            offset=self._offset or 0,
         )
+        return self._table._execute_query(query).read_all()
 
     def rerank(self, reranker: Reranker) -> LanceEmptyQueryBuilder:
         """Rerank the results using the specified reranker.

python/python/lancedb/remote/table.py

@@ -327,10 +327,6 @@ class RemoteTable(Table):
             - and also the "_distance" column which is the distance between the query
             vector and the returned vector.
         """
-        # empty query builder is not supported in saas, raise error
-        if query is None and query_type != "hybrid":
-            raise ValueError("Empty query is not supported")
-
         return LanceQueryBuilder.create(
             self,
             query,

python/python/tests/test_remote_db.py

@@ -197,6 +197,23 @@ def test_query_sync_minimal():
         assert data == expected
 
 
+def test_query_sync_empty_query():
+    def handler(body):
+        assert body == {
+            "k": 10,
+            "filter": "true",
+            "vector": [],
+            "columns": ["id"],
+        }
+
+        return pa.table({"id": [1, 2, 3]})
+
+    with query_test_table(handler) as table:
+        data = table.search(None).where("true").select(["id"]).limit(10).to_list()
+        expected = [{"id": 1}, {"id": 2}, {"id": 3}]
+        assert data == expected
+
+
 def test_query_sync_maximal():
     def handler(body):
         assert body == {

python/python/tests/test_table.py

@@ -892,10 +892,15 @@ def test_empty_query(db):
     table = LanceTable.create(db, "my_table2", data=[{"id": i} for i in range(100)])
     df = table.search().select(["id"]).to_pandas()
     assert len(df) == 10
+    # None is the same as default
     df = table.search().select(["id"]).limit(None).to_pandas()
-    assert len(df) == 100
+    assert len(df) == 10
+    # invalid limist is the same as None, wihch is the same as default
     df = table.search().select(["id"]).limit(-1).to_pandas()
-    assert len(df) == 100
+    assert len(df) == 10
+    # valid limit should work
+    df = table.search().select(["id"]).limit(42).to_pandas()
+    assert len(df) == 42
 
 
 def test_search_with_schema_inf_single_vector(db):