fix(kernel): handle empty exact catalog filters

Signed-off-by: Vu Anh Phung <vu.phung@databricks.com>

Author: vuanhphung

src/databricks/sql/backend/kernel/client.py

@@ -25,7 +25,7 @@
 import logging
 import threading
 import uuid
-from typing import Any, Dict, List, Optional, Set, TYPE_CHECKING, Union
+from typing import Any, Dict, List, Optional, Set, Tuple, TYPE_CHECKING, Union
 
 from databricks.sql.backend.databricks_client import DatabricksClient
 from databricks.sql.backend.kernel._errors import (
@@ -125,13 +125,30 @@ def _catalog_or_none(value: Optional[str]) -> Optional[str]:
     treats the catalog as an exact identifier for SHOW COLUMNS. ``None``
     remains the only absent filter. Other strings must be preserved as real
     filters; in particular, an empty string matches nothing just as it does
-    on the Thrift backend.
+    on the Thrift backend. Non-empty whitespace-only strings remain invalid
+    and are left for kernel validation.
     """
     if value is None or value in ("%", "*"):
         return None
     return value
 
 
+def _exact_catalog_and_pattern(
+    catalog: Optional[str], pattern: Optional[str]
+) -> Tuple[Optional[str], Optional[str]]:
+    """Adapt an empty exact catalog to an empty subordinate pattern.
+
+    At ``KERNEL_REV``, ``Identifier("")`` is invalid while
+    ``LikePattern("")`` means match-nothing. Schema and column metadata
+    take an exact catalog, so represent an empty catalog as all catalogs
+    constrained by an empty schema pattern. This preserves empty-catalog
+    semantics without passing an invalid identifier to the kernel.
+    """
+    if catalog == "":
+        return None, ""
+    return _catalog_or_none(catalog), pattern
+
+
 def _is_staging_statement(operation: str) -> bool:
     """True iff ``operation`` is a volume/staging statement (PUT / GET /
     REMOVE).
@@ -918,9 +935,12 @@ def get_schemas(
         if self._kernel_session is None:
             raise InterfaceError("get_schemas requires an open session.")
         try:
+            catalog, schema_pattern = _exact_catalog_and_pattern(
+                catalog_name, schema_name
+            )
             stream = self._kernel_session.metadata().list_schemas(
-                catalog=_catalog_or_none(catalog_name),
-                schema_pattern=schema_name,
+                catalog=catalog,
+                schema_pattern=schema_pattern,
             )
             return self._make_result_set(stream, cursor, self._synthetic_command_id())
         except Exception as exc:
@@ -975,9 +995,12 @@ def get_columns(
             # row's `TABLE_CAT` is correctly attributed. Matches the
             # Thrift backend's `getColumns(null, …)` behaviour from
             # the user's perspective.
+            catalog, schema_pattern = _exact_catalog_and_pattern(
+                catalog_name, schema_name
+            )
             stream = self._kernel_session.metadata().list_columns(
-                catalog=_catalog_or_none(catalog_name),
-                schema_pattern=schema_name,
+                catalog=catalog,
+                schema_pattern=schema_pattern,
                 table_pattern=table_name,
                 column_pattern=column_name,
             )

tests/unit/test_kernel_client.py

@@ -1616,6 +1616,28 @@ def test_get_schemas_preserves_empty_pattern():
     list_schemas.assert_called_once_with(catalog="main", schema_pattern="")
 
 
+def test_get_schemas_empty_catalog_uses_empty_pattern():
+    """Avoid passing an empty exact ``Identifier`` to the kernel."""
+    c = _make_client()
+    c._kernel_session = MagicMock()
+    list_schemas = c._kernel_session.metadata.return_value.list_schemas
+    list_schemas.return_value = _stream_with_schema()
+    cursor = MagicMock()
+    cursor.arraysize = 100
+    cursor.buffer_size_bytes = 1024
+
+    c.get_schemas(
+        session_id=MagicMock(),
+        max_rows=1,
+        max_bytes=1,
+        cursor=cursor,
+        catalog_name="",
+        schema_name="ignored",
+    )
+
+    list_schemas.assert_called_once_with(catalog=None, schema_pattern="")
+
+
 def test_get_tables_preserves_empty_patterns():
     c = _make_client()
     c._kernel_session = MagicMock()
@@ -1643,9 +1665,64 @@ def test_get_tables_preserves_empty_patterns():
     )
 
 
-@pytest.mark.parametrize("filter_value", ["", "   "])
-def test_get_columns_preserves_blank_filters(filter_value):
-    """Blank strings remain filters instead of becoming match-all ``None``."""
+def test_get_columns_preserves_empty_patterns():
+    c = _make_client()
+    c._kernel_session = MagicMock()
+    list_columns = c._kernel_session.metadata.return_value.list_columns
+    list_columns.return_value = _stream_with_schema()
+    cursor = MagicMock()
+    cursor.arraysize = 100
+    cursor.buffer_size_bytes = 1024
+
+    c.get_columns(
+        session_id=MagicMock(),
+        max_rows=1,
+        max_bytes=1,
+        cursor=cursor,
+        catalog_name="main",
+        schema_name="",
+        table_name="",
+        column_name="",
+    )
+
+    list_columns.assert_called_once_with(
+        catalog="main",
+        schema_pattern="",
+        table_pattern="",
+        column_pattern="",
+    )
+
+
+def test_get_columns_empty_catalog_uses_empty_pattern():
+    """An empty catalog matches nothing without constructing ``Identifier("")``."""
+    c = _make_client()
+    c._kernel_session = MagicMock()
+    list_columns = c._kernel_session.metadata.return_value.list_columns
+    list_columns.return_value = _stream_with_schema()
+    cursor = MagicMock()
+    cursor.arraysize = 100
+    cursor.buffer_size_bytes = 1024
+
+    c.get_columns(
+        session_id=MagicMock(),
+        max_rows=1,
+        max_bytes=1,
+        cursor=cursor,
+        catalog_name="",
+        schema_name="ignored",
+        table_name="table",
+        column_name="column",
+    )
+
+    list_columns.assert_called_once_with(
+        catalog=None,
+        schema_pattern="",
+        table_pattern="table",
+        column_pattern="column",
+    )
+
+
+def test_get_columns_preserves_whitespace_for_kernel_validation():
     c = _make_client()
     c._kernel_session = MagicMock()
     list_columns = c._kernel_session.metadata.return_value.list_columns
@@ -1659,17 +1736,17 @@ def test_get_columns_preserves_blank_filters(filter_value):
         max_rows=1,
         max_bytes=1,
         cursor=cursor,
-        catalog_name=filter_value,
-        schema_name=filter_value,
-        table_name=filter_value,
-        column_name=filter_value,
+        catalog_name="   ",
+        schema_name="   ",
+        table_name="   ",
+        column_name="   ",
     )
 
     list_columns.assert_called_once_with(
-        catalog=filter_value,
-        schema_pattern=filter_value,
-        table_pattern=filter_value,
-        column_pattern=filter_value,
+        catalog="   ",
+        schema_pattern="   ",
+        table_pattern="   ",
+        column_pattern="   ",
     )