fix(mcp): normalize column names to fix time series filter prompt issue (#37187)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

tests/unit_tests/mcp_service/explore/tool/test_generate_explore_link.py

@@ -35,6 +35,7 @@ from superset.mcp_service.chart.schemas import (
     TableChartConfig,
     XYChartConfig,
 )
+from superset.mcp_service.common.error_schemas import DatasetContext
 
 logging.basicConfig(level=logging.DEBUG)
 logger = logging.getLogger(__name__)
@@ -706,3 +707,151 @@ class TestGenerateExploreLink:
             assert result.data["form_data"].get("x_axis") == "date"
             # Verify datasource field format: "{dataset_id}__table"
             assert result.data["form_data"].get("datasource") == "1__table"
+
+
+class TestGenerateExploreLinkColumnNormalization:
+    """Tests that generate_explore_link normalizes column names.
+
+    This verifies the fix where user-provided column names in wrong case
+    (e.g., 'order_date') are normalized to the canonical dataset name
+    (e.g., 'OrderDate') before being used in form_data.
+    """
+
+    @patch(
+        "superset.mcp_service.chart.validation.dataset_validator.DatasetValidator._get_dataset_context"
+    )
+    @patch("superset.daos.dataset.DatasetDAO.find_by_id")
+    @patch(
+        "superset.mcp_service.commands.create_form_data.MCPCreateFormDataCommand.run"
+    )
+    @pytest.mark.asyncio
+    async def test_xy_chart_x_axis_normalized_in_form_data(
+        self,
+        mock_create_form_data,
+        mock_find_dataset,
+        mock_get_context,
+        mcp_server,
+    ):
+        """x-axis column name in wrong case is normalized in form_data."""
+        mock_create_form_data.return_value = "norm_test_key_1"
+        mock_find_dataset.return_value = _mock_dataset(id=18)
+        mock_get_context.return_value = DatasetContext(
+            id=18,
+            table_name="Vehicle Sales",
+            schema="public",
+            database_name="examples",
+            available_columns=[
+                {"name": "OrderDate", "type": "DATE", "is_temporal": True},
+                {"name": "Sales", "type": "FLOAT", "is_numeric": True},
+            ],
+            available_metrics=[],
+        )
+
+        config = XYChartConfig(
+            chart_type="xy",
+            x=ColumnRef(name="orderdate"),
+            y=[ColumnRef(name="sales", aggregate="SUM")],
+            kind="line",
+        )
+        request = GenerateExploreLinkRequest(dataset_id="18", config=config)
+
+        async with Client(mcp_server) as client:
+            result = await client.call_tool(
+                "generate_explore_link", {"request": request.model_dump()}
+            )
+
+            assert result.data["error"] is None
+            # x-axis should be normalized from 'orderdate' to 'OrderDate'
+            assert result.data["form_data"]["x_axis"] == "OrderDate"
+
+    @patch(
+        "superset.mcp_service.chart.validation.dataset_validator.DatasetValidator._get_dataset_context"
+    )
+    @patch("superset.daos.dataset.DatasetDAO.find_by_id")
+    @patch(
+        "superset.mcp_service.commands.create_form_data.MCPCreateFormDataCommand.run"
+    )
+    @pytest.mark.asyncio
+    async def test_filter_column_normalized_in_form_data(
+        self,
+        mock_create_form_data,
+        mock_find_dataset,
+        mock_get_context,
+        mcp_server,
+    ):
+        """Filter column name in wrong case is normalized in adhoc_filters."""
+        mock_create_form_data.return_value = "norm_test_key_2"
+        mock_find_dataset.return_value = _mock_dataset(id=18)
+        mock_get_context.return_value = DatasetContext(
+            id=18,
+            table_name="Vehicle Sales",
+            schema="public",
+            database_name="examples",
+            available_columns=[
+                {"name": "OrderDate", "type": "DATE", "is_temporal": True},
+                {"name": "Sales", "type": "FLOAT", "is_numeric": True},
+            ],
+            available_metrics=[],
+        )
+
+        config = XYChartConfig(
+            chart_type="xy",
+            x=ColumnRef(name="orderdate"),
+            y=[ColumnRef(name="sales", aggregate="SUM")],
+            kind="line",
+            filters=[
+                FilterConfig(column="orderdate", op=">", value="2023-01-01"),
+            ],
+        )
+        request = GenerateExploreLinkRequest(dataset_id="18", config=config)
+
+        async with Client(mcp_server) as client:
+            result = await client.call_tool(
+                "generate_explore_link", {"request": request.model_dump()}
+            )
+
+            assert result.data["error"] is None
+            form_data = result.data["form_data"]
+            # x-axis normalized
+            assert form_data["x_axis"] == "OrderDate"
+            # filter subject normalized to match x-axis
+            adhoc_filters = form_data.get("adhoc_filters", [])
+            assert len(adhoc_filters) == 1
+            assert adhoc_filters[0]["subject"] == "OrderDate"
+
+    @patch(
+        "superset.mcp_service.chart.validation.dataset_validator.DatasetValidator._get_dataset_context"
+    )
+    @patch("superset.daos.dataset.DatasetDAO.find_by_id")
+    @patch(
+        "superset.mcp_service.commands.create_form_data.MCPCreateFormDataCommand.run"
+    )
+    @pytest.mark.asyncio
+    async def test_normalization_fallback_when_dataset_not_found(
+        self,
+        mock_create_form_data,
+        mock_find_dataset,
+        mock_get_context,
+        mcp_server,
+    ):
+        """When dataset context is unavailable, original names pass through."""
+        mock_create_form_data.return_value = "norm_test_key_3"
+        mock_find_dataset.return_value = _mock_dataset(id=99)
+        mock_get_context.return_value = None
+
+        config = XYChartConfig(
+            chart_type="xy",
+            x=ColumnRef(name="orderdate"),
+            y=[ColumnRef(name="sales", aggregate="SUM")],
+            kind="line",
+        )
+        request = GenerateExploreLinkRequest(dataset_id="99", config=config)
+
+        async with Client(mcp_server) as client:
+            result = await client.call_tool(
+                "generate_explore_link", {"request": request.model_dump()}
+            )
+
+            assert result.data["error"] is None
+            # original names should pass through unchanged
+            assert result.data["form_data"]["x_axis"] == "orderdate"