[BUGFIX] Databricks Fix Type Translation - `ExpectColumnValuesToBeI…

…nTypeList` and `ExpectColumnValuesToBeInType` (#10791)

great_expectations/expectations/core/expect_column_values_to_be_in_type_list.py

@@ -458,7 +458,7 @@ def _validate_pandas(  # noqa: C901, PLR0912
     def _validate_sqlalchemy(self, actual_column_type, expected_types_list, execution_engine):
         if expected_types_list is None:
             success = True
-        elif execution_engine.dialect_name == GXSqlDialect.SNOWFLAKE:
+        elif execution_engine.dialect_name in [GXSqlDialect.SNOWFLAKE, GXSqlDialect.DATABRICKS]:
             success = isinstance(actual_column_type, str) and any(
                 actual_column_type.lower() == expected_type.lower()
                 for expected_type in expected_types_list

great_expectations/expectations/core/expect_column_values_to_be_of_type.py

@@ -412,7 +412,7 @@ def _validate_sqlalchemy(self, actual_column_type, expected_type, execution_engi
 
         if expected_type is None:
             success = True
-        elif execution_engine.dialect_name == GXSqlDialect.SNOWFLAKE:
+        elif execution_engine.dialect_name in [GXSqlDialect.SNOWFLAKE, GXSqlDialect.DATABRICKS]:
             success = (
                 isinstance(actual_column_type, str)
                 and actual_column_type.lower() == expected_type.lower()

great_expectations/expectations/metrics/util.py

@@ -414,16 +414,19 @@ def get_sqlalchemy_column_metadata(  # noqa: C901
             )
 
         dialect_name = execution_engine.dialect.name
-        if dialect_name == GXSqlDialect.SNOWFLAKE:
+        if dialect_name in [GXSqlDialect.SNOWFLAKE, GXSqlDialect.DATABRICKS]:
             # WARNING: Do not alter columns in place, as they are cached on the inspector
             columns_copy = [column.copy() for column in columns]
             for column in columns_copy:
                 column["type"] = column["type"].compile(dialect=execution_engine.dialect)
-            return [
-                # TODO: SmartColumn should know the dialect and do lookups based on that
-                CaseInsensitiveNameDict(column)
-                for column in columns_copy
-            ]
+            if dialect_name == GXSqlDialect.SNOWFLAKE:
+                return [
+                    # TODO: SmartColumn should know the dialect and do lookups based on that
+                    CaseInsensitiveNameDict(column)
+                    for column in columns_copy
+                ]
+            else:
+                return columns_copy
 
         return columns
     except AttributeError as e:

tests/integration/data_sources_and_expectations/expectations/test_expect_column_values_to_be_in_type_list.py

@@ -1,9 +1,14 @@
 import pandas as pd
 import pytest
 import sqlalchemy.types as sqltypes
+from packaging import version
 
 import great_expectations.expectations as gxe
+from great_expectations.compatibility.databricks import DATABRICKS_TYPES
 from great_expectations.compatibility.snowflake import SNOWFLAKE_TYPES
+from great_expectations.compatibility.sqlalchemy import (
+    sqlalchemy as sa,
+)
 from great_expectations.core.result_format import ResultFormat
 from great_expectations.datasource.fluent.interfaces import Batch
 from tests.integration.conftest import parameterize_batch_for_data_sources
@@ -379,3 +384,210 @@ def test_success_complete_snowflake(
     assert isinstance(result_dict["observed_value"], str)
     assert isinstance(expectation.type_list, list)
     assert result_dict["observed_value"] in expectation.type_list
+
+
+@pytest.mark.parametrize(
+    "expectation",
+    [
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="STRING", type_list=["STRING"]),
+            id="STRING",
+        ),
+        # SqlA Text gets converted to Databricks STRING
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="TEXT", type_list=["STRING"]),
+            id="TEXT",
+        ),
+        # SqlA UNICODE gets converted to Databricks STRING
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="UNICODE", type_list=["STRING"]),
+            id="UNICODE",
+        ),
+        # SqlA UNICODE_TEXT gets converted to Databricks STRING
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="UNICODE_TEXT", type_list=["STRING"]),
+            id="UNICODE_TEXT",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="BOOLEAN", type_list=["BOOLEAN"]),
+            id="BOOLEAN",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(
+                column="DECIMAL", type_list=["DECIMAL", "DECIMAL(10, 0)"]
+            ),
+            id="DECIMAL",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="DATE", type_list=["DATE"]),
+            id="DATE",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="TIMESTAMP", type_list=["TIMESTAMP"]),
+            id="TIMESTAMP",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(
+                column="TIMESTAMP_NTZ", type_list=["TIMESTAMP_NTZ"]
+            ),
+            id="TIMESTAMP_NTZ",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="FLOAT", type_list=["FLOAT"]),
+            id="FLOAT",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="INT", type_list=["INT"]),
+            id="INT",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(column="TINYINT", type_list=["TINYINT"]),
+            id="TINYINT",
+        ),
+        pytest.param(
+            gxe.ExpectColumnValuesToBeInTypeList(
+                column="DECIMAL", type_list=["DECIMAL", "DECIMAL(10, 0)"]
+            ),
+            id="DECIMAL",
+        ),
+        # SqlA Time gets converted to Databricks STRING,
+        # but is not supported by our testing framework
+        # pytest.param(
+        #     gxe.ExpectColumnValuesToBeInTypeList(column="TIME", type_list=["STRING"]),
+        #     id="TIME",
+        # ),
+        # SqlA UUID gets converted to Databricks STRING,
+        # but is not supported by our testing framework.
+        # pytest.param(
+        #     gxe.ExpectColumnValuesToBeInTypeList(column="UUID", type_list=["STRING"]),
+        #     id="UUID",
+        # )
+    ],
+)
+@parameterize_batch_for_data_sources(
+    data_source_configs=[
+        DatabricksDatasourceTestConfig(
+            column_types={
+                "STRING": DATABRICKS_TYPES.STRING,
+                "TEXT": sqltypes.Text,
+                "UNICODE": sqltypes.Unicode,
+                "UNICODE_TEXT": sqltypes.UnicodeText,
+                "BIGINT": sqltypes.BigInteger,
+                "BOOLEAN": sqltypes.BOOLEAN,
+                "DATE": sqltypes.DATE,
+                "TIMESTAMP_NTZ": DATABRICKS_TYPES.TIMESTAMP_NTZ,
+                "TIMESTAMP": DATABRICKS_TYPES.TIMESTAMP,
+                "FLOAT": sqltypes.Float,
+                "INT": sqltypes.Integer,
+                "DECIMAL": sqltypes.Numeric,
+                "SMALLINT": sqltypes.SmallInteger,
+                "TINYINT": DATABRICKS_TYPES.TINYINT,
+                # "TIME": sqltypes.Time,
+                # "UUID": sqltypes.UUID,
+            }
+        )
+    ],
+    data=pd.DataFrame(
+        {
+            "STRING": ["a", "b", "c"],
+            "TEXT": ["a", "b", "c"],
+            "UNICODE": ["\u00e9", "\u00e9", "\u00e9"],
+            "UNICODE_TEXT": ["a", "b", "c"],
+            "BIGINT": [1111, 2222, 3333],
+            "BOOLEAN": [True, True, False],
+            "DATE": [
+                "2021-01-01",
+                "2021-01-02",
+                "2021-01-03",
+            ],
+            "TIMESTAMP_NTZ": [
+                "2021-01-01 00:00:00",
+                "2021-01-02 00:00:00",
+                "2021-01-03 00:00:00",
+            ],
+            "TIMESTAMP": [
+                "2021-01-01 00:00:00",
+                "2021-01-02 00:00:00",
+                "2021-01-03 00:00:00",
+            ],
+            "DOUBLE": [1.0, 2.0, 3.0],
+            "FLOAT": [1.0, 2.0, 3.0],
+            "INT": [1, 2, 3],
+            "DECIMAL": [1.1, 2.2, 3.3],
+            "SMALLINT": [1, 2, 3],
+            # "TIME": [
+            #     sa.Time("22:17:33.123456"),
+            #     sa.Time("22:17:33.123456"),
+            #     sa.Time("22:17:33.123456"),
+            # ],
+            # "UUID": [
+            #      uuid.UUID("905993ea-f50e-4284-bea0-5be3f0ed7031"),
+            #      uuid.UUID("9406b631-fa2f-41cf-b666-f9a2ac3118c1"),
+            #      uuid.UUID("47538f05-32e3-4594-80e2-0b3b33257ae7")
+            #  ],
+        },
+        dtype="object",
+    ),
+)
+def test_success_complete_databricks(
+    batch_for_datasource: Batch, expectation: gxe.ExpectColumnValuesToBeInTypeList
+) -> None:
+    result = batch_for_datasource.validate(expectation, result_format=ResultFormat.COMPLETE)
+    result_dict = result.to_json_dict()["result"]
+
+    assert result.success
+    assert isinstance(result_dict, dict)
+    assert isinstance(result_dict["observed_value"], str)
+    assert isinstance(expectation.type_list, list)
+    assert result_dict["observed_value"] in expectation.type_list
+
+
+if version.parse(sa.__version__) >= version.parse("2.0.0"):
+    # Note: why not use pytest.skip?
+    # the import of `sqltypes.Double` is only possible in sqlalchemy >= 2.0.0
+    # the import is done as part of the instantiation of the test, which includes
+    # processing the pytest.skip() statement. This way, we skip the instantiation
+    # of the test entirely.
+    @pytest.mark.parametrize(
+        "expectation",
+        [
+            pytest.param(
+                gxe.ExpectColumnValuesToBeInTypeList(
+                    column="DOUBLE", type_list=["DOUBLE", "FLOAT"]
+                ),
+                id="DOUBLE",
+            )
+        ],
+    )
+    @parameterize_batch_for_data_sources(
+        data_source_configs=[
+            DatabricksDatasourceTestConfig(
+                column_types={
+                    "DOUBLE": sqltypes.Double,
+                }
+            )
+        ],
+        data=pd.DataFrame(
+            {
+                "DOUBLE": [1.0, 2.0, 3.0],
+            },
+            dtype="object",
+        ),
+    )
+    def test_success_complete_databricks_double_type_only(
+        batch_for_datasource: Batch, expectation: gxe.ExpectColumnValuesToBeInTypeList
+    ) -> None:
+        """What does this test and why?
+
+        Databricks mostly uses SqlA types directly, but the double type is
+        only available after sqlalchemy 2.0. We therefore split up the test
+        into 2 parts, with this test being skipped if the SA version is too low.
+        """
+        result = batch_for_datasource.validate(expectation, result_format=ResultFormat.COMPLETE)
+        result_dict = result.to_json_dict()["result"]
+
+        assert result.success
+        assert isinstance(result_dict, dict)
+        assert isinstance(result_dict["observed_value"], str)
+        assert isinstance(expectation.type_list, list)
+        assert result_dict["observed_value"] in expectation.type_list

tests/integration/data_sources_and_expectations/expectations/test_expect_column_values_to_be_of_type.py

@@ -66,13 +66,12 @@ def test_success_for_type__INTEGER(batch_for_datasource: Batch) -> None:
     assert result.success
 
 
-@pytest.mark.xfail
 @parameterize_batch_for_data_sources(
     data_source_configs=[DatabricksDatasourceTestConfig()],
     data=DATA,
 )
 def test_success_for_type__Integer(batch_for_datasource: Batch) -> None:
-    expectation = gxe.ExpectColumnValuesToBeOfType(column=INTEGER_COLUMN, type_="Integer")
+    expectation = gxe.ExpectColumnValuesToBeOfType(column=INTEGER_COLUMN, type_="INT")
     result = batch_for_datasource.validate(expectation)
     assert result.success