datahub/metadata-ingestion/tests/unit/test_glue_source.py

import json
from pathlib import Path
from typing import Any, Dict, Optional, Tuple, Type, cast
from unittest.mock import patch

import pytest
from botocore.stub import Stubber
from freezegun import freeze_time

from datahub.configuration.common import ConfigurationError
from datahub.ingestion.api.common import PipelineContext
from datahub.ingestion.extractor.schema_util import avro_schema_to_mce_fields
from datahub.ingestion.run.pipeline import Pipeline
from datahub.ingestion.sink.file import write_metadata_file
from datahub.ingestion.source.aws.glue import GlueSource, GlueSourceConfig
from datahub.ingestion.source.state.checkpoint import Checkpoint
from datahub.ingestion.source.state.sql_common_state import (
    BaseSQLAlchemyCheckpointState,
)
from datahub.metadata.com.linkedin.pegasus2avro.schema import (
    ArrayTypeClass,
    MapTypeClass,
    RecordTypeClass,
    StringTypeClass,
)
from datahub.utilities.hive_schema_to_avro import get_avro_schema_for_hive_column
from tests.test_helpers import mce_helpers
from tests.test_helpers.state_helpers import (
    run_and_get_pipeline,
    validate_all_providers_have_committed_successfully,
)
from tests.test_helpers.type_helpers import PytestConfig
from tests.unit.test_glue_source_stubs import (
    databases_1,
    databases_2,
    get_bucket_tagging,
    get_databases_response,
    get_dataflow_graph_response_1,
    get_dataflow_graph_response_2,
    get_jobs_response,
    get_object_body_1,
    get_object_body_2,
    get_object_response_1,
    get_object_response_2,
    get_object_tagging,
    get_tables_response_1,
    get_tables_response_2,
    tables_1,
    tables_2,
)

FROZEN_TIME = "2020-04-14 07:00:00"
GMS_PORT = 8080
GMS_SERVER = f"http://localhost:{GMS_PORT}"


def glue_source(platform_instance: Optional[str] = None) -> GlueSource:
    return GlueSource(
        ctx=PipelineContext(run_id="glue-source-test"),
        config=GlueSourceConfig(
            aws_region="us-west-2",
            extract_transforms=True,
            platform_instance=platform_instance,
            use_s3_bucket_tags=True,
            use_s3_object_tags=True,
        ),
    )


column_type_test_cases: Dict[str, Tuple[str, Type]] = {
    "char": ("char", StringTypeClass),
    "array": ("array<int>", ArrayTypeClass),
    "map": ("map<string, int>", MapTypeClass),
    "struct": ("struct<a:int, b:string>", RecordTypeClass),
}


@pytest.mark.parametrize(
    "hive_column_type, expected_type",
    column_type_test_cases.values(),
    ids=column_type_test_cases.keys(),
)
def test_column_type(hive_column_type: str, expected_type: Type) -> None:
    avro_schema = get_avro_schema_for_hive_column(
        f"test_column_{hive_column_type}", hive_column_type
    )
    schema_fields = avro_schema_to_mce_fields(json.dumps(avro_schema))
    actual_schema_field_type = schema_fields[0].type
    assert type(actual_schema_field_type.type) == expected_type


@pytest.mark.parametrize(
    "platform_instance, mce_file, mce_golden_file",
    [
        (None, "glue_mces.json", "glue_mces_golden.json"),
        (
            "some_instance_name",
            "glue_mces_platform_instance.json",
            "glue_mces_platform_instance_golden.json",
        ),
    ],
)
@freeze_time(FROZEN_TIME)
def test_glue_ingest(
    tmp_path: Path,
    pytestconfig: PytestConfig,
    platform_instance: str,
    mce_file: str,
    mce_golden_file: str,
) -> None:
    glue_source_instance = glue_source(platform_instance=platform_instance)

    with Stubber(glue_source_instance.glue_client) as glue_stubber:
        glue_stubber.add_response("get_databases", get_databases_response, {})
        glue_stubber.add_response(
            "get_tables",
            get_tables_response_1,
            {"DatabaseName": "flights-database"},
        )
        glue_stubber.add_response(
            "get_tables",
            get_tables_response_2,
            {"DatabaseName": "test-database"},
        )
        glue_stubber.add_response("get_jobs", get_jobs_response, {})
        glue_stubber.add_response(
            "get_dataflow_graph",
            get_dataflow_graph_response_1,
            {"PythonScript": get_object_body_1},
        )
        glue_stubber.add_response(
            "get_dataflow_graph",
            get_dataflow_graph_response_2,
            {"PythonScript": get_object_body_2},
        )

        with Stubber(glue_source_instance.s3_client) as s3_stubber:
            for _ in range(
                len(get_tables_response_1["TableList"])
                + len(get_tables_response_2["TableList"])
            ):
                s3_stubber.add_response(
                    "get_bucket_tagging",
                    get_bucket_tagging(),
                )
                s3_stubber.add_response(
                    "get_object_tagging",
                    get_object_tagging(),
                )

            s3_stubber.add_response(
                "get_object",
                get_object_response_1(),
                {
                    "Bucket": "aws-glue-assets-123412341234-us-west-2",
                    "Key": "scripts/job-1.py",
                },
            )
            s3_stubber.add_response(
                "get_object",
                get_object_response_2(),
                {
                    "Bucket": "aws-glue-assets-123412341234-us-west-2",
                    "Key": "scripts/job-2.py",
                },
            )

            mce_objects = [wu.metadata for wu in glue_source_instance.get_workunits()]

            glue_stubber.assert_no_pending_responses()
            s3_stubber.assert_no_pending_responses()

            write_metadata_file(tmp_path / mce_file, mce_objects)

    # Verify the output.
    test_resources_dir = pytestconfig.rootpath / "tests/unit/glue"
    mce_helpers.check_golden_file(
        pytestconfig,
        output_path=tmp_path / mce_file,
        golden_path=test_resources_dir / mce_golden_file,
    )


def test_underlying_platform_takes_precendence():
    source = GlueSource(
        ctx=PipelineContext(run_id="glue-source-test"),
        config=GlueSourceConfig(aws_region="us-west-2", underlying_platform="athena"),
    )
    assert source.platform == "athena"


def test_platform_takes_precendence_over_underlying_platform():
    source = GlueSource(
        ctx=PipelineContext(run_id="glue-source-test"),
        config=GlueSourceConfig(
            aws_region="us-west-2", platform="athena", underlying_platform="glue"
        ),
    )
    assert source.platform == "athena"


def test_underlying_platform_must_be_valid():
    with pytest.raises(ConfigurationError):
        GlueSource(
            ctx=PipelineContext(run_id="glue-source-test"),
            config=GlueSourceConfig(
                aws_region="us-west-2", underlying_platform="data-warehouse"
            ),
        )


def test_platform_must_be_valid():
    with pytest.raises(ConfigurationError):
        GlueSource(
            ctx=PipelineContext(run_id="glue-source-test"),
            config=GlueSourceConfig(aws_region="us-west-2", platform="data-warehouse"),
        )


def test_without_underlying_platform():
    source = GlueSource(
        ctx=PipelineContext(run_id="glue-source-test"),
        config=GlueSourceConfig(aws_region="us-west-2"),
    )
    assert source.platform == "glue"


def get_current_checkpoint_from_pipeline(
    pipeline: Pipeline,
) -> Optional[Checkpoint]:
    glue_source = cast(GlueSource, pipeline.source)
    return glue_source.get_current_checkpoint(
        glue_source.stale_entity_removal_handler.job_id
    )


@freeze_time(FROZEN_TIME)
def test_glue_stateful(pytestconfig, tmp_path, mock_time, mock_datahub_graph):
    test_resources_dir = pytestconfig.rootpath / "tests/unit/glue"

    deleted_actor_golden_mcs = "{}/glue_deleted_actor_mces_golden.json".format(
        test_resources_dir
    )

    stateful_config = {
        "stateful_ingestion": {
            "enabled": True,
            "remove_stale_metadata": True,
            "fail_safe_threshold": 100.0,
            "state_provider": {
                "type": "datahub",
                "config": {"datahub_api": {"server": GMS_SERVER}},
            },
        },
    }

    source_config_dict: Dict[str, Any] = {
        "extract_transforms": False,
        "aws_region": "eu-east-1",
        **stateful_config,
    }

    pipeline_config_dict: Dict[str, Any] = {
        "source": {
            "type": "glue",
            "config": source_config_dict,
        },
        "sink": {
            # we are not really interested in the resulting events for this test
            "type": "console"
        },
        "pipeline_name": "statefulpipeline",
    }

    with patch(
        "datahub.ingestion.source.state_provider.datahub_ingestion_checkpointing_provider.DataHubGraph",
        mock_datahub_graph,
    ) as mock_checkpoint:
        mock_checkpoint.return_value = mock_datahub_graph
        with patch(
            "datahub.ingestion.source.aws.glue.GlueSource.get_all_tables_and_databases",
        ) as mock_get_all_tables_and_databases:
            tables_on_first_call = tables_1
            tables_on_second_call = tables_2
            mock_get_all_tables_and_databases.side_effect = [
                (databases_1, tables_on_first_call),
                (databases_2, tables_on_second_call),
            ]

            pipeline_run1 = run_and_get_pipeline(pipeline_config_dict)
            checkpoint1 = get_current_checkpoint_from_pipeline(pipeline_run1)

            assert checkpoint1
            assert checkpoint1.state

            # Capture MCEs of second run to validate Status(removed=true)
            deleted_mces_path = "{}/{}".format(tmp_path, "glue_deleted_mces.json")
            pipeline_config_dict["sink"]["type"] = "file"
            pipeline_config_dict["sink"]["config"] = {"filename": deleted_mces_path}

            # Do the second run of the pipeline.
            pipeline_run2 = run_and_get_pipeline(pipeline_config_dict)
            checkpoint2 = get_current_checkpoint_from_pipeline(pipeline_run2)

            assert checkpoint2
            assert checkpoint2.state

            # Validate that all providers have committed successfully.
            validate_all_providers_have_committed_successfully(
                pipeline=pipeline_run1, expected_providers=1
            )
            validate_all_providers_have_committed_successfully(
                pipeline=pipeline_run2, expected_providers=1
            )

            # Validate against golden MCEs where Status(removed=true)
            mce_helpers.check_golden_file(
                pytestconfig,
                output_path=deleted_mces_path,
                golden_path=deleted_actor_golden_mcs,
            )

            # Perform all assertions on the states. The deleted table should not be
            # part of the second state
            state1 = cast(BaseSQLAlchemyCheckpointState, checkpoint1.state)
            state2 = cast(BaseSQLAlchemyCheckpointState, checkpoint2.state)
            difference_urns = list(
                state1.get_urns_not_in(type="*", other_checkpoint_state=state2)
            )

            assert len(difference_urns) == 1

            urn1 = (
                "urn:li:dataset:(urn:li:dataPlatform:glue,flights-database.avro,PROD)"
            )

            assert urn1 in difference_urns