Implement KedroDataCatalog CLI tests (#4456)

* Draft for TestCatalogFactoryCommands

Signed-off-by: Elena Khaustova <[email protected]>

* Added template for TestCatalogListCommand

Signed-off-by: Elena Khaustova <[email protected]>

* Updated list command to print _LazyDataset actual type

Signed-off-by: Elena Khaustova <[email protected]>

* Test fixture parametrization

Signed-off-by: Elena Khaustova <[email protected]>

* Implemented tests for TestCatalogCreateCommand

Signed-off-by: Elena Khaustova <[email protected]>

* Added rank factories tests

Signed-off-by: Elena Khaustova <[email protected]>

* Updated test_catalog.py to use fixtures

Signed-off-by: Elena Khaustova <[email protected]>

* Removed test_kedro_data_catalog.py

Signed-off-by: Elena Khaustova <[email protected]>

* Duplicated fake project cli

Signed-off-by: Elena Khaustova <[email protected]>

* Removed duplicasted fixture

Signed-off-by: Elena Khaustova <[email protected]>

* Added a clarification about setting fixture params

Signed-off-by: Elena Khaustova <[email protected]>

* Returned fake_project_cli_parametrized

Signed-off-by: Elena Khaustova <[email protected]>

* Returned trial fail loop

Signed-off-by: Elena Khaustova <[email protected]>

---------

Signed-off-by: Elena Khaustova <[email protected]>

kedro/framework/cli/catalog.py

@@ -14,6 +14,7 @@
 from kedro.framework.project import pipelines, settings
 from kedro.framework.session import KedroSession
 from kedro.io.data_catalog import DataCatalog
+from kedro.io.kedro_data_catalog import _LazyDataset
 
 if TYPE_CHECKING:
     from pathlib import Path
@@ -64,7 +65,6 @@ def list_datasets(metadata: ProjectMetadata, pipeline: str, env: str) -> None:
 
     session = _create_session(metadata.package_name, env=env)
     context = session.load_context()
-
     try:
         data_catalog = context.catalog
         datasets_meta = data_catalog._datasets
@@ -126,7 +126,10 @@ def _map_type_to_datasets(
     """
     mapping = defaultdict(list)  # type: ignore[var-annotated]
     for dataset_name in filterfalse(is_parameter, datasets):
-        ds_type = datasets_meta[dataset_name].__class__.__name__
+        if isinstance(datasets_meta[dataset_name], _LazyDataset):
+            ds_type = str(datasets_meta[dataset_name]).split(".")[-1]
+        else:
+            ds_type = datasets_meta[dataset_name].__class__.__name__
         if dataset_name not in mapping[ds_type]:
             mapping[ds_type].append(dataset_name)
     return mapping

tests/framework/cli/conftest.py

@@ -27,6 +27,7 @@
 from kedro.framework.cli.starters import create_cli
 from kedro.framework.project import configure_project, pipelines, settings
 from kedro.framework.startup import ProjectMetadata
+from kedro.io import KedroDataCatalog
 
 REPO_NAME = "dummy_project"
 PACKAGE_NAME = "dummy_package"
@@ -116,7 +117,9 @@ def fake_kedro_cli():
 
 @fixture(scope="module")
 def fake_project_cli(
-    fake_repo_path: Path, dummy_config: Path, fake_kedro_cli: click.CommandCollection
+    fake_repo_path: Path,
+    dummy_config: Path,
+    fake_kedro_cli: click.CommandCollection,
 ):
     old_settings = settings.as_dict()
     starter_path = Path(__file__).resolve().parents[3]
@@ -126,7 +129,64 @@ def fake_project_cli(
     )
     # Delete the project logging.yml, which leaves behind info.log and error.log files.
     # This leaves logging config as the framework default.
-    (fake_repo_path / "conf" / "logging.yml").unlink()
+    try:
+        (fake_repo_path / "conf" / "logging.yml").unlink()
+    except FileNotFoundError:
+        pass
+
+    # NOTE: Here we load a couple of modules, as they would be imported in
+    # the code and tests.
+    # It's safe to remove the new entries from path due to the python
+    # module caching mechanism. Any `reload` on it will not work though.
+    old_path = sys.path.copy()
+    sys.path = [str(fake_repo_path / "src"), *sys.path]
+
+    import_module(PACKAGE_NAME)
+    configure_project(PACKAGE_NAME)
+    yield fake_kedro_cli
+
+    # reset side-effects of configure_project
+    pipelines.configure()
+
+    for key, value in old_settings.items():
+        settings.set(key, value)
+    sys.path = old_path
+
+    # configure_project does imports that add PACKAGE_NAME.pipelines,
+    # PACKAGE_NAME.settings to sys.modules. These need to be removed.
+    # Ideally we would reset sys.modules to exactly what it was before
+    # running anything, but removal of distutils.build.commands from
+    # sys.modules mysteriously makes some tests for `kedro micropkg package`
+    # fail on Windows, Python 3.7 and 3.8.
+    for module in list(sys.modules.keys()):
+        if module.startswith(PACKAGE_NAME):
+            del sys.modules[module]
+
+
+# We use `None` as the first parameter since the default `DATA_CATALOG_CLASS`
+# set in `settings.py` is `DataCatalog`. We do not set `DataCatalog` explicitly
+# in fixture params to test loading the default class.
+@fixture(scope="module", params=[None, KedroDataCatalog])
+def fake_project_cli_parametrized(
+    fake_repo_path: Path,
+    dummy_config: Path,
+    fake_kedro_cli: click.CommandCollection,
+    request,
+):
+    # TODO: remove parametrization after removing old catalog as KedroDataCatalog will be default
+    default_catalog = request.param
+    old_settings = settings.as_dict()
+    starter_path = Path(__file__).resolve().parents[3]
+    starter_path = starter_path / "features" / "steps" / "test_starter"
+    CliRunner().invoke(
+        fake_kedro_cli, ["new", "-c", str(dummy_config), "--starter", str(starter_path)]
+    )
+    # Delete the project logging.yml, which leaves behind info.log and error.log files.
+    # This leaves logging config as the framework default.
+    try:
+        (fake_repo_path / "conf" / "logging.yml").unlink()
+    except FileNotFoundError:
+        pass
 
     # NOTE: Here we load a couple of modules, as they would be imported in
     # the code and tests.
@@ -137,6 +197,8 @@ def fake_project_cli(
 
     import_module(PACKAGE_NAME)
     configure_project(PACKAGE_NAME)
+    if default_catalog is not None:
+        settings.set("DATA_CATALOG_CLASS", default_catalog)
     yield fake_kedro_cli
 
     # reset side-effects of configure_project

tests/framework/cli/pipeline/test_pipeline.py

@@ -204,7 +204,7 @@ def test_catalog_and_params(
 
         with KedroSession.create() as session:
             ctx = session.load_context()
-        assert isinstance(ctx.catalog._datasets["ds_from_pipeline"], CSVDataset)
+        assert isinstance(ctx.catalog._get_dataset("ds_from_pipeline"), CSVDataset)
         assert isinstance(ctx.catalog.load("ds_from_pipeline"), DataFrame)
         assert ctx.params["params_from_pipeline"] == params_dict["params_from_pipeline"]