simplify last failing tests

tests/conftest.py

@@ -305,9 +305,9 @@ def threadsafe_data_dir(tmp_path_factory):
 
 
 @pytest.fixture(scope="session")
-def nimbus(threadsafe_data_dir):
+def nimbus(threadsafe_data_dir, worker_id):
     return _nimbus(
-        data_dir=threadsafe_data_dir,
+        data_dir=default_cache_dir if worker_id == "master" else threadsafe_data_dir,
         repo=helpers.TESTDATA_REPO_URL,
         branch=helpers.TESTDATA_BRANCH,
     )

tests/test_ensembles.py

@@ -26,7 +26,6 @@
 
 from xclim import ensembles
 from xclim.indices.stats import get_dist
-from xclim.testing.helpers import TESTDATA_BRANCH
 
 
 # sklearn's KMeans doesn't accept the standard numpy Generator, so we create a special fixture for these tests
@@ -38,9 +37,7 @@ def random_state():
 
 
 class TestEnsembleStats:
-    def test_create_ensemble(
-        self, open_dataset, ensemble_dataset_objects, threadsafe_data_dir
-    ):
+    def test_create_ensemble(self, open_dataset, ensemble_dataset_objects, nimbus):
         ds_all = []
         for n in ensemble_dataset_objects["nc_files_simple"]:
             ds = open_dataset(n, decode_times=False)
@@ -62,10 +59,7 @@ def test_create_ensemble(
         ens1 = ensembles.create_ensemble(ds_all, realizations=reals)
 
         # Kinda a hack? Alternative is to open and rewrite in a temp folder.
-        files = [
-            threadsafe_data_dir / TESTDATA_BRANCH / "EnsembleStats" / Path(f).name
-            for f in ensemble_dataset_objects["nc_files_simple"]
-        ]
+        files = [nimbus.fetch(f) for f in ensemble_dataset_objects["nc_files_simple"]]
         ens2 = ensembles.create_ensemble(dict(zip(reals, files)))
         xr.testing.assert_identical(ens1, ens2)
 

tests/test_testing_utils.py

@@ -40,9 +40,15 @@ def file_md5_checksum(f_name):
         return hash_md5.hexdigest()
 
     @pytest.mark.requires_internet
-    def test_open_testdata(self):
+    def test_open_testdata(
+        self,
+    ):
+        from xclim.testing.helpers import default_cache_dir, default_testdata_version
+
         ds = helpers.open_dataset(
-            Path("cmip5/tas_Amon_CanESM2_rcp85_r1i1p1_200701-200712"), engine="h5netcdf"
+            Path("cmip5/tas_Amon_CanESM2_rcp85_r1i1p1_200701-200712.nc"),
+            cache_dir=default_cache_dir.joinpath(default_testdata_version),
+            engine="h5netcdf",
         )
         assert ds.lon.size == 128
 

xclim/testing/helpers.py

@@ -85,27 +85,6 @@
     $ env XCLIM_TESTDATA_BRANCH="my_testing_branch" pytest
 """
 
-PREFETCH_TESTING_DATA = bool(os.getenv("XCLIM_PREFETCH_TESTING_DATA"))
-"""Indicates whether the testing data should be downloaded when running tests.
-
-Notes
------
-When running tests multiple times, this flag allows developers to significantly speed up the pytest suite
-by preventing sha256sum checks for all downloaded files. Proceed with caution.
-
-This can be set for both `pytest` and `tox` by exporting the variable:
-
-.. code-block:: console
-
-    $ export XCLIM_PREFETCH_TESTING_DATA=1
-
-or setting the variable at runtime:
-
-.. code-block:: console
-
-    $ env XCLIM_PREFETCH_TESTING_DATA=1 pytest
-"""
-
 CACHE_DIR = os.getenv("XCLIM_DATA_DIR", default_cache_dir)
 """Sets the directory to store the testing datasets.
 
@@ -128,11 +107,13 @@
 
 
 __all__ = [
-    "PREFETCH_TESTING_DATA",
+    "CACHE_DIR",
     "TESTDATA_BRANCH",
+    "TESTDATA_REPO_URL",
     "add_example_file_paths",
     "assert_lazy",
     "default_cache_dir",
+    "default_testdata_version",
     "generate_atmos",
     "nimbus",
     "open_dataset",