fix: availability checks for basins

CHANGELOG

@@ -1,3 +1,5 @@
+0.54.2
+ - fix: availability checks for basins
 0.54.1
  - fix: properly check for basin availability before getting features
 0.54.0

dclab/rtdc_dataset/core.py

@@ -244,19 +244,22 @@ def _get_basin_feature_data(
                 if basin_type is not None and basin_type != bn.basin_type:
                     # User asked for specific basin type
                     continue
-                if feat in bn.features:
+                try:
                     # There are all kinds of errors that may happen here.
+                    # Note that `bn.features` can already trigger an
+                    # availability check that may raise a ValueError.
                     # TODO:
                     #  Introduce some kind of callback so the user knows
                     #  why the data are not available. The current solution
                     #  (fail silently) is not sufficiently transparent,
                     #  especially when considering networking issues.
-                    try:
+                    if feat in bn.features:
                         data = bn.get_feature_data(feat)
-                    except BaseException:
-                        # Basin data not available
-                        pass
-                    break
+                        # The data are available, we may abort the search.
+                        break
+                except BaseException:
+                    # Basin data not available
+                    pass
         return data
 
     @staticmethod

tests/test_rtdc_fmt_hdf5_basins.py

@@ -105,6 +105,41 @@ def test_basin_not_available():
         _ = bn.ds
 
 
+def test_basin_nothing_available():
+    h5path = retrieve_data("fmt-hdf5_fl_wide-channel_2023.zip")
+    h5path_small = h5path.with_name("smaller.rtdc")
+
+    # Dataset creation
+    with h5py.File(h5path) as src, h5py.File(h5path_small, "w") as dst:
+        # first, copy all the scalar features to the new file
+        rtdc_dataset.rtdc_copy(src_h5file=src,
+                               dst_h5file=dst,
+                               features="scalar")
+        # Next, store the basin information in the new dataset
+        bdat = {
+            "type": "file",
+            "format": "hdf5",
+            "paths": [
+                "fake.rtdc",  # fake path
+            ]
+        }
+        blines = json.dumps(bdat, indent=2).split("\n")
+        basins = dst.require_group("basins")
+        with RTDCWriter(dst, mode="append") as hw:
+            hw.write_text(basins, "my_basin", blines)
+        # sanity checks
+        assert "deform" in dst["events"]
+        assert "image" not in dst["events"]
+
+    h5path.unlink()
+
+    # Now open the scalar dataset and check whether basins missing
+    with new_dataset(h5path_small) as ds:
+        assert "image" not in ds
+        assert not ds.features_basin
+        _ = ds["index"]
+
+
 def test_basin_features_path_absolute():
     """Create a dataset that refers to a basin in a relative path"""
     h5path = retrieve_data("fmt-hdf5_fl_wide-channel_2023.zip")

tests/test_rtdc_fmt_s3_basin.py

@@ -112,6 +112,9 @@ def test_basin_not_available(url):
     # Open the dataset and check whether basin is missing
     with new_dataset(h5path) as ds:
         assert not ds.features_basin
+        # This is a very subtle test for checking whether invalid basins
+        # are just ignored:
+        _ = ds["index"]
 
     # Also test that on a lower level
     bn = S3Basin("https://example.com/nonexistentbucket/nonexistentkey")