fix: correctly identify intra-dataset basins via basin identifiers

Author: paulmueller

CHANGELOG

@@ -1,6 +1,8 @@
 1.0.5
+ - fix: correctly identify intra-dataset basins via basin identifiers
  - fix: list all features available in intra-dataset basins
  - fix: disable simple name-based basin check for intra-dataset basins
+ - setup: bump dclab to 0.66.0
 1.0.4
  - fix: resolve self-referencing basin creation in background job
 1.0.3

ckanext/dc_serve/jobs.py

@@ -225,7 +225,7 @@ def _get_intra_dataset_upstream_basins(res_dict, ds) -> list[dict]:
             else:
                 basin_map = ds[map_data][:]
 
-            # check whether the file name is in the list of resources.
+            # check whether the basin is in the list of resources.
             for res in pkg.resources:
                 if _is_basin_of_dataset(ds, res, bn_dict):
                     # upstream resource ID
@@ -300,6 +300,9 @@ def _is_basin_of_dataset(ds,
                          ):
     """Check whether a CKAN resource is a basin for a dataset
 
+    Return True when `resource_basin` is a basin of `ds`
+    as described by `basin_dict`.
+
     Parameters
     ----------
     ds:
@@ -313,26 +316,30 @@ def _is_basin_of_dataset(ds,
     ds_runid = ds.get_measurement_identifier()
     with get_dc_instance(resource_basin.id) as ds_bn:
         bn_runid = ds_bn.get_measurement_identifier()
+    bn_runid_dict = basin_dict.get("identifier")
 
-    if not bn_runid:
+    if not bn_runid or not bn_runid_dict:
         # No run identifier specified -> no basin inference possible.
         return False
 
+    if bn_runid != bn_runid_dict:
+        # Basin-resource mismatch.
+        # This resource is not the measurement referenced in this basin.
+        return False
+
     mapping_is_same = basin_dict.get("mapping", "same") == "same"
 
     if mapping_is_same and bn_runid == ds_runid:
         # This is an ideal case. Both run identifiers match and the mapping
         # is identical.
-        print(f"same {bn_runid} {ds_runid}")
         return True
 
-    elif not mapping_is_same and ds_runid.rsplit("_", 1)[0] == bn_runid:
-        # This is a slightly more complicated cases. If you split the
-        # run identifiers with an underscore (dclab 0.65.0 and dcnum 0.25.11
-        # correctly support this), then the referrer has a run identifier
-        # that starts with that of the basin.
-        print(f"basinmap {bn_runid} {ds_runid}")
+    elif not mapping_is_same and ds_runid.startswith(bn_runid):
+        # This is a slightly more complicated cases. If the mapping
+        # is not "same", this means that we have filtered datasets.
+        # For a filtered dataset, the identifier is the identifier of
+        # its basin plus a unique string.
         return True
 
-    # None of the above cases matched. This is not a basin of ds.
+    # None of the above cases matched. This is not a direct basin of ds.
     return False

pyproject.toml

@@ -28,7 +28,7 @@ classifiers = [
 license = {text = "GNU Affero General Public License v3 or later (AGPLv3+)"}
 dependencies = [
     "ckan>=2.10.4, <3",
-    "dclab>=0.65.0",
+    "dclab>=0.66.0",
     "dcor_shared>=1.0.3",
 ]
 dynamic = ["version"]