Fix classification of attributes as new/overridden and issue with custom spec keys #503
Conversation
e.g. an NWB file with NWBGroupSpec, NWBDatasetSpec, etc. using 'neurodata_type_def' instead of 'data_type_def' searches for 'neurodata_type_def' in the cached hdmf-common schema which contains 'data_type_def'
rly
changed the title
Codecov Report
@@ Coverage Diff @@
## dev #503 +/- ##
==========================================
- Coverage 85.47% 84.48% -1.00%
==========================================
Files 41 38 -3
Lines 8213 7700 -513
Branches 1770 1657 -113
==========================================
- Hits 7020 6505 -515
- Misses 842 861 +19
+ Partials 351 334 -17
Continue to review full report at Codecov.
|
|
This should resolve the error in PyNWB mentioned here: #550 (comment) |
|
|
||
| def is_many(self): | ||
| return self.quantity not in (1, ZERO_OR_ONE) | ||
|
|
||
| @classmethod | ||
| def get_data_type_spec(cls, data_type_def): | ||
| def get_data_type_spec(cls, data_type_def): # unused |
There was a problem hiding this comment.
I assume 'unused' means they are never called anywhere. Are you leaving these here just in case?
There was a problem hiding this comment.
Yeah, this is a note-to-self. Reading the code, one might think it would be used in every data_type spec, but actually it is not used anywhere.
Motivation
Fix #501, address part of #502, and fix another related bug where cached HDMF schema loaded from a NWB file were not being resolved correctly.
When extracting HDMF out of PyNWB, a few keys were renamed to be more general, e.g., 'neurodata_type_def' -> 'data_type_def', and subclasses of
GroupSpec,DatasetSpec, andSpecNamespacein HDMF were created in PyNWB. These NWB-specific subclasses are used when reading the hdmf-common schema cached within a file, but the cached schema uses 'data_type_def' instead of 'neurodata_type_def'. This prevented hdmf-common specs from being resolved correctly so that attributes, datasets, groups, and links defined by a base type within hdmf-common were in some cases not inherited by a type that extends the base type.I also refactored parts of spec.py, added comments, and added additional tests to push code coverage of spec.py from 79% to 83% and namespace.py from 79% to 90%.
Checklist
flake8from the source directory.