Mismatched fixed name and included name causes errors on read/validate

[rly]

Description

In HDMF currently, if a container child data type is defined with a fixed name and included as a subgroup of
some group, and the subgroup is specified with a different name, then after writing to disk, the subgroup keeps the fixed name rather than the subgroup name. The data cannot be read and validation fails.

Example:

groups:
- data_type_def: TypeA
  name: a
- data_type_def: TypeB
  groups:
  - data_type_inc: TypeA
    name: inner_a

In HDMF:

a = TypeA()  # a.name = 'a'
b = TypeB(name='b', inner_a=a)
b.inner_a  # returns a and a.name is still 'a'
with HDF5IO('test.h5', 'w') as io:
    io.write(b)  # no errors
with HDF5IO('test.h5', 'r') as io:
    read_b = io.read()  # fails, __init__: missing argument 'inner_a'
# test.h5 has subgroup named 'a'

Environment

Python Executable: Python
Python Version: Python 3.8
Operating System: Windows
HDMF Version: dev

Checklist

• Have you ensured the bug was not already reported ?
• Have you included a brief and descriptive title?
• Have you included a clear description of the problem you are trying to solve?
• Have you included a minimal code snippet that reproduces the issue you are encountering?
• Have you checked our Contributing document?

[rly]

This bug appears when users write NWB extensions and add name to both their data type definitions and included type specs, as in https://github.com/akshay-jaggi/ndx-photometry/

[alessandratrapani]

In catalystneuro/ndx-holographic-stimulation#7 we are defining new neurodata type following the best practice as in hdmf-dev/hdmf-schema-language#14
But if we get this error when trying to add a data type with a different name:

ValueError: Field 'site' on PatternedOptogeneticSeries must be named 'site'.

Schema:

- neurodata_type_def: PatternedOptogeneticStimulusSite
  neurodata_type_inc: OptogeneticStimulusSite
  doc: An extension of OptogeneticStimulusSite to include the geometrical representation
    for the stimulus.
  attributes:
  - name: effector
    dtype: text
    doc: Light-activated effector protein expressed by the targeted cell (eg. ChR2)
    required: false
- neurodata_type_def: PatternedOptogeneticSeries
  neurodata_type_inc: TimeSeries
  doc: An extension of OptogeneticSeries to include the spatial patterns for the photostimulation.
  links:
  - doc: link to the patterned stimulus site
    name: site
    target_type: PatternedOptogeneticStimulusSite

Example

        stim_site = PatternedOptogeneticStimulusSite(
            name="StimulusSite",
            device=device,
            description=site_description,
            excitation_lambda=stimulation_wavelenght,  # nm
            effector=effector,
            location=location,
        )
        self.nwbfile.add_ogen_site(stim_site)
        photostimulation = PatternedOptogeneticSeries(
            name=series_name,
            description=series_description,
            data=data,
            timestamps=timestamps,
            site=stim_site,
        )

The error occurs when using hdmf 3.11.0 and pynwb 2.5.0

[oruebel]

I assume the error is related to the definition of the link with the fixed name site that points to PatternedOptogeneticStimulusSite. As far as I know, links should be allowed to have a different name than the target they point, but I'm not sure how this needs to be handled in the code. My guess that this may be a bug in the `get_class(..)' method. @rly what do you think?

[alessandratrapani]

Ah sorry, another important piece of information: this only happens using hdmf 3.11 not 3.10

[oruebel]

Ah sorry, another important piece of information: this only happens using hdmf 3.11 not 3.10

Thanks for the additional info. Just a hunch, but my guess is that this may then be related to the bug fixes in this PR #800 that were release in 3.11. @rly thoughts?

[rly]

Yes, that's a bug introduced in #800. Sorry about that. I'll work on a fix Monday/Tuesday.