incf-nidash / nidmresults Goto Github PK

• move testing to GHA
• update to avoid any failing or flaky test
• update packaging (setup.py + setuptools -> pyproject.toml + hatchling)
• use pytest for testing and get coverage
• enforce code style: black + flake8 + precommit
• upgrade code to use more modern python features: https://github.com/asottile/pyupgrade
• incorporate refactoring from sourcery: Remi-Gau#1

As discussed with @mih at cmaumet#2 (comment), we need to implement "how to properly inject an ID namespace in the NIDM context. But for now search&replace is good enough."

nidmresults lacks documentation. It would be great to create an examples folder containing jupyter notebooks.

The first example could be a demonstration of how to read a NIDM pack and retrieve the statistic map (example datasets are available at: https://neurovault.org/collections/1425/).

Check if whenever we create a new PR in this repo we could run the FSL's exporter tests to check that both modules are still compatible.

The python file https://github.com/incf-nidash/nidm/blob/master/nidm/nidm-results/test/TestResultDataModel.py should be moved over here as all the NIDM-Results exporter repositories now depend on the nidmresults module.

After discussion at NIDM sprint we decided to implement easy to use functions for common NIDM Results queries. This should lower the access barrier for new developer. Those functions can live in the nidmresults package. They could immediately replace some sparql queries in NeuroVault and nidmviewer (maintained by @vsoch).

get_statmaps() https://github.com/NeuroVault/NeuroVault/blob/0ba7f42a11d703ee608317fe7f0c24fd62d48fc5/neurovault/apps/statmaps/nidm_results.py#L64
as well as https://github.com/vsoch/nidmviewer/blob/882f5201354135ff4dc6ec8739621a94341d0fef/nidmviewer/sparql.py#L43
(this could be split into more functions if that makes sense)

get_coordinates()
https://github.com/vsoch/nidmviewer/blob/882f5201354135ff4dc6ec8739621a94341d0fef/nidmviewer/sparql.py#L28

nidm-results.owl (released versions) should be included in the nidmresults library so that we can look up labels and avoid dependency on nidm repository.

Let's make our project more welcoming!

First steps, adding:

• Open source license --> #47
• README
• Contributing guidelines
• Code of conduct

Some resources: Open source guide, Mozilla open leadership training: write a great project readme and README checklist.

As agreed in incf-nidash/nidm-specs#378, provn serialisation should be removed.

Reading is quite slow... is there anything we can do to have read faster. Maybe by updating the queries / checking which query if run when / relying on the JSON when available.

get_excursion_set_maps() returns a dictionary and get_statistic_maps() returns a list

(instead of nidm version in the exported document).

I would like to suggest a modification to owl_reader.py. Right now any modifications to the nidm-experiment owl or import files fails due to the string cast here. The files pass the Travis checking if the cast to string is removed. This was discussed awhile ago, but it was never changed. IIRC the reason for the cast was not clear at the time. @cmaumet is there any reason not to make this modification?

(so that we can version control differences in exported files)

Just discussed with @nicholst: when an error is thrown (e.g. no number of subjects for a group analysis), the created temporary nidm folder should be deleted before returning.