ATM we have quite a lot of "heterogeneity" in how we do validation. We have multiple layers
pynwb
bidsschematools
nwbinspector
dandischema
dandi-cli glueing: uses above + adding more (zarr checking) - relies on content to be available, not just validation of extracted metadata
dandi-archive: invokes validations of dandischema, validates extracted metadata, not relying on content being present
We also have two "dataset layouts": DANDI and BIDS, with DANDI being our "ad-hoc" layout which is instrumented in code in dandi-cli.
NB We might move DANDI layout into dandischema
We have https://www.dandiarchive.org/handbook/135_validation/ which outlines to user only some, very limited, set of validations. We should get some more detailed description of "validation framework" and "DANDI layout" here.
edit: and then worth getting some meeting/presentation for DANDI team to sync our knowledge etc on all these aspects.
ATM we have quite a lot of "heterogeneity" in how we do validation. We have multiple layers
pynwb
bidsschematools
nwbinspector
dandischema
dandi-cli glueing: uses above + adding more (zarr checking) - relies on content to be available, not just validation of extracted metadata
dandi-archive: invokes validations of dandischema, validates extracted metadata, not relying on content being present
We also have two "dataset layouts": DANDI and BIDS, with DANDI being our "ad-hoc" layout which is instrumented in code in dandi-cli.
NB We might move DANDI layout into dandischema
We have https://www.dandiarchive.org/handbook/135_validation/ which outlines to user only some, very limited, set of validations. We should get some more detailed description of "validation framework" and "DANDI layout" here.
edit: and then worth getting some meeting/presentation for DANDI team to sync our knowledge etc on all these aspects.