Integrity protections matching loaded data with `paramset` - Githubissues

datajoint / element-calcium-imaging

DataJoint Element for multi-photon calcium imaging analysis with CaImAn, Suite2p, and EXTRACT

https://datajoint.com/docs

MIT License

9 stars 40 forks source link

Integrity protections matching loaded data with `paramset` #88

Open CBroz1 opened 1 year ago

CBroz1 commented 1 year ago

Current

A user could

imaging.ProcessingParamSet.insert_new_params(
    processing_method='suite2p', 
    paramset_idx=1, 
    params=my_params_dict,
    paramset_desc='my description'
)
imaging.ProcessingTask.insert1(dict(
    **scan_key,
    paramset_idx=1, 
    processing_output_dir='/my/dir'
))

If my_dir exists, we load the results regardless of whether or not the output in this directory matches paramset_idx=1.
If the paramset were wrong, this would result in a mismatch of params and analyses which is not aligned with general DataJoint integrity protection principles.
If my_dir was a previous analysis, but the user might mistakenly load instead of trigger a new analysis.

Future

Minor fix: when loading, check that every item specified in the paramset is true of the relevant directory.
Major fix: position 'load vs. trigger' upstream of ProcessingParamset and always trigger insert_new_params when loading existing results.

CBroz1 commented 1 year ago

@kabilar - I'm reviewing outstanding issues to label a subset as 'good-first-issue'. If you can propose a path forward on this, it may qualify for that label.