Downgrade absence of sub- folders to warning

[yarikoptic]

I could be wrong but I think there is no strict and agreed upon requirement to require having sub- subfolders in BIDS specification. Although odd, thus warranting a warning, I think it is not strictly forbidden, and e.g. a BIDS dataset only with stimuli/ or derivatives/ or sourcedata might still be well legit BIDS dataset.

Check was added in b9ea963 likely just copying from prior node version of bids-validator. Asking @effigies for setting me straight ;)

[effigies]

The legacy validator errored on the absence of subject directories, so this was written for consistency with that.

I don't see the use-case for subject-less datasets. Calling a directory that contains opaque sourcedata/ and derivatives/ directories a BIDS dataset seems counter to the goal of BIDS, which is legible neuroimaging data. What's your goal?

With my OpenNeuro hat on, removing basic "are there any data?" checks, combined with mechanisms to shield data from validation, is just asking for turn into an arbitrary data hosting service. So this is just a note that, if this is demoted to a warning, we will need to promote it back to an error, and should probably go ahead and promote it just in case. cc @nellh

[yarikoptic]

I don't see the use-case for subject-less datasets. Calling a directory that contains opaque sourcedata/ and derivatives/ directories a BIDS dataset seems counter to the goal of BIDS, which is legible neuroimaging data. What's your goal?

immediate, although not mandatory if we do add dedicated DatasetType:

• Add DatasetType="project" and rework existing "layout" example into a proper BIDS dataset #1861
  and inline with that -- it could be neuro*-related (BIDS is beyond MRI for a while already) dataset of stimuli/, or purely phenotype/ data and IMHO could still be quite legit without "neuroimaging" or per subject data. Also group level results in derivatives etc.

But also it is a matter of consistency: if there is such a rule it should be clearly described also in the specification, which I think is not the case ATM.

Indeed it might require openneuro specific configuration, but I thought you already have one anyways.