Create CLI

[effigies]

I would recommend a typer CLI

[effigies]

@ubdbra001 This might be a good gradual start.

You could consider pulling the file tree implementation in https://github.com/bids-standard/python-validator/pull/2/files#diff-46eb2e58d73ce99806349f810eca529c3cf4dd2b23ab74dc1dc410f03462e3fd out and iterating on that to identify good/bad files.

https://github.com/bids-standard/bids-examples/ should have a bunch of valid filenames, and there are some bad datasets that we'd expect to fail in https://github.com/bids-standard/bids-error-examples.

I haven't had time to do much other curation since the last time we spoke, and will be traveling the rest of the week, so I wanted to ping you before I go even more silent. :-)

[ubdbra001]

Great, I'll have a look into it tomorrow and Monday. I'll be on leave for a week myself after Mon, so probably won't make any decent headway until mid-Feb.

[ubdbra001]

The CLI should be able to handle single files as well as directories, right?

So, if I'm using the examples and pass something like:

python -m bids_validator ./ds000117/

It should walk the whole directory to ensure the filenames follow the bids spec

And if I pass:

python -m bids_validator ./ds000117/sub-01/ses-meg/sub-01_ses-meg_scans.tsv

It should just check that one file?

[effigies]

I would not prioritize handling single files, but I don't see a problem with it.

I'll note that once we go beyond filename validation, validating individual files will probably look like:

1. Find the dataset root and construct the file tree
2. Populate the dataset context
3. Disable dataset-level checks
4. Only visit the listed files

This is because pretty much the only thing in BIDS that can be validated context-free is filenames.

Not sure if this affects your design, but just looking a bit ahead.

[ubdbra001]

De-prioritising individual files would actually make things a bit easier. So, I'll focus on parsing directories.

I'll note that once we go beyond filename validation, validating individual files will probably look like:

1. Find the dataset root and construct the file tree
2. Populate the dataset context
3. Disable dataset-level checks
4. Only visit the listed files

I'll bear it in mind. I don't think it'll affect the design as it stands, and if it does I'm sure you'll spot it when it comes to reviewing the PR.