snakebids/resources/bids_tags.json update

[LouFel]

Hello,

First of all, I wanted to thank you for developing your snakebids tool. I am a post-graduate intern trying to build a PET/MR pipeline and it has been really helpful.

After running:

config.update(
    snakebids.generate_inputs(
        bids_dir=config["bids_dir"],
        pybids_inputs=config["pybids_inputs"],
        derivatives=config["derivatives"],
        participant_label=config["participant_label"],
        exclude_participant_label=config["exclude_participant_label"]
    )
)

In my dataset, the PET images have a "trc" label (to specify the tracer used)
I noticed that the "trc" wildcard was missing in the updated config, in input_wildcards.

As I understand the way the inputs generation works is:

• the generate_inputs() function generate the snakemake inputs,
• _get_lists_from_bids() called inside populate input_path, input_lists, input_zip_lists, and input_wildcards,
• then along the process _process_layout_wildcard() is called which calls read_bids_tags()

The read_bids_tags() function checks compatibility for pybids as BIDS entities label differs from their key tags (e.g. Tracer -> trc), and I noticed that it is always using the snakebids/resources/bids_tags.json within _process_layout_wildcard() (function call is empty, and bids_tags.json is the default file).

Finally, inside the bids_tags.json file the pair (label/key) "tracer": "trc" is missing, so I figured my issue might come from here. I might be wrong but adding the missing pair "tracer":"trc" inside the bids_tags.json should fix the problem.

Thanks in advance for your help 😄

[pvandyken]

This is actually a pybids issue, the library that does the path parsing. It will only read entities that it knows about, as defined in a config file (see here for the default config). Pybids allows custom config files, but as it turns out, Snakebids currently doesn't support these.

Adding basic support would be very straightforward: we can just add another parameter to generate_inputs called config and pass it transparently on to BIDSLayout. I'm going to make an issue from this. @tkkuehn , @akhanf, @kaitj, any thoughts?

[LouFel]

Never mind then, thank you @pvandyken !

[pvandyken]

@LouFel The update's been merged. You can use it by installing the latest version of snakebids directly off snakebids. You'll need to provide your own custom config file for pybids; you should just be able to base it off the default ones.

[LouFel]

Thank you @pvandyken , it's perfectly working. Just a little warning if you want to use this functionality in the future:
In the pybids_config.json file, the name of the entities specified seems to matter (case by case), let me explain with an example:

{
    "name": "tracer",
    "pattern": "[_/\\\\]+trc-([a-zA-Z0-9]+)"
}

  pet:
    filters:
      suffix: 'pet'
      extension: '.nii.gz'
      datatype: 'pet'
    wildcards:
      - subject
      - session
      - tracer
      - reconstruction
      - run

Here with the tracer entity added in the pybids_config.json file, with the wildcard entry specified in the config.yml file as tracer, it returns this error:

'NoneType' object is not subscriptable
File "PATH_TO_PROJECT/workflow/Snakefile", line 13, in <module>
File "PATH_TO_VIRTUALENV/lib/python3.9/site-packages/snakebids/core/input_generation.py", line 187, in generate_inputs
File "PATH_TO_VIRTUALENV/lib/python3.9/site-packages/snakebids/core/input_generation.py", line 478, in _get_lists_from_bids
File "PATH_TO_VIRTUALENV/lib/python3.9/site-packages/snakebids/core/input_generation.py", line 385, in _process_layout_wildcard

Whereas with an entity like reconstruction that follows exactly the same logic:

{
    "name": "reconstruction",
    "pattern": "[_/\\\\]+rec-([a-zA-Z0-9]+)"
}

With the wildcard defined as reconstruction in the config.yaml file, it works perfectly fine...
Simply replacing the name and wildcard by "trc" fix the issue for the tracer entity. So it seems to me that the problem is entity related, but I don't really understand why.
Maybe I am missing something, but I let you know just in case.

[akhanf]

Ah yes that's what we use bids_tags.json for.. I've made a quick PR to update that too #139

[pvandyken]

For future reference, the issue here is that, in your tracer example, the name of the entity: tracer, and the actual tag found on the file path: trc, are different. Snakebids has no way of knowing the two are meant to be equivalent, other than the bids_tags.json file which catalogues those equivalencies. Unfortunately, there's no easy way for users to update the bids_tags.json file directly in their installed library, which obviously isn't reproducible. We've discussed fixing this before though... @akhanf, I have some ideas, I'm going to open a new discussion item for them.

[akhanf]

Ah yes you're right @pvandyken, my memory of bids_tags.json wasn't very good!
I take back my suggestion to harmonize that element across snakebids and pybids..