Fix how build_path deals with run entity

Is there a good reason for “run” to parse as a integer (instead of a string)? It means that layout.build_path does not return something matching the original…

data_path = os.path.join(get_test_data_path(), 'ds005')
layout = BIDSLayout(data_path)
f = layout.get(suffix = 'bold')[1]
rebuilt_path = layout.build_path(f.entities)
assert f.filename == os.path.basename(rebuilt_path)

orig_filename 		sub-12_task-mixedgamblestask_run-02_bold.nii.gz
recontructed to 	sub-12_task-mixedgamblestask_run-2_bold.nii.gz

[effigies]

Might be worth entering into the record, I almost always see run-01 instead of run-1 in existing datasets, which, given my sampling pool, means a good chunk of OpenFMRI/OpenNeuro.

I suspect many people in practice use OpenNeuro as a model for their BIDS datasets, rather than reading the spec with a pedantic eye, so this practice is bound to spread. I’m not sure that we can really put our foot down here; the spec is permissive if fairly negative: (“In contrast to other labels, run and echo labels MUST be integers. Those labels MAY include zero padding, but this is NOT RECOMMENDED to maintain their uniqueness.”)

I’m not really sure that there is a good pybids solution, but if you’re confident in the naming schemes of your dataset, you can do the following programmatically:

layout = BIDSLayout(data_path)
layout.path_patterns = [path.replace('run-{run}', 'run-{run:02d}')
                        for path in layout.path_patterns]
...

This obviously isn’t going to match a dataset that doesn’t zero-pad. You could get fancy and do a regex to count the number of digits… But then you’re really running into the generality issues that we’ve avoided by refusing to accommodate zero-padding.

That will use your customized config instead of the BIDS-compliant one. I can’t guarantee that everything will work properly throughout the codebase if you start treating run_index as a string, but nothing jumps out at me as an obvious pain point.

Layouts should be fine. IIRC Analysis has the assumption that run is an int baked in pretty heavily.

[tyarkoni]

There seems to be agreement that the solution is for users to either rename their files or use one of the above workarounds, so I’m closing this.