[bug] `BIDSFile.get_associations()` returns wrong files for some entities

[MaximeBICMTL]

Greetings,

I think I found a bug with the BIDSFile.get_associations() method, see the reproduction method below.

Version

Python 3.11
PyBIDS 0.22.0 (also reproduces on PyBIDS 0.17.0)

Reproduction

I have the following BIDS dataset:

/bids/
└── sub-SUB/
    └── ses-SES/
        ├── func/
        │   ├── sub-SUB_ses-SES_task-rest_echo-1_bold.nii.gz
        │   ├── sub-SUB_ses-SES_task-rest_echo-1_bold.json
        │   ├── sub-SUB_ses-SES_task-rest_echo-1_part-phase_bold.nii.gz
        │   └── sub-SUB_ses-SES_task-rest_echo-1_part-phase_bold.json
        │
        └── anat/
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_inv-1_MP2RAGE.nii.gz
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_inv-1_MP2RAGE.json
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-1_MP2RAGE.nii.gz
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-1_MP2RAGE.json
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_inv-2_MP2RAGE.nii.gz
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_inv-2_MP2RAGE.json
            ├── sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-2_MP2RAGE.nii.gz
            └── sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-2_MP2RAGE.json

And the following short script to get the associated files of each NIfTI file:

from bids import BIDSLayout
import sys

layout = BIDSLayout(sys.argv[1])
for f in layout.get(extension=["nii", "nii.gz"]):
    print(f.path)
    for assoc in f.get_associations():
        print("  ", assoc.path)

This outputs the following:

/bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_bold.nii.gz
   /bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_bold.json
   /bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_part-phase_bold.nii.gz
/bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_part-phase_bold.nii.gz
   /bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_bold.json
   /bids/sub-SUB/ses-SES/func/sub-SUB_ses-SES_task-rest_echo-1_bold.nii.gz
/bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_inv-1_MP2RAGE.nii.gz
   /bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_inv-1_MP2RAGE.json
   /bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-1_MP2RAGE.nii.gz
/bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_inv-2_MP2RAGE.nii.gz
   /bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_inv-2_MP2RAGE.json
   /bids/sub-SUB/ses-SES/anat/sub-SUB_ses-SES_acq-05mm_rec-deface_run-2_inv-2_MP2RAGE.nii.gz

As we can see, the JSON sidecars returned for acquisitions with the part and run entities are not the correct ones.

EDIT: Actually this issue might be a duplicate of #530, though I guess this issue may be a more direct description?

[effigies]

Yes, this probably is #530. However, it's probably worth noting that strictly speaking, dropping run-1 and part-mag is your issue, so your dataset is falling foul of

4. There MUST NOT be multiple metadata files applicable to a data file at one level of the directory hierarchy.

from Common principles - The Inheritance Principle - Rules.

The validator has specifically chosen to implement exact matching because situations like this are very difficult to avoid in the derivatives case, but perhaps there should be validation errors instead.

[effigies]

xref bids-standard/bids-specification#1583

[MaximeBICMTL]

Oh interesting, thank you for your answer, I will communicate it to the owners of the BIDS dataset.

However, this behavior and validation absence does seem like a footgun to me. If possible, I would still advocate to update BIDSFile.get_associations() behavior and raise validation errors for the non-derivative case.