BEP: make dir optional for other than fmap modalities

[yarikoptic]

ATM, "Phase encoding direction" (dir) is a required field for fmap(epi). But it is relevant for other sequences (fmap, dwi) but is not available in BIDS specification for them. To overcome this specification people resort to list the value in a flexible and omni-present acq field:

$> datalad -c datalad.search.index-egrep-documenttype=all search path:.*_acq-ap.*
search(ok): /home/yoh/datalad/openfmri/ds000221/sub-010001/ses-02/func/sub-010001_ses-02_task-rest_acq-AP_run-01_bold.json (file)
...
search(ok): /home/yoh/datalad/openfmri/ds000244/sub-01/ses-00/func/sub-01_ses-00_task-ArchiSocial_acq-ap_bold.nii.gz (file)
...

So it seems that demand is there and I do not see a reason why dir should not be allowed to be used in those cases.

[chrisgorgo]

I am a bit skeptical:

• it makes the filenames potentially longer
• creates two ways to describe the same thing (acq-PA and dir-PA)

I'm also not sure there would be such a big benefit from this change in terms of usability.

[yarikoptic]

Length stays the same since len(acq) == len(dir), unless there is also actually something else to say about acq.
acq is too generic/non-specific, dir meaning is consistent with current use in fmap, and potentially would be used for the same purpose.
To me it is about clarity - again acq can't be interpreted pretty much, while dir could

[kaczmarj]

@chrisfilo - i agree with @yarikoptic. dir should be valid for all epi, or it shouldn't be a tag at all.

it's confusing to have dir mean phase encoding direction for field map, and then acq potentially mean phase encoding direction for dwi, bold, etc. i am bids-ifying a study that uses AP and PA direction for dwi and bold, and it's inconsistent to specify the phase encoding direction with dir- for the field maps but acq- for dwi and bold, when they represent the same information.

[chrisgorgo]

Please make a pull request - we can see if anyone feels strongly against this.