Sidecar fields should specify their data type (e.g. float or int) if not a string

[yarikoptic]

My use case was to look at the MagneticFieldStrength field. That one is a RECOMMENDED one in https://bids-specification.readthedocs.io/en/stable/04-modality-specific-files/01-magnetic-resonance-imaging-data.html#scanner-hardware and "Corresponds to DICOM Tag 0018,0087 Magnetic Field Strength". So it seems (not spelled out) that it should be a numeric value.

Absence of clarity results in cases where either people specify non numeric value (e.g. 3T for MagneticFieldStrength) where numeric value is expected:

$> datalad -f '{parentds}' -c datalad.search.index-egrep-documenttype=all search bids.MagneticFieldStrength:3T | uniq 
/home/yoh/datalad/openfmri/ds000217
/home/yoh/datalad/openneuro/ds001525

or have it as a string "3" in the json (don't have an immediate example in the wild but the result of listing known values via datalad suggests that:

$> datalad search --show-keys full | grep -A2 'Field.*Strength' 
bids.FieldStrength
 in  2 datasets
 has 1 unique values: 3.0
--
bids.MagneticFieldStrength
 in  104 datasets
 has 8 unique values: 1.5, 2.8936200141907, 3, 1.494, 6.98, 7, '3T', '3'
--
dicom.MagneticFieldStrength
 in  19 datasets
 has 1 unique values: 3.0
--
dicom.Series.MagneticFieldStrength
 in  15 datasets
 has 1 unique values: 3.0
[WARNING] In this search mode, the reported list of metadata keys may be incomplete 

NB ignore those non-compliant hits using FieldStrength ;-)

So I think that it would be valuable to enhance all lists of Fields with information about the expected data types, and make bids-validator verify correct specification (the issue for the MagneticFieldStrenth is https://github.com/bids-standard/bids-validator/issues/851).

[sappelhoff]

I agree that being explicit about the expected data type for a given JSON field is important.

The validator does currently already take care of this for some fields using JSON schemas ... see this example: https://github.com/bids-standard/bids-validator/blob/0f55b8f4b2f6e608a8a3cb0aee93fa732e898307/bids-validator/validators/json/schemas/eeg.json#L18

it'd be good to consistently advertise the expected data type in the spec and validate it.

[tsalo]

This is related to #533 (adding type information to the documentation and schema) and #221 (adding type information to the sidecars directly).