unspecified status of extra content in the JSON files

[Remi-Gau]

This relates to an issue raised on the bids-validator but that should probably be discussed here before we decide to change the specs: https://github.com/bids-standard/bids-validator/issues/1044

TL;DR

For most JSON files the validator does not allow extra content to be added. This is not mentioned in the specs and probably should be.

---

Problem

For example, the validator will throw an error if this json file for an iEEG file.

{
    "Instructions": "do this",
    "PowerLineFrequency": 50,
    "SamplingFrequency": 2400,
    "SoftwareFilters": "n/a",
    "TaskName": "implanted Target Practice",
    "iEEGReference": "Cz",
    "extraInfo": {
        "nestedExtraInfo": "something extra"
    }
}

But the specs does not mention that users cannot do that.

From the validator side, it allows for stronger validation (see here).

From the community and user side, this is can seen as a way to try to incite users to ask for their extraInfo to be supported by the specs.

Suggestion

• makes this rule and its motivations explicit in the specs.
• change the error message on the validator to point users to open issues to have extra content supported by the specs

Related issue on the validator

https://github.com/bids-standard/bids-validator/issues/1074

[satra]

i don't think the spec should restrict the set of properties in the metadata file. for example, over the years heudiconv has allowed putting in more metadata into the json files than supported by bids. in general the validator could operate in strict (restricted to schema) or non strict mode (validates only the properties in the schema).

since each dataset carries a bids version, one can always validate with respect to that version of the schema, rather than the current state of the spec. this could also be a choice the validator makes (i.e. to validate against a specific version of the schema).

allowing schema + schemaless is more general and perhaps should not be discouraged (see: https://www.martinfowler.com/articles/schemaless/)

[sappelhoff]

I like the restriction of metadata files for the reasons I outlined in https://github.com/bids-standard/bids-validator/issues/1044 (note: I didn't read the slides linked by Satra yet)

So I would be reluctant to allow any amount of metadata for Ephys ... however given that additional arbitrary metadata was allowed for MRI for a long time now, I would be fine to keep it that way.

In any case, documenting this explicitly would be a welcome addition to the spec. (after more discussion has happened in this issue)

since each dataset carries a bids version, one can always validate with respect to that version of the schema, rather than the current state of the spec. this could also be a choice the validator makes (i.e. to validate against a specific version of the schema).

yes, this is an often requested feature that would indeed be sensible to have

[Remi-Gau]

Just realized that the followings are in the specs when it comes to the data dictionary for TSV files. I wonder why this is not partly the reason why I assumed this applied to all JSON files

If a data dictionary is provided, it MAY contain one or more fields describing the columns found in the TSV file (in addition to any other metadata one wishes to include that describe the file as a whole).

Bold is mine.

https://bids-specification.readthedocs.io/en/stable/02-common-principles.html#tabular-files

See also:

Additional metadata may be included as in any TSV file to specify, for example, the units of the recorded time series.
https://bids-specification.readthedocs.io/en/stable/04-modality-specific-files/06-physiological-and-other-continuous-recordings.html#physiological-and-other-continuous-recordings

[neuromechanist]

To throw my support in relaxing this and provide an example: I am finishing the curation of the Child Mind HBN dataset, and for the movie-watching tasks I made the following description:

{
  "InstitutionName": "Child Mind Institute",
  "InstitutionAddress": "101 E 56th St, New York, NY 10022",
  "Manufacturer": "Magtism EGI",
  "ManufacturersModelName":"128-channel GSN 200 v.2.1",
  "TaskDescription": {
    "GeneralDescription": "Participants viewed four short, age-appropriate video clips taken from television and movies. (Prior to this task, parents were given the opportunity to review the full list of clips and exclude any video clips they deemed unsuitable for their children; no parents had any objections to the clips.).",
    "VideoDescription":{
      "Rating": "No parental guideline rating",
      "Description": "Mother gives her son a poppy as a present, but the poppy has a missing limb.",
      "Rationale": "The clip is included to probe for interpersonal empathy functioning.",
      "Link": "https://youtu.be/3XA0bB79oGc",
      "Length": "3:23"
    }
},
  "Instructions": "Just maintain fixation on the central spot at all times. Press to begin. First, we have to measure the position of your eyes. Just follow the circle with your eyes."
}

This fails because additional tags are not allowed, and everything under TaskDescription is additional.

I understand that not allowing additional properties helps with more stringent validation and prevents people from making typos or making mistakes with the key names.

A happy medium could probably be enforcing the explicit properties on the first level but relaxing the additional properties on the second level and so on. Also, having an ExtraInfo key can be a good choice.

[effigies]

This fails because TaskDescription is defined to be a string, not because extra fields are not permitted. Extra fields have been permitted for MRI data the entire time.

[neuromechanist]

Very true, I thought since the same exception is thrown, it might be the same condition.

FWIW, opening the additional properties or allowing key-value objects where there is unregulated string would resolve this issue.

For my datasets, I just slap all the info together in a single string to become compliant, but I think these additional info are worhty especially when there are thousands of subjects.

[VisLab]

Maybe a warning rather than an error can be issued if extra fields are included. This would alert users if they made a typo, but still allow additional information to be included.

[effigies]

Just a note that the schema validator treats all sidecars uniformly, as we do not construct one omnibus JSON schema, but a collection of partial schemas.

An additional check could be devised to look for unvalidated sidecar fields, and I look forward to the contribution. :-)

As to this, I believe we could extend the unspecified data text to include unspecified metadata:

https://bids-specification.readthedocs.io/en/stable/common-principles.html#unspecified-data

Non-standard files and directories should be named with care. Future BIDS efforts may standardize new entities and suffixes, changing the meaning of filenames and setting requirements on their contents or metadata. Validation and parsing tools MAY treat the presence of non-standard files and directories as an error, so consult the details of these tools for mechanisms to suppress warnings or provide interpretations of your filenames.

We could just as easily write:

Non-standard metadata should be named with care. Future BIDS efforts may standardize new metadata fields, changing the meaning or setting constraints on their values.

[effigies]

I believe @yarikoptic had a proposal to reserve the X- prefix to permit nonstandard fields to be used with no possibility of clashing with future standardization. You could also imagine a DICOM- prefix that dcm2niix and friends could use for fields that are not yet standardized.

[yarikoptic]

+1 on X- prefix. Here is an issue for that

• Custom prefix (X-) for arbitrary metadata in .json files bids-2-devel#80

Please chime in! there

Could be x-dicom- prefix indeed ;-)