Make project_role ontology field an array

[ESapenaVentura]

For which schema is a change/update being suggested?
module/contact

What should the change/update be?

• Change project_role field in contact module from object to array of objects.
• Change project_role property name to project_roles to adhere to the style guide.

What new field(s) need to be changed/added?

For contact module:

• project_role wanted type: array (Accepts more than 1 value)
• project_role property name change: project_roles

Why is the change requested?

Talking with @zperova, we need the project_role ontology field to be an array of objects. For some projects, some contributors will be fulfilling more than one type of role, and our metadata schema needs to reflect that.

[ESapenaVentura]

Thinking about it, it might not be a major change. You're turning it into an array, so there is backwards compatibility, right?

[lauraclarke]

It would be good to understand the specific use case where one contributor has more than one role

As the field is changing type I think that defacto means this is a major change

[malloryfreeberg]

It is a major change because the JSON will go from:

  "project_role": {
        "text": "blah",
        "ontology": "blah:0000",
        "ontology_label": "blah"
    },

to

  "project_roles": [
      {
        "text": "blah",
        "ontology": "blah:0000",
        "ontology_label": "blah"
      }
    ],

which is not backwards compatible. Note the addition of square brackets to indicate a JSON array.

EDIT: Also, if we make the field an array, we should make the field name plural.

[diekhans]

Shouldn't this field be named roles if it is a vector?

[malloryfreeberg]

Field names are plural if they are JSON arrays. Currently, the field is an object (points to the ontology module) and therefore singular. If we do change the field to be an array of objects, then yes, according to the Style Guide, the field name should be changed to be plural, as well.

[ESapenaVentura]

Then, we leave it blocked until further notice.

(By the way, good catch with the plural, didn't think about it)

[ESapenaVentura]

@lauraclarke The specific use case is for nilsson's dataset inside the HDCA sweden pilot project, where one of the contributors asked for more than one position for the dataset (Experimental scientist and administrator). @zperova is the primary wrangler, so she probably has more info about this than I do.

[zperova]

@lauraclarke I also had correspondence with Linnarsson lab with comments that these roles do not capture in full what the people do for the project as "these days 4 out of 5 scientists are both computational and experimental"

[daniwelter]

Can I just point out that enums and ontology fields are mutually exclusive? I think based on the comments above, we just want to make this field an array of ontology terms. Could someone please update the title and primary issue description to clarify?

[samanehsan]

This rename/type change will require changes to the metadata-api library as well: https://github.com/HumanCellAtlas/metadata-api/blob/5106533b0b0216bc1b8b45b8da69d8ee09fde3bf/src/humancellatlas/data/metadata/api.py#L148