elements: urls.value should not be validated as URI

[jacquerie]

We currently have

inspire-schemas/inspire_schemas/records/elements/url.yml

Lines 10 to 14 in afcc205

	value:
	description: |-
	:MARC: ``8564_u``
	format: uri
	type: string

but this is too strict, since we use this field to store information like

<datafield tag="856" ind1="4" ind2=" ">
  <subfield code="u">@particleist</subfield>
  <subfield code="y">TWITTER</subfield>
</datafield>

which would become

{
    "urls": [
        {
            "description": "TWITTER",
            "value": "@particleist"
        }
    ]
}

but @particleist isn't a URI.

[jacquerie]

Thing is, it makes a lot of sense to validate URLs as URIs, while it makes less sense to store there what are essentially external system identifiers. So my proposal is to relax temporarily this field to unblock inspirehep/inspire-next#2992 and inspirehep/inspire-next#2904 ASAP, and then restrict again when we have another place in the schema to put this information.

[michamos]

Not sure I understand. The content of those fields are not URLs but external system identifiers, so they should be moved there. Note that this is already foreseen in commit e5a1ff4 (part of #182), that adds twitter and linkedin to the allowed author ids. In the meantime, those authors will fail migration. What's wrong with that?

[jacquerie]

In the meantime, those authors will fail migration. What's wrong with that?

One of the authors that fails migration is part of our demo set of records, which means we can't bump the schemas (which is happening in commit inspirehep/inspire-next@2afcbb0 of inspirehep/inspire-next#2992), which is blocking the manual merger from being merged.

[michamos]

can't you patch the demo records then? It seems cleaner to me to do a temporary patch at the point of failure rather than here.

[jacquerie]

can't you patch the demo records then? It seems cleaner to me to do a temporary patch at the point of failure rather than here.

👍

Will do like you suggest, and thanks for the weekend answer...