doi registrant code is too restrictive in the schema

[schwehr]

In extensions/scientific/json-schema/schema.json

"sci:doi": {
          "type": "string",
          "title": "Data DOI",
          "pattern": "^(10[.][0-9]{4,}(?:[.][0-9]+)*/(?:(?![%\"#? ])\\S)+)$"
        }, 

This is too narrow: [0-9]{4,}

https://www.doi.org/overview/DOI_article_ELIS3.pdf

a unique alphanumeric string assigned to an organization
that wishes to register DOI names (four digit numeric codes
are currently used though this is not a compulsory syntax).
The registrant code is assigned through a DOI registration
agency, and a registrant may have multiple-registrant
codes. 

https://www.doi.org/doi_handbook/2_Numbering.html#2.2.2

The registrant code is a unique string assigned to a registrant.

So my best guess at what the doi regex should be is this based on the alphanumeric statement in the pdf.

          "pattern": "^(10[.][0-9a-zA-Z]+(?:[.][0-9a-zA-Z]+)*/(?:(?![%\"#? ])\\S)+)$"

So this should be a valid doi if the prefix was registered: 10.123abc.foo.bar/issn.1476-4687/this/is/nuts

I'm not sure what the suffix part of the pattern will match: (?:(?![%\"#? ])\\S)+)

https://json-schema.org/understanding-json-schema/reference/regular_expressions.html

[m-mohr]

That seems to be a valid point. The original reg exp we used is from here: https://www.regextester.com/93795
Maybe we can also learn a bit from here, although they also just use numbers: https://www.crossref.org/blog/dois-and-matching-regular-expressions/
Would you be up to contribute a new version in a PR?

[schwehr]

I will definitely cook up a PR. I have been exploring DOIs more while working on the scientific extension in stac-utils/pystac#199. e.g. https://gist.github.com/schwehr/22ce6080eb9e730ef04fccfa25072e3a

I am wondering if anyone has valid DOIs that fail to validate?

[m-mohr]

With the CrossRef article in mind, I have relaxed the DOI regex to ^10\.[0-9a-zA-Z]{4,}/[^\s]+$ That should capture mostly all regexp out there, I think. See PR #964