draft-06 Support

[khakulov]

Is it in plan to support draft-06?

in my Project i need to use propertyNames to validate object keys agains regex pattern.
but propertyNames was introduced in draft-06.

for now i am using patternProperties with "additionalProperties": False together, but the error is confusing because of "schema does not allow additional properties"

[Julian]

It's definitely planned yep, just a bit behind. Patches definitely welcome.

Branch notes:

• Finish getting all the Draft 6 tests passing
• Draft3's validate still does the True/False sentinel checking
• Fix the type checker skipped tests
• Clean up validators / meta_schemas in tests
• Validator.DEFAULT_TYPES is now only present on the class, not on instances
• Do something about FormatChecker not being able to distinguish the two time formats
• Ensure error properties are properly set for the False schema
• Ensure error properties are properly set for contains
• Upstream tests for non-strings
• if/then/else is missing tests for boolean schemas
• idn-hostname doesn't actually check it's a hostname
• Kill warnings from tests that aren't testing deprecated behavior
• Ensure tests aren't making network requests
• Check whether the (failing) ref tests should be backported to draft3
• is_any is uncovered / Draft3's type special cases any / Should it be redefinable?
• TestValidationErrorMessages + Draft6Validator
• idn-email doesn't actually allow idn
• Should there be a to_schema argument to create
• Fix messages support for boolean schemas
• Check for consolidations
• func = FormatChecker.cls_checks(draft3, raises)(func) should probably not do the assignment in jsonschema._format._checks_draft
• Consider whether there should be an easy way to evade int/bool inheritance ridiculousness
• What should validates do about $id vs id
• Provide an API for validating URI Templates python-hyper/uritemplate#33
• Run coverage
• Document that iso8601 is back-compatted only for older drafts
• What should validator_for do about False and True?
• Make sure True and False don't work in Draft4
• Check that nowhere claims that the type of a schema is dict anymore
• Provide an API for detecting invalid JSON Pointers stefankoegl/python-json-pointer#20
• validates + "$id" vs "id"
• address the default in validate
• check for more optional tests
• create + bool metaschema
• Investigate whether any tests can be backported to draft4/3 to ensure no functionality leaks backwards

[bsmithers]

I've just raised a PR (#370) addressing the currently failing tests.

To continue the conversation in the appropriate place, are there particular components of the above checklist that you'd appreciate some help with? Or indeed anything new you have realised that needs work before draft6 support is complete?

[fireundubh]

FYI: Draft 7 was recently published.

[vlad-khramov]

Hi, what tasks from #337 (comment) are still undone?

[Julian]

@quantum13 so the main one is to update the draft-06 branch with @bsmithers' type checker changes. I was hoping/planning to get a release out first just containing those, and then to get to merging them into the draft 6 stuff.

Out of curiosity for anyone following this ticket, do y'all have interest specifically in draft 6, or would jumping directly to draft 7/8 be just as acceptable?

[Julian]

I think python-hyper/uritemplate#36 is still open as well, so we'd need to help out with getting that one merged there.

[adamcharnock]

As a follower of this ticket: @Julian, I'd be happy to jump to 7/8.

[Julian]

@adamcharnock thanks.

(Want to by the way say to all those following this ticket that I wish these things would get done faster but unfortunately open source gets an exceedingly small fraction of my time these days with all of the other fun things in life... Really am hoping to get out a release with newer draft support within the next 2-3 weeks though.)

[handrews]

As a follower of this ticket: @Julian, I'd be happy to jump to 7/8.

@adamcharnock draft-07 is a pretty small incremental step that is backwards-compatible with draft-06, so it should be relatively fast to add, and near-trivial to support both. draft-08 will be more substantial, so it would be really good to at least get to draft-06 (and past draft-04) rather than wait for the even larger draft-08.

[adamchainz]

wrong adam

[diogobaeder]

My two cents: I think it's a better idea to move to draft-06 first, as this is a more natural step than jumping that version. Of course, some people might get unhappy that it's moving slower to draft-07, but it's a less surprising move, and it copes better with supporting all drafts by the end of the day.