[Fix] order: codify invariants from docs into config schema

[Xunnamius]

Updating schema wrt #3127 (comment)

EDIT: depends on #3129 for its added tests

[ljharb]

This PR seems to have too much stuff in it - but either way, if a rule or option is already released, its schema can’t be tightened further.

[Xunnamius]

@ljharb It seems like it has a lot because it's on top of the consolidateIslands PR (rather than main so the schema changes can run against the latest tests), but the two new commits are tiny and only touch the schema and two tests. I can pull the lines around dependencies/consolidateIslands one commit up into the consolidateIslands PR if desired :)

EDIT: I can also rebase onto main and decouple from the other PRs, but not hitting all the new tests plus no direct tests of the schema makes me a bit nervous.

As for the tightening of existing options, I assume you're talking about lines 865 and 866 for pathGroupsExcludedImportTypes? I can revert those. I did add a comment to the groups schema warning that additional validation happens elsewhere, and added two newlines to the named schema for consistency, but those changes should be just cosmetic.

[ljharb]

In that case, I'll wait to review this since it needs to wait on #3129

[ljharb]

these changes are non-breaking iff they exclude things that the rule already checks.

… which would also imply there's some code that can be removed, since the checks are not guaranteed to be done by eslint?

[ljharb]

no need for the comment. but this can be verified to have numbers in it, and no duplicates, yes?

[Xunnamius]

The secondary validation has its own error messages. I believe you're saying to move the duplicates/existence validation logic up to the schema? That would change the error messages and perhaps other behavior. For instance, there are currently unit tests that check for the current error messages.

Would that count as tightening the schema of a released option?

[ljharb]

Changed error messages are fine; test cases that always error, changing to "removed test cases because it's now unreachable" is also fine, I think.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.19%. Comparing base (4f145a2) to head (30b4dcf).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3152      +/-   ##
==========================================
- Coverage   95.17%   92.19%   -2.98%     
==========================================
  Files          83       83              
  Lines        3688     3690       +2     
  Branches     1331     1331              
==========================================
- Hits         3510     3402     -108     
- Misses        178      288     +110     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.