Regression in renovate regex comprehension

[Callek]

Similar to #516 and with a similar workaround I can't update to the newest (0.31.0) check-jsonschema via pre-commit due to this regression in renovate.

Example renovate.json

{
  "$schema": "https://docs.renovatebot.com/renovate-schema.json",
  "customManagers": [
    {
      "description": "Update rust toolchain declarations of dtolnay/rust-toolchain",
      "customType": "regex",
      "fileMatch": [
        "\\.github/(?:workflows|actions)/.+\\.ya?ml$",
        "(^|/)action\\.ya?ml$"
      ],
      "matchStrings": [
          "# renovate: rust-toolchain\\s*toolchain\\:\\s*(?<currentValue>\\d+\\.\\d+(\\.\\d+)?)"
      ],
      "depNameTemplate": "rust",
      "packageNameTemplate": "rust-lang/rust",
      "datasourceTemplate": "github-tags"
    }
  ]
}

Verification that its fixed with nonunicode

$ check-jsonschema --version
check-jsonschema, version 0.31.0
$ check-jsonschema --builtin-schema renovate renovate.json 
Schema validation errors were encountered.
  renovate.json::$.customManagers[0].matchStrings[0]: '# renovate: rust-toolchain\\s*toolchain\\:\\s*(?<currentValue>\\d+\\.\\d+(\\.\\d+)?)' is not a 'regex'
$ check-jsonschema --builtin-schema renovate --regex-variant nonunicode renovate.json 
ok -- validation done

Environment Details

• MacOS Sequoia
• Bash + Python3.12

(Though I don't think the environment matters in this case)

[sirosen]

Thanks for filling this!

I'm starting to regret not doing more research before choosing Unicode mode as the default. With these issues cropping up, I'm doubting that decision a bit.

To keep churn in the project to a minimum, I'll most likely mark the renovate hook as nonunicode in the next release (possibly today, if I can get some time).

I want to look at why that pattern fails though -- it's not obvious to me why the Unicode mode regex engine fails it. Maybe something about the capture group?

[sirosen]

I just checked using my browser and the issue with that regex is that there's an improper escape, \:.
Nonunicode mode will, indeed, accept that expression because it is less strict.

This might be evidence of the 0.31.0 update catching a (trivial) mistake in your usage? But if the regex works in renovate then they're probably not evaluating regexes in Unicode mode. So I think that (since renovate is implemented in JS), I'll proceed with the change in hook config, to try to match their implementation.

[Callek]

the issue with that regex is that there's an improper escape, :

Good catch, but i agree with your planned outcome anyway.