Allow diagnostic severity to be configured for rule groups

[tschaub]

I was surprised to see "correctness" diagnostics were emitted with "warning" severity and was hoping to find an easy way to configure all correctness rules to run at the "error" level instead.

I see that the CLI can be run with --error-on-warnings to treat all "warning" diagnostics as errors. It could be nice to have more control over which groups were treated as errors.

For example, I was hoping I could do something like this:

{
  "linter": {
    "enabled": true,
    "rules": {
      "correctness": {
        "all": "error"
      },
      "nursery": {
        "all": true
      }
    }
  }
}

Would it be possible for the linter.rules.[group].all to be configured with a RulePlainConfiguration in addition to a boolean?

[Conaclos]

I was surprised to see "correctness" diagnostics were emitted with "warning" severity and was hoping to find an easy way to configure all correctness rules to run at the "error" level instead.

All recommended rules emit errors. Only noUndeclaredVariables and useHookAtTopLevel are not recommended and in the correctness group. Are you referring to these rules?

We are planning to change this by using error for correctness and warning for other groups. We could also allow configuring the severity at a group level. It is still under discussion.

[tschaub]

Thanks for the detail, @Conaclos. I didn't look through all the rules to see which were warning and which were error. I noticed that both undeclared and unused variables were only generating warnings.

Does the documentation show the diagnostic level for these rules? I may be missing it, but it wasn't obvious to me how to determine this.

[Conaclos]

On the documentation, the default severity of a rule corresponds to the severity of the diagnostics emitted in the examples.