Mismatch in allowed chars for validator.id in definition v/s suppress trait

[gosar]

It is possible to define a validator with the id containing the dot . character.

metadata validators = [
    {
        name: "CamelCase",
        id:"Camel.Case"
    },

However, trying to add a suppress trait for this event ID, like

@suppress(["Camel.Case"])

leads to

Error validating trait `suppress`.0: String value provided for `smithy.api#suppress$member` must match regular expression: ^[_a-zA-Z][A-Za-z0-9]*$

[mtdowling]

Good catch.

Given that validator suppressions are so restrictive, I think it's fair to clean up the spec and now constrain validator IDs to match suppressions, but also expand what suppressions support match what's been used in practice inside Amazon. I think we should update the suppression trait to allow "."; it should be the same ABNF Smithy uses for shape ID namespaces:

^[_a-zA-Z][A-Za-z0-9_]*(\.[_a-zA-Z][A-Za-z0-9_]*)*$

Then add validation when loading validators to ensure they comply with this regex (see https://github.com/awslabs/smithy/blob/main/smithy-model/src/main/java/software/amazon/smithy/model/shapes/ShapeId.java#L87). We could repeat the ABNF in the validation spec to not overly couple the concepts.

[mtdowling]

Actually, I think the more backward compatible change with less potential impact on existing users is to remove the constraints on the @suppress trait rather than constrain validation event IDs. You can define a suppression today using the suppressions metadata key, and that doesn't have any constraints on it. That plus the ID of a validation event make me think the @suppress trait is the outlier and should be updated instead to something more permissive like a non-empty string.