feat: implement ValidatorValuer interface feature

[thommeo]

Problem

In order to have custom type validations for generic types, every possible type has to be registered with RegisterCustomTypeFunc to return the underlying type. It is especially relevant for nullable types e.g. sql.Null[T].

Discussed here: #1232

Fixes Or Enhances

• Adds a mechanism for validating generic types without using RegisterCustomTypeFunc by introducing ValidatorValuer interface.

Example:

type Nullable[T any] struct {
	Data T
}

func (t Nullable[T]) ValidatorValue() any {
	return t.Data
}

type Config struct {
	Name string `validate:"required"`
}

type Record struct {
	Config Nullable[Config] `validate:"required"`
}

func TestValidatorValuerInterface2(t *testing.T) {
	s := Record{
		Config: Nullable[Config]{},
	}
	v := New()
	errs := v.Struct(s)
	t.Fatalf("Error: %v", errs.Error())
}

// Key: 'Record.Config.Name' Error:Field validation for 'Name' failed on the 'required' tag
// Notice: no Data in the Record.Config.Name

Make sure that you've checked the boxes below before you submit PR:

• Tests exist or have been written that cover this particular change.

@go-playground/validator-maintainers

[coveralls]

coverage: 73.661% (+0.006%) from 73.655%
when pulling 62dc169 on thommeo:feat/validation-valuer
into 72c4503 on go-playground:master.

[zemzale]

The good looks good.

My only concerns would maybe be the naming of the interface and it's function.

[zemzale]

I think we can drop the Validator, Outside this functions it's validator.ValidatorValuer which stutters.

validator.Valuer should be fine

[thommeo]

The function name should probably stay ValidatorValue to not conflict with the sql.Valuer interface. Consequently, I'd suggest that validator valuer interface should stay ValidatorValuer to match the function name it expects, but I renamed it to Valuer anyway. The positive here is that we have validator.Valuer now which looks cleaner. 👍