Use traits for validation rules

[tqwewe]

When trying to derive Validate on a struct, it seems like the derive macro filters what types you can use manually.

Wouldn't it make more sense to instead have a trait for each validation rule, allowing users to implement validation rules for their custom types?

For example, this does not work:

#[derive(Validate)]
struct Data {
  #[validate(length(min = 5, message = "must have 5+ characters"))]
  name: Mutable<String>,
}

But I'd like to be able to do something like this to resolve it:

impl validator::rules::ValidateLength for Mutable<String> {
  fn validate_length(&self, min: Option<u64>, max: Option<u64>) -> Result<(), ValidationError> {
    // ...
  }
}

[Keats]

How would you the trait look like? I don't think it would work without variadic generics

[tqwewe]

I'm not sure why generics may be needed.
I was imagining something like:

trait ValidateLength {
  fn validate_length(&self, min: Option<u64>, max: Option<u64>) -> bool;
}

trait ValidateEmail {
  fn validate_email(&self) -> bool;
}

trait ValidateUrl {
  fn validate_url(&self) -> bool;
}

trait ValidateContains {
  fn validate_contains(&self, s: &str) -> bool;
}

...

And these traits would be implemented for some standard types where appropriate, eg impl ValidateEmail for String, etc.

[Keats]

Yeah my bad I was thinking the wrong thing yesterday. This would be nice but how do you envision it working in the proc macro? You still need to search for patterns in the fields types to know whether they are supported or try to di it anyway and get compilation errors in invisible generated code which isn't great.

[tqwewe]

I don't think you need to do any kind of checking of the types in the proc macro, just let Rust tell the user "ValidateEmail is not supported for type Foo". Proc macro would just expand to <Foo as validator:: ValidateEmail >::validate_email(), and Rust would error if it's not implemented for the type.

[Keats]

If the error message is not too bad it could work. It's a pretty huge change though so it would need to come as a PR so it can be tested properly.

[pintariching]

I have opened #225 to see if I'm going in the right direction. For now I've just implemented if for length`