Go. no validations during generation but not build/dep errors about protovalidate

[elliotmjackson]

Thanks for clarifying. For the performance considerations and the mechanics behind protovalidate, you can refer to our performance documentation. The current CEL-based implementation in protovalidate does indeed use reflection for validation.

Regarding your inquiry about VTProto-like unrolling to avoid reflection: it's entirely possible for someone to create a version that translates the CEL AST into concrete code. However, such an implementation is not in our immediate plans.

For now, if you're comfortable with it, you can continue using PGV. It's not being deprecated. However, we believe that protovalidate offers more advantages for most use cases.

I'll be closing this issue now. If you have any further questions or concerns, please don't hesitate to reopen or create a new issue.

[delaneyj]

I moved everything to https://github.com/bufbuild/protoc-gen-validate and changed tags/options/etc and see correct validation code generated.

if val := m.GetInputCount(); val <= 0 || val > 4096 {
		err := HyperParametersValidationError{
			field:  "InputCount",
			reason: "value must be inside range (0, 4096]",
		}
		if !all {
			return err
		}
		errors = append(errors, err)
	}

	if val := m.GetOutputCount(); val <= 0 || val > 4096 {
		err := HyperParametersValidationError{
			field:  "OutputCount",
			reason: "value must be inside range (0, 4096]",
		}
		if !all {
			return err
		}
		errors = append(errors, err)
	}
	```

[elliotmjackson]

Hi @delaneyj, i think theres some confusion here. There is no generated code with protovalidate, simply define the rules on your proto message like you've done, then pass the message through to the protovalidate.Validate method. protovalidate will handle the rest and return an error of the message does not conform to the defined rules.

I think what youve done here is define buf.validate.* options and then attempted to generate using the protoc-gen-validate plugin, which will not be able to interpret the options defined, resulting in no rules being generated for the type.

[elliotmjackson]

Hi @delaneyj, just checking in on this issue. Please let us know if you're still facing the problem, or if there's any additional information you can provide to assist us. Thanks!

[delaneyj]

I haven't ported back yet, so its using reflection at runtime for validation. Is there plans for something like VTProto were that can get unrolled to avoid the reflection?