Named @RestForm and error mapping of the violations

[jmini]

The @RestForm annotation is accepting a value which is the name of the parameter.

So it is valid to name the the parameter:

@POST
public void procced(@RestForm("u") @NotBlank String userName) {
    // do something with the form parameters
}

But in this case the the validation is executed (hasError() returns true) but the error is placed on the "userName" key. I would have expected to have them on "u".

Is that expected?

The returned violations here do not seems to contains this information:

quarkus-renarde/runtime/src/main/java/io/quarkiverse/renarde/util/MyValidationInterceptor.java

Lines 34 to 35 in 7533edc

	Set<ConstraintViolation<Object>> violations = executableValidator.validateParameters(ctx.getTarget(),
	ctx.getMethod(), ctx.getParameters());

So the mapping to the lastNode key seems difficult to change:

quarkus-renarde/runtime/src/main/java/io/quarkiverse/renarde/util/Validation.java

Lines 101 to 110 in 7533edc

	public void addErrors(Set<ConstraintViolation<Object>> violations) {
	for (ConstraintViolation<Object> violation : violations) {
	Iterator<Node> iterator = violation.getPropertyPath().iterator();
	String lastNode = null;
	while (iterator.hasNext()) {
	lastNode = iterator.next().getName();
	}
	addError(lastNode, violation.getMessage());
	}
	}

[FroMage]

Ah yes, good point. We need to add logic to find the original form name. If you're up for making a PR, I can help you.

[jmini]

I just implemented the unit test in a10cd25, but I would need some pointer on "logic to find the original form name" because right now I have no clue how it is supposed to be done.

[FroMage]

Unfortunately, we don't have any logic in Quarkus REST to reverse-map parameters back to their HTTP form name (or query, or header, or cookie).

It should be doable for simple cases, though, in MyValidationInterceptor, because when you get a validation error on method parameters, the constraint violation's property path is {<method name>, <parameter name>} so you can ignore the first element, only keep the last member (as we do now) and then iterate the method parameters to find the parameter with the right name using ctx.getMethod().getParameters().

Once you get the right parameter, you can load its annotations and if you see a @RestForm or a @FormParam annotation with a non-null value, you can use that for the error key.

This won't work for bean params, though, but we can add support for them later, let's not over-engineer this for now.

[jmini]

This won't work for bean params, though, but we can add support for them later, let's not over-engineer this for now.

This is an other potential issue, but right now if you use FormData class approach (which every field annotated with @RestForm as suggested in this section), according to my tests there is no validation.

[FroMage]

Are you sure? Did you add @Valid to the endpoint parameter?

[jmini]

Thank you for the pointer, I was not aware of the @Valid annotation.

So for the second example of the page https://docs.quarkiverse.io/quarkus-renarde/dev/advanced.html#_the_endpoint (with the FormData). This applies the validations:

@Consumes(MediaType.MULTIPART_FORM_DATA)
@POST
public void complete(@RestQuery String confirmationCode,
        @Valid FormData form) {
    // do something with the form parameters
}

public static class FormData {
        @RestForm @NotBlank @Length(max = Util.VARCHAR_SIZE) String userName;
        @RestForm @NotBlank @Length(min = 8, max = Util.VARCHAR_SIZE) String password;
        @RestForm @NotBlank @Length(max = Util.VARCHAR_SIZE) String password2;
        @RestForm @NotBlank @Length(max = Util.VARCHAR_SIZE) String firstName;
        @RestForm @NotBlank @Length(max = Util.VARCHAR_SIZE) String lastName;
}

---

With the same limitation as the one discussed in this issue:

public static class FormData {
        @RestForm("foo") @NotBlank @Length(max = Util.VARCHAR_SIZE) String userName;
        //...
}

The validation works, but is set on errors map with the key "userName" and not "foo".

[FroMage]

OK, so same problem, I suspected.