> On 20 Mar 2018, at 13:24, Guillaume Smet wrote:
>
> Maybe we could have some gain by avoiding the parameter name generation in
> most cases but I'm not sure it's worth making the code less readable.
Readability is the root of all evil. I read that somewhere on the
This is a spec-level API, defined in ValidatorContext:
https://github.com/beanvalidation/beanvalidation-api/blob/master/src/main/java/javax/validation/ValidatorContext.java#L77
So unless we change the rules (there might be some leeway now when
formalizing compat rules for Jakarta EE), we cannot
Let's hypotethically assume that some scenario needs this. I'm thinking of
an application which as different compilation units, so some module might
have been compiled with different conventions and need different name
resolvers.
Would you enforce such a use case to use different, independently
Can't you a priori apply your optimization and have a if ( perInstance
!= perFactory ) backdoor? Meaning I don't really get why you cannot
apply your optimization by default and fall back to a slow codepath when
the value is overridden.
Emmanuel
On Mon 18-03-19 17:47, Guillaume Smet wrote:
>Hi,
Hi,
So I know we like to have API compatibility discussions these days, so
let's start a new one.
FWIW, this discussions doesn't come out of the blue, it's based on
discussions we had for one of Marko's PRs.
In HV, you can set the parameter name provider at the VF level and you can
also