RFR: 8262889: Compiler implementation for Record Patterns

[Maurizio Cimadamore]

On Fri, 6 May 2022 10:51:33 GMT, Jan Lahoda <jlahoda at openjdk.org> wrote:

>> I now believe that the check is needed to properly classify patterns based on the type of the i-th component. That said, not sure this should be a subtyping check, or a type equality
>
> A good question. Consider code like:
> 
>     private void test(R r) {
>         switch (r) {
>             case R(A a, A v) -> {}
>             case R(B b, A v) -> {}
>             case R(I i, B v) -> {}
>         }
>     }
>     final class A implements I {}
>     sealed interface I permits A, B {}
>     final class B implements I {}
>     record R(I i1, I i2) {}
> 
> 
> The switch is exhaustive - all the possible combinations are covered. When checking the first component, the code will categorize the patterns like:
> 
> A -> R(A a, A v), R(I i, B v)
> B -> R(B b, A v), R(I i, B v)
> I   -> R(I i, B v)
> 
> this categorization is done using the subtype check, so that `R(I i, B v)` will appear in the list for `A`. When checking the second component, the possibility for `I` is not exhaustive (does not cover `A` in the second component), but the possibilities for `A` and `B` are exhaustive, and they together cover `I`.

Ah - makes sense of course - I "just" needed a more convoluted example :-)

-------------

PR: https://git.openjdk.java.net/jdk/pull/8516