RFR: JDK-8242478: compiler implementation for records (Second Preview)

[Tagir Valeev]

Hello!

Please clarify whether local enums and interfaces is planned to be a
preview feature or standard feature in Java 15. Also, I don't see any
version check in Check.java or JavacParser.java. Shouldn't it still display
an error if --release=14 is used? I'm not experienced in javac source code,
so please sorry if I'm missing something trivial.

With best regards,
Tagir Valeev.

чт, 30 апр. 2020 г., 22:17 Vicente Romero <vicente.romero at oracle.com>:

> Hi all,
>
> Please review the patch for the second preview of records at [1], the bug
> entry is at [2] the related spec can be found at [3], the JEP is at [4].
> This patch implements the following changes:
>
> - local enums and interfaces, along with local records
> - removed possibility of final modifier for record components
> - removed requirement that canonical constructor must be public. Any
> access modifier must provide at least as much access as the record class.
> If a canonical constructor is implicitly declared, then its access modifier
> is the same as the record class
> - check that there is a iff relation between a varargs record component
> and the corresponding parameter in the canonical constructor so:
>
> record R(int... i) { R(int... i) {...}} is allowed but:
> record R(int... i) { R(int[] i) {...}} is not
>
> - new case for using @Override annotation to declare that a method is an
> accessor method for a record component
>
> These are the main changes, plus:
> - there is one liner change to java.io.ObjectStreamClass, this is related
> to the fact that now the canonical record don't necessarily has to be
> public thus the use of Class::getDeclaredConstructor
> - there is a one liner change concerning the flags of the generated
> toString method, this is a bug fix, toString didn't have the same flags as
> hashCode and equals
> - I have added several tests to enforce other rules in the spec like the
> one saying that in a compact constructor fields have to be initialized in
> the same order in which the corresponding record component has been
> declared, plus some additional tests
> - test RecordCompilationTests is executed twice now, first in a mode that
> is equivalent to the original test and then using an "empty" annotation
> processor. This is because some regressions have been discovered in the
> past when an annotation processor is present. I have also made some changes
> to the libraries this test uses.
>
> Thanks,
> Vicente
>
> [1] http://cr.openjdk.java.net/~vromero/8242478/webrev.00
> [2] https://bugs.openjdk.java.net/browse/JDK-8242478
> [3]
> http://cr.openjdk.java.net/~gbierman/records2/20200428/specs/records-jls.html
> [4] https://bugs.openjdk.java.net/browse/JDK-8242303
>