class.c: Define a :writer attribute, applicable to scalar fields only

[leonerd]

Permits automatic creation of writer accessor methods, in addition to the reader ones we already had.

---

• This set of changes requires a perldelta entry, and it is included.

[tonycoz]

It makes deparse unhappy:

$ ./perl -Ilib -MO=Deparse,Foo::set_x -Mfeature=class -e 'class Foo { field $x :writer; }'
class is experimental at -e line 1.
field is experimental at -e line 1.
use feature 'class';
{
    package Foo;
    ;
}
unexpected OP_METHSTART at lib/B/Deparse.pm line 1677.
package Foo;
sub set_x {
    XXX;
    die sprintf("Too many arguments for subroutine at %s line %d.\n", (caller)[1, 2]) unless @_ <= 1;
    die sprintf("Too few arguments for subroutine at %s line %d.\n", (caller)[1, 2]) unless @_ >= 1;
    $x = $_[0];
    return $self;
}
-e syntax OK

but that looks like a general deparse problem with methods:

$ ./perl -Ilib -MO=Deparse,Foo::set_x -Mfeature=class -e 'class Foo { field $x; method set_x { $x = $_[0]; } }'
class is experimental at -e line 1.
field is experimental at -e line 1.
method is experimental at -e line 1.
unexpected OP_METHSTART at lib/B/Deparse.pm line 1677.
use feature 'class';
{
    package Foo;
    ;
    sub set_x : method {
        XXX;
        $x = $_[0];
    }
}
-e syntax OK

You asked about identifier validation in #p5p, but:

$ ./perl -Ilib -Mfeature=class -E 'class Foo { field $x :reader :writer(a c); } my $m = "a c"; say Foo->new()->$m("abc")->x'
class is experimental at -e line 1.
field is experimental at -e line 1.
abc

but I expect :reader has the same problem.

Similarly:

$ ./perl -Ilib -Mfeature=class -E 'class Foo { field $x :writer( ); } say "<$_>" for keys %Foo::'
class is experimental at -e line 1.
field is experimental at -e line 1.
<__ANON__>
<new>

(no method defined, but no error reported)

[leonerd]

On grounds that a bunch of these issues also affect the :reader attribute as well, I think I will first go fix them in a separate PR before this, and then come back to this one after that's done.

[leonerd]

Now with a guard against invalid method names, and a test for the non-scalar error message.

The deparse issues remain, but that also applies to existing methods in general, including :reader generated ones. We should look at that later.