It works like this (tested in 5.8.5):

#!/usr/bin/perl -w
BEGIN {
    print "Compile time\n";
}

package Foo;

use fields qw(some_attribute);

sub new {
    return fields::new(@_);
}

sub test {
    my Foo $self = shift;  # note the my Foo $self
    $self->{does_not_exist} = 1;
}

package main;

my Foo $foo = Foo->new();
print "run time\n";
$foo->test();

print "ok\n";
__OUTPUT__
Compile time
No such pseudo-hash field "does_not_exist" in variable $self of type F
+oo at test.pl line 16.
[download]

Compile time
run time
No such pseudo-hash field "does_not_exist" at test.pl line 16.
[download]

Programming Perl 3 and the warnings also mention pseudo-hashes, but since perl v5.8.x 5.9 they've been replaced by restricted hashes in the "fields" pragma.

updated: fields.pm still uses pseudohashes in perl 5.8.x

It looks like a NOP to me....

C:\>perl -MO=Deparse -e"package Foo; my Foo $bar=42"
package Foo;
my $bar = 42;
-e syntax OK
C:\>
[download]

The other example is invalid syntax, dunno what you think it should do.

This doesn't show up at Deparse because it only does error checking at compile time. It doesn't actually change the code. See my comment below.

edited: "because IT only..."

"What should it profit a man, if he should win a flame war, yet lose his cool?"

I have a hard time believing it is a NOP. I actually first saw it right at the top of the synopsis section of the fields doc:

[snip]

package Foo;
use fields qw(foo bar _Foo_private);
sub new {
    my Foo $self = shift;

[snip]
[download]

As for the attributes example, the attributes doc claims the example is syntactically valid. I realize it is not supposed to "do" anything (other than, I would hope, compile); I'm just trying to grok attributes and fields and ran into these stumbling blocks.

Brad

Does this convince you?

C:\>perl -MO=Concise,-ascii -e"package Foo;my $bar = 23;"
6  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(Foo 1 -e:1) v ->3
5     <2> sassign vKS/2 ->6
3        <$> const[IV 23] s ->4
4        <0> padsv[$bar:1,2] sRM*/LVINTRO ->5
-e syntax OK

C:\>perl -MO=Concise,-ascii -e"package Foo;my Foo $bar = 23;"
6  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(Foo 1 -e:1) v ->3
5     <2> sassign vKS/2 ->6
3        <$> const[IV 23] s ->4
4        <0> padsv[$bar:1,2] sRM*/LVINTRO ->5
-e syntax OK
[download]

Looks pretty convincing to me (there is no difference between the op trees, thus it is a null op).

MJD says "you can't just make shit up and expect the computer to know what you mean, retardo!"

I run a Win32 PPM repository for perl 5.6.x and 5.8.x -- I take requests (README).

** The third rule of perl club is a statement of fact: pod is sexy.

I have a hard time believing it is a NOP

It is syntactically valid because it compiles. QED. It apparently does nothing accoridng to Deparse. QED also is it not? This is also valid syntax, left over from previous perls.....

C:\>perl -MO=Deparse -e"package Foo; $Foo'bar=42"
package Foo;
$bar = 42;
-e syntax OK
C:\>
[download]

cheers

tachyon

AFAIK the my Dog $spot syntax has never been deprecated (see this thread). Pseudohashes have, but the fields pragma works with restricted hashes instead starting perl 5.9.
"What should it profit a man, if he should win a flame war, yet lose his cool?"

It never really worked anyway, as Damian was quick to point out. If you forget to add it somewhere, it blows the whole thing.