I've been playing around with Perl over the last couple of days to see what I can do with references. I really get annoyed with having to test everything to see if it is blessed before testing to see if it matches the interface that I'm expecting:
if (blessed( $arg ) && $arg->isa('Some::Class')) {
# ...
}
So what I've been trying to do is get it so that references are automagically blessed into a class that shares the name of their reference type:
% perl -Mmagicrefs -MData::Dumper -e 'print Dumper( {} )'
$VAR1 = bless( {}, 'HASH' );
And it all works, so now I can do:
use lib './lib';
use strict;
use magicrefs; ## turn on magic references
I'm guessing the chances of this going into the core are pretty slim however :-)
my $c = {
Auto => 'blessed',
refs => 'are',
fun => '.'
};
print $c->keys->join(','),"\n";
## or ##
if ($c->keys->grep( sub { /^[A-Z]/ } )->size == 1) {
print "right\n";
} else {
print "wrong\n";
}
## or ##
my %hash = $c->deref;
## or with the magic of deparse
print sub { 1; }->code(), "\n";
Re:UNIVERSAL::isa
james on 2003-06-10T11:36:55
I know, I was using it as a small, trite example. The actual problem is that its easier to refactor away conditions if everything is an object - which is my real dislike. I felt going into why I like to refactor away conditions was probably beyond the scope of the journal entry, which was more about having HASH/ARRAY/SCALAR/... objects, and the nice things that you can do once you have them.Re:UNIVERSAL::isa
rafael on 2003-06-10T11:50:37
And by the way are you going to upload magicrefs to some comprehensive perl archive network ?Re:UNIVERSAL::isa
james on 2003-06-10T11:54:01
It's essentially a patch against pp.c in the perl source. I could put it on CPAN, but it seems like a strange place to put a patch. I think it might be possible to do something similar with optimizer.pm, but I need to talk to sky a little about that.Re:UNIVERSAL::isa
rafael on 2003-06-10T12:14:59
Ah. You're evil, then;-) If this is a core patch, then I don't understand what's in magicrefs.pm -- is it a lexical pragma ? What core function(s) are you patching, and to do what ? Re:UNIVERSAL::isa
james on 2003-06-10T12:26:28
magicrefs.pm is a pragma similar to sort, ie, not lexical.
Essentially it alters S_refto in pp.c. magicrefs.pm sets a hint that is examined in S_refto to determine whether or not we should bless all references that are created (with the exception of those that are SvREADONLY). magicrefs also use's the various reftype modules.
Re:UNIVERSAL::isa
rafael on 2003-06-10T12:49:27
OK, I can now see how optimizer.pm can help -- by inserting a bless op before the refgen/srefgen ops.Re:UNIVERSAL::isa
malte on 2003-06-10T15:04:43
You could probably do this by overloading the dereferencing operators.
Obviously, at the expend of acceptable performance.
blessed which you've just broken.
Let's fix that
(untested code, of course)sub BasicType::a_sub_not_likely_to_be_here { 0 }
@HASH::ISA=qw(BasicType);
@ARRAY::ISA=qw(BasicType);
...