Filehandle of the current sourcecode, pointing after last code line?

[LanX]

DATA is - as some of you already know - the filehandle of the currently executed sourcecode, just pointing behind the last code-line after compilation. (see SelfLoader and perldata for details)

Question : Is there a way to get this filehandle without specifying a __DATA__ or __END__ section, i.e. without reopening $0 and always pointing behind the last codeline after compilation (even if there is already a __DATA__ section defined).

Otherwise if I have defined DATA filehandles in various packages, can I check which file (if existent) is associated with it?

Or if there is no pure Perl way to do it, is it possible to inspect this info with help of the B backend or other magic?

For motivation see Re^2: Why shows B::Deparse __END__ as __DATA__? (BUG #1) following.

[Anonymous Monk]

Question : Is there a way to get this filehandle without specifying a __DATA__ or __END__ section, i.e. without reopening $0 and always pointing behind the last codeline after compilation (even if there is already a __DATA__ section defined)

No, I don't think so, only stuff like open DATA, '<', \'this works too';

Otherwise if I have defined DATA filehandles in various packages, can I check which file (if existent) is associated with it?

No, I don't think so, you cannot get filename from filehandle, there is no api and no guarantee there exists a filename, but %INC ought to have package/filename associations ...

Or if there is no pure Perl way to do it, is it possible to inspect this info with help of the B backend or other magic?

I don't know but I doubt it :)

if (tmp && tmp != KEY___DATA__ && tmp != KEY___END__

http://perl5.git.perl.org/perl.git/blob?f=t/io/data.t

[LanX]

Thanx, that's what I expected... :-/

The only way I see to fix the bug in Deparse is to compare the possible DATA handles (which are already at the right position) with the content if $0 / __FILE__ .

Not elegant though...

[shmem]

Just fully qualify your file handles.

# file Foo.pm
package Foo;
1;
__DATA__
Foo
foo foo foo

# file Bar.pm
package Bar;

1;
__DATA__
print Bar. I told you so.

Last, the script...

#!/usr/bin/perl

use Foo;
use Bar;
print while <DATA>;
print "Foo:\n";
print while <Foo::DATA>;
print "Bar:\n";
print while <Bar::DATA>;

__DATA__
so this works.

__END__
so this works.
Foo:
Foo
foo foo foo

Bar:
print Bar. I told you so.

If you can read from the different DATA filehandles, you've got them. ;)

Parting from there, you can get at the file name via fileno and /proc/$$/fd/ (on linux, that is):

#!/usr/bin/perl

use Foo;
use Bar;
print while <DATA>;
system "ls -l /proc/$$/fd";
print "Foo: ",fileno Foo::DATA, "\n";
print while <Foo::DATA>;
print "Bar: ",fileno Bar::DATA, "\n";
print while <Bar::DATA>;

__DATA__
so this works.

[LanX]

Did you read the "motivation" part? :)

update

I.e. I have no control about the inspected code and it has to be cross platform...

[shmem]

Did you read the "motivation" part? :)

Yes, I did. And I can't see any bug there regarding DATA. Maybe I don't get the point.

package Quux;
our $foo = "blorflydick";
print "in Quux\n";
print while <DATA>;
print $foo,$/;
1;
package blorflydick;
print $foo,$/;
__DATA__

blorf, blorf.

#!/usr/bin/perl

use Foo;
use Bar;
use Quux;
print while <DATA>;
print "Foo (",fileno Foo::DATA, ")\n";
print while <Foo::DATA>;
print "Bar (",fileno Bar::DATA, ")\n";
print while <Bar::DATA>;
print "Symbol table Quux::\n";
print "$_ => $Quux::{$_}\n" for sort keys %Quux;
print "Symbol table blorflydick::\n";
print "$_ => $blorflydick::{$_}\n" for sort keys %blorflydick::;
print "Quux (",fileno Quux::DATA, ")\n";
print while <Quux::DATA>;
print "ok, trying blorflydick\n";
print "blorflydick (",fileno blorflydick::DATA, ")\n";

print while <blorflydick::DATA>;
package Baz;
__DATA__
so this works.
__END__
in Quux
blorflydick
blorflydick
Foo (4)
Foo
foo foo foo

Bar (5)
print Bar. I told you so.
Symbol table Quux::
Symbol table blorflydick::
DATA => *blorflydick::DATA
Quux ()
ok, trying blorflydick
blorflydick (6)

blorf, blorf.

I think of the DATA filehandle as an alias to the fully qualified DATA filehandle with similar semantics as our, with the difference that it isn't file scoped like our variables are.

So, the blorflydick::DATA filehandle points to Quux.pm which is not a bug IMHO.