desire to pass file handles by value

[mandog]

I'd like to pass file handles to subs by value. The FAQ gave me a recipe but the position of the file gets changed.

#!/usr/bin/perl -w 

use strict;
use 5.6.1;

sub find {
    local *FH = shift;
    my $find_me = shift;
    my $count   = 0;

    my $tell = tell(FH);
    while (<FH>) {
        $count++ if (/$find_me/);
    }

    return "count: $count tell: $tell\n";
}

open OUT, '>', '/tmp/tmp.txt' or die "$!\n";
print OUT while (<DATA>);
close OUT or die "$!\n";

open IN, '/tmp/tmp.txt' or die "$!\n";

print find( *IN, 'a' );
print find( *IN, 'a' );
print find( *IN, 'd' );
print find( *IN, 'e' );

#count: 3 tell: 0
#count: 0 tell: 13
#count: 0 tell: 13
#count: 0 tell: 13


__DATA__
a
a
a
c
d 
e

[Roger]

I think you *need* to rewind the file by calling seek FH, 0, 0; in the subroutine find.

The file operation in Perl is ultimately linked to the underlying unix file handles. When you read from a Perl file handle, Perl uses and updates the underlying unix file handle, thus giving the side effect.

I think ideally you are looking for the Perl's equivalent to the C stdio's dup function, to duplicate/copy an existing file handle into a new file handle (not just an alias to the existing file handle, but a new independent file handle), and operate on the second file handle in the subroutine. I am not aware of such facility in Perl 5 (my lack of research perhaps), but I think Perl 6 has implimented the dup function for the File Object.

Update:
Ok, I just learned that you can duplicate a file handle in Perl with open NEW, "<&OLD";. I quickly came up with the following code -



#!/usr/bin/perl -w
use strict;

sub find {
    local *IN = shift;
    my $find_me = shift;
    my $count   = 0;

    # duplicate existing file handle
    open F2, '<&', *IN or die "Can not duplicate file handle";
    my $tell = tell(F2);
    while (<F2>) {
        $count++ if (/$find_me/);
    }

    return "count: $count tell: $tell\n";
    close F2;
}

open OUT, '>', 'tmp.txt' or die "$!\n";
print OUT while (<DATA>);
close OUT or die "$!\n";

open IN, 'tmp.txt' or die "$!\n";

print find( *IN, 'a' );
print find( *IN, 'a' );
print find( *IN, 'd' );
print find( *IN, 'e' );

__DATA__
a
a
a
c
d
e

The output is still -


#count: 3 tell: 0
#count: 0 tell: 13
#count: 0 tell: 13
#count: 0 tell: 13

That didn't work either! Ok, that taught me a lession - my assumption on the duplicated file handle could be wrong. I need some re-education.

Fellow monks, could you please tell me what is wrong with the duplicated file handle? Is perl actually creating a second independent file handle? Am I doing the right thing at all?

Thanks!

[mandog]

Thanks for the help. For now, I'm going to record the orignal file position on entry to the sub & restore it on exit.

#...
my $pos=tell(FH);
#...
seek(FH,$pos,0);
return

The file handle duplicating thing looks interesting though...

update: fixed dumb semantic error

[jdporter]

Now if only you could localize a variable that represents the file position, you wouldn't have to manually save/restore its value. Hmmm....

    package Tie::Scalar::FHPos;
    use FileHandle;
    sub TIESCALAR
    {
        my( $pkg, $fh ) = @_;
        bless [$fh], $pkg;
    }
    sub FETCH
    {
        my( $self ) = @_;
        $self->[0]->tell
    }
    sub STORE
    {
        my( $self, $pos ) = @_;
        $self->[0]->seek($pos,0);
    }

Test:

open F, "< foo.dat";
our $pos;
tie $pos, 'Tie::Scalar::FHPos', \*F;

{
 local $pos = 20;
 print "$pos: ", scalar(<F>);
}
print "$pos: ", scalar(<F>);

Cool!

[nevyn]

Update: Ok, I just learned that you can duplicate a file handle in Perl with open NEW, "<&OLD";. I quickly came up with the following code -

From a Linux man page...

After successful return of dup or dup2, the old and new descriptors may be used interchangeably. They share locks, file position pointers and flags; for example, if the file position is modified by using lseek on one of the descriptors, the position is also changed for the other.

[Anonymous Monk]

The file operation in Perl is ultimately linked to the underlying unix file handles. When you read from a Perl file handle, Perl uses and updates the underlying unix file handle, thus giving the side effect.

Well not on windows, or other non-unix operating systems ....

[pg]

Besides what Roger said, if this is all what you want to do, you can:

use Data::Dumper;
use strict;
use warnings;

my @lines = <DATA>;
my $count = {};
map {chomp;$count->{$_} ++} @lines;
print Dumper($count);

__DATA__
a
a
a
c
d 
e