Sorry, I wrote a bad example, my mistake. Please look at the complete code, I am trying to build hash of hashes where each folder is a key. Early I did hash of hashes where keys are full path to folder, but it wasn't pretty.

#!/usr/bin/perl
use strict;
use warnings;
use Data::Dumper;


my %filesystem;
my $ref = $filesystem{root} ||= {};


traverse(".","");

sub traverse{
        my ($folder,@path) = @_;
        opendir DIR,$folder;
        my @list = grep {!/^\.{1,2}$/} readdir DIR; # exclude . and ..
        closedir DIR;
        foreach my $entry(@list){
                next if ( -l $folder."/". $entry); # skip links
                $ref = $$ref{$_} ||= {} for @path;
                if( -d $folder."/". $entry){
                    traverse($folder."/".$entry,@path,$entry);
                }else{
                    $$ref{$folder} = $entry;
                };
        };
};


print Dumper(\%filesystem);
[download]

if I run it it eat all my RAM, so I think I use $ref from your example wrong

It seems like you are trying to build up a deeply nested data structure which looks a lot like your directory tree. However, if all you need is a simple array of full paths to files, take a look at File::Find::Rule, which recursively traverses the filesystem.

You need to decide what you want your data structure to look like before you do anything else. Right now, you're trying to store every non-directory file in directory at the same key as the directory itself.

Try running your script with some extra print statements and you can see what it is doing as it progresses. I expect the data structure it is building is not what you want and may help explain why you are using so much memory.

Read more... (1366 Bytes)

You should think carefully about the line $ref = $$ref{$_} ||= {} for @path; in your subroutine.

Also, your subroutine is recursive. It keeps changing the global variable $ref. Think about what $ref is referencing and how this changes as your program progresses.