The first argument of split is either a compiled regexp or a regexp string. | is a special character in regular expressions, so it must be escaped if you mean it literally.

Some solutions:

split(/\|/, ...)
[download]

my $re = qr/\|/;
split($re, ...)
[download]

my $re = "\\|";  # \|
split($re, ...)
[download]

my $str = "|";             # |
my $re = quotemeta($str);  # \|
split($re, ...)
[download]

In your case, the last alternative is the desired one. You want to convert a text into a regexp, and quotemeta will do that for you.

my $sep = $arr[2];
split(quotemeta($sep), ...)
[download]

hey Thanx to both of you...

Your second split is attempting to use ~ as the match character, not |. Did you intend to use $sep? Actually, even that won't do what you want because you need to quote | (it is a regex meta character). Something like:

@arr = split /\Q$sep\E/, $var;
[download]

my @arr = split /\Q$ele\E/, $var;
[download]

BTW, you have $ele when you should have $sep, unless you plan on splitting on the "~" character. You did say in your question that you wanted split on the "|" character?

In this case you probably don't need the \E since you are quoting the entire regex anyway.

#!/usr/bin/perl
# vim: sw=4

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

my $var = "21|23|24~|34|45|56~";

my @arr = split("", $var);

my $sep = $arr[2];
my $ele = $arr[8];

my @ar = split(/\Q$sep\E/, $var);

print Dumper \@ar;
bruce:1:~/tmp $ ./p.pl
$VAR1 = [
          '21',
          '23',
          '24~',
          '34',
          '45',
          '56~'
        ];
bruce:1:~/tmp $
[download]