They are an exception in that they are not subject to use strict;, so it's best to avoid them.

Thanks for the response, That makes sense... I did that, but what in case if you want to change these default variables??? Any clue???

if there is some strange reason you cannot lexify your other usages of $a and $b, then you can consider using a named ($$)-prototyped sort subroutine, and perl will pass the sort variables in @_ instead of package variables.

    sub mysort ($$) { $_[0]->{key} <=> $_[1]->{key} }
    my @sorted = sort mysort @unsorted;
[download]

But it's also a good idea to avoid them since they aren't meaningful names. I think one of the most important parts of writing code is choosing good names for variables (and functions, and all name-able things). I don't mean needlessly long names, but appropriate names (name length should in general be loosely proportional to variable scope - if that gives you long names you want smaller scope for your variables).

Yes, you have to stop and think about a good name, but it makes the code much more readable. Compare: $page->render(); with $tmp->do_it().

It's also important to change names if you change the code, so they stay accurate. (e.g. change a variable from $employee to $person if you make a routine more general).

Sorting a hash table? Do you mean sorting keys in a hash? Or?

If you have lexicalized $a and $b (with my), you won't be able to use them in sort in the same or in the child block:

km@krot:/home/km> perl -le 'my $a = "q"; my $b = "x"; print join "-", 
+sort { $a cmp $b } qw(foo bar baz); print $a; print $b;'
Can't use "my $a" in sort comparison at -e line 1.
[download]

If you have lexicalized $a and $b in a separate block, you can use $a and $b in sort (note that both $a and $b are empty in last prints):

km@krot:/home/km> perl -le '{my $a = "q"; my $b = "x";} print join "-"
+, sort { $a cmp $b } qw(foo bar baz); print $a; print $b;'
bar-baz-foo
[download]

Please read perldoc -f sort and perlvar. I'd avoid use of $a and $b altogether.