Book HomePerl CookbookSearch this book

4.14. Sorting an Array Numerically

Problem

You want to sort a list of numbers, but Perl's sort (by default) sorts alphabetically in ASCII order.

Solution

Use Perl's sort function and the <=> numerical comparison operator:

@sorted = sort { $a <=> $b } @unsorted;

Discussion

The sort function takes an optional code block, which lets you replace the default alphabetic comparison subroutine with your own. This comparison function is called each time sort has to compare two values. The values to compare are loaded into the special package variables $a and $b, which are automatically localized.

The comparison function should return a negative number if $a ought to appear before $b in the output list, 0 if they're the same and their order doesn't matter, or a positive number if $a ought to appear after $b. Perl has two operators that behave this way: <=> for sorting numbers in ascending numeric order, and cmp for sorting strings in ascending alphabetic order. By default, sort uses cmp-style comparisons.

Here's code that sorts the list of PIDs in @pids, lets the user select one, then sends it a TERM signal followed by a KILL signal. We use a code block that compares $a to $b with <=> to sort numerically:

# @pids is an unsorted array of process IDs
foreach my $pid (sort { $a <=> $b } @pids) {
    print "$pid\n";
}
print "Select a process ID to kill:\n";
chomp ($pid = <>);
die "Exiting ... \n" unless $pid && $pid =~ /^\d+$/;
kill('TERM',$pid);
sleep 2;
kill('KILL',$pid);

If you use $a <=> $b or $a cmp $b, the list will be sorted in ascending order. For a descending sort, all we have to do is swap $a and $b in the sort subroutine:

@descending = sort { $b <=> $a } @unsorted;

Comparison routines must be consistent; that is, they should always return the same answer when called with the same values. Inconsistent comparison routines lead to infinite loops or core dumps, especially in older releases of Perl.

You can also say sort SUBNAME LIST where SUBNAME is the name of a comparison subroutine returning -1, 0, or +1. In the interests of speed, the normal calling conventions are bypassed, and the values to be compared magically appear for the duration of the subroutine in the global package variables $a and $b. Because of the odd way Perl calls this subroutine, it may not be recursive.

A word of warning: $a and $b are set in the package active in the call to sort, which may not be the same as the one that the SUBNAME function passed to sort was compiled in! For example:

package Sort_Subs;
sub revnum { $b <=> $a }

package Other_Pack;
@all = sort Sort_Subs::revnum 4, 19, 8, 3;

This will silently fail (unless you have -w in effect, in which case it will vocally fail), because the sort call sets the package variables $a and $b in its own package, Other_Pack, but the revnum function uses its own package's versions. This is another reason why in-lining sort functions is easier, as in:

@all = sort { $b <=> $a } 4, 19, 8, 3;

For more on packages, see Chapter 10, Subroutines.

See Also

The cmp and <=> operators in perlop (1) and Chapter 2 of Programming Perl; the kill, sort, and sleep functions in perlfunc (1) and Chapter 3 of Programming Perl; Recipe 4.15


Previous: 4.13. Finding All Elements in an Array Matching Certain CriteriaPerl CookbookNext: 4.15. Sorting a List by Computable Field
4.13. Finding All Elements in an Array Matching Certain CriteriaBook Index4.15. Sorting a List by Computable Field

Library Navigation Links

Copyright © 2002 O'Reilly & Associates. All rights reserved.