Extracting comma-separated integers with Perl

A friend writes asking which of

my @data = ( $data =~ m|(-?\d+),(-?\d+),(\-?\d+)\r| );


($data) = (split "\r", $data);
my @data = split ',', $data;

would be better for extracting integers from a line of input, and my reply is below.

The best approach depends on a few factors. Who generates the data you’re processing? How much slop in the input format do you need to accommodate? How flexible do you need to be with respect to future changes in the input, e.g., extra fields, different types, and so on?

The CR (\r) in your input is a red flag. Are you running on a Unix platform with input generated on Windows? Can you be more specific about “possibly some other stuff” after the comma-separated numbers?

Perl’s $/ special variable can handle oddball line endings. Its default value varies with what makes sense for the current platform: e.g., \n on Unix and \r\n on Windows. Macs introduce another twist (see Newlines in the perlport documentation), but I assume you’re not using that platform.

On Windows for files opened in text mode (the default), the C library silently transforms CR followed by LF into LF, so if this matches your setup, I’m surprised you’re seeing the \r at all.

Say your input is plain text generated on Windows, and you’re running on Linux. Then you’d process it with code of the form

$/ = "\r\n";
while (defined($data = <>)) {

Remember that chomp removes the value of $/ from the end of the target.

As for extracting the data, Randal Schwartz (author of Learning Perl, a.k.a. the llama book) has a rule of thumb:

Use capturing or m//g when you know what you want to keep.

Use split when you know what you want to throw away.

I first saw this useful guideline in Regular Expression Mastery by Mark Dominus.

If this is a quick-and-dirty utility, I’d be inclined to write

@data = split /\s*,\s*/, $data;

This allows for and removes any whitespace around the commas.

If it’s important to nail down the format (maybe as a sanity check that you’re in the section of the config file where you think you are), you could write

if ($data =~ /^\s*(-?\d+)\s*,\s*(-?\d+)\s*,\s*(-?\d+)\s*$/) {
    my($x,$y,$z) = ($1,$2,$3);
else {
    die "$0: $ARGV:$.: unexpected format";

Note the use of $1 and friends inside the conditional only. Always, always, always protect uses of capture variables with conditionals.

The pattern is just at the annoyance threshold of repetition and illegibility. Perl version 5.10 opens up nice possibilities with named capture buffers:

#! /usr/bin/env perl
use strict;
use warnings;
use 5.10.0;
my $record = qr/
    (?<num>(?&n)) (?&sep) (?<num>(?&n)) (?&sep) (?<num>(?&n))
        (?<n> -? \d+)
        (?<ws> \s* )
        (?<sep> (?&ws) , (?&ws))
while (<DATA>) {
    if (/$record/) {
        my($x,$y,$z) = @{ $-{num} };
        print "got: x=$x, y=$y, z=$z\n";
    else {
        warn "$0: line $.: no match\n";

Its output:

got: x=1, y=2, z=3
got: x=4, y=5, z=6
./prog: line 3: no match

Notice its use of the special %- hash that records all captures named “num” in this case. With (DEFINE), subpatterns get meaningful names, and the /x modifier allow for judicious use of whitespace inside the pattern.