How can I do alpha numeric sort in Perl? - perl

Question

I have a file which looks like this:
80,1p21
81,19q13
82,6p12.3
83,Xp11.22
84,3pter-q21
86,3q26.33
87,14q24.1-q24.2|14q24|14q22-q24
88,1q42-q43
89,11q13.1
90,2q23-q24
91,12q13
92,2q22.3
93,3p22
94,12q11-q14
95,3p21.1
97,14q24.3
98,2p16.2
And I want to sort them based on the second column. And the first column should change accordingly too. When you use the 'sort' command in Perl, it doesn't do it because it says it's not numeric. Is there a way to sort things alpha numerically in Perl?

Answer 1

If you read the documentation for sort, you'll see that you don't need to do a numeric sort in Perl. You can do string comparisons too.
#sorted = sort { $a cmp $b } #unsorted;
But that still leaves you with a problem as, for example, 19q will sort before 6p. So you can write your own sort function which can make whatever transformations you want before doing the comparison.
#sorted = sort my_complex_sort #unsorted;
sub my_complex_sort {
# code that compares $a and $b and returns -1, 0 or 1 as appropriate
# It's probably best in most cases to do the actual comparison using cmp or <=>
# Extract the digits following the first comma
my ($number_a) = $a =~ /,(\d+)/;
my ($number_b) = $b =~ /,(\d+)/;
# Extract the letter following those digits
my ($letter_a) = $a =~ /,\d+(a-z)/;
my ($letter_b) = $b =~ /,\d+(a-z)/;
# Compare and return
return $number_a <=> $number_b or $letter_a cmp $letter_b;
}

Answer 2

#!/usr/bin/env perl
use strict;
use warnings;
my #datas = map { /^(\d+),(\d*)(.*)$/; [$1, $2, $3]; } <DATA>;
my #res = sort {$a->[1] <=> $b->[1] or $a->[2] cmp $b->[2]} #datas;
foreach my $data (#res) {
my ($x, $y, $z) = #{$data};
print "$x,$y$z\n";
}
__DATA__
80,1p21
81,19q13
82,6p12.3
83,Xp11.22
84,3pter-q21
86,3q26.33
87,14q24.1-q24.2|14q24|14q22-q24
88,1q42-q43
89,11q13.1
90,2q23-q24
91,12q13
92,2q22.3
93,3p22
94,12q11-q14
95,3p21.1
97,14q24.3
98,2p16.2

Answer 3

I actually found the answer to this. The code looks a bit complicated though.
#!/usr/bin/env perl
use strict;
use warnings;
sub main {
my $file;
if (#ARGV != 1) {
die "Usage: perl hashofhash_sort.pl <filename>\n";
}
else {
$file = $ARGV[0];
}
open(IN, $file) or die "Error!! Cannot open the $file file: $!\n";
my #file = <IN>;
chomp #file;
my ($entrez_gene, $loci, $chr, $band, $pq, $band_num);
my (%chromosome, %loci_entrez);
foreach my $line (#file) {
if ($line =~ /(\d+),(.+)/) {
# Entrez genes
$entrez_gene = $1;
# Locus like 12p23.4
$loci = $2;
if ($loci =~ /^(\d+)(.+)?/) {
# chromosome number alone (only numericals)
$chr = $1;
if ($2) {
# locus minus chromosome number. If 12p23.4, then $band is p23.4
$band = "$2";
if ($band =~ /^([pq])(.+)/) {
# either p or q
$pq = $1;
# stores the numericals. for p23.4, stores 23.4
$band_num = $2;
}
if (exists $chromosome{$chr}) {
if (exists $chromosome{$chr}{$pq}) {
push (#{$chromosome{$chr}{$pq}}, $band_num);
}
else {
$chromosome{$chr}{$pq} = [$band_num];
}
}
else {
$chromosome{$chr}{$pq} = [$band_num];
}
}
}
}
} # End of foreach loop
foreach my $key (sort {$a <=> $b} keys %chromosome) {
my %seen = ();
foreach my $key2 (sort {$a cmp $b } keys %{$chromosome{$key}}) {
my #unique = grep { ! $seen{$_}++ } #{$chromosome{$key}{$key2}};
my #sorted = sort #unique;
foreach my $element (#sorted) {
my $sorted_locus = "$key$key2$element";
if (exists $loci_entrez{$sorted_locus}) {
foreach my $element2 (#{$loci_entrez{$sorted_locus}}) {
print "$element2,$sorted_locus\n";
}
}
}
}
}
} # End of main
main();

Answer 4

In the very general case, the question is ambiguous on what to do with integers that are equal but written differently, because of the possibility of leading zeros. The following comparison function (for sort) allows one to consider the lexicographic order as soon as one doesn't have different integers. This is the same as zsh's numeric sort.
sub alphanumcmp ($$)
{
my (#u,#v);
if ((#u = $_[0] =~ /^(\d+)/) &&
(#v = $_[1] =~ /^(\d+)/))
{
my $c = $u[0] <=> $v[0];
return $c if $c;
}
if ((#u = $_[0] =~ /^(.)(.*)/) &&
(#v = $_[1] =~ /^(.)(.*)/))
{
return $u[0] cmp $v[0] || &alphanumcmp($u[1],$v[1]);
}
return $_[0] cmp $_[1];
}
For instance, one would get the following sorted elements:
a0. a00. a000b a00b a0b a001b a01. a01b a1. a1b a010b a10b a011b a11b
Note 1: The use of <=> assumes that the numbers are not too large.
Note 2: In the question, the user wants to do an alphanumeric sort on the second column (instead of the whole string). So, in this particular case, the comparison function could just be adapted to ignore the first column or a Schwartzian transform could be used.