sort_doc_topics

#!/usr/bin/env perl 
#===============================================================================
#
#         FILE: sort-doc-topics
#
#        USAGE: ./sort-doc-topics
#
#  DESCRIPTION: read mallet doc-topics output from stdin, write a tab-delimited 
#  file with the topic proportion data ordered by 0-indexed topic number to 
#  stdout. I haven't found a fast way to do this in R, so here it is in perl 
#  (fast).
#
#  If the columns of the doc-topics output are 0, 1, 2, ...
#  and there are 100 topics,
#  then the topic numbers are at positions 2, 4, ... 200
#  and the proportions are at 3, 5, ... 201
#  but topics themselves are numbered 0 .. 99
#  (position 0 is the doc number and position 1 is the doc name)
#
#  This script discards the initial comment line from the mallet output
#  as well as the trailing tab stop at the end of every data line
#
#      OPTIONS: ---
# REQUIREMENTS: ---
#         BUGS: ---
#        NOTES: ---
#       AUTHOR: Andrew Goldstone (agoldst), andrew.goldstone@gmail.com
# ORGANIZATION: Rutgers University, New Brunswick
#      VERSION: 1.0
#      CREATED: 07/24/2012 17:12:29
#     REVISION: ---
#===============================================================================
use v5.14;                                  # entails strict, unicode_strings 
use autodie;
use utf8;                                   # source code itself is in utf-8
use warnings;
use warnings FATAL => "utf8";               # Unicode encode errors are fatal
use open qw( :std :utf8 );                  # utf8 layer for open, stdin/out


# Read and discard the header
my $line = <STDIN>;

$line = <STDIN>;


# Number of topics is: (number of tab delimeters - 2) / 2
# since mallet adds a tab at the end of every line
# So count tabs
my $n = (($line =~ tr/\t//) - 2) / 2;

do {
    my @row = split /\t/, $line;

    # hash saying which topics are in which columns of this row

    my %indexmap = ();
    for(my $i = 2;$i <= $n * 2;$i += 2) {
        $indexmap{$row[$i]} = $i;
    }

    # keep the first two columns
    
    print "$row[0]\t$row[1]";

    # now output the topic proportions in order

    for (my $i = 0;$i <= $n - 1;$i++) {
        print "\t", $row[$indexmap{$i} + 1];
    }

    print "\n";
} while($line = <STDIN>);