Mercurial > repo
view share/construct_grams.pl @ 8065:591b1467ccdf
<int-e> le/rn paste/"Paste" is a short story by Henry James. Its contents has been cut into pieces and distributed over numerous tin boxes on the World Wide Web, little pearls of wisdom buried among ordinary pastes.
author | HackBot |
---|---|
date | Sun, 15 May 2016 13:14:57 +0000 |
parents | e037173e0012 |
children |
line wrap: on
line source
#!/usr/bin/perl use v5.10; use strict; use warnings; use utf8; use Storable 'store'; use Getopt::Long; my $n = 4; my %grams; my %fs; my %seen; my $encoding = "UTF-8"; my $filter = qr/^[\p{Alphabetic}\p{Dash_Punctuation}\p{Connector_Punctuation}']+$/; sub parse(_) { my ($f) = @_; while(my $line = <$f>) { my $word = lc ((split /[^\S\240]/, $line)[0]); chomp $word; next if $seen{$word}++ || $word !~ $filter; $fs{length $word}++; $word = ' ' x ($n-1) . "$word "; for(my $i = 0; $_ = substr($word, $i, $n); $i++) { last unless length == $n; $grams{substr($_, 0, $n-1)}->{substr($_, $n-1, 1)}++; } } } sub main { my $target_mod = "Default"; GetOptions( 'm|module=s' => \$target_mod, 'e|encoding=s' => \$encoding, 'f|filter=s' => \$filter ) or exit 1; $target_mod =~ s/(^|[-_ ])(.)/\u$2/g; $filter = qr/$filter/i; print "Constructing $target_mod dataset from $encoding\n"; print "Filter: $filter\n"; for (@ARGV) { print "Reading $_...\n"; open my $f, "<:encoding($encoding)", $_; parse $f; close $f; } store [\%grams, \%fs], "$target_mod"; } main unless caller;