#!/usr/bin/env perl
use feature ':5.10';
use strict;
use warnings;
use File::Slurp qw(slurp);
use YAML::Syck qw(Dump Load LoadFile DumpFile);
BEGIN {
    $YAML::Syck::Headless = 1;
    $YAML::Syck::SortKeys = 1;
}
use WWW::Mechanize;
use HTML::TableParser::Grid;
use Pod::Usage ();
use Getopt::Long ();
use Data::Dump 'dump';
use File::Spec::Functions qw(catfile);
use Storable;
use autodie;

=head1 NAME

merge-from-translatewiki - Get new translations from L<http://translatewiki.net> and selectively merge them with ours

=head1 SYNOPSIS

    # Run this normally, hopefully...
    merge-from-translatewiki --locales-dir=config/locales

    # Diff the existing files:
    config/locales$ for i in $(ls *yml | grep -v en.yml); do perl ../../script/locale/diff --dump-flat $i > $i.0 ;done

    # Merge and find out what changed:
    rails_port$ perl script/locale/merge-from-translatewiki --locales-dir config/locales

    # Or, more complexy:
    rails_port$ for i in $(svn st config/locales/ | egrep '^M|\\?' | awk '{print $2}' | grep 'yml$'); do rm -v $i; done && svn up config/locales && perl script/locale/merge-from-translatewiki --locales-dir config/locales && svn st config/locales
 
    # Diff:
    config/locales$ for i in $(ls *yml | grep -v en.yml); do perl ../../script/locale/diff --dump-flat $i > $i.1 ;done && for i in $(ls *yml | grep -v en.yml); do diff -ru $i.*; done

=head1 DESCRIPTION

Translatewiki's export process L<is
broken|http://trac.openstreetmap.org/ticket/2305>. This script imports
new messages from it while tiptoeing around known bugs.

=head1 OPTIONS

=over

=item -h, --help

Print this help message.

=item --locales-dir

The locales dir we'll merge stuff into. F<config/locales> by default.

=item --only-new

Only import translations that don't exists for us yet.

=item --cache

Write a L<Storable> cache for things downloaded from Translatewiki and
use it if it exists.

=back

=head1 AUTHOR

E<AElig>var ArnfjE<ouml>rE<eth> Bjarmason <avarab@gmail.com>

=cut

# Get the command-line options
Getopt::Long::Parser->new(
    config => [ qw< bundling no_ignore_case no_require_order pass_through > ],
)->getoptions(
    'h|help' => \my $help,
    'locales-dir=s' => \(my $locales_dir = 'config/locales'),
    'only-new' => \my $only_new,
    'cache' => \my $cache,
) or help();

# On --help
help() if $help;

help() unless $locales_dir and -d $locales_dir;

###
### Main
###

### Get Translatewiki data
my %translatewiki_languages = translatewiki_languages();

# Don't process English from Translatewiki
delete $translatewiki_languages{en};

#say Dump \%translatewiki_languages;

my @translatewiki_languages_codes = keys %translatewiki_languages;
my %translatewiki_translations = get_translatewiki_translations(@translatewiki_languages_codes);

#say Dump \%translatewiki_translations;

### Get our existing data
my %my_translations;
my @my_yaml_files = glob catfile($locales_dir, '*.yml');
for my $my_yaml_file (@my_yaml_files) {
    my $basename = basename($my_yaml_file);
    my $tw_lang = lc $basename;

    say STDERR "Loading my translation $tw_lang ($my_yaml_file)";
    $my_translations{$tw_lang} = load_and_flatten_yaml(scalar slurp($my_yaml_file));
}

say "loaded my translations";

## Write out merged data
for my $translatewiki_lang (sort @translatewiki_languages_codes) {
    my $rails_lang = $translatewiki_lang; $rails_lang =~ s/(?<=-)(\w+)/\U$1\E/;
    my $out_file = catfile($locales_dir, $rails_lang . '.yml');

    unless (-f $out_file) {
        # No translation like this exists
        say STDERR "$rails_lang has no existing translation. Importing as-is from Translatewiki to $out_file";
        my $expanded = expand_hash($translatewiki_translations{$translatewiki_lang});
        my $out = +{ $rails_lang => $expanded };
        spit_out($out_file, $out);
    } elsif (ref $my_translations{$translatewiki_lang} eq 'HASH' and not $only_new) {
        say STDERR "$rails_lang has existing translations. Merging the old translation with the new Translatewiki one";

        # Get the data
        my %tw = %{ $translatewiki_translations{$translatewiki_lang} };
        my %me = %{ $my_translations{$translatewiki_lang} };
        my %en = %{ $my_translations{en} };
        # Use %tw to start with
        my %new = %tw;

        ### Merge stuff

        ## These keys shouldn't be removed but are due to
        ## Translatewiki fail (they were missing in the original
        ## import)
        my @url_keys = qw(
                             browse.relation_member.entry
                             changeset.changeset.id
                             geocoder.search_osm_namefinder.suffix_suburb
                             html.dir
                             layouts.intro_3_bytemark
                             layouts.intro_3_ucl
                             layouts.project_name.h1
                             layouts.project_name.title
                             printable_name.with_version
                             site.edit.anon_edits
                             layouts.help_wiki_url
                             layouts.shop_url
                             notifier.gpx_notification.failure.import_failures_url
                             notifier.signup_confirm_plain.the_wiki_url
                             notifier.signup_confirm_plain.wiki_signup_url
                             trace.edit.visibility_help_url
                             trace.trace_form.help_url
                             trace.trace_form.visibility_help_url
                     );

        for my $key (@url_keys) {
            if ( exists $me{$key} and not exists $new{$key} ) {
                $new{$key} = $me{$key} if $me{$key} ne $en{$key};
            }
        }

        ## When foo exists in this file but only foo.one, foo,other
        ## etc in English or the original file we don't want to throw away what we have
        my @plural_keys = qw( zero one many few other two );

        while (my ($me_k, $me_v) = each %me) {
            if (not exists $tw{ $me_k } and 
                not exists $en{ $me_k } and
                (
                    exists $en{ $me_k . '.zero' } or
                    exists $en{ $me_k . '.one' } or
                    exists $en{ $me_k . '.many' } or
                    exists $en{ $me_k . '.few' } or
                    exists $en{ $me_k . '.other' } or
                    exists $en{ $me_k . '.two' })) {
                #say STDERR "Bringing back nuked plural form '$me_k' Setting it to '$me{ $me_k }'";
                $new{ $me_k } = $me{ $me_k };
            }
        }

        # Both arrays and strings are supported in the site key. Avoid removing e.g.:
        #   -site.key.table.entry.school: 學校；大學
        # Just because en.yml has site.key.table.entry.school.0 and site.key.table.entry.school.1
        while (my ($me_k, $me_v) = each %me) {
            next unless $me_k =~ /^site\.key\.table\.entry/;
            next if $me_k =~ /\.\d+$/;

            if (ref $en{ $me_k } eq 'ARRAY' and not ref $me{ $me_k }) {
                $new{ $me_k } = $me{ $me_k };
            }
        }

        # There are a bunch of keys on Translatewiki that are
        # equivalent to English for some reason. Probably because they
        # were there at import time. Nuke them.
        while (my ($new_k, $new_v) = each %new) {
            if (exists $en{ $new_k } and $en{ $new_k } eq $new_v) {
                #say "Purging dupe in $rails_lang: $new_k=$new_v";
                delete $new{ $new_k };
            }
        }

        my $expanded = expand_hash( \%new );
        my $out = +{ $rails_lang => $expanded };
        spit_out($out_file, $out);
    } elsif (not $only_new) {
        die "Internal error on $translatewiki_lang";
    }
}

sub spit_out
{
    my ($file, $data) = @_;
    my $yaml_out = Dump $data;
    
    open my $fh, ">", $file;
    print $fh $yaml_out;
    close $fh;
}

#
# YAML stuff
#

sub mark_utf8
{
    my ($hash) = @_;

    # Mark as UTF-8
    map { if (ref $_ eq 'ARRAY') { map { utf8::decode($_) } @$_ } else {  utf8::decode($_) } } values %$hash;
}

sub iterate
{
    my ($hash, @path) = @_;
    my @ret;
        
    while (my ($k, $v) = each %$hash)
    {
        if (ref $v eq 'HASH')
        {
             push @ret => iterate($v, @path, $k);
        }
        else
        {
            push @ret => join(".",@path, $k), $v;
        }
    }

    return @ret;
}

sub expand_hash
{
    my ($flat_hash) = @_;
    my %new_hash;
    while (my ($k, $v) = each %$flat_hash) {
        #say "Inserting $k=$v";
        insert_string_deep(\%new_hash, $k, $v);
    }

    \%new_hash;
}

# Fails under strict in certain cases:
## Inserting browse.start_rjs.object_list.history.type.way=Vía [[id]]
## Inserting activerecord.models.relation_tag=Etiqueta de la relación
## Inserting browse.changeset_details.has_nodes.one=Tiene el siguiente {{count}} nodo:
## Can't use string ("Tiene {{count}} nodos:") as a HASH ref while "strict refs" in use at script/locale/merge-from-translatewiki line 234.
# Line 234 = my $p = \$h; $p = \$$p->{$_} for split /\./, $ks;

# sub insert_string_deep_X {
#     my ($h, $ks, $v) = @_;
#     my $p = \$h; $p = \$$p->{$_} for split /\./, $ks;
#     $$p = $v;
# }

sub insert_string_deep
{
    my ($hash, $key, $value) = @_;

    my @key = split /\./, $key;
    my $h = $hash;

    my $i = 0;
    for my $k (@key) {
        $i ++;
        if ($i == @key) {
            $h->{$k} = $value;
        } else {
            if (ref $h->{$k}) {
                $h = $h->{$k};
            } else {
                $h = $h->{$k} = {};
            }
        }
    }
}

#
# Get language from Translatewiki
#

sub get_translatewiki_translations
{
    my @languages = @_;

    my $cache_file = "/tmp/merge-from-translatewiki.storable";
    if ($cache) {
        if (-f $cache_file) {
            my $c = retrieve($cache_file);
            return %$c;
        }
    }

    my %translatewiki_languages;
    my $all_count = scalar @languages;
    say "Translatewiki has $all_count languages I'm about to get";
    my $count = 0;
    for my $lang (@languages) {
        $count ++;
        say STDERR "Getting language $count/$all_count ($lang) from Translatewiki";
        my $yaml = get_language_from_translatewiki($lang);

        my $flat_data = load_and_flatten_yaml($yaml);

        $translatewiki_languages{$lang} = $flat_data;
    }

    if ($cache) {
        store \%translatewiki_languages, $cache_file;
    }

    return %translatewiki_languages;
}

sub get_language_from_translatewiki
{
    my ($lang) = @_;
    my $mech = WWW::Mechanize->new;

    $mech->get("http://translatewiki.net/w/i.php?title=Special%3ATranslate&task=export-to-file&group=out-osm-site&language=$lang");

    die "Couldn't get lang $lang lang from Translatewiki" unless $mech->success;

    return $mech->content;
}

#
# from language list
#

sub translatewiki_languages
{
    my $mech = WWW::Mechanize->new;

    $mech->get('http://translatewiki.net/wiki/Translating:OpenStreetMap/stats/trunk/site');

    die "Couldn't get translatewiki table" unless $mech->success;

    my $content = $mech->content;
    my ($sortable) = $content =~ m[(<table class="sortable.*</table>)]s;

    my @table = parse_language_table($sortable);

    # Just get the codes
    map { $_->{code} => $_->{language} } @table;
}

sub parse_language_table
{
    my ($table) = @_;

    my $parser = HTML::TableParser::Grid->new($table);
    my @rows;
    for my $n (0 .. $parser->num_rows - 1) {
        my %row;
        @row{qw(code language done fuzzy)} = $parser->row($n);
        mark_utf8(\%row);

        push @rows => \%row;
    }

    @rows;
}

#
# Misc
#

sub basename
{
    my $name = shift;
    $name =~ s[\..*?$][];
    $name =~ s[.*/][];
    $name;
}

sub load_and_flatten_yaml
{
    my ($yaml) = @_;

    my $data = Load($yaml);

    # Remove the root $lang => key
    my @keys = keys %$data;
    die "YAML data had more than 1 root key" if @keys != 1;
    $data = $data->{$keys[0]};

    # Flatten it
    my $flat_data = { iterate($data) };

    mark_utf8($flat_data);

    $flat_data;
}

#
# Help
#

sub help
{
    my %arg = @_;

    Pod::Usage::pod2usage(
        -verbose => $arg{ verbose },
        -exitval => $arg{ exitval } || 0,
    );
}