mirror of
https://git.lyx.org/repos/lyx.git
synced 2024-11-24 10:40:48 +00:00
5439f4e554
While extracting names of a translator from the html page it can happen that the perl's HTML::Parser module splits the "text".
412 lines
9.4 KiB
Perl
Executable File
412 lines
9.4 KiB
Perl
Executable File
#! /usr/bin/env perl
|
|
# -*- mode: perl; -*-
|
|
|
|
# file getTranslators.pl
|
|
# This file is free software; you can redistribute it and/or
|
|
# modify it under the terms of the GNU General Public
|
|
# License as published by the Free Software Foundation; either
|
|
# version 2 of the License, or (at your option) any later version.
|
|
#
|
|
# This software is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
# General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU General Public
|
|
# License along with this software; if not, write to the Free Software
|
|
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
#
|
|
# Copyright (c) 2015 Kornel Benko <kornel@lyx.org>
|
|
|
|
use strict;
|
|
|
|
package IdentityParse;
|
|
use base "HTML::Parser";
|
|
use locale;
|
|
use POSIX qw(locale_h);
|
|
use feature 'fc';
|
|
|
|
setlocale(LC_CTYPE, "");
|
|
setlocale(LC_MESSAGES, "en_US.UTF-8");
|
|
$ENV{LC_ALL} = "C"; # set language for the output of command 'msgfmt'
|
|
|
|
# Prototypes
|
|
sub insertEntry(\%$$);
|
|
sub start($$$$$);
|
|
sub text($$);
|
|
sub end($$$);
|
|
sub actual_tag();
|
|
sub convertlang($);
|
|
sub scanPoFile($);
|
|
sub row_init(\%);
|
|
sub row_valid(\%);
|
|
sub phantomscript();
|
|
|
|
my $p = new IdentityParse;
|
|
$p->strict_names(0);
|
|
|
|
my $lyxurl = "http://www.lyx.org/I18n-trunk";
|
|
my $podir = "./po"; # script starts in top binary dir
|
|
my $jsfile = "$podir/lyxtranslators.js";
|
|
|
|
my %langs = ( # translation to make language spec unique
|
|
"pt" => "Portuguese",
|
|
"pt_BR" => "Portuguese (Brazilian)",
|
|
"ar" => "Arabic",
|
|
"bg" => "Bulgarian",
|
|
"ca" => "Catalan",
|
|
"cs" => "Czech",
|
|
"da" => "Danish",
|
|
"de" => "German",
|
|
"el" => "Greek",
|
|
"en" => "English",
|
|
"es" => "Spanish",
|
|
"eu" => "Basque",
|
|
"fi" => "Finnish",
|
|
"fr" => "French",
|
|
"gl" => "Galician",
|
|
"he" => "Hebrew",
|
|
"hu" => "Hungarian",
|
|
"ia" => "Interlingua",
|
|
"id" => "Indonesian",
|
|
"it" => "Italian",
|
|
"ja" => "Japanese",
|
|
"ko" => "Korean",
|
|
"sk" => "Slovak",
|
|
"nb" => "Norwegian (Bokmål)",
|
|
"nl" => "Dutch",
|
|
"nn" => "Norwegian (Nynorsk)",
|
|
"pl" => "Polish",
|
|
"ro" => "Romanian",
|
|
"ru" => "Russian",
|
|
"sr" => "Serbian",
|
|
"sv" => "Swedish",
|
|
"tr" => "Turkish",
|
|
"uk" => "Ukrainian",
|
|
"zh_CN" => "Chinese (simplified)",
|
|
"zh_TW" => "Chinese (traditional)",
|
|
"Simplified" => "Chinese (simplified)",
|
|
"Simplified Chinese" => "Chinese (simplified)",
|
|
"Traditional Chinese" => "Chinese (traditional)",
|
|
);
|
|
|
|
open(FO, '>', "$jsfile");
|
|
print FO &phantomscript();
|
|
close(FO);
|
|
|
|
my %status = ();
|
|
my @tag = (); # stack for active html-tags
|
|
my %page_row = (); # entries for mail, name pofile, language (in the actual row or po-file)
|
|
my @translation_entries = qw(language mail name pofile);
|
|
|
|
my %list = (); # collected list of rows, entry key is language
|
|
|
|
my $errors = 0;
|
|
if (open(my $fh, "phantomjs $jsfile|")) {
|
|
$p->parse_file($fh);
|
|
print "Parsed \"$lyxurl\"\n";
|
|
}
|
|
else {
|
|
$errors++; # cannot parse html file
|
|
print "ERROR: Program \"phantomjs\" to parse \"$lyxurl\" could not be executed\n";
|
|
}
|
|
|
|
if (opendir(DI, $podir)) {
|
|
my $po_count = 0;
|
|
while (my $po = readdir(DI)) {
|
|
if ($po =~ /\.po$/) {
|
|
my $res = &scanPoFile("$po");
|
|
if ($res == 0) {
|
|
print "No valid entry found in \"$po\"\n";
|
|
$errors++;
|
|
}
|
|
else {
|
|
$po_count += $res;
|
|
}
|
|
}
|
|
}
|
|
closedir(DI);
|
|
if ($po_count < 1) {
|
|
print "ERROR: No correct po-files in directory \"$podir\" found\n";
|
|
$errors++;
|
|
}
|
|
else {
|
|
print "Found $po_count po-files with valid translator entry\n";
|
|
}
|
|
}
|
|
else {
|
|
print "Directory for po-files ($podir) missing\n";
|
|
$errors++; # PO directory not found, so cannot check
|
|
}
|
|
|
|
for my $lang (sort keys %list) {
|
|
for my $rentry (@{$list{$lang}}) {
|
|
my $prefix = sprintf("(%03d%) ", "$rentry->{fract}");
|
|
if (defined($rentry->{error})) {
|
|
$errors++;
|
|
$prefix .= sprintf("%-24s", "$rentry->{error}:");
|
|
}
|
|
else {
|
|
$prefix .= sprintf("%24s", "");
|
|
}
|
|
my $msg = sprintf("%-24s%-10s", "$lang:", "$rentry->{po},");
|
|
print "$prefix$msg mail = \"$rentry->{name}\" <$rentry->{mail}>\n";
|
|
}
|
|
}
|
|
|
|
if ($errors > 0) {
|
|
exit(1);
|
|
}
|
|
else {
|
|
exit(0);
|
|
}
|
|
###################################################################
|
|
# Insert collected row values in %list
|
|
sub insertEntry(\%$$)
|
|
{
|
|
my ($rrow, $cycle, $fract) = @_;
|
|
|
|
# Convert mangled mail
|
|
$rrow->{mail} =~ s/ pound /\@/;
|
|
$rrow->{mail} =~ s/ dot /\./g;
|
|
$rrow->{mail} =~ s/ underscore /_/g;
|
|
my %entry = ();
|
|
my $language = $rrow->{language};
|
|
$entry{mail} = $rrow->{mail};
|
|
$entry{name} = $rrow->{name};
|
|
$entry{po} = $rrow->{pofile};
|
|
$entry{cycle} = $cycle;
|
|
# Check, if entry already exists
|
|
if (! defined($list{$language})) {
|
|
$list{$language} = [];
|
|
}
|
|
my $found = 0;
|
|
my $empty = 1;
|
|
my $name1 = undef;
|
|
for my $rentry (@{$list{$language}}) {
|
|
$empty = 0;
|
|
if ($cycle == 2) {
|
|
if (! defined($rentry->{fract})) {
|
|
$rentry->{fract} = $fract;
|
|
}
|
|
}
|
|
next if (fc($rentry->{mail}) ne fc($rrow->{mail})); # char case does not matter in mail strings
|
|
next if ($rentry->{po} ne $rrow->{pofile});
|
|
$name1 = $rentry->{name};
|
|
$found = 1;
|
|
last;
|
|
}
|
|
if ($cycle == 1) {
|
|
push(@{$list{$language}}, \%entry);
|
|
}
|
|
else {
|
|
$entry{fract} = $fract;
|
|
if ($empty) {
|
|
if ($fract > 40) {
|
|
$entry{error} = "Missing in page";
|
|
push(@{$list{$language}}, \%entry);
|
|
}
|
|
}
|
|
elsif (! $found) {
|
|
$entry{error} = "Different mail in po";
|
|
push(@{$list{$language}}, \%entry);
|
|
}
|
|
else {
|
|
# found, but maybe incorrect name?
|
|
if ($name1 ne $rrow->{name}) {
|
|
$entry{error} = "Different name in po";
|
|
push(@{$list{$language}}, \%entry);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
#######################################################################
|
|
# Routines called from parse_file(): start(), text(), end().
|
|
sub start($$$$$)
|
|
{
|
|
my ($self, $tag, $attr, $attrseq, $origtext) = @_;
|
|
|
|
push(@tag, $tag);
|
|
if ($tag eq "tr") { # new table row
|
|
&row_init(\%page_row);
|
|
for my $k (keys %status) {
|
|
$status{$k} = 0;
|
|
}
|
|
}
|
|
$status{"Tag_" . $tag} = $status{"Tag_" . $tag} + 1;
|
|
|
|
if ($tag eq "a") {
|
|
if (defined($attr->{class})) {
|
|
if ($attr->{class} eq "urllink") {
|
|
if ($status{Tag_td} == 6) {
|
|
if ($attr->{href} =~ /^mailto:(.*)$/) {
|
|
$page_row{mail} = $1;
|
|
$page_row{mail} =~ s/\%20/ /g;
|
|
}
|
|
}
|
|
elsif ($status{Tag_td} == 1) {
|
|
if ($attr->{href} =~ /f=po\/([a-z][a-z](_[A-Z][A-Z])?\.po)$/) {
|
|
$page_row{pofile} = $1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
sub text($$)
|
|
{
|
|
my ($self, $text) = @_;
|
|
|
|
if ($status{Tag_td} == 1) {
|
|
if (&actual_tag() eq "a") {
|
|
if ($text =~ /^[A-Z][a-z]+( .+)?$/) {
|
|
$page_row{language} = &convertlang($text);
|
|
}
|
|
}
|
|
}
|
|
if ($status{Tag_td} == 6) {
|
|
if (&actual_tag() eq "a") {
|
|
$page_row{name} .= $text; # '.=' because text can be splitted
|
|
}
|
|
elsif (&actual_tag() eq "td") { # name without associated e-mail
|
|
$page_row{name} .= $text;
|
|
}
|
|
}
|
|
}
|
|
|
|
sub end($$$)
|
|
{
|
|
my ($self, $tag, $origtext) = @_;
|
|
|
|
while (my $t = pop(@tag)) {
|
|
last if ($t eq $tag);
|
|
}
|
|
if ($tag eq "tr") {
|
|
# check row entry for completeness
|
|
return if (! &row_valid(\%page_row));
|
|
&insertEntry(\%page_row, 1);
|
|
}
|
|
}
|
|
|
|
sub actual_tag()
|
|
{
|
|
return undef if (@tag == 0);
|
|
return($tag[$#tag]);
|
|
}
|
|
|
|
sub convertlang($)
|
|
{
|
|
my ($ilang) = @_;
|
|
$ilang =~ s/\s+$//;
|
|
if (defined($langs{$ilang})) {
|
|
return($langs{$ilang});
|
|
}
|
|
else {
|
|
return($ilang);
|
|
}
|
|
}
|
|
|
|
sub scanPoFile($)
|
|
{
|
|
my ($pofile) = @_;
|
|
my %po_row;
|
|
my ($translated, $fuzzy, $untranslated) = (0, 0, 0);
|
|
|
|
if (open(FM, "msgfmt -c --statistics $podir/$pofile 2>&1 |")) {
|
|
while (my $l = <FM>) {
|
|
if ($l =~ s/^(\d+)\s+translated messages.\s*//) {
|
|
$translated = $1;
|
|
}
|
|
if ($l =~ s/^(\d+)\s+fuzzy translations.\s*//) {
|
|
$fuzzy = $1;
|
|
}
|
|
if ($l =~ s/^(\d+)\s+untranslated messages//) {
|
|
$untranslated = $1;
|
|
}
|
|
}
|
|
close(FM);
|
|
}
|
|
return 0 if ($translated == 0);
|
|
my $fract = int(($translated * 100)/($translated+$fuzzy+$untranslated));
|
|
if (open(FI, "$podir/$pofile")) {
|
|
&row_init(\%po_row);
|
|
$po_row{pofile} = $pofile;
|
|
my $i = 0;
|
|
while (my $l = <FI>) {
|
|
last if ($l =~ /^"[A-Z].*:/);
|
|
}
|
|
my $inserted = 0;
|
|
while (my $l = <FI>) {
|
|
last if ($l !~ /^"/);
|
|
chomp($l);
|
|
if ($l =~ s/^"Last-Translator:\s//) {
|
|
while ($l !~ />\\n"$/) {
|
|
$l =~ s/"$//;
|
|
my $extraline = <FI>;
|
|
chomp($extraline);
|
|
$extraline =~ s/^"//;
|
|
$l .= $extraline
|
|
}
|
|
if ($l =~ /^([^<]*)<([^>]*)>/) { # allow empty mail
|
|
$po_row{mail} = $2;
|
|
($po_row{name} = $1) =~ s/\s+$//;
|
|
$i += 2;
|
|
}
|
|
}
|
|
elsif ($l =~/^"Language:\s*([^\\]+)\\n/) {
|
|
$po_row{language} = &convertlang($1);
|
|
$i += 1;
|
|
}
|
|
if (&row_valid(\%po_row)) {
|
|
&insertEntry(\%po_row, 2, $fract);
|
|
$inserted++;
|
|
last;
|
|
}
|
|
}
|
|
close(FI);
|
|
return($inserted);
|
|
}
|
|
else {
|
|
return(0);
|
|
}
|
|
}
|
|
|
|
###########################################################
|
|
# handling of row entries
|
|
|
|
sub row_init(\%)
|
|
{
|
|
my ($rrow) = @_;
|
|
%{$rrow} = ();
|
|
$rrow->{mail} = ""; # Allow for empty mail
|
|
}
|
|
|
|
sub row_valid(\%)
|
|
{
|
|
my ($rrow) = @_;
|
|
for my $k (@translation_entries) {
|
|
return 0 if (! defined($rrow->{$k}));
|
|
}
|
|
return(1);
|
|
}
|
|
|
|
# used by phantomjs command to output the refered html page
|
|
sub phantomscript()
|
|
{
|
|
return "var page = require(\"webpage\").create();
|
|
var url = \"$lyxurl\";
|
|
page.open(url,
|
|
function (status) {
|
|
var f = function () {
|
|
var html = page.evaluate(function () { return document.documentElement.innerHTML });
|
|
console.log(html);
|
|
phantom.exit();
|
|
};
|
|
setTimeout(f, 5000);
|
|
}
|
|
);
|
|
";
|
|
}
|