Let's consider this source code:

<div>
    <p><?php __('plain'); ?></p>
    <p><?php __('DATE \a\t TIME'); ?></p>
    <p><?php __("DATE \a\\t TIME"); ?></p>
    <p><?php __("DATE \\a\\t TIME"); ?></p>
    <p><?php __("FIELD\tFIELD"); ?></p>
</div>

Here we have only 3 different strings, but the php extractor finds these strings:

msgid "plain"
msgstr ""

msgid "DATE \\a\\t TIME"
msgstr ""

msgid "DATE \\a\\\\t TIME"
msgstr ""

msgid "DATE \\\\a\\\\t TIME"
msgstr ""

msgid "FIELD\\tFIELD"
msgstr ""

Hi guys.

I have one problem with your project and php ver. 5.5. When i used 5.3 or 5.4 - all works fine, but when i updated php to 5.5 - native translator (class GettextTranslator, vendor\gettext\gettext\src\GettextTranslator.php) stop working for me. If i use php Translator (vendor\gettext\gettext\src\Translator.php) - all is ok again, but i need exactly in native translator.

Test string:

Original string: Контакты
Translated string (en): Contacts
Translated string (de): Kontakte

Files structure:

index.php
data
  \- locales
     \- de
        \- LC_MESSAGES
           |- messages.mo
           \- messages.po
     \- en
        \- LC_MESSAGES
           |- messages.mo
           \- messages.po

Test code:

$translator = new \Gettext\GettextTranslator();
$translator->setLanguage('de')->loadDomain(
  'messages',
  'data/locales'
);
die($translator->gettext('Контакты'));

First test:

$ php -v
PHP 5.4.45 (cli) (built: Sep  2 2015 23:48:30)
Copyright (c) 1997-2014 The PHP Group
Zend Engine v2.4.0, Copyright (c) 1998-2014 Zend Technologies

Output:

Kontakte

All is ok. Second test:

$ php -v
PHP 5.5.33 (cli) (built: Mar  2 2016 15:19:21)
Copyright (c) 1997-2015 The PHP Group
Zend Engine v2.5.0, Copyright (c) 1998-2015 Zend Technologies

Output:

Контакты

My string is not translated. Test code was not changed.

Gettext installed by composer and source has no any changes by me:

{
  "require": {
    "php": ">=5.5",
    "gettext/gettext": "3.*@dev",
  },
}

Server software:

• Apache 2.4
• PHP 5.3, 5.4, 5.5
• Windows

Can you fix this or say me how can i do it?

Current Twig parser operates by

1. Transforming Twig as PHP code = tokenize() + parse() + render()
2. Using a PhpCode extractor (and its specific configuration about function name)

The disadvantages are:

• Twig rendered PHP code can be transformed/wrapped in call_user_func() making that gettext functions undetectable by PhpCode extractor. (See for example timber/timber#1753).
• Can't handle templates making use of custom Twig extensions (very common)

This patchset implements an extractor that:

• Parses Twig generated AST tree (= tokenize()+parse())
• Recursively iterates over node three to find function gettext calls.

Advantages:

• Operating sooner, at the AST level, Twig expressions like {{ __("foo", "domain") }} aren't yet wrapped.
• More robust because it directly iterates over the AST from Twig parser instead of looking at PHP source-code.
• Supports initialized $twig environment, thus supporting user-defined extensions?
• Possibly more efficient.

ToDos:

1. $options['twig'] being a static variable, we can NOT initialize it in a row with different parameters. Would get a LogicException: Unable to add function "action" as extensions have already been initialized. Solution: don't use static variables for this.

2. Translation/Notes comments are not yet supported.

3. Timber needs a tweak to avoid a fatal error: Add, to lib/Timber.php

  function add_action() { }
  function apply_filters() {}

The issue is that when loaded through composer Timber constructor forcefully calls init() which make use of WP functions.

When generating PO files, this line always prepends an empty translation:

msgid ""
msgstr ""
"Project-Id-Version: \n"
"Report-Msgid-Bugs-To: \n"
...

Because of this, if I have an empty text in my gettext or __ function, e.g. __(''), it will be replaced by Project-Id-Version: Report-Msgid-Bugs-To: ... and the entire header.

What is the purpose of this empty translation? Can it be removed?

I run a few tests again a PHP project with ~4300 translatable strings, comparing the results of xgettext and those of the PhpCode extractor.

I only found two problems:

1. xgettext (correctly) does not extracts strings like the me in __($avoid['me']). I think that the correct solution would be to do not extract strings with variables (since it does not have any sense), but gettext stops as soon as it find a variable (for instance from __('Stop at the variable'.$var.'!') it extracts Stop at the variable) - so I'd adopt this same approach for consistency.

2. Consider this example:

__(
  'line1'
 .' line2')

xgettext (correctly) extracts 'line1 line2', but the PhpCode extractor only extracts 'line1'.

I added some tests here to highlight the above problems.

@oscarotero Could you wait a few days before publishing a new release? IMHO some of the plural rules that we took from http://localization-guide.readthedocs.org/en/latest/l10n/pluralforms.html are not correct. I discovered two discrepancies but it seems that nobody is taking care of them, so IMHO that repository is not well maintained. I'm going to write a script that takes the rules from CLDR and converts them to gettext. It's not that easy, since CLDR rules refer to decimals and negative numbers, whereas gettext works with unsigned integers, but I'm working on it.

When adding string to the Translations instance, we call Translations::offsetSet that in turns calls Translation:is many times. If we load 5.000 strings, its inner cycle calls Translation:is 12.497.500 times (5000*(5000-1)/2). What about adding an option to skip this check?

With the code of this pull request, the checks are skipped by default for the .po and .mo importers (but it's configurable by writing Gettext\Extractors\Mo::$skipIntegrityChecksOnImport.

This is the "least-invasive" way to add unit IDs to the existing toolset, without potentially introducing problems with uniqueness (XLIFF doesn't require uniqueness of IDs in any way).

Let's say we have a test.po file like this:

msgid ""
msgstr ""
"Project-Id-Version: \n"
"POT-Creation-Date: \n"
"PO-Revision-Date: \n"
"Language: it\n"
"Language-Team: \n"
"Last-Translator: \n"
"Plural-Forms: nplurals=2; plural=(n != 1);\n"
"MIME-Version: 1.0\n"
"Content-Type: text/plain; charset=UTF-8\n"
"Content-Transfer-Encoding: 8bit\n"

msgid "1 child"
msgstr ""

msgid "1 child"
msgid_plural "2 children"
msgstr[0] ""
msgstr[1] ""

If we compile it with msgfmt we have:

test.po:17: duplicate message definition...
test.po:15: ...this is the location of the first definition
msgfmt: found 1 fatal error

So, we may not have in .po files two strings with same context+msgid but different plurals.

I think that we should apply two changes:

1. Revert https://github.com/oscarotero/Gettext/pull/46
2. Add an alternative method to Translations::find that does not take in account the plural form

Hi! I trying to create a sample project in PHP with gettext integration for localization, but I can't manage to create the right architecture in order to use it.

I tried the different examples in the README file but I don't see how to link everything together. Is there a complete and concrete example somewhere? I searched the Web and everything I found is kind of outdated.

I only want to have the possibility to generate po files from php files and use them for localization changing the language from a variable.

For example if I have a index.php file:

<?php

echo _("Hello world!")

How should I use php-gettext to extract this string into a po file in a structured locale folder and use it back to translate the page according to the specified language?

Thanks in advance for your help :smile: I'm kind of lost.

How it works - VueJs tempaltes basically are valid HTML. We extract template part (then extract dynamic attributes which are JS expressions and extract expressions from within template elements) and script part and parse them both as a regular JS.

In previous versions, one could use $translation->addReference() to store arbitrary key-value payload in Translation objects. Since the $line parameter wasn't typehinted, one could use arbitrary string keys, turning the references into a key-value store.

In more current versions, to which I am trying to upgrade, the API is $translation->getReferences()->add(), and the ->add() method requires that the $line parameter be an int. Code that used references as payload store can no longer work.

What is the recommended way to store arbitrary key-value payload in Translation objects now?

If I have a .po file with non-string lines in references, created by previous version of my code, will the current version of this library be able to load these?

Hello,

First of all thanks for the work on this library, highly appreciated :)

Second, I would like to be able to find translations by their ids. Currently it is only possible to search by translations. https://github.com/php-gettext/Gettext/blob/master/src/Translations.php#L169

May I propose a PR to implement the changes? It should be fairly easy to implement.

Cheers,

It is possible to make use of msgctxt context into the .po file definition? There is no clear method from native functions to get the translation of a context either. Or I haven't found.

The method Translattion::withContext ($s) { .. } / Translation:getContext ($s) { .. } would meet this?

Thanks,

Hi,

in version 4.6.1 library work fine with extracting translations from template literals in JS files, but after upgrade to 4.8 library doesnt extract translations from template literals for example new version doesnt extract "Foo" from string Text ${__("Foo")} bar but older version does.

Unfortunately I didn`t found solution or some clues to resolve this problem...

Thanks for response.

First of all, thank you for a library. Second, I don't know how to load multiple *.mo translation files and use them with textdomain context.

This is my code so far:

use Gettext\Translator;
use Gettext\Loader\MoLoader;

$lang = 'sk' // set by a side function
define('ROOT', '/path/to/root');

function __( string $string, string $textdomain = null ) {

    static $t = null;

    global $basic_lang;

    $translation = ROOT . '/languages/' . $lang . '.mo';

    if (!file_exists( $translation )) {
        return $string;
    }

    if ($t === null) {

        $loader = new MoLoader();
        $t = Translator::createFromTranslations( $loader->loadFile( $translation ) );
    }

    //return $textdomain ? $t->dgettext( $textdomain, $string ) : $t->gettext( $string );
    return $t->gettext( $string );

}

//echo __( 'Buy a ticket', 'reservation' ); // does not work right now, how to make it work?
echo __( 'Buy a ticket' ); // works

I guess that mode code samples would be helpful for others too.

As per https://github.com/wp-cli/i18n-command/pull/217, WP-Cli i18n-command still use v4 (which does not provide Twig scanner, among others) because it stick to package supporting PHP 5.6. What is itself a consequence of wp-cli requirements regarding WP core which tend to support old version of PHP for a long time.

I wonder why exactly, in October 2019, f6ca83d7d7c6 bumped the mimimum requirements to PHP 7.2? Isn't the codebase/testsuite compatible with PHP 5.6 at all? And if it is, is lowering the min version back to 5.6 feasible?