best practice for internationalizing white-labeled PHP code - php

Question

I have a code base that can be customized for different customers. This is handled, in most instances, by loading a class with constants specific to each customer, but with the same class name. Class name collision is managed, and, obviously, only one is loaded at a time.
class Customer () {
const CUST_NAME = 'Alpha Corp.';
}
class Customer () {
const CUST_NAME = 'Beta, Inc.';
}
For each customer, I also want the code to be internationalized. So, a program can draw its displayed text from a language-specific source, but also still be dynamic with regard to customer-specific information. Assume the customer-specific information is indifferent toward i18n. This reduces the number of files from one per customer per language, to just one per language. The desired result is to implement something similar to
// in separate language files:
$greeting = 'Hello. Welcome to ';
$greeting = 'Hola. Bienvenido a ';
// in front-end
$greeting = get $greeting in desired language;
echo $greeting, Customer::CUST_NAME, "\n";
The best solution I can devise to meet these requirements is an i18n class (or family of classes), which would handle multiple customers and/or multiple languages. However, it is execution-expensive to call methods, as opposed to string literals or constants. The methods would combine translations from constants or external sources, and combine them with values from the Customer class. (I am stuck with PHP 5.2 for now, so the niceties of heredoc for class properties/constants is unavailable.)
class i18n_en () {
const GREETING = 'Hello. Welcome to ';
static public function getGreeting () {
return self::GREETING . Customer::CUST_NAME;
}
}
Alternatively, I can write a script for a "template approach", in which I maintain a template file of the class, with text placeholders. The customer-specific files are generated at the time of customer creation from these templates. It would be easy to generate or re-generate files as needed, but I'm back to needing a separate file for each customer for each language. Thus, it doesn't meet my need as nicely.
Surely, I'm not the first to face this problem. Can anyone offer alternatives or best practices? Since the code will execute many times more than I'll create a new customer, I prefer run-time efficiency to ease of maintenance. Of course, a great solution will offer both.

Answer 1

There are two tried and true methods for PHP internationalization.
The most prevalent is to create huge language arrays like so:
define('MSG__GREETING', 1);
$lang['en_US'] = array(MSG__GREETING => 'Hello, and welcome to ');
$lang['fr_FR'] = array(MSG__GREETING => 'Bonjour, et bienvenue à ');
$selectedLang = 'fr_FR';
echo $lang[$selectedLang][MSG__GREETING] . 'Fhloston Paradise';
Unfortunately, this gets very tedious very quickly.
The ideal method, which I've used numerous times, is the accepted method of internationalization for Linux apps: il8n, via PHP's gettext extension.
With this method, you basically end up doing stuff like this:
setenv('LANG=fr_FR');
echo _('Hello, and welcome to ') . 'Fhloston Paradise';
and then in il8n files (called .po) you have each translation. It's actually much easier to maintain and extend over the long run, especially since you can just email your .po files to various translators and they just fill in the blanks, as it were... no coding skills necessary.
Here's a tutorial to get you started: http://kunststube.net/frontback/

Answer 2

Personally, when handling internationalization I have always used templates, specifically Smarty - http://www.smarty.net/ and create language based template. You can also use translation keys in a database, ex table with key, language, translation columns which can be cached this way you aren't calling tons of methods. Also, you can cache the output of all the translation methods in memory using memcached. Tons of ways. Hope some of these help a little.