I'm in the process of rewriting one of my games and I want to add multiple-language support.
Currently I'm making one big file for each language with one big array for each piece of text that I want to display.

Is this a bad approach?
Should I split the array into many, smaller arrays (for each site one)?
Or should I do it via database?

yeah, I have one array which is the same for every language, just the values are different.

I speak fluently swiss-german (duh), german (again, duh!), english and french. I can speak some italian, but not as much as I'd like. Then I can speak PHP und JavaScrip

To translate the game into more languages I thought about letting the users do it. I'll provide either the english (needs to be written) or the german (base) text and let the users work from there on.

Sorry for dragging out an old topic, but I did a search and this is the best place.

The method that I'm currently using is the data is input into a database with an added 'lang' column giving the 2 character language code (I'm not messing with accents at the moment, but easy enough to change if necessary).

It took me about a day to convert all of the existing data into utf-8 format and a few moments to install the mb_string extension but now all the data can be stored in a single location with quick indexing / searching.

My search was originally to see if forum users were offering to translate games into other languages but I figured I'd share another technique

The question was how to store the data. I didn't know if anyone had a better method. I'm actually reading the http accept header which is supposed to indicate the user's preferred language choices. I default to english if none of the requested languages are supported.

And, yes, I sanitize the input because it is possible to modify the request headers.

Now I'm really late answering this, but I thought I'd give it shot since it might be of help to others.

First of all, the one-file-per-lang-array approach sure works, but it's nothing but trouble.
The size isn't the problem, but the maintainability is. Take it from me, I've done it once and I'll never do it again.

What you should look at is one of the following:
either you use gettext. (http://php.net/gettext) which is a nice standard that is used widely online. There's multiple GUI tools out there which makes it easier to edit/change existing collections. Although it might be a bit tricky to implement if you have done it before.

if you use XML with XSL(t) as templating then gettext doesn't do the the trick. Basically what you do is to save a xml-file for each language with a standard way to store a phrase/word/whatever. In PHP (if you using that) you simply load the language xml file, and append it to your XML document.

Yours,
ninto

Nope, you summed it up pretty good.

The only thing I'd add is that INI files can often be better than .txt and XML, since XML is kinda bulky and .txt's force you to hardcode the array numbers everywhere.

if you do use txt's though, use something like this instead of opening it...
<?php
// + LineArray ( FileName : string ) : string array
public function LineArray( $FileName )
{
return array_map( 'rtrim' , file( $FileName ) ) ;
} 
?>

(All I did was copy that from my framework(ish) project, but it should be able to give you a line array with the least amount of overhead that will work in the way that Mdshare posted.

Nearly more than 5 years all of my new code has multilang capability.

This was my current implementation. I had extensions system in my code. Each extension has own language file per language.

Like

language_en.php
<?php
//--- Language: en
 $_lang = array (
&#160; '_extension_system_' => '_extension_system_',
&#160; '_plugin_system::checkin_' => '_plugin_system::checkin_',
&#160; '_plugin_system::extension_manager_' => '_plugin_system::extension_manager_',
&#160; '_plugin_system::index_' => '_plugin_system::index_',
&#160; '_plugin_system::user_manager_' => '_plugin_system::user_manager_',
&#160; '_user_email_' => '_user_email_',
&#160; '_user_login_' => '_user_login_',
&#160; '_user_logout_' => '_user_logout_',
&#160; '_user_name_' => '_user_name_',
&#160; '_user_password_' => '_user_password_',
);
?>


Usage was simple

<div>
<?=$GLOBALS['lang']['_plugin_system::checkin_']?>
</div>

I had some util functions to check all my php files and extract all $GLOBALS['lang']['_whatever_'] (or $this->lang['_whatever_']); ) data, also I include old language file and merge all keys. Then write array like native php file (example was auto generated file).

Plain dead simple. Plus opcode compatible. one include than everthing in memory.

One big array for every language leads too much memory usage. One common array for language needs and load only one language per request.

Regards


And also. Using XML or function based methods just increase your cpu load.


And also (sorry for that)

Load your language keys in GLOBAL space (with $GLOBALS or public static method) you have to access them easly. Load them when you need. And if they where too many language file. First concat them, then include one file.

And another common practice was using CONSTANTS. I did not go that way.

- are there any tricks/tips for increasing the maintainability for variable approach?

If you use INI's for storing the array, you have the array keys as strings instead of ints.



EN.ini
Username=Username
Password=Password
Help=Help

DE.ini
Username=Benutzername
Password=Paßwort
Help=Hilfe

<?php 
$text = parse_ini_file( 'DE.ini' , false );
?>

<?php echo $text['Username']; ?> <input type="text" ...
<?php echo $text['Password']; ?> <input type="text" ...


   

(however, if it's a huge project the array might get too bulky)

Damn language barrier.

As you know, I'm very over zealous about performance. 

Problem is when your code starts to include too many files, performance starts to drop. And of course using an opcode cache (APC, Xcache etc) will solve problem.

And also, my current programming model has extensions/plugins model. On front page I may load 20 different plugins from 6 different extensions. that means 6 different language file include (per extension). Concatting those files into one file and include one language file, help performance (of course under heavy load).

This approach reduces my I/O overload. After moving to MVC model (no not something like CI or Cake, mine just seperated files for plugin). Each plugin loads 4 files for programming + 1 one for views. And you know 20 * 4 = 80 was too much (of course in my twisted mind).

Also, My game plans always on one big server.


One of my current implementation.

<?php
/**
* k5_get_language_file#
* gets Requested language file
*
* Variables: $o['mbase']:&#160; modlue base
* $o['lang_key']: current language key (en, tr, de etc)
*
* @author Sancar Saran <saran@delifisek.net>
* @version 1.0
*/
function k5_get_language_file($o) {
$file = $o['mbase'].'language/language_'.$o['lang_key'].'.php';
if(is_file($file)) : require_once($file); endif;
if(isset($_lang) and is_array($_lang)) : $GLOBALS['k5']['lang'] = array_merge($GLOBALS['k5']['lang'],$_lang); endif;
}
?>

and

Re: Barrikor

And you have to pay parse_ini_file for each request.

Regards

Just a thought... Awhile back I spent a LOT of time and energy into getting one of my sites multi-language. In the end I realized a couple of things.

First that for the size of my game it was a stupid move because the number of players from other countries didn't make a large percentage, and even after there were 5 languages available in our post popular non-us audiences, there was only a little growth from those places. In the end the game would have been healthier had I just worked on other features instead.

Second, note that it does add a layer of complication across the board for all of your future coding. You will be paying for it in small amounts forever. Not that it's a problem if you have enough users to support it, but if nobody ends up using it, it can be a permanent slowdown for your development, albiet small.

Third, after the fact someone pointed out to me that there were already a number of translation libraries out there and that although I had some good coding experience in the making of ours, it would have saved a lot of time not to re-invent the wheel.

http://pear.php.net/package/Translation2

Just my three cents...

Not sure what you mean by library or function, a bit if both I guess. Basically I went with the approach Zeggy first mentioned. I went with arrays for each page.

Basically the setup that I ended up with stored the language mappings in an associative array. The array was then serialized and stored in a flat file under a subdirectory for each language. A function was then called to load the language array on the page, first loading the english version in case the translation was incomplete, then overwriting it with the other language.

So each page had something like:

$langArr = loadLang($thisPage);

After that each string to be output to the page has to come from the array, so to echo hello would be something like this:

echo $langArr['hello'];

Note that the 'hello' value could be anything, probably a paragraph or even a basic HTML segment.

I also built a control panel to allow language admins to edit the mappings. Once that was done the upkeep is just in dealing with the extra variables on the page and keeping the language files up to date. It's not much, but when you are a single developer only working part time on the game, it is a level of headache. I didn't mean to make it sound like a complete nightmare once it is setup, just personally I wish I had spent the time developing either a new game or adding something that would have been of value to the majority of my players instead.

As for whether I can think of a better approach, nope not really. Using something that is already tested and proven will put you a leg forward for sure, but it is something you will always have to maintain.

(in theory, we still lack someone who went that route and could confirm how it works in practice...)

I'd like to mention that I have both used and implemented gettext.
My main point about this approach is that there's lots of GUI tools out there making i18n less stressful on me, as the coder.

I'm a bit short on time, but when I get home I could perhaps make a small guide if you want to?

Yours,
ninto

I recently tried doing both array based translation and gettext. If I had to I could live with arrays method, but when it comes to ease of use gettext wins without one second of doubt Not sure about performance through.