Multilingual site with database and PHP

Question 1

As I mentioned in my comment above, I would recommend putting the language in the URL mainly for search engine optimization. Admittedly, it can take a bit more work to set up, but I feel the benefits outweigh the cost.

For example, I have a hotel website with English, Spanish, Chinese and French. The URL structure is like this:

/    <- Main page in English
/es/ <- Main page in Spanish
/zh/ <- Main page in Chinese
/fr/ <- Main page in French

Then for sub-pages I do similarly:

/pagename/
/es/pagename/
/zh/pagename/
/fr/pagename/

Then, this is how I redirect the other languages to the correct scripts:

# Spanish
RewriteRule ^es/(.*/)?$ $1/index.php?lang=es [NC,L]

# French
RewriteRule ^fr/(.*/)?$ $1/index.php?lang=fr [NC,L]

# Chinese
RewriteRule ^zh/(.*/)?$ $1/index.php?lang=zh [NC,L]

This works for rewriting the index.php. For other things, I specify them explicitly, usually using a "friendly" URL. For example the "thank you" page for our Contact Us page:

# Contact Us thank you page
RewriteRule ^([a-z_]+)?/?contact/thankyou$ /contact/thankyou.php?lang=$1 [L]

I also used to do the following to rewrite URLs that contained parameters, but I think doing it like this might be deprecated, not too sure (I have it in the code, but since I don't use parameters, it doesn't get triggered):

RewriteCond %{query_string} ([^=]*)=(.*)
RewriteRule ^es/(.*)$ $1/index.php?lang=es&%1=%2 [NC,L]

However, the hotel offers tours and we have a main /tours/ page and I wanted to have a friendly URL for each of the individual tours, like /tours/56/waterfall-hike (with the tour ID and a slug from the tour name), so this handles the rewriting of the tours:

# Rewrite tours
RewriteRule ^([a-z_]+)?/?tours/([0-9]+)/reserve$ /tours/tour_reserve.php?lang=$1&id=$2 [L]
RewriteRule ^([a-z_]+)?/?tours/([0-9]+)/(.*) /tours/tour_text.php?lang=$1&id=$2&urlstr=$3 [L]

# Tours thank you page
RewriteRule ^([a-z_]+)?/?tours/thankyou$ /tours/thankyou.php?lang=$1 [L]

I just need to verify with PHP that the slug string provided is correct and if not do a 301 redirect to the correct URL based on the ID. I use this to calculate it:

function getTourURL($tour_name) {
    // Transliterate non-ascii characters to ascii
    $str = trim(strtolower($tour_name));
    $str = iconv('UTF-8', 'ASCII//TRANSLIT', $str);

    // Do other search and replace
    $searches = array(' ', '&', '/');
    $replaces = array('-', 'and', '-');
    $str = str_replace($searches, $replaces, $str);

    // Make sure we don't have more than one dash together
    $str = preg_replace("/(-{2,})/", "-", $str );

    // Remove all invalid characters
    $str = preg_replace("/[^A-Za-z0-9-]/", "", $str );

    // Done!
    return $str;
}

Then the only other difficult thing is switching languages without just redirecting them back to the home page for that language (yuck!). This is how I did it:

// First define the languages used:
$langs = array(
    'en' => array(
        'lang_code' => 'en_US',
        'locale' => 'en_US.UTF-8',
        'base_url' => '/',
        'name' => 'English',
    ),
    'es' => array(
        'lang_code' => 'es_MX',
        'locale' => 'es_MX.UTF-8',
        'base_url' => '/es/',
        'name' => 'Español',
    ),
    'fr' => array(
        'lang_code' => 'fr_FR',
        'locale' => 'fr_FR.UTF-8',
        'base_url' => '/fr/',
        'name' => 'Français',
    ),
    'zh' => array(
        'lang_code' => 'zh_CN',
        'locale' => 'zh_CN.UTF-8',
        'base_url' => '/zh/',
        'name' => '中国的',
    ),
);
define('LOCALE', $_GET['lang']);

// Then get the path to the current script after the language code
$path = (LOCALE == 'en') ? '' : LOCALE.'/';

$parsed_url = parse_url($_SERVER['REQUEST_URI']);
$path = ltrim(str_replace('/'.LOCALE , '', $parsed_url['path']), '/');

define('REDIRECT_SCRIPT', $path);

// Then I put all the links into an array to be displayed in the menu
foreach ($langs as $lang => $arr) {
    if ($lang == LOCALE) {
        continue;
    }
    $link = (isset($lang_override[$lang]))
        ? $lang_override[$lang]
        : $arr['base_url'] . REDIRECT_SCRIPT;
    $lang_subs[] = array(
        'name' => '<div class="'.$lang.'"></div> '.$langs[$lang]['name'],
        'link' => $link,
        'lang' => $lang,
    );
}

This probably won't work for you out of the box, but should at least give you a starting point. Try a print_r($lang_subs); to see what it contains and adapt it to your site's design.

Question 2

Maybe I'm old fashioned but I've always had a lang folder with files for each languages (lang.en.php, lang.fr.php, lang.es.php, and so on).

In each file I've got an array, like this one:

$langarray = array("login" => "...", "connect" => "...", "logout" => "...");

Eventually with real stuff in them... works better :}

And then, depending on the $_SESSION variable, I include the right file. You can even stock "en" and include lang.'.$_SESSION['lang'].'.php.

It seems slightly better not having to query SQL for that kind of thing but less easy to maintain. I think this is the type of problem where nobody's wrong.

Question 3

Yes, saving the language code in the session is better than constantly passing around a parameter.

No, storing the translated text in the session doesn't make sense because then you are storing the same text over and over in memory per user. Better to implement database caching or have a PHP file to include for the translation table than to store it in the session.

Instead of making up numeric codes for languages, you really should use the standard letter abbreviations that are part of the HTML spec. Browsers send preferred languages in order of preference as a header called Accept-Language. To avoid making the user click a language choice, you could read that list from the header and iterate through it until you find the first language you support. But always good to give the user some manual way to change it.

Zend framework has some functions for dealing with languages but I've never used it.