php - Transliterate any convertible utf8 char into ascii equivalent

Question

Is there any good solution out there that does this transliteration in a good manner?

I've tried using iconv(), but is very annoying and it does not behave as one might expect.

• Using //TRANSLIT will try to replace what it can, leaving everything nonconvertible as "?"
• Using //IGNORE will not leave "?" in text, but will also not transliterate and will also raise E_NOTICE when nonconvertible char is found, so you have to use iconv with @ error suppressor
• Using //IGNORE//TRANSLIT (as some people suggested in PHP forum) is actually same as //IGNORE (tried it myself on php versions 5.3.2 and 5.3.13)
• Also using //TRANSLIT//IGNORE is same as //TRANSLIT

It also uses current locale settings to transliterate.

WARNING - a lot of text and code is following!

Here are some examples:

$text = 'Regular ascii text + ????? + ??ü? + é?ě??? + ??€ + $ + ? + @';
echo '<br />original: ' . $text;
echo '<br />regular: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
//> regular: Regular ascii text + ????? + ???ss + ?????? + ae?EUR + $ + ? + @

setlocale(LC_ALL, 'en_GB');
echo '<br />en_GB: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
//> en_GB: Regular ascii text + cczs? + aouss + eeeeee + ae?EUR + $ + ? + @

setlocale(LC_ALL, 'en_GB.UTF8'); // will this work?
echo '<br />en_GB.UTF8: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
//> en_GB.UTF8: Regular ascii text + cczs? + aouss + eeeeee + ae?EUR + $ + ? + @

Ok, that did convert ? ? ? ? ? ü ? é ? ě ? ? ? and ?, but why not ? and ??

// now specific locales
setlocale(LC_ALL, 'hr_Hr'); // this should fix croatian ?, right?
echo '<br />hr_Hr: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
// wrong > hr_Hr: Regular ascii text + cczs? + aouss + eeeeee + ae?EUR + $ + ? + @

setlocale(LC_ALL, 'sv_SE'); // so this will fix swedish ??
echo '<br />sv_SE: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
// will not > sv_SE: Regular ascii text + cczs? + aouss + eeeeee + ae?EUR + $ + ? + @

//this is interesting
setlocale(LC_ALL, 'de_DE');
echo '<br />de_DE: ' . iconv("UTF-8", "ASCII//TRANSLIT", $text);
//> de_DE: Regular ascii text + cczs? + aeoeuess + eeeeee + ae?EUR + $ + ? + @
// actually this is what any german would expect since ? ? ü really is same as ae oe ue

Lets try with //IGNORE:

echo '<br />ignore: ' . iconv("UTF-8", "ASCII//IGNORE", $text);
//> ignore: Regular ascii text + + + + + $ + + @
//+ E_NOTICE: "Notice: iconv(): Detected an illegal character in input string in /var/www/test.server.web/index.php on line 49"

// with translit?
echo '<br />ignore/translit: ' . iconv("UTF-8", "ASCII//IGNORE//TRANSLIT", $text);
//same as ignore only> ignore/translit: Regular ascii text + + + + + $ + + @
//+ E_NOTICE: "Notice: iconv(): Detected an illegal character in input string in /var/www/test.server.web/index.php on line 54"

// translit/ignore?
echo '<br />translit/ignore: ' . iconv("UTF-8", "ASCII//TRANSLIT//IGNORE", $text);
//same as translit only> translit/ignore: Regular ascii text + cczs? + aouss + eeeeee + ae?EUR + $ + ? + @

Using solution of this guy also does not work as wanted: Regular ascii text + YYYYY + aous + eYYYeY + aoY + $ + ? + @

Even using PECL intl Normalizer class (which is not awailable always even if you have PHP > 5.3.0, since ICU package intl uses may not be available to PHP i.e. on certain hosting servers) produces wrong result:

echo '<br />normalize: ' .preg_replace('/p{Mn}/u', '', Normalizer::normalize($text, Normalizer::FORM_KD));
//>normalize: Regular ascii text + cczs? + aou? + eeeeee + ??€ + $ + ? + @

So is there any other way of doing this right or the only proper thing to do is to do preg_replace() or str_replace() and define transliteration tables yourself?

// appendix: I have found on ZF wiki debate from 2008 about proposal for Zend_Filter_Transliterate but project was dropped since in some languages it is not possible to convert (i.e. chinese), but still for any latin- and cyrilic-based language IMO this option should exist.

Answer 1

The toAscii() function of PatchworkUtf8 does exactly this, see:

https://github.com/nicolas-grekas/Patchwork-UTF8/blob/master/src/Patchwork/Utf8.php

It leverages iconv and intl's Normalizer to remove accents, split ligatures and do many other generic transliterations.