Multibyte for ucfirst function

[gnutix]

Description

In the same vein as #12459, I was wondering if it wouldn't make sense to implement an mb_ucfirst function ?

See https://stackoverflow.com/questions/2517947/ucfirst-function-for-multibyte-character-encodings/58915632#58915632 for examples.

What do you think ? /cc @youkidearitai

[youkidearitai]

Hi. @gnutix !

I have a question. ucfirst function is convert only alphabet. However, that examples seems use mb_strtoupper and mb_strtolower. I understand that ucfirst is convenience to English. Is that other languages convenience to mb_strtoupper and mb_strtolower from first code point ?

Personally, I think make sense to implement mb_ucfirst function (of course, I think required an RFC).

[gnutix]

I don't think I can answer your question (I'm not even sure I really understand it). Could you maybe provide some example of what you're talking about ?

[youkidearitai]

Could you maybe provide some example of what you're talking about ?

Please see Examples #2 of mb_strtoupper.

<?php
$str = "Τάχιστη αλώπηξ βαφής ψημένη γη, δρασκελίζει υπέρ νωθρού κυνός";
$str = mb_strtoupper($str, 'UTF-8');
echo $str; // Prints ΤΆΧΙΣΤΗ ΑΛΏΠΗΞ ΒΑΦΉΣ ΨΗΜΈΝΗ ΓΗ, ΔΡΑΣΚΕΛΊΖΕΙ ΥΠΈΡ ΝΩΘΡΟΎ ΚΥΝΌΣ
?>

I'm talking about "Τάχιστη αλώπηξ βαφής ψημένη γη, δρασκελίζει υπέρ νωθρού κυνός" of first word "T". I said effect to ucfirst if not latin text.

Anyway, I discuss to Internals. Just a moment, please.

[tstarling]

I just wanted to mention that title case is generally a better choice for the first character of a word than upper case. In T292552 we discovered that Unicode 11 upper case maps Georgian characters to their Mtavruli equivalents, but Mtavruli characters are never used to capitalise the first letter of a word or sentence, so applying Unicode 11 upper case to the first character of a Georgian string looks ridiculous. There's no title case mapping for those characters, so converting the first letter to title case produces a better result for Georgian.

For more familiar characters, like ASCII letters, there is no difference between title case and upper case mappings.

Ligatures like lj have both title case (Lj) and upper case (LJ) mappings. Title case is better for initial capitalisation, except that ideally you don't want to map the upper case form down to its title case form, LJ -> Lj, so that mb_ucfirst() on an all-caps string stays all caps.

[youkidearitai]

@tstarling Thanks for advice. I see. Would I implement like below if it's PHP code?

<?php
$encoding = "UTF-8";
$str = "abc";
$str = mb_convert_case(mb_substr($str, 0, 1, $encoding), MB_CASE_TITLE, $encoding) . mb_substr($str, 1, null, $encoding);
var_dump($str);
?>

[tstarling]

$str = mb_convert_case(mb_substr($str, 0, 1, $encoding), MB_CASE_TITLE, $encoding) . mb_substr($str, 1, null, $encoding);

That's basically what I did, but it converts uppercase ligatures to title case rather than preserving them. ucfirst() preserves all-caps strings, so by analogy, it makes sense for mb_ucfirst() to preserve all-caps ligatures. To do that, you would need a new case conversion mode with a new table in unicode_data.h which is the same as title case except that it omits uppercase ligatures.

Maybe that would be overkill. At Wikimedia we're doing something similar to ucfirst() on about 400 million natural language page titles across 320 languages, which could be considered a worst case scenario for complexity. And yet we're not actually hitting a problem with uppercase ligatures. Nobody is using them.

Documentation of mb_ucfirst() will be tricky. ucfirst() says it will "make a string's first character uppercase" which is not a correct description of what I'm proposing here. But at least what I'm proposing does not mangle texts in the Georgian script.

[youkidearitai]

@tstarling Thanks for further asking! I'll investigate.
Sorry edited: Point to $encoding to UTF-8, I forgot changed default character encoding.

[youkidearitai]

ucfirst() preserves all-caps strings, so by analogy, it makes sense for mb_ucfirst() to preserve all-caps ligatures. To do that, you would need a new case conversion mode with a new table in unicode_data.h which is the same as title case except that it omits uppercase ligatures.

Hmm... This implementation is seems to need allocate giant memory.

I reading 5 Implementation Guidelines of unicode.org, Moreover I just have one idea, use to php_unicode_totitle_raw in php_unicode.c.

If any wrong, feel free to comment.

[youkidearitai]

Memo: Draft of an RFC

PHP RFC: Multibyte for ucfirst, lcfirst functions, mb_ucfirst mb_lcfirst

• Version: 0.1
• Date: 2024-01-14
• Author: Yuya Hamada(https://github.com/youkidearitai), youkidearitai@gmail.com
• Status: Draft
• First Published at: TBD

Introduction

PHP does not have a multibyte equivalent of ucfirst, lcfirst functions. It is possible to get close enough behavior below:

function mb_ucfirst(string $str, ?string $encoding = null): string
{
    return mb_strtoupper(mb_substr($str, 0, 1, $encoding), $encoding) . mb_substr($str, 1, null, $encoding);
}

function mb_lcfirst(string $str, ?string $encoding = null): string
{
    return mb_strtolower(mb_substr($str, 0, 1, $encoding), $encoding) . mb_substr($str, 1, null, $encoding);
}

However adding a pre-built functions to do this will implobe the readability and clarify of PHP code. And it will standardize how it is done as it can be tricky.

Proposal

Add mb_ucfirst function, mb_lcfirst function.

function mb_ucfirst(string $string, ?string $encoding = null): string

function mb_lcfirst(string $string, ?string $encoding = null): string

According to research about Unicode, some (natural) lanugage doesn't may expected behavior, please deal with it in userland if any wrong. Because (natural) languages is a lot of exists, it is difficult to deal in mbstring.

Backward Incompatible Changes

This could break a function existing in userland with the same name.

Proposed PHP Version(s)

next PHP 8.x

RFC Impact

To SAPIs

To SAPIs Will add the aforementioned functions to all PHP environments.

To Existing Extensions

Adds mb_ucfirst(), mb_lcfirst() to the mbstring extension.

To OpCache

No effect.

New Constants

No new constants.

php.ini Defaults

No changed php.ini settings.

Open Issues

#13075

Future Scope

This section details areas where the feature might be improved in future, but that are not currently proposed in this RFC.

Voting

Implementation

TBD

Rejected Features

Keep this updated with features that were discussed on the mail lists.

[nielsdos]

These are now implemented via #13161 thanks to @youkidearitai :-)