Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:122322
Message-ID: <7e42d1eb-4924-4a57-a4c1-412a665a4496@wikimedia.org>
Date: Wed, 7 Feb 2024 14:56:40 +1100
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Content-Language: en-US
To: Ayesh Karunaratne <ayesh@php.watch>
Cc: youkidearitai <youkidearitai@gmail.com>,
 php internals <internals@lists.php.net>
References: <CAEPPVa3JWXGLZh9rEE7WGa+nbbfFr-Lv4B7TX4F2TX3OSw6K+Q@mail.gmail.com>
 <CANS-=pcRFH=1y8O6N-R2Ty_Y-xid_i6NM05BPcnKGEdPBhTi_Q@mail.gmail.com>
 <CANLbj-pYFZj5M+A9=kQW9JobnM7ip0s6FqXA5z5qew6QGRDpfQ@mail.gmail.com>
 <CAEPPVa35rq3=UatBKNKNz9aH=FAcc5Fv3yXtUx-b-+uc8W-ppQ@mail.gmail.com>
 <29b0a205-8903-4ae8-b1e4-45db846fee7f@wikimedia.org>
 <CANLbj-rPHhnJiEOABEN-=yJHVvS6rbFOt7jUP4ei_=V4+aka2g@mail.gmail.com>
In-Reply-To: <CANLbj-rPHhnJiEOABEN-=yJHVvS6rbFOt7jUP4ei_=V4+aka2g@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Subject: Re: [PHP-DEV][VOTE][RFC] mb_ucfirst and mb_lcfirst functions
From: tstarling@wikimedia.org (Tim Starling)

On 7/2/24 13:43, Ayesh Karunaratne wrote:
>
> Hi Tim,
> Now that the RFC is restarted, could you mention some examples in 
> Georgian that might be good test cases?
>
> I was thinking there might be some good test cases in Turkish, but 
> couldn't find any. The RFC has examples 
> (https://github.com/php/php-src/pull/13161) in Vietnamese, but they 
> are correct for both "uppercase first character" and titlecase 
> conversions.

Any Georgian word would do. Your ASCII test case is "abc". The 
Georgian equivalent for that would be "აბგ" (ani bani gani, U+10D0 
U+10D1 U+10D2) which should remain the same after passing through 
mb_ucfirst(). Compare mb_strtoupper("აბგ") -> "ᲐᲑᲒ" (U+1C90 U+1C91 
U+1C92).

On the task I mentioned that ligatures are also affected. I gave the 
example mb_ucfirst("ǉ") -> "ǈ", that is, U+01C9 -> U+01C8. You could 
add a test case for that. Compare mb_strtoupper("ǉ") -> "Ǉ" (U+01C7).

To repeat my rationale -- we can view ucfirst() either through a 
technical lens (convert the first character of a string to upper case) 
or through a natural language lens (convert a string to sentence case, 
with the initial letter capitalised per local conventions). I am 
arguing to make mb_ucfirst() be a natural language extension of 
ucfirst(), because applying the technical extension would produce 
results that look quite jarring in a natural language context.

There are some edge cases which are not quite right. To really do a 
good job, a new case map will be needed. But if we document it as 
being for natural language, and set the right expectations, we can fix 
the edge cases later.

-- Tim Starling