Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:128045
Precedence: bulk
MIME-Version: 1.0
References: <CAEPPVa0hODpRVdGMuS30n-UFFg9sb4nZ29Z4L1B9qJP6VQbokg@mail.gmail.com>
 <3b7b0362-537e-5e2f-4345-63d0b7ef0964@php.net>
In-Reply-To: <3b7b0362-537e-5e2f-4345-63d0b7ef0964@php.net>
Date: Tue, 15 Jul 2025 16:05:30 +0900
Message-ID: <CAEPPVa1RGbMOjARNf6yhisoVBDXCoh6ZEojsKP_eb-dSAkxEcQ@mail.gmail.com>
Subject: Re: [PHP-DEV][DISCUSSION] Add locale and strength for grapheme functions
To: php internals <internals@lists.php.net>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
From: youkidearitai@gmail.com (youkidearitai)

2025=E5=B9=B47=E6=9C=8814=E6=97=A5(=E6=9C=88) 19:22 Derick Rethans <derick@=
php.net>:
>
> On Wed, 9 Jul 2025, youkidearitai wrote:
>
> > Hi, Internals
> >
> > I changed below the RFC.
> > - https://wiki.php.net/rfc/grapheme_add_locale_for_case_insensitive
> > Pull request is below:
> > - https://github.com/php/php-src/pull/18792
> >
> > Change point is below:
> > - Add a strength for grapheme_* functions
> >   - Affect to all over the world characters, ex: Ideographic Variation
> > Sequence(IVS)
> >   - Use Collator object const values.
>
> These settings are indeed important for these functions, but I can't get
> around the fact that it makes these APIs really cluttered and
> complicated =E2=80=94 something that many functions in the grapheme_ / in=
tl
> extension already suffer from.
>
> Is this API really the best way?
>
> > $locale parameter is not change anything. Because I could not find any =
way.
>
> It seems that I came to a similar conclusion, but locales are much more
> complicated than just languageCode_regionCode (for example, see
> https://github.com/derickr/php-text/blob/main/tests/text-contains.phpt#L2=
5)
>
> You also don't really need a strength argument, as you can 'encode' that
> in the locale name, like: 'nb_NO-u-ks-primary' (I know, it's rather ugly
> and the list of options is vast:
> https://www.unicode.org/reports/tr35/tr35-collation.html#Common_Settings
>
> cheers,
> Derick

Hi, Derick

Thank you very much for response.

> Is this API really the best way?

I reconsidered the function signature based on what you said.

> It seems that I came to a similar conclusion, but locales are much more
> complicated than just languageCode_regionCode (for example, see
> https://github.com/derickr/php-text/blob/main/tests/text-contains.phpt#L2=
5)
>
> You also don't really need a strength argument, as you can 'encode' that
> in the locale name, like: 'nb_NO-u-ks-primary' (I know, it's rather ugly
> and the list of options is vast:
> https://www.unicode.org/reports/tr35/tr35-collation.html#Common_Settings

Indeed, since strength can be specified in the locale,
I thought it would be better to specify it in the locale rather than
as a parameter for strength.

For example, The grapheme_* functions can detect difference for IVS.
```
$ sapi/cli/php -r 'var_dump(grapheme_levenshtein("\u{908A}",
"\u{908A}\u{E0101}", locale: "ja_JP-u-ks-identic"));'
int(1)
$ sapi/cli/php -r 'var_dump(grapheme_levenshtein("\u{908A}",
"\u{908A}\u{E0101}"));'
int(0)
$ sapi/cli/php -r 'var_dump(grapheme_strpos("\u{908A}", "\u{908A}\u{E0101}"=
));'
int(0)
$ sapi/cli/php -r 'var_dump(grapheme_strpos("\u{908A}",
"\u{908A}\u{E0101}", locale: "ja_JP-u-ks-identic"));'
bool(false)
```

Since ideographic characters also have identities (e.g., names), we
would like to make IVS compatible with them.
However, it should be simple, so we should compromise somewhere.

Regards
Yuya


--=20
---------------------------
Yuya Hamada (tekimen)
- https://tekitoh-memdhoi.info
- https://github.com/youkidearitai
-----------------------------