Hi Tom,

Well this will not be a 'burn it to the ground' option ( maybe ) but to be honestly just wondering there shouldn't be a much more elegant solution to whole Unicode text support inside C++ then what's now it seems to be trying to fix a huge mess created over long period of time. Like if all container classes inside STL were not created with templated allocators, so all code did it manually and then now when it was needed to change/add new allocators, people would need to change/add on all existing code base changes to make new allocators work ....

So my suggestion and I will put a simple example here is to create a templated character type ( lets call it encoded character or echar in short ) that will hold an encoding template inside.

template < class  ENCODING >
concept   CHAR_ENCODING = requires( ENCODING::type * _charType, char32_t * _charUtf32, size_t size )
{
    typename ENCODING::type;
    { ENCODING::Encode( _charType , size, _charUtf32, size ) } -> std::same_as<bool>;
    { ENCODING::Decode( _charUtf32, size, _charType , size ) } -> std::same_as<bool>;
};
//-/////////////////////////////////////////////////////////////////////////////////////////////////////-//
#pragma pack(push, 1)
//-/////////////////////////////////////////////////////////////////////////////////////////////////////-//
template<CHAR_ENCODING Encoding>
struct echar
{
    using ENCODING  = Encoding;
    using TYPE      = Encoding::type;
    TYPE            encoding_char;
};
//-/////////////////////////////////////////////////////////////////////////////////////////////////////-//
#pragma pack(pop)
//-/////////////////////////////////////////////////////////////////////////////////////////////////////-//

Then creating encoding and new encoded character is simple and even easily added by average c++ user :

class ASCII_ENCODING
{
public:
    using type = char;
    // Required by concept for all Encodings 
    static constexpr bool Encode( char*     pData, const size_t sizeData, const char32_t* pSource, const size_t sizeSource )
    { return false; };
    static constexpr bool Decode( char32_t* pData, const size_t sizeData, const char*     pSource, const size_t sizeSource )
    { return false; };
};  static_assert( CHAR_ENCODING<ASCII_ENCODING> == true );
class UTF08_ENCODING
{
public:
    using type  = char8_t;
    // Required by concept for all Encodings 
    static constexpr bool Encode( char8_t*  pData, const size_t sizeData, const char32_t* pSource, const size_t sizeSource )
    { return false; };
    static constexpr bool Decode( char32_t* pData, const size_t sizeData, const char8_t*  pSource, const size_t sizeSource )
    { return false; };
};  static_assert( CHAR_ENCODING<UTF08_ENCODING> == true );

This is just basic simple encoding concept as it would required a lots of more static constexpr function for sure ( like to check for conversion size, validate text, compare texts case sensitive or case insensitive .... ) and all this functions can be required or optional and we can check at compile time for it availability and even test some of them at compile time as would require them all to be constexpr ....

Then its simple to create new encoded char types:

using ascii_char    = echar<ASCII_ENCODING>;
static_assert( sizeof(ascii_char) == sizeof(char   ));
using utf08_char    = echar<UTF08_ENCODING>;
static_assert( sizeof(utf08_char) == sizeof(char8_t));

And usage atm is not perfect:

const char*             text1   =   "ASCII text";
const char8_t*          text2   = u8"UTF08 text";
const ascii_char* ascii_text    = reinterpret_cast<const ascii_char*>(text1);
const utf08_char* utf08_text    = reinterpret_cast<const utf08_char*>(text2);

But it's even work, so the idea is basically that for all other Standard Library Classes if they work with echar type ( strings, views, format, print ... ) all will work perfectly, even with encoding the user will add in his code...

So even things like utf16_roman_char string will auto convert to utf16_korean_char string if requested ( 1st conversion to utf32 then 2nd conversion back to utf16 with korean encoding ). Ofc its not best performance wise to do double conversion but ofc we can always add constexpr function that can even optimize this and if check on compile time is true use them like for example :

// Optional for faster encoding // 
inline constexpr bool FastEncode( std::span<utf08_char> Dest, const std::span<ascii_char> Source ) { return false; };
inline constexpr bool FastEncode( std::span<ascii_char> Dest, const std::span<utf08_char> Source ) { return false; };

Full example on compile explorer : https://godbolt.org/z/azEzKceeq so people could grasp the idea of this approach to the whole unicode problem. And this approach is to put whole Unicode standard rule mess out of Standard Library code into single templated class per encoding, and make all our classes inside Library work perfectly now when they know encoding type of a character, and even adding new encoding types for a user will be simple and how constexpr will be forced will make it possible to check for lots of issue at compile time and report to user when creating this new encoding types...

How it would be simple solution if it would be possible to templated simple types so like :

char = echar<ASCII_ENCODING>; ( maybe better char = echar<CHAR_ENCODING>; for back support as people used char as utf08 char )

char8_t = echar<UTF08_ENCODING>;

char16_t = echar<UTF16_DEFAULT_ENCODING>;

char32_t = echar<UTF32_ENCODING>;

wchar = echar<UTF16_DEFAULT_ENCODING> for Windows

wchar = echar< UTF32_ENCODING > for Mac

and then make all classes work with echar<> type instead of supporting all these 5 types ....

Best regards to all :)

Zoran Sibalic.

On Tue, Oct 24, 2023 at 8:40 AM <sg16-request@lists.isocpp.org> wrote:

Send SG16 mailing list submissions to
sg16@lists.isocpp.org

To subscribe or unsubscribe via the World Wide Web, visit
https://lists.isocpp.org/mailman/listinfo.cgi/sg16
or, via email, send a message with subject or body 'help' to
sg16-request@lists.isocpp.org

You can reach the person managing the list at
sg16-owner@lists.isocpp.org

When replying, please edit your Subject line so it is more specific
than "Re: Contents of SG16 digest..."

Today's Topics:

1. Agenda for the 2023-10-25 SG16 telecon (Tom Honermann)
2. Re: Agenda for the 2023-10-25 SG16 telecon (Jens Maurer)

----------------------------------------------------------------------

Message: 1
Date: Tue, 24 Oct 2023 01:11:20 -0400
From: Tom Honermann <tom@honermann.net>
To: SG16 <sg16@lists.isocpp.org>, Alisdair Meredith
<alisdairm@me.com>, Jonathan Wakely <cxx@kayari.org>, Charles Barto
<chbarto@microsoft.com>, Mark de Wever <koraq@xs4all.nl>
Subject: [SG16] Agenda for the 2023-10-25 SG16 telecon
Message-ID: <22f6f509-63c5-4728-96c3-241562ffa940@honermann.net>
Content-Type: text/plain; charset="utf-8"; Format="flowed"

SG16 will hold a telecon on Wednesday, October 25th, at 19:30 UTC
(timezone conversion
<https://www.timeanddate.com/worldclock/converter.html?iso=20231025T193000&p1=1440&p2=tz_pt&p3=tz_mt&p4=tz_ct&p5=tz_et&p6=tz_cest>).

The agenda follows.

* charN_t, char_traits, codecvt, and iostreams:
o P2873R0: Remove Deprecated Locale Category Facets For Unicode
from C++26 <https://wg21.link/p2873r0>
o LWG 3767: codecvt<charN_t, char8_t, mbstate_t> incorrectly added
to locale <https://wg21.link/lwg3767>
o LWG 2959: char_traits<char16_t>::eof is a valid UTF-16 code unit
<https://wg21.link/lwg2959>
+ SG16 #32: std::char_traits<char16_t>::eof() requires
uint_least16_t to be larger than 16 bits
<https://github.com/sg16-unicode/sg16/issues/32>
o SG16 #33: A correct codecvt facet that works with basic_filebuf
can't do UTF conversions
<https://github.com/sg16-unicode/sg16/issues/33>

Hang on, this is going to be a bumpy ride.

When char16_t and char32_t were added for C++11, the standard library
was extended to support corresponding specializations of
std::char_traits ([char.traits.general]p1
<http://eel.is/c++draft/char.traits.general#1>) and std::basic_string
([string.classes.general]p1
<http://eel.is/c++draft/string.classes#general-1>). Curiously, type
aliases were added for specializations of the std::fpos ([iosfwd.syn]
<http://eel.is/c++draft/iosfwd.syn#lib:fpos>) class template (but only
in the synopsis) and support for these types was added for the
std::codecvt ([tab:locale.category.facets]
<http://eel.is/c++draft/locale.category#tab:locale.category.facets>) and
std::codecvt_byname ([tab:locale.spec]
<http://eel.is/c++draft/locale.category#tab:locale.spec>) locale facets,
but not for any of the other locale facets nor for iostreams in general.
Support for these types was added to std::basic_string_view
([string.view.synop] <http://eel.is/c++draft/string.view.synop>) and
std::filesystem::path ([fs.path.type.cvt]p2
<http://eel.is/c++draft/fs.path.type.cvt#2>) in C++17, but no additional
support was ever extended to iostreams. The status quo is thus that the
standard requires implementations to provide some fragments (std::fpos,
std::codecvt, and std::codecvt_byname) of iostream support for these
types despite there being no use of these type aliases and
specializations in the standard; implementations are not required to
support streams of char16_t or char32_t.

std::char_traits is used by both the string library (e.g.,
std::basic_string) and iostreams. However, the string library only
depends on some of the std::char_traits members; it does not make use of
the int_type member type alias nor any of the member functions that
depend on that type (eof(), ?not_eof(), ?to_char_type(), ?to_int_type(),
?eq_int_type()). Per LWG 2959 <https://wg21.link/lwg2959> and SG16 #32
<https://github.com/sg16-unicode/sg16/issues/32>, the specified
std::char_traits<char16_t> specialization has a defect; all char16_t
values are valid code unit values, but the int_type member type alias is
defined as uint_least16_t (the same underlying type as char16_t) and it
is thus unable to hold a distinct value for EOF. The obvious fix is to
use a larger type for int_type, but that would result in an ABI break. I
recently asked the ABI review group if there are any known tricks they
could deploy to mitigate an ABI break, but no direct solutions were
identified; a suggestion to provide an alternative type for
std::char_traits<char16_t> that programmers would have to explicitly use
instead of the broken specialization was offered. That is an option, but
since the problematic int_type member is not actually used by any
functionality the standard requires implementors to provide, an ABI
break in this case might have little practical consequence.

When char8_t was added for C++20 via P0482R6 (char8_t: A type for UTF-8
characters and strings) <https://wg21.link/p0482>, I failed to
understand the intended purpose for which std::codecvt was added to the
standard. My impression of it at the time was that it was a poorly
designed general transcoding facility; I failed to appreciate its
significance as a locale facet as used by iostreams. This resulted in
two mistakes:

1. I deprecated the following specializations (and their use as locale
category facets):
std::codecvt<char16_t, char, std::mbstate_t>
std::codecvt<char32_t, char, std::mbstate_t>
std::codecvt_byname<char16_t, char, std::mbstate_t>
std::codecvt_byname<char32_t, char, std::mbstate_t>
2. I added the following specializations as required locale category
facets (adding the specializations themselves is arguably not a
mistake, but adding them as locale category facets is):
std::codecvt<char16_t, char8_t, std::mbstate_t>
std::codecvt<char32_t, char8_t, std::mbstate_t>
std::codecvt_byname<char16_t, char8_t, std::mbstate_t>
std::codecvt_byname<char32_t, char8_t, std::mbstate_t>

Note that std::codecvt facets are only used by std::basic_filebuf which
only ever converts to and from elements of type char; the facets that
convert to and from char8_t are not substitutable for that purpose.

P2873R0 <https://wg21.link/p2873r0>, which SG16 already approved (or,
rather, did not object to) during the 2023-05-26 SG16 meeting
<https://github.com/sg16-unicode/sg16-meetings#may-24th-2023>, now seeks
to remove the deprecated specializations. LWG 3767
<https://wg21.link/lwg3767> tracks addressing the incorrect addition of
the char8_t specializations as locale facets.

Arguably, P0482R6 <https://wg21.link/p0482> should have added the
following specializations as locale facets:

* std::codecvt<char8_t, char, std::mbstate_t>
* std::codecvt_byname<char8_t, char, std::mbstate_t>

The only specification for std::codecvt_byname in the standard is the
synopsis in [locale.codecvt.byname]
<http://eel.is/c++draft/locale.codecvt.byname>; there is no other
wording present.

As mentioned, the standard does not require implementations to provide
iostream support for the charN_t types. However, implementations may do
so as an extension. If they do, then, per [filebuf.general]p7
<http://eel.is/c++draft/input.output#filebuf.general-7>, specializations
of std::codecvt<charN_t, char, std::mbstate_t> are required to be
available via a call to std::use_facet() for the imbued locale. In which
case, per the standard, the status of the necessary specializations are:

* std::codecvt<char8_t, char, std::mbstate_t> # Not specified.
* std::codecvt<char16_t, char, std::mbstate_t> # Deprecated.
* std::codecvt<char32_t, char, std::mbstate_t> # Deprecated.

If it is desirable to provide a better foundation for iostream support
of the charN_t types, either for a future version of the standard, or
for implementations that want to provide such support as an extension,
we could undeprecate the previously deprecated specializations and add
the missing one for char8_t. Since iostreams does not support charN_t in
the standard today and since the char16_t and char32_t specializations
have already been deprecated for two release cycles, perhaps it is even
reasonable to change their behavior so that they convert to and from the
locale encoding rather than UTF-8. This would remove the existing
inconsistency with the corresponding char and wchar_t specializations
that was part of the motivation for their deprecation in the first place
(see the discussion of codecvt in the Motivation section of P0482R6
<https://wg21.link/p0482r6#motivation>).

However, an endeavor to improve the situation for iostreams and charN_t
next runs into SG16 #33
<https://github.com/sg16-unicode/sg16/issues/33>; std::basic_fstream
does not support the UTF-8 and UTF-16 encodings for the "internal" side
of a std::codecvt conversion because std::basic_filebuf requires that,
per [locale.codecvt.virtuals]p4
<http://eel.is/c++draft/locale.codecvt#virtuals-4> and its related
footnote <http://eel.is/c++draft/locale.codecvt#footnote-246>,
"internal" characters are mapped 1-N to "external" characters. This is
an existing issue for std::basic_fstream<wchar_t> with UTF-16 data.

The Microsoft and libstdc++ standard library implementations appear to
support iostreams with charN_t types; at least on the surface. Libc++
intentionally does not provide definitions for charN_t specializations
of locale facets that are not required by the standard and this suffices
for basic usage to provoke compilation errors. I have not yet
investigated to what extent the Microsoft and libstdc++ implementations
work as might be expected. My impression is that, where they do produce
expected results, it is serendipity at work. See
https://godbolt.org/z/6T7hebY33 for a bit of fun (testing on Windows
requires changes to use an actual zero valued file since Windows doesn't
provide a builtin analog for /dev/zero, but in that case, MSVC produces
an executable that behaves as might be expected).

I haven't looked hard, but I have not yet identified any code in the
wild that uses iostreams with charN_t types. One would think that, if
any project did, it would be ICU. I confirmed that ICU, despite its use
of char16_t, makes no attempt to use it with iostreams.

So where is this all going? I see three general options that can be
pursued to resolve these various issues.

1. We can fix these issues, despite the acknowledged ABI impact, so
that the standard no longer actively hiders support for iostreams
with the charN_t types. Optionally, we could further explore
requiring such support in the standard (doing so would require
adding charN_t support to more locale facets).
2. We can declare that iostreams will never support the charN_t types
in the standard and deprecate and remove the fragments of such
support that are present. Implementations could of course provide
support as an extension if they so desire.
3. We can admit things are broken, choose to do nothing about it, and
close the related LWG issues while chanting sorry-not-sorry.

The above issues are sufficiently complicated that I believe a paper is
warranted regardless of the direction that we favor. I'm signing up to
write that paper since I'm responsible for some of the mess. I do not
intend to poll any directions in this meeting; rather, the focus is to
ensure that the issues are well understood, to discuss decisions we
could make and their potential consequences, and to generally collect
information that will lead to a better paper.

Responses provided before the meeting to identify other existing related
issues or considerations would be appreciated. Ideal responses do not
include the phrase "burn it all to the ground".

Tom.
-------------- next part --------------
HTML attachment scrubbed and removed

------------------------------

Message: 2
Date: Tue, 24 Oct 2023 08:40:17 +0200
From: Jens Maurer <jens.maurer@gmx.net>
To: sg16@lists.isocpp.org, Alisdair Meredith <alisdairm@me.com>,
Jonathan Wakely <cxx@kayari.org>, Charles Barto
<chbarto@microsoft.com>, Mark de Wever <koraq@xs4all.nl>
Cc: Tom Honermann <tom@honermann.net>
Subject: Re: [SG16] Agenda for the 2023-10-25 SG16 telecon
Message-ID: <c6b06c90-f3ed-4b00-96a6-3eedbf144f79@gmx.net>
Content-Type: text/plain; charset=UTF-8

Hi Tom,

On 24/10/2023 07.11, Tom Honermann via SG16 wrote:
> Hang on, this is going to be a bumpy ride.

Thanks for the write-up.

We should ask the implementers whether their basic_stream support
for charN_t is intentional or accidental, and investigate a little
more whether it works.

Hyrum's law suggests that this will be used in the wild.

My opinion:

Let's make sure the ground is clear for a future extension
to charN_t for basic_stream, but let's not try to address
any of the deeper troubles (in particular the 1:N mapping for
basic_fstream). In particular:

Let's fix "int_type". The ABI of the standard library
itself will not be broken, we just risk ABI breakage
of user components, I think?

Let's deprecate

std::codecvt<char16_t, char8_t, std::mbstate_t>
std::codecvt<char32_t, char8_t, std::mbstate_t>
std::codecvt_byname<char16_t, char8_t, std::mbstate_t>
std::codecvt_byname<char32_t, char8_t, std::mbstate_t>

Those might come back when a proper solution arrives.

std::codecvt<char16_t, char, std::mbstate_t> # Deprecated.
std::codecvt<char32_t, char, std::mbstate_t> # Deprecated.

"Since iostreams does not support charN_t in the standard today and since the char16_t and char32_t specializations have already been deprecated for two release cycles, perhaps it is even reasonable to change their behavior so that they convert to and from the locale encoding rather than UTF-8."

That might work for

std::codecvt<char32_t, char, std::mbstate_t>

but

std::codecvt<char16_t, char, std::mbstate_t>

runs afoul of the 1:N mapping issue, unless on a platform where everything
fits into 16-bit Unicode, right?

Best to leave those functions alone; I'm also ok with removing them.

Jens

------------------------------

Subject: Digest Footer

SG16 mailing list
SG16@lists.isocpp.org
https://lists.isocpp.org/mailman/listinfo.cgi/sg16

------------------------------

End of SG16 Digest, Vol 48, Issue 7
***********************************