Making strings of arbitrary character types

Making strings of arbitrary character types

[Nick NoSpam]

I read the section: "Making strings of arbitrary character
types" (http://gcc.gnu.org/onlinedocs/libstdc
++/21_strings/howto.html#5) ; very helpful.

With GCC 3.3.2, the section is spot on--trying something like
basic_string<unsigned char>--yields linker errors about missing
char_traits for unsigned char.

But with GCC 3.4.2, basic_string<unsigned char> works out of the box.

Does 3.4.2 provide a specialization for this type (and possibly others)?
If so, is it intentional (ie. can I expect it in future versions)?  Or
should I specialize char_traits for portability?

-- 

Regards,
Nick G.
http://www.wheeloyum.com

Re: Making strings of arbitrary character types

[Paolo Carlini]

Hi,

>But with GCC 3.4.2, basic_string<unsigned char> works out of the box.
>
>Does 3.4.2 provide a specialization for this type (and possibly others)?
>If so, is it intentional (ie. can I expect it in future versions)?  Or
>should I specialize char_traits for portability?
>
Hey, this is Free Software, no mysteries! ;)

Have a look to include/bits/char_traits.h, there are plenty of 
interesting comments
and info. You will learn that a "generalization" of char_traits is now 
provided, which
often will work out of the box for many POD types besides char and wchar_t.
However, you will also learn that some typedef and eof() are likely to 
be wrong,
in general (are Ok for unsigned char, by the way, and other "simple" cases)

Indeed, this is an *intentional* improvement, and you can expect it to 
be present at
least in the whole 3.4.x series and the forthcoming 4.0.x series. You 
are also right,
however, that the fully portable solution involves providing your own 
char_traits
(probably, you can master that by studying the above file!)

Thanks,
Paolo.

Re: Making strings of arbitrary character types

[Nick NoSpam]

On Wed, 2004-12-01 at 10:40 +0100, Paolo Carlini wrote:
> Hi,
> 
> Have a look to include/bits/char_traits.h, there are plenty of 
> interesting comments
> and info. You will learn that a "generalization" of char_traits is now 
> provided, which
> often will work out of the box for many POD types besides char and wchar_t.
> However, you will also learn that some typedef and eof() are likely to 
> be wrong,
> in general (are Ok for unsigned char, by the way, and other "simple" cases)
Great; thanks!  Any chance this generalization can be used on 3.3.x or
does the compiler have issues?

> Indeed, this is an *intentional* improvement, and you can expect it to 
> be present at
> least in the whole 3.4.x series and the forthcoming 4.0.x series. You 
> are also right,
> however, that the fully portable solution involves providing your own 
> char_traits
> (probably, you can master that by studying the above file!)
Sounds like this is the route I'll have to take as portability is high
on the priority list.

Thanks for your help and the quick response!

-- 

Regards,
Nick G.
http://www.wheeloyum.com

Re: Making strings of arbitrary character types

[Nick NoSpam]

On Wed, 2004-12-01 at 11:01 +0100, Paolo Carlini wrote:
> In principle, you can try: basic_string itself should be ok, but you
> may see problems in other areas (like iostreams, fstreams),
Right you are; I just ran across problems w/ ofstream using 3.3.x.

> which, at the time, were not completely "char_traits"-general, so to
> speak...
Are you referring to GCC's implementation?

--

Regards,
Nick G.
http://www.wheeloyum.com

Re: Making strings of arbitrary character types

[Paolo Carlini]

Nick NoSpam wrote:

>>which, at the time, were not completely "char_traits"-general, so to
>>speak...
>>    
>>
>Are you referring to GCC's implementation?
>  
>
Right, 3.4.x is vastly better in this area (and we don't have any open 
bug ;)

Paolo.

Re: Making strings of arbitrary character types

[Nick NoSpam]

On Wed, 2004-12-01 at 11:22 +0100, Paolo Carlini wrote:
> Right, 3.4.x is vastly better in this area (and we don't have any open 
> bug ;)
I hope this isn't the first.  Compiling the program below w/ GCC 3.4.2
and running produces:

terminate called after throwing an instance of 'std::bad_cast'
  what():  St8bad_cast
Aborted


--- CODE BEGIN ---

#include <string>
#include <iostream>
#include <fstream>

typedef unsigned char UINT8;
typedef std::basic_string<UINT8> ByteArray;

void writeBytes()
{
	ByteArray bytes(30, (UINT8)'A');

	std::basic_ofstream<ByteArray::value_type> fout("test.dat",
std::ios::binary | std::ios::out);
	fout << bytes;		// Exception thrown here
}

int main(int argc, char *argv[])
{
	writeBytes();
	return(0);
}

--- CODE END ---

Basically, the exception is thrown because the locale is not set.

Backtrace:
#0  std::__check_facet<std::codecvt<unsigned char, char, __mbstate_t> >
(__f=0x0) at localefwd.h:187
#1  0x0804cfe3 in std::basic_filebuf<unsigned char,
std::char_traits<unsigned char> >::xsputn (this=0xbfffef74,
__s=0x805005c 'A' <repeats 30 times>, __n=30) at fstream.tcc:507
#2  0x0804ac5e in std::basic_streambuf<unsigned char,
std::char_traits<unsigned char> >::sputn (this=0xbfffef74, __s=0x805005c
'A' <repeats 30 times>, __n=30) at streambuf:421
#3  0x0804a2af in std::basic_ostream<unsigned char,
std::char_traits<unsigned char> >::_M_write (this=0xbfffef70,
__s=0x805005c 'A' <repeats 30 times>, __n=30) at ostream:288
#4  0x08049867 in std::operator<< <unsigned char,
std::char_traits<unsigned char>, std::allocator<unsigned char> >
(__out=@0xbfffef70, __str=@0xbffff0a0) at ostream.tcc:665
#5  0x08049276 in writeBytes () at main.cpp:13
#6  0x0804930b in main (argc=1, argv=0xbffff134) at main.cpp:18


GCC -v:
Reading specs from /usr/lib/gcc/i686-pc-linux-gnu/3.4.2/specs
Configured with: /var/tmp/portage/gcc-3.4.2/work/gcc-3.4.2/configure
--enable-version-specific-runtime-libs --prefix=/usr
--bindir=/usr/i686-pc-linux-gnu/gcc-bin/3.4
--includedir=/usr/lib/gcc/i686-pc-linux-gnu/3.4.2/include
--datadir=/usr/share/gcc-data/i686-pc-linux-gnu/3.4
--mandir=/usr/share/gcc-data/i686-pc-linux-gnu/3.4/man
--infodir=/usr/share/gcc-data/i686-pc-linux-gnu/3.4/info
--with-gxx-include-dir=/usr/lib/gcc/i686-pc-linux-gnu/3.4.2/include/g
++-v3 --with-local-prefix=/usr/local --host=i686-pc-linux-gnu
--disable-multilib --enable-nls --without-included-gettext
--enable-__cxa_atexit --enable-clocale=gnu --enable-shared
--with-system-zlib --disable-checking --disable-werror
--disable-libunwind-exceptions --with-gnu-ld --enable-threads=posix
--enable-languages=c,c++ --disable-libgcj
Thread model: posix
gcc version 3.4.2

-- 

Regards,
Nick G.
http://www.wheeloyum.com

Re: Making strings of arbitrary character types

[Paolo Carlini]

Nick NoSpam wrote:

>On Wed, 2004-12-01 at 11:22 +0100, Paolo Carlini wrote:
>  
>
>>Right, 3.4.x is vastly better in this area (and we don't have any open 
>>bug ;)
>>    
>>
>I hope this isn't the first.
>
No, it isn't ;)

You have to prepare a suited codecvt facet, able to convert during I/O 
from/to
you internal ("in memory") type (unsigned char) to the external ("on 
disk") type
(char): unfortunately, there is no such as thing as a "general" codecvt...

I suggest taking inspiration from the snippet present in our testsuite, 
in particular:

    testsuite/27_io/basic_filebuf/seekoff/wchar_t/12790*.cc
    + testsuite_character.h

Paolo.

Re: Making strings of arbitrary character types

[Paolo Carlini]

Paolo Carlini wrote:

> I suggest taking inspiration from the snippet present in our 
> testsuite, in particular:
>
>    testsuite/27_io/basic_filebuf/seekoff/wchar_t/12790*.cc

Sorry, of course I meant (no wchar_t involved!):

    testsuite/27_io/basic_filebuf/seekoff/12790*.cc
    + testsuite_character.h

Paolo.

Re: Making strings of arbitrary character types

[Nick NoSpam]

On Wed, 2004-12-01 at 18:58 +0100, Paolo Carlini wrote:
> You have to prepare a suited codecvt facet, able to convert during I/O 
> from/to
> you internal ("in memory") type (unsigned char) to the external ("on 
> disk") type
> (char): unfortunately, there is no such as thing as a "general" codecvt...
That clears it up.  Thanks for all your help!

-- 

Regards,
Nick G.
http://www.wheeloyum.com