Re: can orbit/corba string support unicode

[Michael Meeks <michael ximian com>]

Hi Gilbert,

On Mon, 2002-12-02 at 08:42, Gilbert Fang wrote:
>  I wonder how orbit map the corba string (CORBA_char) in c, and how it 
> implement the CORBA_string_alloc(length).

	Ok - CORBA_char is just a typedef for 'char *' to meet the mapping
spec; ie. it's a plain C string, and can be used interchangably as such.

> If I want to implement a method which has a parameter --inout string, 
> how could I deal with it in chinese or japanese character?

	Generally inout parameters are tricky; if you have one you want to do:

	CORBA_free (*inOutStr);
	*inOutStr = CORBA_string_dup ("Foo");

	See ORBit2/test/everything for some torturous tests of various sorts.

> Generally, for eg, one chinese char has 2 byte . So if I have a string 
> with 2 chinese char, should I use
> CORBA_string_alloc(2) or CORBA_string_alloc(4)?

	Neither - they both allocate a string of length 2 or 4 characters.

	I can't _Emphasize_ this enough:

		Abandon all hope of excessive multi-byte character 
	string complexity, Gnome(2) uses a single scheme: UTF8, all
	strings are encoded as UTF8, they thus all have 'char *'
	pointers, which point at arrays of 8bit entities, encoding UCS4.

		For more information read '$ man utf-8'.

	Please, please - forget about wchar_t, unichar_t, UCS-2, UTF-16, and
all other abominations. Simply focus on UTF8 and how pleasant life is
with it.

> I guess that CORBA_char is char , is that right?
> If I  do
>  CORBA_char * astr=CORBA_string_alloc(2);
> can I
>   strcpy(astr, "HI");
> then
>   Echo_echoString(astr);

	No need; you can do :

	Echo_echoString (obj, "hi", ev);

	since it's a native C type - the mapping is designed that way. Also,
for every 'alloc' there must be a 'free' ;-) in this case CORBA_free
'smart' frees every ORB allocated type.

> Every one is saying that I can find answer in c mapping spec, but 
> actually c mapping spec has not described  the detail fo one corba  
> implementation.  It does not say what is the really type of corba_long, 
> it depends on how orbit implement it.  And it does not say how 
> translate/co-op the raw c type with the corba type.

	Sigh - I think the mapping does specify this; but reading the mapping
is acutely painful. I'd just read the ORBit2 regression tests in
test/everything if you ever have a query.

	Oh - and finally - learn to love utf-8 ;-)

	HTH,

		Michael.

-- 
 mmeeks@gnu.org  <><, Pseudo Engineer, itinerant idiot