Re: a better libtds

On Sun, 21 Aug 2011 11:53:51 +0200
Frediano Ziglio <[email protected]> wrote:

> And also dblib does not support wide characters.

True, but it could.  I think db-lib works well enough now that we could
call it 1.0; within the subset we intended to implement, it does
everything the vendors' implementations do.  The specifications
handed us don't address Unicode because Sybase & Microsoft have both
deprecated db-lib.  Therefore we're free to extend in any way that
doesn't break compatibility.  

DB-Library defines some parameters and returned strings as type
CHAR.  So, dbcolname() must return CHAR*, and "everyone knows" that's a
single-byte encoding.  Other functions, such as dbuse() take a char*.
Still others e.g. dbbind() use BYTE*.  

The answer?  All functions returning character data (either by
returning a pointer or by filling a client-provided buffer) should use
the encoding specified by DBSETLCHARSET() or "client charset" in
freetds.conf.  Regardless of char*, CHAR*, or BYTE*, if DBSETLCHARSET()
indicates UCS2, the client gets UCS2.  That includes the strings passed
to the message & error handlers.  

I think that covers it.  Wide characters are currently unsupported in
any existing db-lib, including ours.  Redefining character pointers
won't affect any existing applications, and new ones get one consistent
change.  

I don't see any need to duplicate the function names with W or A, the
way ODBC does.  

--jkl