Re: User-specified encoding for postgresql-socket [Was CCL postgresql-socket with UTF-8]

On 06/12/2011 12:00 PM, Kevin Rosenberg wrote:
> Otto Diesenbacher wrote:
>> so - here is a patch against postgresql-socket-api.lisp to have UTF-8
>> support with CCL (tested withh CCL64/Linux and CCL32/Windows) and
>> postgresql-socket (against clsql-20110522-git, current quicklisp-dist).
> Thanks for the patch, it's a good patch, and I/m releasing CLSQL 4.3.3
> with your patch.
>
> There's a better solution that should take someone just a few hours to
> implement. Here's what I added to postgresql-socket-api.lisp. Patches
> gratefully accepted :)
>
> ;; FIXME: The file has code specific to sb-unicode and CCL
> ;; to assume UTF8 encoded strings.
> ;; Best fix would be to use the user-specified encoding that is now
> ;; stored in the database object and use the UFFI 2.x encoding functions
> ;; to convert strings to/from octet vectors. This allows encoding
> ;; other than UTF8 and also works on all CL implementations that
> ;; support wide character strings

Still don't have the better implementation you outlined above but that 
patch broke non-CCL lisps slightly.

CCL doesn't support a null terminated write, so it needs to write an 
extra 0 byte. That should have been limited to CCL as the other codepath 
is already taking care of tis. Fixed this by adding a ccl conditional 
before the write-byte call.

Patch available at:
git pull git://github.com/UnwashedMeme/clsql.git master