On Wed, 18 Jul 2001, Tom Lane wrote:
> Reinhard Max <max@suse.de> writes:
> > On Wed, 18 Jul 2001, Bruce Momjian wrote:
> >> Do you have any idea how this will work with earlier TCL versions?
>
> > It won't. If pgtcl is supposed to still be able to compile with older
> > versions of Tcl, the changes have to be made a compile time option.
>
> Please do that and resubmit the patch.
OK, I'll pack the new stuff inside #ifdef TCL_UTF8 and define that if
the Tcl version is 8.1 or greater.
Can anybody tell me, if that TCL_ARRAYS stuff is still good for
something? If I could remove it, TCL_UTF8 would be less complex at
some places, because I'd only have to cover two instead of four cases.
BTW, I think the proposed Patch doesn't go far enough as it assumes
the database (client) encoding is identical to the system encoding by
using NULL as the first argument for Tcl_UtfToExternalDString and
Tcl_ExternalToUtfDString. I think it should instead either use the
database's encoding for the conversion to be correct or set
PostgreSQL's client encoding to UNICODE so that no conversion would be
needed. Unfortunately, I don't have the time to do that at the moment.
> We really don't want to give up backwards compatibility just yet.
How far do you want it to be backwards compatible? If >= 8.0 is OK,
I'll possibly overwork libpq later this year to use Tcl's object
interface. I expect at least some performance gain out of this.
cu
Reinhard
