Andrew Dunstan wrote:
Joe Conway wrote:
Andrew Dunstan wrote:
Are we really sure that this isn't a solution in search of a problem?
The need for value-per-call is real (examples mentioned down-thread)
and was anticipated from day one of the SRF implementation (in fact
the first patch I wrote was value-per-call, not materialize). But when
we realized that value-per-call was not going to work very well for
any PL *except* C-functions, we switched to SFRM_Materialize as the
only supported mode, with SFRM_ValuePerCall left as a
to-be-coded-later option (see SetFunctionReturnMode in execnodes.h).
Personally I think it is worth having SFRM_ValuePerCall even if only C
functions can make use of it.
Yeah, makes plenty of sense for C funcs. I don't think there's an
argument about that. But for that we don't need any threading
infrastructure.
Oh sure -- sorry I wasn't clear. I wasn't trying to support the idea of
threading so much as the idea that value-per-call itself has merit for a
number of use cases.
Joe
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match