Re: [Dbdpg-general] benchmarking old Pg and DBD::Pg

g == greg(at)turnstep(dot)com writes:

g> Let's slow down and back up a bit. First, what version of
g> PostgreSQL are you using to compile DBD::Pg, and what version
g> are you connecting to? Many of the things being discussed
g> have already been solved if you are running a modern version
g> of PG (e.g. 7.4 or greater).

g> The binding is already handled quite well: we use PQprepare
g> whenever possible. The default of type "VARCHAROID" is mostly
g> cosmetic, so that some other things work: we actually send
g> a type "unknown" (0) if no type is specified for a placeholder
g> via bind_param. (see the "defaultval" of the placeholder
g> struct). Older servers may benefit from the VARCHAROID change
g> however, so I will see about making that change.

g> The stuff in types.c could certainly use some cleaning up, but
g> it is also not really used anymore - we let the server do all
g> the quoting for us now via PQexecPrepared and PQexecParams.

This is consistent with what I'm seeing. That is, I see no
performance change using Vlad's patch to DBD-Pg-1.41. What I am
seeing, however, is my code that uses the old Pg interface is much
faster than my version that uses DBI/DBD::Pg. But this is probably in
the way I'm using it. For example, there are places where I should be
using execute_array() instead of wrapping a loop around execute().

--
Brandon