Hi,

Interestingly I suspect we're seeing similar problems with Sybase.
When I manually implement a select instead of using Torque's save()
I seem to get about twice the performance.

Can you tell me precisely what you changed?
I'd like to try it with our system.

Thanks

Joe

On 12/06/06, Tassos Bassoukos <[EMAIL PROTECTED]> wrote:

Greetings,
I've encountered an interesting performance disparity when using
torque, specifically requests from torque took 40+ms to complete
(based on the timing in BasePeer), while the database reported
completion in 1-2ms.

After some digging, it looks like a village-postgresql mismatch, in
the sense that village caches the metadata for all columns in a query,
but postgresql will emit an extra (internal) query to fetch two
metadata options: for the nullAllowed and autoIncrement properties.

That is, for a select with 6 fields, a total of 13 round-trips to the
server are made.

Now comes the interesting bit: These attributes aren't used anywhere.
I've removed these attributes (and their accessors), and everything
compiled and plays fine....

Tassos Bassoukos
P.S.: I'm not subscribed to the list, so you may want to CC me for a
fast response.
P.P.S.: This is all on stock torque 3.2

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Reply via email to