Re: Cursors: getting the number of tuples; moving backwards

On Fri, Nov 01, 2002 at 10:03:17PM -0500, Tom Lane wrote:
> MOVE does execute the query, it just doesn't ship the tuples to the
> client. This would save some formatting overhead (no need to run
> the datatype I/O conversion procedures), but unless you have a slow
> network link between client and server I would not expect it to be
> "much" faster ...

It must be the fact that the computer is quite old : Cyrix 6x86 166Mhz.
( this is not the deplyoment machine ).

Using MOVE is about 5 times faster in my case :
For 150784 tuples in the table, FETCH-ing took about 1m30 ,
while MOVE-ing took only about 17sec.

| Real | User | Sys
-------------------------------------------------------------------
select * from PRODTEST | 1m30.843s | 0m42.960s | 0m1.720s
-------------------------------------------------------------------
declare cursor... + FETCH | 1m32.835s | 0m42.680s | 0m1.780s
-------------------------------------------------------------------
declare cursor... + MOVE | 0m17.215s | 0m0.030s | 0m0.030s
-------------------------------------------------------------------
( i used commands like: time psql -f test.sql db_rex
to get those timings )

The difference must be smaller on fast machines.

So i guess that my computer is pretty good when it comes to finding
performance problems in applications ;-)

Bye,
Adrian Maier
(am(at)fx(dot)ro)