Why not create a table with cols a, b, c and d. Where you insert a row for each combination and key and index abc then return d?
ons. 24. feb. 2021, 21:15 skrev Paul van der Linden <paul.doskabouter@xxxxxxxxx>:
Thanks for all the suggestions,When the server is not in use for mission-critical work, I'll definitely going to do some testing based on your ideas.Will let you know what comes out of thatCheers,PaulOn Thu, Feb 18, 2021 at 6:49 PM hubert depesz lubaczewski <depesz@xxxxxxxxxx> wrote:On Thu, Feb 18, 2021 at 10:24:25AM -0700, Michael Lewis wrote:
> [1]https://www.depesz.com/2010/03/18/profiling-stored-proceduresfunctions/
>
> Thanks for this reference. I enjoy your blog, but haven't made the time to read all the archives somehow. Stuff doesn't stick very
> well when it isn't yet "needed" info besides.
> I have seen overhead from 'raise notice' in small functions that are sometimes called many thousands of times in a single query, but
> hadn't done the test to verify if the same overhead still exists for raise debug or another level below both client_min_messages
> and log_min_messages. Using your examples, I saw about .006 ms for each call to RAISE DEBUG with a client/log_min as notice/warning.
Sure, this overhead is definitely possible, but kinda besides the point
- there will be some slowdowns in other places, and it will be good to
track them.
That's why I suggested to do it on small sample of data.
Best regards,
depesz
