Terry Yapt wrote:
> You are right.  I have noted that wrong query because you have wrote 
> about timing of both queries but, if queries give us different results, 
> then the 'timing' will be different and not comparable either.
>   

Exactly right, and that was one problem I had with the test data - 
because it was being created randomly, it took a little while to verify 
via simpler SQL queries that the answers coming back were in fact 
correct and that the other queries were right.

> Glad to see we are got the best with the LEFT JOIN approach.  If I have 
> another crazy idea I will tell you.
>   

Great :)

I think probably the way forward is to use the current statement we 
have, which seems to be good enough (on my test data it takes about five 
seconds to complete over a million documents; I don't consider that to 
be quite good enough, but a. there is still all the start-up overhead in 
there which will account for some of that and b. I suspect a real 
workload is going to be kinder to the query than the not-very-random 
random data inserted into my test db).

The plan then would be to get 0.4 out relatively quickly with the new 
store, and start trying to build some real-life mail loads in the store 
database and try queries against that. I think for most users the 
performance will be fine with 0.4.

Thanks,

Alex.


_______________________________________________
Bongo-devel mailing list
[email protected]
https://mail.gna.org/listinfo/bongo-devel

Reply via email to