Geographical indexing

[Ruslan Zasukhin]

On 4/28/09 12:16 AM, "Thomas Flemming" <tf at ttqv.com> wrote:

Hi Thomas,

> thanks for testing.
> 
>> You think this is slow?
> Dont worry, Valentina IS fast and I like it in the meantime very much!
> 
> I was just trying to find out whats the best for my very specific Select.
> And still its too slow, but this is why these kind of things are better done
> with a rtree. I think I will go for an external rtree which gives me the
> RecIDs of the respective records.
> This one looks promising:
> http://trac.gispython.org/spatialindex/wiki

Really interesting.

I will download it to see how it fits to us.
 
> I also wanted to be sure, that there are no bugs in the net-wrapper.

Not found except that leak.

> I still have another case for the database with hundred-thousands of simple
> selects like: "Select * from tbl where a=x"
> I will try this with FindValue as soon as the memoryleak is fixed.

Yes, FindValue() is the fastest way.
Zero parser overhead.
 
For example, we have hear bench of Valentina for inserts.

100K inserts with 6-8 fields

1) INSERT INTO T(f1, ...f8) VALUES (1,2,...8)           32 sec

2) INSERT INTO T VALUES (1,2,...8)                      23 sec

3) INSERT INTO T(f1, ...f8) VALUES (:1,:2,...:8)       7.5 sec

4) API way, Table.AddRecord                            1.4 sec

----------
Reported by you that FindRange() make leaks

FindValue also?



-- 
Best regards,

Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc

Valentina - Joining Worlds of Information
http://www.paradigmasoft.com

[I feel the need: the need for speed]