Hi Tim,

Didn't mean to distract you from ElevateDB :D

I've just tried SuperFastDistinct demo and generate 1 million records, it
would be great if ElevateDB is faster, it only produce 2k records per second
and down to 1k when the table is > 500.000 even using transaction.

You can try other db, and they perform faster.

Your other promised features are just great

Regards,

Surjanto

Surjanto,

<< Didn't mean to distract you from ElevateDB :D

I've just tried SuperFastDistinct demo and generate 1 million records, it
would be great if ElevateDB is faster, it only produce 2k records per second
and down to 1k when the table is > 500.000 even using transaction. >>

What is the SuperFastDistinct demo ?

We'll discuss performance further once ElevateDB is in beta.  The
performance is better, specifically the append and delete performance.

--
Tim Young
Elevate Software
www.elevatesoft.com

Tim


Having seen the post I had a look and as a guess its simply grabbing the trees out of the index and returning the unique root values. The demo creates a table with x rows and 50 unique indices and returns those 50 unique indices - great for populating a lookup combobox very fast.

Roy Lambert

"Roy Lambert" <roy.lambert@skynet.co.uk> wrote in message news:D7B9FDB3-57CB-4055-82EA-2E53C51BBFE6@news.elevatesoft.com...
> Tim
>
>
> Having seen the post I had a look and as a guess its simply grabbing the trees out of the index and returning the unique root
> values. The demo creates a table with x rows and 50 unique indices and returns those 50 unique indices - great for populating a
> lookup combobox very fast.
>
> Roy Lambert
>

I actually fail to see it usefulness at all.

The code to do what it does is trivial since it merely does
a variation of  A Set Index Go Top  B Set Key C FindNearest Key+x'FF' D If not EOF Goto B
to position itself through each unique index value

I've done this a squillion times
I am sure their product works fine
but at about $10 per line of code
thats a bloody expensive product.

gF

Tim,

Benchmarks: http://www.grebarsys.com/html/graphs.html
Demo Program: http://www.grebarsys.com/filedownloads/SFDDemo.zip

If you have some time, you can try create 1 million records and compare it
creation speed to others.

Regards,


Surjanto

GregF


Or if you prefer SQL a simple SELECT DISTINCT fieldname FROM table would work, but their method is a lot faster which I suppose is the benefit.

Roy Lambert

"Roy Lambert" <roy.lambert@skynet.co.uk> wrote in message news:22FD4D41-1A46-4298-8534-955A092D2124@news.elevatesoft.com...
> GregF
>
>
> Or if you prefer SQL a simple SELECT DISTINCT fieldname FROM table would work, but their method is a lot faster which I suppose is
> the benefit.
>
> Roy Lambert
>

the whole point is that it avoids using distinct
by using a variation of the technique in my
previous post.

The code to do this is so trivial and is so
one dimensional in that its only usefulness
is where the fields are the high order part of a key
that it would seem (even at USD90) to be
so extraordinarily overpriced (since it only replaces
about nine lines of code) that has no use.

gF

Greg,

<< The code to do what it does is trivial since it merely does
a variation of  A Set Index Go Top  B Set Key C FindNearest Key+x'FF' D If
not EOF Goto B
to position itself through each unique index value >>

It's a lot more difficult also when you start throwing non-US-ASCII and
international collation issues in the mix.  It's not so easy in such a case
to simply increment a character value like it is with ASCII characters, and
the "next" value must be found by reading the index in a raw format if you
want to avoid scanning the entire bottom of the index tree.

--
Tim Young
Elevate Software
www.elevatesoft.com