Mauricio,

<< Well, it seems that I am alone in this fight, so I will retreat, at least
for now >>

Sorry to be such a hard-ass about this, but I'm trying to move EDB in a
certain direction, and putting the index statistics back in would move it in
the opposite direction.

--
Tim Young
Elevate Software
www.elevatesoft.com

Hello Tim,

> << Let's say that i'm building a custom grid ... Do you have any idea how to
> know which it's the position of the current record ? >>
>
> The only way would be to navigate the dataset and count the rows.

Wouldn't be faster if EDB creates a fast temporary "count" index
in order to make the query/table sequenced ?

--
Charalabos Michael - [Creation Power] - http://www.creationpower.gr

Hello Mauricio,

> <<No, it wouldn't.  You don't understand the index format, so I can understand
> why you would think this, but the index format and the way the index costs
> are estimated now do not allow for turning statistics on or off.  They
> simply aren't there, nor will they ever be there.>>
>
> Well, it seems that I am alone in this fight, so I will retreat, at least for now

Well, i'm still hate the 3 scroll bars state think and
as you can see i'm pushing Tim for a ... "solution" without
adding statistics ...

--
Charalabos Michael - [Creation Power] - http://www.creationpower.gr

Michael,

<< Well, i'm still hate the 3 scroll bars state think and as you can see i'm
pushing Tim for a ... "solution" without adding statistics ... >>

There is no solution without statistics, unless you care to simply never see
another live dataset in EDB.  In that case, I can provide counts all day
long.

--
Tim Young
Elevate Software
www.elevatesoft.com

Michael,

<< Wouldn't be faster if EDB creates a fast temporary "count" index in order
to make the query/table sequenced ? >>

Umm, that's what the statistics did, essentially.  Besides, what you're
proposing is creating a shadow index for every index order in a given
dataset, which as you can imagine is a non-starter.

Trust me on this guys/gals, I've been doing this for a long time and I've
thought this through very thoroughly.  There is no solution to the issue
apart from using index statistics.  To see this for yourselves, go try to
find another database engine besides DBISAM that provides logical sequence
numbers of *live* datasets without counting rows.

--
Tim Young
Elevate Software
www.elevatesoft.com

Hello Tim,

> Umm, that's what the statistics did, essentially.  Besides,
> what you're proposing is creating a shadow index for every
> index order in a given dataset, which as you can imagine is
> a non-starter.

Yeap but why non-starter ? It will defently cause a delay
on large datasets but it will not on small datasets.

Take for an example QuantumGrid when it loads the dataset
into memory in order to have accurate scrollbar. It's pretty
fast! EDB can do it with the only difference that it will not
store the index into memory (can't imaging loading million records
into memory! but instead on the hard disk.

> Trust me on this guys/gals, I've been doing this for a long
> time and I've  thought this through very thoroughly.  There is
> no solution to the issue apart from using index statistics.  To
> see this for yourselves, go try to find another database engine
> besides DBISAM that provides logical sequence numbers of *live*
> datasets without counting rows.

I trust you and i have a feeling that you have some kind
solution in your genius head ...

> There is no solution without statistics, unless you care to simply never see
> another live dataset in EDB.  In that case, I can provide counts all day
> long.

Here's what i'm thinking. EDB creates a "Shadow/Hidden" index
upon the query execution. The query will still be live (so that
we can edit etc) but it will cause a delay for creation the
"shadow" index. This on the fly and as an option. Eg, if someone
wants this then it will have to enable it or something but by
default will be off.

So ... EDB will work like (Table.Open or Query.Open):
1) Open the query internally
2) Makes the "Shadow" index
3) Now we can see/use the data.

Soooo, the developer will choose.

Please, put it in your plan if it's possible. (even as not high
priority)

Thank you!

--
Charalabos Michael - [Creation Power] - http://www.creationpower.gr

Michael,

<< Yeap but why non-starter ? It will defently cause a delay on large
datasets but it will not on small datasets. >>

Then what's the point ?  If you're using a small dataset, then just count
the rows yourself and display the row number value.  We're not particularly
interested in adding features to EDB that are only useful under 10,000 rows.

<< Take for an example QuantumGrid when it loads the dataset into memory in
order to have accurate scrollbar. It's pretty
fast! EDB can do it with the only difference that it will not store the
index into memory (can't imaging loading million records
into memory! but instead on the hard disk. >>

Sorry, but you're not seeing the whole picture.  There's a *lot* more to
this than what you're discussing.  You haven't even touched on the issue of
live datasets and change detection/concurrency issues.  Do you propose that
this index be rebuilt every time changes are detected in the source dataset
?

<< I trust you and i have a feeling that you have some kind solution in your
genius head ... >>

There is no solution to this issue other than index statistics.

<< Here's what i'm thinking. EDB creates a "Shadow/Hidden" index upon the
query execution. The query will still be live (so that
we can edit etc) but it will cause a delay for creation the "shadow" index.
This on the fly and as an option. Eg, if someone wants this then it will
have to enable it or something but by default will be off. >>

And changes by other users ?  Do we just ignore them and allow the logical
row numbers in this index to become corrupted and useless ?

--
Tim Young
Elevate Software
www.elevatesoft.com