On Tue, 2007-03-27 at 13:24 -0600, John Stanton wrote:
> > Another reason for my puzzlement -- although I love SQLite, my
> > expectations are based on using Foxpro for many years. Foxpro's indexing
> > speed for a problem like this is about 10 - 20 times faster. And I've
> > never come across a Foxpro database where the indexing took longer than
> > the loading -- and Foxpro is blazingly fast at loading. So I assumed
> > that every DBMS would be faster at indexing than loading. (Both use
> > B-trees for indexes, so I believe it's a meaningful comparison. But
> > maybe in this case the single-file architecture of SQLite works against
> > it; Foxpro uses a binary format for its B-trees.)
> > 
> > In other words, since SQLite is so fast at some things, I expect it to
> > be fast at all things. Is this unreasonable? Is it optimized for fast
> > retrieval and not indexing?
> > 
> > 
> > Thanks!
> You are comparing an ACID RDBMS with rollback and commit with a much 
> simpler situation.  For example we developed a data storage software 
> product which was of the same generation as Foxpro. I wrote a fast 
> indexing program which would create a 10 million entry B-Tree index in 
> less than a minute on a very slow machine, but it had minimal features, 
> unlike Sqlite.

I do all the indexing in a single transaction, which I thought would
have turned off any rollback potential, and also saved time. 


> I assume that if Foxpro had all the features you want now you would not 
> be changing from it.

Of course. I'm just trying to learn what I don't know  :-)


> Have you thought of doing the index creation as a background process 
> unseen by the user?

Actually, it already is a separate process launched on the server by the
web app. But since users can't use the system until the index is built,
I show them the progress of the indexing so at least they have something
to look at!


Thanks,
Stephen
 

Stephen Toney
Systems Planning
[EMAIL PROTECTED]
http://www.systemsplanning.com


-----------------------------------------------------------------------------
To unsubscribe, send email to [EMAIL PROTECTED]
-----------------------------------------------------------------------------

Reply via email to