[Nitro] Idea for improving performance
reid.thompson at ateb.com
Tue Nov 22 13:33:08 EST 2005
Sam Smoot wrote:
> It doesn't mark it for Deletion in the SQL sense. A better way to
> describe it might be that it's marked for GarbageCollection. Either
> way an Update query combines the existing tuple with the new data. IOW
> you don't need to submit the full tuple.
> This idea for improving performance is actually very sound, and
> implemented by "most" major O/R Mappers around. It's called a
> UnitOfWork (http://www.martinfowler.com/eaaCatalog/unitOfWork.html).
So what is being stated is that currently Og( or an underlying layer ),
given a table with Columns N-P, when asked to perform an update
affecting only P, generates SQL along the lines of '... update
table_blah set N=Nold, M=Mold, O=Oold, P=Pnew ...' instead of '...
update table_blah set P=Pnew ...'
and that given the above is true, that by using the second form, we know
that Og ( and/or underlying layers ) will perform faster by not having
to handle the non-changing values
More information about the Nitro-general