Re: [Gnumed-devel] Gnumed on Windows: the pain of pyPgSQL

[Tim Churches]

Karsten Hilbert wrote:

> For me the one reason to switch would be to gain
> performance. Psycopg is said to be the one that's fastest.
> So, *if at all*, I'd consider that one.
>  
Is DB-API adaptor performance an issue with GNUmed, since you are only 
loading and updating the records for one patient at a time. I doubt that 
teh speed of the database adaptor has any measurable impact on that. If 
the database is slow, it may be because of the schema design or 
inadequate indexing. The database adaptor speed *does* matter if you are 
doing large aggregate queries of 100,000 of records at a  time - in 
which case I agree, pyPgSQL is a bit slow. But even in our public health 
surveillance application we are getting update speed in transactions of 
several thousand records per second, and we honly have several thousand 
records at a time to update, so changing to a faster adaptor might save 
us perhaps a second at each update (every hour). In GNUmed's case, you 
might save a few milliseconds per transaction. But the best thing to do 
is to measure the speed difference in osm real-life transations.

Tim C