I haven't done any speed optimization or even profiled the code, so that looks really good to me. I think optimization might help with parameter binding and type conversion.
Did your tests include parameters? Could you run your tests against PyGreSQL from which pypg is derived? That would show a clearer picutre of the speed difference between C/Python for this application. Randall Stephane Bortzmeyer wrote: > On Tue, Jul 12, 2005 at 11:41:43PM -0500, > Randall Smith <[EMAIL PROTECTED]> wrote > a message of 52 lines which said: > > >>I put in ident support and Unix socket support as well as taking the >>username from the process when it is not specified. > > > OK, it works much better now. > > Performance tests can be done and they seem to show that pypg is > slower but you may decide by yourself if it is *too* slow: > > % python test.py > Result with <module 'psycopg' from > '/usr/lib/python2.3/site-packages/psycopgmodule.so'> > 100 repetitions: > INSERT: 0.4 msec/pass > SELECT (with join): 0.6 msec/pass > DELETE: 0.3 msec/pass > > % python test.py > Result with <module 'pypg' from > '/usr/local/lib/python2.3/site-packages/pypg/__init__.pyc'> > 100 repetitions: > INSERT: 0.7 msec/pass > SELECT (with join): 1.1 msec/pass > DELETE: 0.6 msec/pass _______________________________________________ DB-SIG maillist - DB-SIG@python.org http://mail.python.org/mailman/listinfo/db-sig
