hi mark, you found something new, that isn't the bottleneck we saw.
regards, gerhard 2013/4/13 Romain Manni-Bucau <[email protected]> > It can be done at startup for sure, either use the same scanning than in > tomee or more cleverly use default scanning allowing it to be extended for > needed cases (noone in real life). > Le 13 avr. 2013 18:47, "Mark Struberg" <[email protected]> a écrit : > > > Romain, I think it cannot only be done at startup. What about all the > > manually executed getBeans() + resolve() + getReference() ? > > > > But all those just use the injected BeanManager where we can check it. > > > > > > LieGrue, > > strub > > > > > > ----- Original Message ----- > > > From: Romain Manni-Bucau <[email protected]> > > > To: Mark Struberg <[email protected]>; [email protected] > > > Cc: > > > Sent: Saturday, 13 April 2013, 14:50 > > > Subject: Re: getBeans() performance > > > > > > It should just be done at startup (it clearly can) then be thrown at > > > runtime if mandatory > > > Le 13 avr. 2013 14:11, "Mark Struberg" <[email protected]> a > > > écrit : > > > > > >> Hi! > > >> > > >> Leo and Gerhard recently found a performance impact by getBeans() > doing > > >> all the validations for the given qualifiers over and over again. > > >> > > >> Imo we can spare this for almost all of our internal getBeans() > > >> invocations. > > >> > > >> > > >> > > >> Wdyt: should we move this check from our BeanManagerImpl to the > > >> InjectableBeanManager? > > >> > > >> Thus if a user (or Extension) gets a BeanManager injected, he will > > always > > >> get the injectable version which will perform the checks. But we > would > > get > > >> rid of the punishment for all our internal calls, e.g. in our > proxies, > > etc. > > >> > > >> LieGrue, > > >> strub > > > >
