-
1. Re: Weld boot performance
dmlloyd Nov 23, 2010 5:42 PM (in response to swd847)Sure, but I'd say get the integration working first so you have a baseline.
Looking at it from another angle, another possible solution is to reduce the total number of classes loaded at boot. Maybe the new annotation processing architecture can help here.
-
2. Re: Weld boot performance
jason.greene Nov 23, 2010 5:46 PM (in response to dmlloyd)David Lloyd wrote:
Looking at it from another angle, another possible solution is to reduce the total number of classes loaded at boot. Maybe the new annotation processing architecture can help here.
Right this is an important distinction. It's likely that a only subset of classes in a deployment will be of interest to weld during deployment.
-
3. Re: Weld boot performance
swd847 Nov 23, 2010 6:05 PM (in response to jason.greene)I may be getting a bit ahead of myself thinking about performance at this stage :-)
The problem is that according to the spec Weld has to fire ProcessAnnotatedType for every class in the bean archive. Weld does have a non-portable extension to limit the classes that are scanned, but in the general case there is not really any way around this. Even if you could build the meta model from other sources without loading the class (which I am not sure is possible), as soon as portable extensions enter the mix you are almost certainly going to have to load the class anyway.