[Moose-dev] Re: verveinej out of memory

wrote:

just when I was going to say that it is working and I corrected the bug you had when parsing ArgoUML :-( (it takes 1 hour to parse it all :-( Something I will have to look into without taking too much memory) Memory is indeed a problem currently and unfortunately there is not much I can do, since a lot of it is used by JDT's parser. There are few possible solutions: - there is a limit to the max. memory the JVM can use, but on my computer iti is at least 2500m. You specified only 2000m, so there is a possibility that the extra 500m will be enough... - incremental parsing: VerveineJ generates an output.mse file. When it starts, it check if such a file already exist. If so, it loads it in memory (silently, this is not a good idea) and new entities will be added to the existing ones. so you can call it several time with various parts of the project to parse. If the classpath is correctly specified, it should be able to do the name resolution correctly. (I am not sure how much this can improve things ...) - I could start to implement options reducing the level of detail of the output: --noFileAnchor, --noInvocation, --noLocalVariable, ... ? If it is an option in this case, it would help. - finally, if all else fails, there is always Markus' theory that they will always build newer, faster, bigger, computers. In 18 months, we will have twice as much memory to run verveineJ ... nicolas PS: I will try your test case ASAP (but it might be difficult this week) ----- Mail original ----- _______________________________________________ Moose-dev mailing list Moose-dev(a)iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev