"alesj" wrote : &quot;adrian(a)jboss.org&quot; wrote : | | Let's go with that, but we need to do so extra work to see whether all this | | caching is (still) effective. | | | But won't this interfere with this: | &quot;adrian(a)jboss.org&quot; wrote : | | Also I think it will cause a memory leak. | | | | You effectively have: | | | | | WeakValueHashMap<ClassInfo, BeanInfo> | | | | | but the ClassInfo could itself get a hard reference on the BeanInfo | | via one of its "attachments", e.g. the caching of SchemaBindings in jbossxb | | against the TypeInfo. | | | I assume you're using a WeakHashMap<ClassInfo, Map> not a WeakValueHashMap as originally proposed? The ClassInfo reference is now weak. anonymous wrote : If not, I'll change it + add some caching tests. :-) Test the contract not the implemenation details. ;-) Otherwise, you'll only have to rewrite the tests when you change the caching alogrithm or some other detail in future - which kind of makes them pointless. See my earlier printCollection code. This tests the contract and would be a valid test with or without caching. Without caching, it would just be that the Bean/ClassInfo instances are different == (but equivalent .equals) at each stage. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4190981#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...