Hello, I’m currently dealing with a Java app using a Groovy DSL and I’m running into some PermGen problems. I'd like to use caching in order to avoid evaluating the scripts again ans again. However I'm struggling with the lack of documentation on this topic.
Here's my situation. A Groovy DSL is received as String and passed on to a new GroovyShell instance like this : GroovyShell shell = new GroovyShell(Thread.currentThread().getContextClassLoader(), binding); Object result = shell.evaluate(sourceCodeAsString); Basically we receive the same DSL until a newer version is pushed. The only difference between the calls is the binding. Each call is given its own shell, contextClassLoader and binding. I do understand that each call to evaluate() will put the compiled script into the class loader. However until a notification of a newer version of the DSL is sent there's no need to recompile the same source over and over. It only makes the class loader bigger. So my question is what are the best practices in order to : - Cache the Groovy DSL until a newer version is detected (in a GroovyClassLoader ?) - Remove the previous version from the cache and from the class loader And also : - Is it possible/acceptable to share a GroovyShell (or a pool of GroovyShell) among the different calls ? (thread safety issues, etc…) - Same question for a GroovyClassLoader (as singleton) Thanks in advance. Damien
