Does that really buy anything over having separate tests? It seems to me that it would just increase clutter inside the test.
I've thought about implementing variable scope for include controllers myself, but that leads you down the load of treating jmeter like a programming language. That got me to thinking whether I really wanted to treat jmeter as a programming language. It's kind of the same reason I don't delve into adding expect/send samplers for ssh connections to the tool. It's really not in-scope for what jmeter is good at. Sure, you could shoe-horn that functionality in, but it just doesn't seem like a terribly good idea. It seems to me that a best practice for jmeter is to keep your test small, simple and focused on one piece of functionality. I've used it for pretty complex functional testing of web applications, and the tests quickly become very difficult to manage and maintain. I'm really leaning toward just wanting a robust test library that I can write tests in with a real programming language for a lot of the functional testing I'm doing. It really wouldn't take a lot of polish around a cppunit or junit framework to get me where I really want to be. But that's really not in the same nieghborhood of jmeter, at all. -- Bruce Ide [email protected]
