Consider adding jMock custom actions to provide side effect functionality

Am I wrong, or is the JMock custom action essentially the same thing as this patch?

JMock uses 'will', I used 'runs'.

jmock: oneOf (mangoTree).pickFruit(with(any(Collection.class))); will(addElements(mango1, mango2));

mocha: mangoTree.expects(:pickfruit).runs{addElements(mango1, mango2)}

To add a use case, I mocked an API and was using some test data (tumblr posts: 5 normal, 4 pics, 3 quotes etc.)

To check things were saving as they should, I needed to mock 'save' and update an external counter every time it was called.

Reposting here from github:

Fopper,

This below is/was my use-case which was solved by Alex's branch:

I just want to make sure a method gets called on an instance. So let's say I have method foo: def foo; o = Object.new; o.some_meth; return o; end # how can I check if some_meth was called on o?

foo() in my case was called deep in some code which I had no access to directly, so I could not rely on any other side-effects to check that it was called. How would I have handled this using mocha as it stands?

@alex: I thought that one key difference was that you are allowing side effects to change anything, not just parameters passed into the stubbed method.

@eydaimon: Can you post a complete example test and implementation that demonstrates what you are trying to do and I'll see if I can help. It's much easier to discuss concrete code rather than descriptions of code.

@alex: to better understand your use case, it would be helpful to see the test and the code e.g. in a Gist

Here's some code pulled from my tumblr-sync code.

http://pastie.org/1290749

I wanted to be able to get actual numbers for saves called etc. so that I could ensure my code was processing it's info properly. 'runs' seemed like a good way to approach that, and possible other situations.

@alex: can you not use the existing Mocha expectation "cardinality" methods e.g. times, at_least, at_most?

I've had a go at an alternative version of your pastie.

I believe this should sufficiently describe it:

https://gist.github.com/fdb0907fcae297006f92

I am using Alex's code here. Not sure how I would proceed otherwise. Since i'm creating a new instance of LogMail, I can't set a flag on it to mark it sent.

Any insight?

@eydaimon: From what I can tell, you want to know that the LogMail#send method has been called. Is that correct? If so, can you not use the following :-

LogMail.any_instance.expects(:send)

This will give you an error along the following lines if the send method is not called :-

not all expectations were satisfied
unsatisfied expectations:
- expected exactly once, not yet invoked: #<AnyInstance:LogMail>.send(any_parameters)

An alternative would be to stub the LogMail.new method and return a mock :-

log_mail = mock('log_mail')
log_mail.expects(:send)
LogMail.stubs(:new).returns(log_mail)

I hope this helps. Cheers, James.

@alex: Did my suggestion help...?

That does help. thank you much James