reset resets only a spy but not a stub
Consider this code:
var sinon = require('sinon');
var stub = sinon.stub();
stub.withArgs(1).returns('Bye');
stub.reset();
stub.returns('Hello');
console.log('stub(1)', stub(1));
it prints
stub(1) Bye
if I change stub.reset(); to stub.resetBehavior(); it works as expected, i.e. prints stub(1) Hello.
But resetBehavior is not documented (#267) and it would be more consistent and natural if reset works for both spies and stubs.
Another option is to change stub.reset(); to stub = sinon.stub();.
But It is not always feasible to create a new stub in each test. Consider this example:
var stub = spy.stub();
var obj = createExpensiveObject(stub);
it('test1', function() {
// setup stub in some way
obj.test1();
});
it('test2', function() {
// setup stub in a different way
obj.test2();
});
dotchev
All 8 comments
This ticket is a bit convoluted. Are you reporting a bug or requesting a change?
Stubs are intentionally cheap to create, so if you want fresh behaviour, just create another one?
mroderick
on 17 Aug 2015
I would consider it a change request, which is
If
reset()is called on a stub, it should reset it to an initial state
One could consider it also a design bug because a stub (being also a spy) has a reset() function but it only resets only the spy part of it.
While stubs are cheap to create, the objects where one needs to inject them might not be so cheap. So just creating a new stub might not be enough if need also to recreate the object where you inject the new stub.
See my example above. I need to use a different stub in each test (which is usually the case). To do this now, I have to call createExpensiveObject for each test, just to pass it the new stub.
dotchev
on 20 Aug 2015
Did this just take away a feature? Before you could have acheived this by calling:
stub.reset();
stub.resetBehvaior();
Now you can only call .resetBehvaior() alone. You cannot reset the call counts without resetting the behavior. This seems like a regression.
fearphage
on 29 Sep 2015
Now you can only call .resetBehvaior() alone. You cannot reset the call counts without resetting the behavior. This seems like a regression.
That's a good point.
Perhaps it would be better if we expanded the api a bit.
// resets both call history and behaviour
stub.reset();
// reset call history
stub.resetHistory();
// reset behaviour
stub.resetBehavior()
Seems reasonable
dotchev
on 30 Sep 2015
@mroderick That seems solid to me.
fearphage
on 1 Oct 2015
So that the idea is not lost, I've created issue #863 for adding .resetHistory to the api.
mroderick
on 1 Oct 2015
This change should be called out in the release notes as a breaking change. This is a change to the expected behavior of .reset().
fearphage
on 1 Oct 2015
Related issues
fearphage
·
4Comments
byohay
·
3Comments
optimatex
·
4Comments
NathanHazout
·
3Comments
akdor1154
·
4Comments
Most helpful comment
That's a good point.
Perhaps it would be better if we expanded the api a bit.