onCall method for building sequential stub behavior

[tf]

The full discussion of this change can be found in #244.

Major changes:

• Calling yields* or callArg* methods on stubs multiple times no longer defines sequential behavior but redefines the behavior.
• New onCall(n), onFirstCall, onSecondCall, onThirdCall methods to define sequential behavior. All stub behavior defining methods (including returns*, throws*, withArgs) can be used. Arbitrary combinations of withArgs and onCall work.

Code organization changes:

• Extract behavior class from parallel arrays in stubs.

[mantoni]

Impressive amount of work!

From looking through the changes it does look good to me. The only thing I noticed are some single quotes mixed with double quotes in the test cases.

I'd volunteer to run my test suites at work using a snapshot build if @cjohansen could provide one?
Maybe here: http://sinonjs.org/releases/sinon-next.js

Thanks @tf, great stuff.

[cjohansen]

http://sinonjs.org/releases/sinon-1.8.0-pre.js
http://sinonjs.org/releases/sinon-1.8.0-ie-pre.js

[mantoni]

That was quick! I've run the current test suites successfully with the pre-release against

• Chrome 30
• Firefox 23
• IE 8
• IE 9

There was one incompatibility that I found: Using yields in a chain does not work anymore.

var stub = sinon.stub();
stub.yields(3).yields(7);

var args = [];
stub(function (arg) { args.push(arg); });
stub(function (arg) { args.push(arg); });

assertEquals([3, 7], args); // throws AssertError: expected [3,7] but was [7,7]

This worked previously and would need to be changed with the pre-release:

stub.onFirstCall().yields(3).onSecondCall().yields(7);

I do like this because it reads a lot better and I'm not sure whether I used an "official" part of the API.

@cjohansen Do you think we need to support the chained yields notation?

[cjohansen]

@cjohansen Do you think we need to support the chained yields notation?

Well, it depends on how much it is used. I'm very reluctant to break people's test suites :/

[mantoni]

Hard to tell. It's not in the documentation and personally, I used it once. But can't speak for others, but I'd be fine with breaking this since we have an alternative notation to achieve the same thing. Could be stated in the release notes as a caveat.

[tf]

If you take a look at my first post in the original discussion #244, this incompatible change actually was the whole point which got me started on this issue. The chaining behavior was introduced in some 1.x minor release and is undocumented. So I'd be very much in favor of changing it back. Please take a look at my first post in #244 again to see my use case which is not possible with the chaining API.

[cjohansen]

You're right. sorry!

[cjohansen]

...and thanks for all your work on this!

[tf]

Great to see this got merged! Is there a recommended process for contributing to the documentation? Shall i prepare a pull request to the doc site repository? How shall the sinon version requirement for the onCall feature be expressed?

[mantoni]

You can send a documentation pull request here: https://github.com/cjohansen/sinon-web
I guess the version will be 1.8.0.