Sinon: One-liner conditional stub returns an unconditional stub

This has probably snuck in with the withArgs modifier. I agree that the current behavior is less than optimal and should probably be changed. However, that would be a _very_ breaking change, and is not likely to happen anytime soon.

The reason for this behavior is the following:

sinon.stub().withArgs("foo").returns("bar").withArgs("fooz").returns("ball")

This is a great illustration of the pitfalls of highly chainable APIs, and in this case,Sinon has unfortunately opted to support the wrong case IMO.

Anyway, as I said, too many people depend on this functionality at this point to allow us to just "fix" it, but we certainly should consider fixing it long term.

I understand the tension and appreciate your candor (agreeing but also favoring stability).

I think in the short-term what I'll do with my own project using Sinon.js is to wrap this API with functions that configure the narrow set of doubles I want to see, which will work around this issue.

The only recommendation I have for you as a maintainer is perhaps to consider an alternative API for setting up a one-and-done stub configuration.

Could this be addressed in a major version bump? I have many of these two-liner instances throughout my test suite that I would much prefer to make one-liners.

And I'll toss in my _suspicion_ that there are more people who would prefer withArgs/returns to return the narrow stub than there are people who do multiple withArgs chaining. I don't think I've ever seen multiple withArgs in a single chain in the wild, though I encounter the two-line method frequently. [confirmation bias alert]

If we can come up with a reasonable name for it (calledWithArgs?), then I can't see why we can't have both styles present in the API. Deprecate the old style in favour of the new hotness, but leave it in until at least the next major version.

Having both styles will be confusing for newcomers, unless we improve documentation around these features, which is probably long overdue anyway.

Aside: I agree method chaining is often the cause of headaches. It's very difficult to incrementally change design, when everything is dependent on every little detail of the API being the same. Since we've gone this far down the rabbit hole, we might as well keep going ...

Hey @cjohansen after looking at your example one more time:

stub = sinon.stub().withArgs("foo").returns("bar").withArgs("fooz").returns("ball")

I tried punching this in to see what it would do:

stub() // => "ball"
stub("foo") // => "ball"
stub("fooz") // => "ball"

Anyone using a long chain like this is also likely to be surprised that the last-returns-in-wins and that it does so without regarding withArgs, right?

After looking at this example I'm struggling to imagine an example of someone relying on a long chain like this one and actually desiring this behavior. If I had to bet, I'd say this change would break a lot of tests for the better (because they're currently false positives from too-permissive of stubs)

Wow, that is interesting. Seeing this I can't even understand what the point of withArgs is? :confused: This is definitely an argument for just making this behave correctly.

Thanks for keeping an open mind to that @cjohansen I really appreciate it. What's the best way to proceed from here? Would you be open to a PR?

This is somewhat related to https://github.com/cjohansen/Sinon.JS/pull/823.

@searls sure, a PR would be great. The current behavior is completely broken for chains defining more than one behavior anyway.

Cool! I'll schedule some time to look at this next week

This chaining works sometimes:

findStub = sandbox.stub()
            findStub
                .withArgs(...)
               .returns(..)
              .withArgs(...)
               .returns(...)

This does not works properly:

findStub = sandbox.stub().withArgs(...)
               .returns(..)
              .withArgs(...)
               .returns(...)

@cjohansen I'm really sorry for being "that guy" but I've shifted gears away from using Sinon.js for the project I'm working on, so I won't be tackling this with a PR on my own. Feel free to close/deprioritize accordingly based on your own interests, of course.

@cjohansen I'm really sorry for being "that guy" but I've shifted gears away from using Sinon.js for the project I'm working on, so I won't be tackling this with a PR on my own. Feel free to close/deprioritize accordingly based on your own interests, of course.

Thank you for getting back on this! I'll apply the help-wanted label, and see if it'll grab the attention of someone with time to take a deeper look into this.

Also, thank you for your investigation into this issue!

pay4bugz ;-)

Whoa whoa. There is a property on 'withArgs' modified stubs called 'parent' which refers back to the root of the stub.

When I do my chaining I just provide explicitly that I want to pass back up the chain.

var stubby = sinon.stub().throws()
.withArgs("foo").returns("bar").parent
.withArgs("foo2").returns("bar2").parent;

Works pretty well for me. Worried your fix will be a breaking change for me :(

@launchcg-ztonia You're depending on implementation details that are not part of the official API. This will almost certainly break at some point.

Here, I want to assert that callback function is called or not. So how can I do it using sinon js. Please suggest.

var send = function (templateId, callback) {

    const request = mailjet
        .post("send")
        .request(params)
    request
        .then((result) => {
            if (typeof callback === 'function') {
                callback(null, result.body);
            }

        })
        .catch((err) => {
            if (typeof callback === 'function') {
                callback(err, null);
            }
        })
    } else {
        callback(err, null);
    }
};

var mailjetClient = require('../../node_modules/node-mailjet/mailjet-client');

sinon.stub(mailjet, 'post').withArgs('send').returns(mailjetClient);
sinon.stub(mailjetClient, 'request').returns(Promise);
I am getting following error:

TypeError: Attempted to wrap undefined property request as function

@apoorva-shah: the javascript output is enough to help you. there is no property called request to stub on that object. and please don't use this issue tracker for usage questions: try stack overflow or our Gitter channel.

It would be great if someone could summarise the result of this discussion as a failing test case added to issues-test.js. Then we would know if we accidentally fix it ;)

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

Any chance this will be re-considered in a near-future release? This still appears to be an issue.

Any chance this will be re-considered in a near-future release?

The issue has been around for years, and we are not seeing much traction for anyone trying to fix it. There is a considerable complexity in the stub api.

I've been working on a new API, that is much simpler #1586. If you're curious, you could try it out and leave feedback on that pull request.

@mantoni 2018, still working great :)