Set wrappedMethod on getters/setters

[rgroothuijsen]

Purpose

This PR fixes #2198 by wrapping accessors if they are present, rather than the property they belong to.

Background

Because the current wrapMethod implementation only assumes the presence of a single method on a property, trying to set both a getter and a setter will cause the former to be overwritten by the latter, and not added as separate get and set methods. Furthermore, restore() is implemented on the property itself, which does not allow the spy.get.restore() and spy.set.restore() outlined in the documentation.

Solution

This problem was addressed by changing wrapMethod to handle multiple methods per property. This includes a check whether any particular method is an accessor or not, and a method-wrapping implementation is chosen based on this. If the method in question is not a "get" or "set" method, the existing implementation will be used.

In order to satisfy ESLint code quality requirements, some restructuring of wrapMethod has been performed as well. To the extent possible, however, the existing structure has been left alone.

• restore() was moved because function declarations are not allowed inside a loop
• Parts of the code have been separated into functions because the complexity of wrapMethod had become too high

How to verify

1. Check out this branch
2. npm install
3. npm test (See included test for details)

[stale]

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.

[mroderick]

Once this is merged, we should reverse the #2270 changes!

[mroderick]

Thank you for diligently fixing this issue and for your patience with waiting for a review 🏆

I've added a few suggestions around using forEach instead of for loops.

[fatso83]

I'll be taking this across the finish line this week (not so heroic as it seems: basically just implementing the feedback).

[mroderick]

Running the tests in browsers, @hexeberlin and I discovered that this PR fails in IE11 and Safari 9

# internet explorer 11:

  1) spy
       sets wrappedMethod on getter and setter:
     AssertionError: [refute.isUndefined] Expected undefined ('undefined') not to be undefined
      at fail (node_modules/@sinonjs/referee/lib/create-fail.js:15)
      at ctx.fail (node_modules/@sinonjs/referee/lib/define-assertion.js:44)
      at assertion (node_modules/@sinonjs/referee/lib/define-assertion.js:62)
      at referee[type][name] (node_modules/@sinonjs/referee/lib/define-assertion.js:89)
      at Anonymous function (test/spy-test.js:350)

  2) stub
       .get
         can restore stubbed setters for functions:
     AssertionError: [assert.equals] undefined expected to be equal to [Function]
      at fail (node_modules/@sinonjs/referee/lib/create-fail.js:15)
      at ctx.fail (node_modules/@sinonjs/referee/lib/define-assertion.js:44)
      at assertion (node_modules/@sinonjs/referee/lib/define-assertion.js:62)
      at referee[type][name] (node_modules/@sinonjs/referee/lib/define-assertion.js:89)
      at Anonymous function (test/stub-test.js:3298)

  3) stub
       .set
         can restore stubbed setters for functions:
     AssertionError: [assert.equals] undefined expected to be equal to [Function]
      at fail (node_modules/@sinonjs/referee/lib/create-fail.js:15)
      at ctx.fail (node_modules/@sinonjs/referee/lib/define-assertion.js:44)
      at assertion (node_modules/@sinonjs/referee/lib/define-assertion.js:62)
      at referee[type][name] (node_modules/@sinonjs/referee/lib/define-assertion.js:89)
      at Anonymous function (test/stub-test.js:3385)


# Safari 9.0

  1) spy
       sets wrappedMethod on getter and setter:
     [refute.isUndefined] Expected undefined ('undefined') not to be undefined
      fail node_modules/@sinonjs/referee/lib/create-fail.js:5
      fail node_modules/@sinonjs/referee/lib/define-assertion.js:44
      assertion node_modules/@sinonjs/referee/lib/define-assertion.js:62
      node_modules/@sinonjs/referee/lib/define-assertion.js:89
      test/spy-test.js:348

Since we will drop IE11 with #2324 and subsequently allow for more modern syntax and probably require ES5 support as minimum, I propose that we delay fixing #2198 until we've had a chance to refactor wrapMethod to more modern JS.

[fatso83]

I propose that we delay fixing #2198 until we've had a chance to refactor wrapMethod to more modern JS.

Not totally sure what should happen, but we have dropped support for IE11 and Safari 9 with Sinon 10, so ... how do we progress from here? I not totally sure what is supposed to happen first.

[mroderick]

I think we should try rebasing it onto the primary branch, and see if that doesn't fix it. I don't currently have the bandwidth to do that, but might be able to look into it on the weekend.

[fatso83]

Rebased: https://github.com/sinonjs/sinon/compare/master...fatso83:SINON-2198?expand=1

Some conflicts, but I think it's ok. All tests pass, at least.

[fatso83]

Resubmitted as #2378 to get it in.