Failed expectation in test with no expectations

I’ve got a test like this:

	describe('.traverse/.traverseVisible/.traverseAncestors', () => {
		test('they traverse up and down the tree', () => {
			a.name = 'parent'
			b.name = 'child'
			c.name = 'childchild 1'
			c.visible = false
			d.name = 'childchild 2'

			b.add(c)
			b.add(d)
			a.add(b)

			a.traverse(function(obj) {
				names.push(obj.name)
			})

		})
	})

It fails on runtime, rtrace and configuration-warnings, but there’s not really any output on rtrace or config branches.

The only somewhat interesting output is on the runtime branch:

❌ src/as/glas/core/Object3D.spec.ts Pass: 1 / 2 Todo: 0 Time: 3.515ms
  Object3D.traverse/.traverseVisible/.traverseAncestors
    ❌ they traverse up and down the tree - 
      [Expected]: null
      [Actual]  : null

The odd part is that if you delete the traverse block, so:

	describe('.traverse/.traverseVisible/.traverseAncestors', () => {
		test('they traverse up and down the tree', () => {
			a.name = 'parent'
			b.name = 'child'
			c.name = 'childchild 1'
			c.visible = false
			d.name = 'childchild 2'

			b.add(c)
			b.add(d)
			a.add(b)
		})
	})

then the test passes.

[willemneal]

You might need to update your local version of as-pect as I didn’t change anything in your setup. I just used a log<string>(obj.name) before and after the push and it only printed the first time, so then I new that it was with how the array was allocated. And from my experience with being burned by empty array literals. The other problems that they had in the past was that since the compiler see it as a literal it lifted it up to the global scope and multiple calls to the a function that declare a literal would still point to that global array. Using the constructor is pretty much always the best way to go.

I can tell from your posts that you are attached to the syntax you’ve become accustom to. However, I think it’s a fair compromise if you get type safety and speed. Also there are others in the works. I’m currently working on implementing iterators so that you can use for (let x of arr) among other useful features that are currently unavailable. I say this not to stop you from making future issues about lack of features as it is good for new people to find, but just to expect to write less flexible syntax.

[willemneal]

I tried this out and it seems this is a runtime issue; I got the following error:

[Describe]: Object3D.traverse/.traverseVisible/.traverseAncestors

    [Fail]: ✖ they traverse up and down the tree

   [Actual]: null
               at node_modules/as-pect/assembly/internal/report/reportActual/__sendActual (wasm-function[106]:115)

 [Expected]: null
               at node_modules/as-pect/assembly/internal/report/reportExpected/__sendExpected (wasm-function[107]:150)

    [Stack]: Error: not a string: 0
               at ~lib/rt/tlsf/insertBlock (wasm-function[25]:342)
               at ~lib/rt/tlsf/reallocateBlock (wasm-function[79]:231)
               at ~lib/rt/tlsf/__realloc (wasm-function[80]:61)
               at ~lib/array/ensureSize (wasm-function[82]:66)
               at ~lib/array/Array<~lib/string/String>#push (wasm-function[97]:28)
               at start:src/as/glas/core/Object3D.spec~anonymous|0~anonymous|1~anonymous|0~anonymous|0 (wasm-function[98]:13)
               at src/as/glas/core/Object3D/Object3D#traverse (wasm-function[99]:11)

  [Result]: ✖ FAIL
   [Tests]: 0 pass, 1 fail, 1 total
    [Todo]: 0 tests
    [Time]: 2.4ms

When I changed the config to "--runtime", ["stub"], it passed. In this particular case I found that the issue was with initializing the array names with an array literal. Changing the line at the top of the file to

const names: string[] = new Array<string>()

made it pass the test. I’ve been burned by empty array literals in the past and I would avoid them. Perhaps you could make as issue with AS letting them know about this.