By turning this into a loop instead of a some we can return early inside the loop rather than trying to return after the some has already gone through all elements of the stack.

Adding virtual node info to the stacking order allows me to use that virtual node to create the stacking context object for the node.

I took this opportunity to do something I've been wanting to do for awhile now which is convert the Color constructor function into a true class. The advantage to this was that I was able to do deepEquals on two Color objects whereas before that was not possible (thus greatly simplifying the tests). Another advantage is that all Color objects now share memory for all their functions thanks to the prototype whereas before each Color object had their own memory instance for each function.

When would this ever not be defined? All calls I can see pass something in. Possibly an empty array, but not null or undefined?

I wanted to make it optional since the foreground color check will use this function to apply opacity but not call the elmStack itself.

We'll want to do some performance testing of getBackgroundStack to see if it's an expensive call that warrants caching the result for when foreground color needs it.

Ran some performance tests using the following code:

performance.clearMarks();
performance.clearMeasures();
performance.mark('start');
stack = axe.commons.color.getBackgroundStack($0);
performance.mark('end');
performance.measure('time', 'start', 'end');
console.log(stack, performance.getEntriesByName('time')[0].duration)

On https://js.tensorflow.org/api/latest/, the first call to color.getBackgroundStack() takes < 1ms, with subsequent calls taking < 0.2ms.

On another page (https://api.freshservice.com/v2/#view_all_vendors), the results were similar with the first call taking 2ms while subsequent calls only took < 0.6ms.

Note that these tests were done after the grid was created so that grid creation didn't throw off the first call.

I also tried calling it on different elements then returning to the first to see if that impacted the time, but it showed similar results.

function randInt(min, max) {
  return ((Math.random() * (max - min + 1)) | 0) + min;
}

const node = $0;
const elms = document.querySelectorAll('*');
for (let i = 0; i < 100; i++) {
    const elm = elms[ randInt(0, elms.length - 1) ];
    const stack = axe.commons.color.getBackgroundStack(elm);
}

performance.clearMarks();
performance.clearMeasures();
performance.mark('start');
stack = axe.commons.color.getBackgroundStack(node);
performance.mark('end');
performance.measure('time', 'start', 'end');
console.log(stack, performance.getEntriesByName('time')[0].duration)  // 0.200