Query API for alfa-dom

[rcj-siteimprove]

Part of #1037

First attempt at creating a query API for querying the DOM. For now, two common queries has been extracted and put into the new Query namespace in alfa-dom and the results are now cached which should give a slight performance improvement.

[Jym77]

Great work. Some small polish.

[Jym77]

We should also use the new cached map for aria-labelledby resolution at

alfa/packages/alfa-aria/src/name.ts

Lines 890 to 903 in b1e42be

	// Since there are a lot of elements in the document, but very few in the
	// aria-labelledby, it is more efficient to grab them in DOM order (in a
	// single traversal) and then sort by tokens order rather than grab the ids
	// one by one in the correct order.
	const references = root
	.elementDescendants()
	.filter(hasId(equals(...ids)))
	.sortWith((a, b) =>
	Comparable.compareNumber(
	// the previous filter ensure that the id exists
	ids.indexOf(a.id.getUnsafe()),
	ids.indexOf(b.id.getUnsafe())
	)
	);

(and it should also simplify that code by not needing to sort since we'll poke the map in the correct order)

[rcj-siteimprove]

We should also use the new cached map for aria-labelledby resolution at

alfa/packages/alfa-aria/src/name.ts

Lines 890 to 903 in b1e42be

	// Since there are a lot of elements in the document, but very few in the
	// aria-labelledby, it is more efficient to grab them in DOM order (in a
	// single traversal) and then sort by tokens order rather than grab the ids
	// one by one in the correct order.
	const references = root
	.elementDescendants()
	.filter(hasId(equals(...ids)))
	.sortWith((a, b) =>
	Comparable.compareNumber(
	// the previous filter ensure that the id exists
	ids.indexOf(a.id.getUnsafe()),
	ids.indexOf(b.id.getUnsafe())
	)
	);

(and it should also simplify that code by not needing to sort since we'll poke the map in the correct order)

There's also a follow up task on #1037 on getting rid of all usages of .elementDescendants() so it would be picked up there too. Should we keep it as part of that follow up task?

[Jym77]

We should also use the new cached map for aria-labelledby resolution at

alfa/packages/alfa-aria/src/name.ts

Lines 890 to 903 in b1e42be

	// Since there are a lot of elements in the document, but very few in the
	// aria-labelledby, it is more efficient to grab them in DOM order (in a
	// single traversal) and then sort by tokens order rather than grab the ids
	// one by one in the correct order.
	const references = root
	.elementDescendants()
	.filter(hasId(equals(...ids)))
	.sortWith((a, b) =>
	Comparable.compareNumber(
	// the previous filter ensure that the id exists
	ids.indexOf(a.id.getUnsafe()),
	ids.indexOf(b.id.getUnsafe())
	)
	);

(and it should also simplify that code by not needing to sort since we'll poke the map in the correct order)

There's also a follow up task on #1037 on getting rid of all usages of .elementDescendants() so it would be picked up there too. Should we keep it as part of that follow up task?

That's a bit different (I think).
The aria-labelledby computation is actually doing a convoluted getElementIdMap, not just a elementDescendants. It is a bit hidden because it is already a bit optimised. But given that this PR is about introducing getElementIdMap (the renaming of elementDescendants is a side effect), I feel this fits here 🤔

The aria-labelledby computation is essentially doing:
document => element descendants => filter by id being one of the wanted one => sort by order of id in the attribute.
The third step is essentially a multiple getElementById and should benefit from the map:
document => idMap => pick the wanted ones, in order.

So, I feel it's more about using the new map that replacing the elementDescendants call (i.e. lust using Query.elementDescendant wouldn't be enough).