ChildNode.nextSibling should be another ChildNode, not just a Node

[jcormont]

TypeScript Version: 3.1.6

Search Terms: ChildNode Node nextSibling

Code

// a basic loop over elements in the DOM
let element = document.body.firstChild;  // is a ChildNode | null
while (element) {
    // ... do something with element
    let sibling = element.nextSibling;  // is a Node | null -- why?
    element = sibling;  // uh-uh!
}

Expected behavior:

Surely a sibling of a child is itself also a child (of the same parent). So siblings should be assignable using the same variable.

Actual behavior:

Error:

Type 'Node | null' is not assignable to type 'ChildNode | null'.
  Type 'Node' is not assignable to type 'ChildNode'.
    Property 'after' is missing in type 'Node'. [2322]

Related Issues:

I think this is what #27453 also refers to, but fails to describe accurately so it has been closed.

In any case this is related to the 'solution' in #24633 which clearly shows the type of nextSibling as Node, not ChildNode.

[weswigham]

PRs welcomed. You can find more information about contributing lib.d.ts fixes at https://github.com/Microsoft/TypeScript/blob/master/CONTRIBUTING.md#contributing-libdts-fixes.

[saschanaz]

Current ChildNode is a hack, it should really be a union of DocumentType | Element | CharacterData.

[Fabianopb]

Updating nextSibling: Node | null to nextSibling: ChildNode | null sounds reasonable.

ChildNode extends Node (which extends EventTarget) so I imagine it won't be a breaking change, just adding the correct type. @saschanaz why do you think ChildNode should be this union type you mentioned?

Btw, I'd be glad to grab this story if you guys haven't started yet!

[Fabianopb]

PR opened, maybe you guys want to continue the conversation in there.

[saschanaz]

why do you think ChildNode should be this union type you mentioned?

No, I think the type of firstChild property should be, not ChildNode itself. ChildNode is by the DOM spec a mixin that adds some child-node-only methods and should only be used in extends part.

microsoft/TypeScript-DOM-lib-generator@5cd35f3 introduced the deviation.

[Fabianopb]

I wonder what was the reason then to make ChildNode extend Node in that commit. As a comparison, ParentNode seems by the spec. Maybe @mhegazy remembers what was the story?

[saschanaz]

I wonder what was the reason then to make ChildNode extend Node in that commit.

I think he just wanted to remove Node & ChildNode.

[Fabianopb]

microsoft/TypeScript-DOM-lib-generator#619 is merged, should this be closed?

[jakebailey]

Just looking at ancient issues; yes, the above was merged and then brought over in #29690.