cdp: dispatch DOM.setChildNodes event for search results

[krichprollsch]

No description provided.

[karlseguin]

Yes.

If it's something we'll need to do often, we could expand the NodeWriter. It already takes a placeholder option, so it could be something like:

.nodes = bc.nodeWriter(parent_node, .{.mode = .child_list})},

[krichprollsch]

We have to dispatch the setChildNodes event for the whole hierarchy, starting from the root until the nodes.
I tried an implementation but I have a leak error with the unit tests.

[karlseguin]

We have to dispatch the setChildNodes event for the whole hierarchy, starting from the root until the nodes. I tried an implementation but I have a leak error with the unit tests.

Really? I was pretty sure it was just the node's sibling, or, maybe more accurately, the node's parent children.

I don't see a leak, both in the CI and locally, it's an unreachable which is being hit.

const parent_node = bc.node_registry.lookup_by_node.get(p) orelse unreachable;

I don't think there's any reason to assume the parent has been seen by the registry before. Always better to use register, which does a getOrPut

const node = try bc.node_registry.register(parser.documentToNode(doc));

After that, you'll get a crash because the session_id isn't set. I would do 2 things here:

1 -
Be defensive and add a check after loading `bc.

var bc = ...
const session_id = bc.session_id orelse return error.SessionIdNotLoaded;

2 - Change the cdp/testing.zig loadBrowserContext and set a default session_id if there's html:

// existing if statement
if (opts.html) |html| {
   // add
   if (bc.session_id == null) bc.session_id = "SID-X";
   ..
}

[krichprollsch]

The issue is fixed. I added a bool into cdp.Node to know if a node has been already dispatched.

Really? I was pretty sure it was just the node's sibling, or, maybe more accurately, the node's parent children.

Yes it is as far as I can see. ..