Make Event.path() return a pre-calculated NodeList rather than calculating it dynamically.

Make Event.path() return a pre-calculated NodeList rather than calculating it dynamically.

The current implementation of Event.path() populates a NodeList dynamically when Event.path() is called.
That has a drawback: If an event handler mutates a DOM tree, the returned value of event.path() will be changed.

This patch pre-calculates a NodeList for each node so that event.path() should not be affected by mutating DOM tree.

BUG=240162
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=150293

[hayato, 2013-05-13 08:53:34]

PTAL.

[dglazkov, 2013-05-13 16:39:30]

Keeping a NodeList on each EventContext and the N^2 populating of these lists
seems really heavy. I am worried that we're going down the wrong "path" with
this API (pardon the pun).

[hayato, 2013-05-14 02:42:06]

On 2013/05/13 16:39:30, Dimitri Glazkov wrote:
> Keeping a NodeList on each EventContext and the N^2 populating of these lists
> seems really heavy. I am worried that we're going down the wrong "path" with
> this API (pardon the pun).

It could be. But I did one optimization so that nodes in the same scope will
share the *same* NodeList. So in practice, it does not take much time.

NodeList is created for each scope, rather than for each node, in the patch.

[dglazkov, 2013-05-14 03:04:37]

okay, lgtm.

[hayato, 2013-05-14 03:47:38]

On 2013/05/14 02:42:06, hayato wrote:
> On 2013/05/13 16:39:30, Dimitri Glazkov wrote:
> > Keeping a NodeList on each EventContext and the N^2 populating of these
lists
> > seems really heavy. I am worried that we're going down the wrong "path" with
> > this API (pardon the pun).
> 
> It could be. But I did one optimization so that nodes in the same scope will
> share the *same* NodeList. So in practice, it does not take much time.
> 
> NodeList is created for each scope, rather than for each node, in the patch.

At the same time, I am afraid if we have a spec which requires that event.path()
should be pre-calculated, it makes an efficient-implementation difficult.

In blink, this patch might be okay, but for other browsers, I am not sure the
spec will be acceptable one.

We should balance somehow.

[hayato, 2013-05-14 03:48:36]

On 2013/05/14 03:04:37, Dimitri Glazkov wrote:
> okay, lgtm.

Thank you. Let me land this patch and watch the perf bot result.

[commit-bot: I haz the power, 2013-05-14 03:48:53]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/hayato@chromium.org/15063004/4007

[commit-bot: I haz the power, 2013-05-14 05:32:38]

Change committed as 150293