289311 - [FIX]Make the HTML content sink always preappend

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect, P1)

Description

[Boris Zbarsky [:bzbarsky]]

This came out in the course of working on bug 286000.  The HTML content sink
currently calls AppendChildTo on nodes when it parses the closing tag, except
when mPreAppend is set.  Since BindToTree is always deep, this means that
content tree construction becomes O(N*D) where D is the depth  of the tree
(since each node gets bound a number of times equal to the number of its ancestors).

The solution is to just switch to always preappending.  This is what the XML
sink does anyway, so it's preferable from that point of view.  It should help
address small the Tp regression introduced by bug 286000.

I checked over our impls of AppendChildTo/InsertChildAt, and it looks like they
should all deal fine with this change (especially, since we did already have
modes where we were preappending).

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Proposed patch

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: Same as diff -w

Comment 3

[Boris Zbarsky [:bzbarsky]]

I think we want to take this for beta2 to deal with this Tp regression....

Comment 4

[Johnny Stenback (:jst)]

Comment on attachment 179865 [details] [diff] [review]
Same as diff -w

sr=jst

Comment 5

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 179865 [details] [diff] [review]
Same as diff -w

Requesting approval for this Tp regression fix.  This has the added bonus of
making our HTML and XML sink do things in the same order.

Comment 6

[Asa Dotzler [:asa]]

Comment on attachment 179865 [details] [diff] [review]
Same as diff -w

a=asa for 1.8a2 checkin.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Fix checked in.  With this patch, the cumulative Tp effect of all the BindToTree
stuff was a slight win (maybe 1%).