209780 - document.getElementById converts ucs2 to utf8 too often

Status: RESOLVED FIXED

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

Description

[Axel Hecht]

This is another side effect of alecf's utf8 atoms. We compare the id atom
of a nsIXMLContent to the ucs2 encoded id value. We should convert that to
utf8 once and compare that. See the code in MatchElementId,
http://lxr.mozilla.org/seamonkey/source/content/xml/document/src/nsXMLDocument.cpp#1022

Comment 1

[Alec Flett]

ooh, this is an easy fix... just make MatchElementId take a UTF8 string instead,
and then fix the caller to create the UTF8 string early on... the real problem
is that GetElementById does a recursive descent on the dom tree. 

I'll even do the work, since I caused the regression.

Comment 2

[Alec Flett]

Attachment: speedup MatchElementId

and here's the speedup
note that I kept both the UTF16 and the UTF8 strings around - the reason is
that for HTML elements, we do a direct string compare out of GetAttr() - so we
need the UTF16 version... but when checking against the atom value, we can use
the UTF8 version.

looking for reviews.

Comment 3

[Alec Flett]

Comment on attachment 125908 [details] [diff] [review]
speedup MatchElementId

And here's the fix, pretty short - requesting reviews from dbaron/jst but would
also accept an r= from axel..

Comment 4

[Johnny Stenback (:jst)]

Comment on attachment 125908 [details] [diff] [review]
speedup MatchElementId

sr=jst

Comment 5

[Alec Flett]

fixed!