310236 - Cache nsAttrValue* for class attribute

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect)

Description

[Brian Ryner (not reading)]

We can speed up nsCSSRuleProcessor::RulesMatching() by only fetching the class
attribute for the element once, instead of on each trip through SelectorMatches().

Comment 1

[Brian Ryner (not reading)]

Attachment: patch

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 197625 [details] [diff] [review]
patch

>Index: content/base/src/nsAttrValue.cpp
>+      nsIAtom* atom = NS_STATIC_CAST(nsIAtom*, GetPtr());

This should be nsIAtom* atom = GetAtomValue();
(It was that way in the code you moved.)

>+        PRInt32 i, count = array->Count();
>+        for (i = 0; i < count; ++i) {

This would be better all inside the for():

for (PRInt32 i = 0, count = array->Count(); i < count; ++i) {

Comment 3

[Brian Ryner (not reading)]

checked in

Comment 4

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

What is the lifetime of the RuleProcessorData? The nsAttrValue pointer is only
valid until someone adds or removes children or attributes from the element.

Comment 5

[Boris Zbarsky [:bzbarsky]]

> What is the lifetime of the RuleProcessorData?

Until we finish resolving style for this node.  It's basically stack-allocated
for the duration of style resolution; no document mutations can happen while one
is alive.

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

It's either stack allocated or heap-allocated and owned by a stack-allocated
object, so things should be fine.  The lifetime is the lifetime of the
nsStyleSet::{ResolveStyleFor,ResolvePseudoStyleFor,ProbePseudoStyleFor} function
call.