193212 - Only use CJK line breaker algorithm for CJK text

Status: RESOLVED INCOMPLETE

(Core :: Layout: Text and Fonts, defect)

Description

[Simon Montagu :smontagu]

This is an idea I experimented with last year, but didn't find that it was much
of a performance win for pageload. Now I want to investigate its impact on
typing and editing performance, where PrepareUnicodeText() consumes a lot of
time (see figures in bug 35296).

I will resurrect my patch, unbitrot it and attach it here. The general strategy
is to set a flag in nsGenericDOMDataNode.cpp when CJK (Chinese, Japanese or
Korean) characters are present, and only use the CJK linebreaker in
nsTextTransformer when this flag is present. Otherwise we will use a simpler
linebreaker which will not have to scan for CJK characters internally.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Will this change make the code simpler?  If not, could we add some refactoring
in the process?

Comment 2

[Simon Montagu :smontagu]

In principle, yes, but in the first instance I want to produce a
simple-as-possible patch for use in performance profiling. 

Comment 3

[Simon Montagu :smontagu]

Attachment: Prototype patch

This patch has at least one serious bug: the nsWesternLineBreaker can end up
being used even on CJK pages, because the first call to
nsDocument::GetLineBreaker may be before the parser has encountered any CJK
characters.

Comment 4

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

It seems to me that bug 255990 suggests pretty much the opposite (and makes more sense to me).

Comment 5

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last 7 months.
:jfkthame, could you have a look please?
For more information, please visit auto_nag documentation.

Comment 6

[Jonathan Kew [:jfkthame]]

The code involved here has all changed; I don't think this is relevant any more.