903912 - Rip out most of the old parser

Status: RESOLVED FIXED

(Core :: DOM: HTML Parser, defect)

Description

[Blake Kaplan (:mrbkap) (inactive)]

The old parser code is unused except for about:blank and yet it sits in the tree, hurting the eyes of passers-by, taking time to compile, and using otherwise-useful space on b2g (about 40-50kb by my measurements). It is much easier to generate blank documents than what we do now.

It would be ideal to switch things over to use the new HTML parser for about:blank, but while we wait, this makes life a little nicer for everybody.

I couldn't figure out a nice way to make small patches out of this. Hopefully it should be pretty easy, though, as the majority of it is simply removing never-reached code.

Comment 1

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Gut CNavDTD and the DTD utils.

This patch also removes some code from nsElementTable.{h,cpp} in order to make
the DTD utils possible. Note that this won't compile alone because I forgot to
change nsHTMLContentSink here (I ended up doing that in the tokenizer removal
patch).

Henry, let me know if you don't want to review this and I'll find someone else to.

Comment 2

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Gut the tokenizer.

This allowed me to get rid of the DTD utils entirely (which involved moving
some stuff to nsElementTable temporarily). Unlike the DTD patch, the tokenizer
literally does nothing now.

Comment 3

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: A small cleanup patch to make the DTD even simpler.

Note the one bug that tryserver found in my patches: If we are given a content
sink that isn't an nsIHTMLContentSink, we need to return
NS_ERROR_HTMLPARSER_STOPPARSING from CNavDTD::BuildModel. This happens due to
(I think) XBL files that don't exist. I haven't tracked down why yet, but I'll
file a followup on fixing that.

Comment 4

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Minify the element table.

There's only one user of nsElementTable: the HTML editor. We should fix that,
but in the meantime, this strips the element table down to answer only the
questions on nsIParserService. I generated most of this diff (the actual table
changes) with a python script that I'm happy to attach here if wanted.

Comment 5

[Blake Kaplan (:mrbkap) (inactive)]

And for the record:

$ git diff master | diffstat
 28 files changed, 650 insertions(+), 12023 deletions(-)

Comment 6

[Henri Sivonen (:hsivonen)]

Comment on attachment 788756 [details] [diff] [review]
Gut CNavDTD and the DTD utils.

I assume that you tested that about:blank still gets a body element generated with this patch applied.

Comment 7

[Henri Sivonen (:hsivonen)]

Comment on attachment 788758 [details] [diff] [review]
A small cleanup patch to make the DTD even simpler.

Never mind. I see that this patch takes care of what I was wondering about in my previous comment.

Comment 8

[Henri Sivonen (:hsivonen)]

Further editor dependencies are handled in bug 729046.

Thank you!

Comment 9

[Blake Kaplan (:mrbkap) (inactive)]

Landed in one squashed patch:

https://hg.mozilla.org/integration/mozilla-inbound/rev/0bcfd14b61e8

Comment 10

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/0bcfd14b61e8