[Edbrowse-dev] css parser

[Edbrowse-dev] css parser

[Karl Dahlke]

[-- Attachment #1: Type: text/plain, Size: 1073 bytes --]

Earlier posts have discussed the possibility, and the advantages, of writing our own css parser and querySelectorAll.
If we are content with handling the high runner cases, and maybe not every corner case, it might not be as hard as first appears.
(We don't handle the unusual cases now anyways.)
css selector syntax is described here.

https://www.w3.org/TR/CSS21/selector.html#grouping

There's a lot going on, but it's straightforward.
An ideal css compiler would produce a 4 dimensional structure.
I have implemented such, and it seems to work on the modest files I've tried.
(I have yet to try the monster css file from stackoverflow.)
Have a look at the comments in startwindow.js line 2105.
If you want to play with it, browse any file with at least one css file,
and in jdb,

v = mw0.cssPieces(mw0.uncomment(document.links[0].data));

Once this is shaken out, cssGather would call this instead of the third party css parser.
Then we would write our own querySelectorAll, then invoke that from within cssApply.
That's the roadmap.

Karl Dahlke

Re: [Edbrowse-dev] css parser

[Adam Thompson]

On Tue, Feb 13, 2018 at 01:41:31PM -0500, Karl Dahlke wrote:
> Earlier posts have discussed the possibility, and the advantages, of writing our own css parser and querySelectorAll.
> If we are content with handling the high runner cases, and maybe not every corner case, it might not be as hard as first appears.
> (We don't handle the unusual cases now anyways.)

Indeed and we've got to do something.  We need some of css but the third-party software just doesn't appear to be good enough... that's a shame.
> css selector syntax is described here.
> 
> https://www.w3.org/TR/CSS21/selector.html#grouping
> 
> There's a lot going on, but it's straightforward.
> An ideal css compiler would produce a 4 dimensional structure.
> I have implemented such, and it seems to work on the modest files I've tried.
> (I have yet to try the monster css file from stackoverflow.)
> Have a look at the comments in startwindow.js line 2105.
> If you want to play with it, browse any file with at least one css file,
> and in jdb,
> 
> v = mw0.cssPieces(mw0.uncomment(document.links[0].data));
> 
> Once this is shaken out, cssGather would call this instead of the third party css parser.
> Then we would write our own querySelectorAll, then invoke that from within cssApply.
> That's the roadmap.

Sounds good, I'll have a play with it and see what I can find.
Hopefully it all works as expected.

Thanks for working on this,
Adam.

Re: [Edbrowse-dev] css parser

[Kevin Carhart]

This is exciting.....!

[Edbrowse-dev] css parser

[Karl Dahlke]

Some stats: on the monster stackoverflow css, their parser finds 3097 descriptors, mine finds 3711.
Who's right? Got me!

Of the 3711, there are 153 that I don't understand or haven't implemented.
Some of them, like @directives, I'm not suppose to implement.
Others I know how to address but it's one thing at a time.
Still 153 out of 3711 isn't bad for version 1.

I think querySelectorAll will come together soon, unless I'm missing something important here.

It's fun to program again, especially in js, where I can wield a regexp like a giant hammer,
where one (somewhat ugly) line of code does the equivalent of a page of C.
Chris remembrers, when we wrote version 1 of edbrowse in perl.

Karl Dahlke