[Edbrowse-dev] css on demand

[Edbrowse-dev] css on demand

[Karl Dahlke]

As promised, the styles that are created in the original document are now on-demand getters.
When you access style it applies the css elements, as though it had done so from the start.
You don't know the difference - except we don't waste all that time populating styles that you never look at.
stackoverflow is back to 12 seconds (assuming you have the nojs to prevent the infinite loop).
I hope other sites will run faster as well.
And now, I need to take a break!

Karl Dahlke

Re: [Edbrowse-dev] css on demand

[Adam Thompson]

On Wed, Feb 14, 2018 at 10:59:01AM -0500, Karl Dahlke wrote:
> As promised, the styles that are created in the original document are now on-demand getters.
> When you access style it applies the css elements, as though it had done so from the start.
> You don't know the difference - except we don't waste all that time populating styles that you never look at.
> stackoverflow is back to 12 seconds (assuming you have the nojs to prevent the infinite loop).
> I hope other sites will run faster as well.
> And now, I need to take a break!

Indeed, actually quicker on my laptop... thanks a lot for sorting this.  I'm now wondering what's with
the infinite loop, will try and work out what's happening but I
fear my js (and jdb in edbrowse) knowledge isn't up to the task.

Cheers,
Adam.

Re: [Edbrowse-dev] css on demand

[Chuck Hallenbeck]

Hi,

>> stackoverflow is back to 12 seconds (assuming you have the nojs to prevent the infinite loop).
>> I hope other sites will run faster as well.


I tried www.amazon.com, and it runs like greased lightning! But while
the resulting screen contains most of what is expected, many links are
not displayed, only their surrounding braces are present.

For example, line 11 is only a string of {} with no link labels shown.

But the speed is fabulous, nice work.

I added "nojs = www.amazon.com" for this improvement to work. When I
comment that line out, it is still unusable, and the load favtor
skyrockets, making the keyboard input unresponsive.

Chuck

-- 
Here too, In Northeast Ohio, The Moon is Waning Crescent (1% of Full)
When your only tool is a hammer, everything looks like a nail.
Sent from Francisco's iPhone.

[Edbrowse-dev] css on demand

[Karl Dahlke]

> many links are not displayed, only their surrounding braces are present.

I finally fixed this one. Trying to clean things up for a new version.
Also amazon comes up for me in just a few seconds with full javascript and no exclusions.

Do you still have the curl https download problem?
Would be nice to understand that one?
Did you check the versions of libcurl and libcurl-devel to make sure they're the same?

Karl Dahlke

[Edbrowse-dev] URL.prototype.slice

[Kevin Carhart]

Not related to the new release.  It could wait or go in.

I pulled up yahoo.com and several others, because I was curious how we're 
doing with the top websites.

Some code in this file is trying to call "slice" on an object of the URL 
class.

execute yui-min.js at 1
yui-min.js line 86: TypeError: undefined not callable (property 'slice' of 
[object Object])
execution complete

We have given it the other string functions like substr, etc, but could we 
also pop this in:

mw0.URL.prototype.slice = function(from, to) {
return this.toString().slice(from, to);
}

It makes the error at 86 go away and a couple of others too.

K

[Edbrowse-dev] URL.prototype.slice

[Karl Dahlke]

Of course.

Karl Dahlke

Re: [Edbrowse-dev] css on demand

[Kevin Carhart]

I was just in startwindow - that's a ton of work!

> When you access style it applies the css elements, as though it had done so from the start.

That's a great idea, which will make a huge difference.

[Edbrowse-dev] css on demand

[Karl Dahlke]

> I was just in startwindow - that's a ton of work!

Yeah, too bad some of it was wasted.   :)
You know the mantra, "A moment of thought is worth a megabyte of programming."
I slightly forgot that, just because I like to program.
I implemented hashing by nodes, hashing by class name, optimizations for certain chains, blah blah blah,
and then I realized I didn't need any of that if I was gathering css elements on demand,
and then I realized I couldn't support it anyways, if querySelectorAll is dynamic,
if it is suppose to query the document tree as it stands right now,
and nodes are always changing, with new nodes added,
so a snapshot cache we build at the start isn't really going to help,
and maintaining that cache as each new node is added etc is rather prohibitive,
so anyways that was 200 tricky lines of code that I wrote, then threw away within 24 hours,
but other than that, you know,
the other 750 lines are good, and represent real progress.
I'm about to delete the jotform stuff, cause I'm pretty sure we're not gonna use it,
and as others have pointed out in this list, if we change our mind it's all in git.

Karl Dahlke

Re: [Edbrowse-dev] css on demand

[Kevin Carhart]

> I'm about to delete the jotform stuff, cause I'm pretty sure we're not gonna use it,

In addition to wielding a regexp like a giant hammer, it can also be 
enjoyable to take a garden shears to the third-party code!