[Edbrowse-dev] suggested additions

[Kevin Carhart <kevin@carhart.net>]

Here are some notes on the 7 things where nasa.gov works if you implement 
them all.  I'd like to get your remarks before making a patch, though I 
don't think any of them will cause any problems.  There's one in decorate 
and the rest are javascript.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(1) give a CSSStyleDeclaration attributes
CSSStyleDeclaration = function(){
         this.element = null;
         this.style = this;
         this.attributes = new Array;
};

Justification:
http://code.jquery.com/jquery.js at line 6787 or thereabouts says
  style.removeAttribute( "filter" );

So it is expecting that a style will have attributes.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(2) Give an iframe a contentWindow in addition to a contentDocument in 
decorate.c:
// special code for frame.contentDocument.
                         if (t->parent->action == TAGACT_FRAME)
                                 set_property_object(t->parent->jv,
                                                     "contentDocument", 
t->jv);
                         if (t->parent->action == TAGACT_FRAME)
                                 set_property_object(t->parent->jv,
                                                     "contentWindow", 
cf->winobj);


Justification: the latest jquery at 6046 or thereabouts says this:

doc = ( iframe[ 0 ].contentWindow || iframe[ 0 ].contentDocument 
).document

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(3) Bring back the following definition for Event.   I am getting this 
from Chris Thatcher's DOM project.  There is a link below.

Event = function(options){
     // event state is kept read-only by forcing
     // a new object for each event.  This may not
     // be appropriate in the long run and we'll
     // have to decide if we simply dont adhere to
     // the read-only restriction of the specification
     this._bubbles = true;
     this._cancelable = true;
     this._cancelled = false;
     this._currentTarget = null;
     this._target = null;
     this._eventPhase = Event.AT_TARGET;
     this._timeStamp = new Date().getTime();
     this._preventDefault = false;
     this._stopPropogation = false;
};

Justification: This is in the env DOM implementation.  Event handlers are 
(often?  or always?)  expecting to be passed an Event object as an 
argument.  You can go to the source if you would like to review this piece 
of code:

git clone https://github.com/thatcher/env-js.git

and then go to env-js/src/event/event.js

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(4) Apply the Event definition from (3) to the addEventHandler and 
attachEvent too.  So the firing of the handler code will accept an Event 
object as an argument.

Inside of the eval() command , this:
a[i]();

Would be changed so that now we would say
var tempEvent = new Event;
tempEvent.type = name of the event that was passed in, like click;
a[i](tempEvent);

Justification: I just found this out from trial and error back when I was 
playing with an edbrowse 3.3.1.  It may be that it's just a jQuery 
convention.  It definitely makes a lot of handlers work correctly when you 
have this.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(5) Implement childNodes.item.  Pages want to be able to reference a set 
of childNodes by an index.  I don't know the best place to put this. 
What I did so far is to define it on document,
document.childNodes.item = function (x) { return this[x]; }

Justification: This is referenced in the nasa.gov file, vendor.js.  (see 
esbrowse.git to deminify and get good line numbers.)
  (vendor.js is minified, annoyingly.  But, if you wish to play along you 
can deminify it using esprima.parse followed by escodegen.generate in 
succession.  I bundled this up at 
https://github.com/KevinCarhart/esbrowse.git)

So, at roughly 8260 in the de-minified line numbering, vendor.js contains 
this line:
           r = r.childNodes.item(t[n]);

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(6)
I also added childNodes.item to the default case of the createElement 
switch statement.  Maybe there is a better way to do this.  Only 
properties are there right now.  Not functions also.  But that's where I 
put it currently:

c.childNodes.item = function (x) { return this[x]; }

Justification:  Same as (5)

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

(7) Following on from Karl's exclusion of @ from the CSS selectors, I 
added some more of these exclusions of CSS constructs that were causing a 
runtime error when I was debugging nasa. 
I used split as a crude form of existence check.  Please change to regular 
expression if you prefer.  This is a proposed section from the 'certain 
modifiers' comment all the way to the querySelectorAll.  (In the CSS code 
in startwindow.)

// certain modifiers not supported in this static view.
// @directives are not selectors.

if(sel.match(/^@/))
continue;

if (sel.substr(0,1) == '}')
{
sel = sel.substr(1,sel.length-1);
}
if (sel.split('-moz-').length > 1)
continue;
if (sel.split('-webkit-').length > 1)
continue;
if (sel.split('-ms-').length > 1)
continue;
if (sel.split(':before').length > 1)
continue;
if (sel.split(':after').length > 1)
continue;
if (sel.split(':active').length > 1)
continue;

// a:link is the same as a.
sel = sel.replace(/:link$/, "");
// :hover :visited etc are dynamic an not relevant here.
if(sel.match(/:(hover|visited|active)$/))
continue;

sel = sel.trim();
a = querySelectorAll(sel);

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

This is all you have to do for nasa.gov to render!!