Color handling with the new CSS parser

[Lars Knoll]

Hi Dave,

did you also use the new color parsing code for HTML attributes? See
HTMLElementImpl::addHTMLColor in our current code base.

I first tried the same as you now did, but this turned out to be too messy and 
broke standards compliance for CSS. 

I'm now using addHTMLColor everywhere and html color attribute is used. 
addHTMLColor is btw 100% IE compatible when it comes to color parsing (at 
least all my tests in khtmltests/html/font/ pass).

Lars

> So Safari's page load tests had a lot of errors with colors with the
> new parser.  The basic issue is that the new parser is strict about
> requiring the # in front of color names, and it also ignores string
> values for colors (so not all of the named colors work).
>
> The following patch fixes all the problems, but I don't really like it.
>   Basically I go ahead and handle all the crazy unit types I can get
> from having a malformed color...
>
> The three examples I had to patch for:
>
> 000099 = a number, but you lose the "0000" so you have to pad back out
> to 6 characters.
> ffffff = a string, so that unit type has to be checked (seems like this
> type needed to be checked
> anyway, but wasn't, which i didn't understand)
> 99ccff = a dimension (not sure why it thought this was a dimension and
> not just a string)
>
> The top of parseColor looks like this for me now:
>
>      QRgb c = khtml::invalidColor;
>      Value *value = valueList->current();
>      if ( value->unit == CSSPrimitiveValue::CSS_NUMBER ) {
>          QString numStr = QString::number( value->fValue );
>          if (numStr.length() < 6) {
>              QString colorStr = QString();
>              int padding = 6 - numStr.length();
>              while (padding) {
>                  colorStr += "0";
>                  padding--;
>              }
>              colorStr += numStr;
>              numStr = colorStr;
>          }
>          c = ::parseColor(numStr);
>      }
>      else if ( value->unit == CSSPrimitiveValue::CSS_RGBCOLOR ||
>                value->unit == CSSPrimitiveValue::CSS_IDENT ||
>                value->unit == CSSPrimitiveValue::CSS_STRING ||
>                value->unit == CSSPrimitiveValue::CSS_DIMENSION )
> 	c = ::parseColor( qString( value->string ));
>      else
> 	...
>
> Is there a way to perhaps fix this with a production in parser.y that
> would be cleaner than this?
>
> dave
>
> _______________________________________________
> Khtml-devel@mail.kde.org
> http://mail.kde.org/mailman/listinfo/khtml-devel