using KHTML without display

[Nom Declavier]

Thanks very much for the xfake suggestion. Suppressing the display is part of what I'm after. The more critical requirement is to dispense with functionality that my application doesn't need. I want the application to be as small and fast as possible. What's the minimal context that allows the DOM, CSS, and KJS classes to be used for parsing and measurement?
  ----- Original Message ----- 
  From: Nom Declavier 
  To: kfm-devel at kde.org 
  Sent: Sunday, May 01, 2005 6:31 PM
  Subject: using KHTML without display


  I'd like to use KHTML to parse HTML/CSS/Javascript, and to deduce the
  sizes and positions of Web page constituents when the page is rendered
  by a KHTML-based browser. But I want to do this without actually
  rendering to any screen, and without invoking more browser functionality
  than I need. My planned application has no graphical user interface. It
  brings about no display. It's all about trees whose nodes may be
  annotated with size and position information. I expect to call getRect()
  frequently.

  So what I really need is DOM::Document and so on. I'll use KApplication,
  KHTMLPart, KHTMLView, and so on, only as I need them to invoke the
  functionality of DOM, CSS, and KJS classes.

  I'm aware of two ways to get a DOM::HTMLDocument from an HTML file, without
  getting into windows and widgets.

  Technique 1 looks like this:

  DOM::HTMLDocument doc;
  doc.setAsync(false);
  doc.load(url);

  Technique 1 has two serious problems. First, when getRect() is called on
  nodes, it produces no useful information. Second, when the program
  exits, either the automatically-invoked destructors bomb, or if I invoke
  destructors myself, they still bomb.

  Technique 2 looks like this, where inputHTMLQString is a QString that's
  read from the HTML file, it doesn't matter how.

  KHTMLPart * pPart = new KHTMLPart();
  pPart->begin();
  pPart->write(inputHTMLQString);
  pPart->end();
  DOM::Document doc = pPart->document();

  Technique 2 has two serious problems. First, when getRect() is called on
  nodes, it produces no useful information, so there's nothing to choose
  between Technique 1 and Technique 2 here. Technique 2 leads to graceful
  destruction, but it brings along by default a very fussy version of the
  HTML parser which wreaks havoc with scripts, among other constituents. I
  can get around the fussy parser, but the way I've done it so far isn't
  pretty.

  If I want to have all of the following:

  calls to getRect() produce useful results
  tolerant parser
  effective destruction

  and I want to have them to the extent possible without without windows
  and widgets, I'm guessing the best technique isn't either of the ones
  I've tried. Best aside, what's a good technique?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20050503/baa122f8/attachment.htm>