Modularity of DOM/JS parts

[David Faure]

On Tuesday 20 March 2007, Paul Giannaros wrote:
> On Tuesday 20 March 2007 11:30, David Faure wrote:
> > On Sunday 18 March 2007, Paul Giannaros wrote:
> > > Hi,
> > >
> > > I've been working on a project to facilitate web navigation/scraping
> > > using a "real" browsing engine to handle (X)HTML/JavaScript/cookies. I
> > > began using KHTMLPart with Xvfb (so that it works on X-less servers -- a
> > > requirement), but that solution is proving to be increasingly finicky and
> > > unextensible. A major problem is how closely it's tied to the graphical
> > > side: there are constantly message boxes asking about cookies,
> > > remembering passwords, kwallet asking things, etc.
> > > Since then I've been looking at different engines to get this to work
> > > the 'proper' way -- by seperating out the DOM/JS stuff from the graphical
> > > side -- and KHTML looks like the nicest.
> > >
> > > How tied are the graphical and non graphical bits of KHTML? I'd need to
> > > use the HTML parsing and DOM stuff, tie it with KJS, and probably handle
> > > cookies myself. The (eventual) goal would be to make it work with
> > > QCoreApplication.
> >
> > All the things you mention are in fact handled by separate programs:
> > kcookiejar for the cookies, kpasswdserver for http passwords, and kwallet
> > for form contents (and kpasswdserver also uses kwallet).
> > So it should be quite easy to replace the calls to those things, or the
> > implementation of those things, to avoid the GUI interaction.
> > The form saving is probably the only one that has the messagebox inside
> > khtml's code - but you can hack it out easily.
> 
> Encouraging. I had a fair read of KHTML's source to get an idea of what's 
> happening yesterday, however, but was disheartened to see that even things 
> like parsing documents require a reference to a KHTMLView -- If I attempt to 
> instanstiate a QWidget subclass when using a QCoreApplication, things will go 
> kaboom, surely?

Yep. Sorry I missed the mention of QCoreApplication in your initial mail.
You can tune down the khtml gui, but with the current design you can NOT use
khtml without any gui support at all. webkit might be a better solution for that,
since they redesigned the khtml code to isolate the platform dependencies.

-- 
David Faure, faure at kde.org, sponsored by Trolltech to work on KDE,
Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).