Re: Scripting in Gnome

[Rodrigo Moya <rodrigo gnome-db org>]

On Fri, 2004-01-30 at 10:16 -0500, Sean Middleditch wrote:
> On Fri, 2004-01-30 at 09:46, Tim Peoples wrote:
> 
> > Along these lines, I've had an idea bouncing around in my head for
> > a few months now.  Somewhat of a marriage of glade (the xml format),
> > javascript, Gtk+/Gnome bindings and some sort of web service protocol
> > (xml-rpc or maybe soap/wsdl) into something I have initially called
> > "Gnomescript".  While I only have a cursory understanding of DBUS, it
> > definately looks like it fits in quite nicely.
> > 
> > I'll admit that I haven't put much work into it after my initial frenzy
> > of binding a few Gtk+ widgets into Mozilla's SpiderMonkey... mostly
> > because I was somewhat leary of whether such a beast was really
> > necessary or not (like was said elswhere, we already have Python and
> >  Perl and ....)
> > 
> > Originally, my idea was leaning more towards "remote applications"
> > somewhat like Mozilla's XUL but much more Gnome specific... but soon
> > expanded that to something more akin to VB on Windows (but without
> > all the ugly baggage the VB carries).  You'll have to admit that the
> > availability of VB on Windows in the early nineties somewhat helped
> > MS achieve their dominance on business desktops.
> > 
> > All that said, I still do have plans to pick this up again... that is,
> > unless I can be convinced that its not worth the effort (which may not
> > be all that hard to do).
> 
> My personal ideas on the matter (which I may or may not have put in my
> Advogato diary - I barely remember yesterday vs 6 months ago...) were to
> offer an API for making clean and *easy* bindings to an application.  I
> haven't looked much at the D-BUS API in recent versions so I don't know
> if it counts as "easy" or as "pain in the ass."
> 
> The wrapper API should be generic enough that it can wrap *any*
> scripting language.  It should support cross-process control (in which
> case D-BUS would be great, altho we also already have ORBit/bonobo) and
> *also* in process scripts.  (so that apps can get inline scriptability
> for "free".)
> 
> Apps would then be able to have a "plugins" menu to load scripts (in
> *any* gnomescript supported language, possibly even including C/C++)
> that would just be inserted into the event loop.  Additionally,
> processes could use the client-side bindings to attach to already
> running processes as COM/AppleScript allows.  So I could write a Python
> script to, say, have Epiphany open up a website and than call the
> Screenshot program to grab a view, then tell Epiphany to attach the page
> to a Mozilla/Gecko bug report, and then repeat for another site.  (This
> particularly over-thought-out example would be useful if you've found a
> big list of sites Gecko chokes on and you want to submit them to the bug
> tracker.)
> 
> Really, the most important part is the API.  If you start with D-BUS and
> work from there, you'll probably just end up with an API that just
> mimics D-BUS, which is a complete waste of time.  The API has to let the
> app register call-in points ("functions") and attach interfaces to
> objects (OO), and then the low-level implementation of how these are
> exported can be *completely* arbitrary; D-BUS for remote execution and
> direct bindings to embedded language APIs for inline scripts.  Or
> whatever.  Just make sure you start at the top and work down -
> specification first, then API, then implementation details.  ;-)
> 
right, an attempt at that is under the gnome-office CVS module. It's
been discussed at length in the gnome-office-list gnome org 

cheers