[mcclim-devel] three questions

[Orm Finnendahl]

Hi Andy,

 thanks for your post, your insight and the explanations saved me
hours! You're right about the confusion on my part about incremental
redisplay. I thought it is a generic term and didn't exactly
understand the way it is implemented in clim. I found out about the
mechanism of updating-output yesterday and got a working example (and
a vague idea, how it works internally). The mechanism is amazing and
can probably save lots of coding time, but it might be a little too
expensive. IIUC a redisplay function has to traverse all output
records in order to find out which ones have changed. That might be
too inefficient when doing redraws of changing records during
pointer-motion, but I'll see about that.

I will also check your suggestions about window redrawing. I found out
that a call to force-output on the pane will do the redraw on my
system (Linux/X11).

--
Orm

Am Dienstag, den 05. Mai 2009 um 04:30:52 Uhr (+0000) schrieb Andy Hefner:
> On Mon, May 4, 2009 at 8:08 AM, Orm Finnendahl
> <o.finnendahl at mh-freiburg.de> wrote:
> > Hi,
> >
> > still working on my specialized browser I'm starting to implement the
> > gui part of the application. Three questions:
> >
> > - Can I update (redraw) a pane without activating it with the mouse?
> >  "(window-refresh pane)" doesn't actually draw unless the window gets
> >  activated in the window-manager).
> 
> That sounds very odd. I could speculate at causes, but McCLIM's
> definition of window-refresh looks suspect anyway. I recommend using
> (repaint-sheet stream (window-viewport stream)) instead of
> window-refresh. Possibly there's a buffering issue, but in most
> circumstances McCLIM takes care of that for you. Try calling
> finish-output on the stream after drawing.
> 
> > - How can I implement scrolling in the pane? The scroll bars are there
> >  but they don't get updated if I draw off the visible region of the
> >  pane.
> 
> After drawing on the pane, at some point the pane has to be resized to
> fit the output. McCLIM tries to do this itself at appropriate times
> (such as at the end of lines of text, after calling the display
> function, or table/graph formatting), but this is somewhat expensive
> so the various drawing functions don't do it automatically. In
> particular, if you're working outside the normal paradigm of display
> functions, you might have to do this manually. One way is to call
> change-space-requirements on the pane with a width determined from the
> bounding rectangle of the stream-output-history. A shorter way is to
> call climi::fit-pane-to-output, which is what mcclim calls internally.
> 
> > - Is there a straightforward way to get a reference to an output
> >  record upon creation of the output record (to store it in the object
> >  which gets presented)?
> 
> with-output-as-presentation returns the output record of the
> presentation. If you mean it's the output record of the drawn lines
> you want, I don't know of a good way of obtaining those as such. What
> I do is to draw things inside with-new-output-record, which will
> return a standard-sequence-output-record containing your drawing, then
> come back and add/remove things from this record or delete it
> altogether.
> 
> > Maybe I'm doing it wrong by trying to access the output records
> > themselves. This is the idea: I want to present graphical objects and
> > change the shape of some of them using incremental redisplay. As far
> > as I see, this should be done by deleting the respective output record
> > and redrawing it with the new properties.
> 
> I do this sort of thing all the time (directly manipulating output
> records) and it's a powerful approach. Incremental redisplay is a
> higher level facility intended to spare you from the lower level
> details of dealing directly with output records, using updating-output
> to handle caching and update of sections of the output. There might be
> some confusion of terminology here. Perhaps using the higher level
> facilities will make things simpler. I personally always go the lower
> level route that you seem to be taking (mostly because I understand
> the output recording code quite well and like to maintain complete
> control), but other software like Climacs and the inspector have used
> incremental redisplay to good effect.