[Cairo] py-cairo example & Context save/restore quesiton

Jesse Andrews jdandr2 at uky.edu
Fri Nov 14 09:07:16 PST 2003

I'll look at it later today.  As far as an example, in the code, I have 
the basic "tree/plant" lsystem coded (but commented).  Set the 
iterations down to 1 or 2 (4 takes FOREVER since save/restore are slow 
at least for me).

At one, for the "plant L system", you should see:


or some rotation of that, but what happens is:


The order that the path is being drawn (in logo psuedo code):

new path

  save state
  restore state


But what is happening on the screen is:

   (un rotate -- I guess this is happening by restore state...)


It seems like the restore does NOT reload the position, but does reload 
the rotation.

As for CVS, sounds good, I'll get you that soon...  (about to take depth 
exam for PhD... this was coded to relax last night)


Carl Worth wrote:

 > On Nov 14, Jesse Andrews wrote:
 >  > I hacked a quick Lindenmayer system for cairo using the 
python-cairo  > port by jamesh.
 > Cool. l-systems are fun. Good to see some done with cairo. I've just
 > committed this as python-cairo/examples/lsystem. If you'd like to
 > maintain this in CVS, just send me an ssh key and a preferred
 > username.
 >  > python-cairo worked pretty good except for the fact that saving 
and  > restoring state in a context seem to take "forver" (in the sense 
that  > doing that hundreds of times in a update took seconds)
 > Hmm... I've never noticed any problems with save/restore performance,
 > but I've probably never done it hundreds of time before
 > either. Looking at the code, most of what it copies is simple data
 > structures and atomic values.
 > It does copy the current path, which can obviously be quite large, so
 > you might want avoid having a large path current path at the time of
 > save if that's possible.
 > Other than that it save the current pen, which is tiny, and it
 > re-creates a copy of the current font. I don't know how heavy the font
 > object is, but since we're dealing with a stack here, it wouldn't be
 > hard at all to do something more efficient there if necessary.
 >  > and the fact that the state doesn't seem to be stored.
 >  > Is the state (rotate/translation/position) stored just as in the 
 > postscript model or is this a different idea?
 > Yes, using cairo.save saves the current matrix, colors, path, etc. and
 > cairo.restore restores them. I've tested this in python-cairo and it
 > does work, (see attached).
 > Oh, one possible point of confusion is that the matrix transforms
 > points as they are added to the path, (eg. at the time of move_to,
 > line_to, curve_to). Further changes to the matrix do not affect points
 > that are already on the path. Could that explain the problem you are
 > seeing?
 > Otherwise, a minimal test case demonstrating the problem would be
 > quite helpful.
 >  > ( Note: python-cairo & cairo are currently out of sync, so 
compiling  > took some commenting out source of python-cairo )
 > I've just removed from python-cairo several things that were recently
 > removed from cairo. Let me know if I missed anything.
 > Thanks,
 > -Carl
 > ------------------------------------------------------------------------
 > #!/usr/bin/env python
 > import gtk
 > import cairo
 > import cairo.gtk
 > def expose_event (widget, event):
 >     cr = cairo.Context ()
 >     cairo.gtk.set_target_drawable (cr, widget.window)
 >     cr.rectangle (10, 10, 50, 50)
 >     # First set things up for a red stroke.
 >     cr.set_rgb_color (0, 0, 1)
 >     cr.set_line_width (8)
 >     # Do blue fill inside save/restore
 >     cr.save ()
 >     cr.rotate (0.16)
 >     cr.set_rgb_color (1, 0, 0)
 >     cr.fill ()
 >     cr.restore ()
 >     # Now back to color/line_width from before save
 >     cr.stroke ()
 > win = gtk.Window()
 > win.set_title('Cairo demo of save/restore')
 > drawingarea = gtk.DrawingArea ()
 > drawingarea.set_size_request (400,100)
 > win.add (drawingarea)
 > win.show_all ()
 > win.connect ('destroy', gtk.mainquit)
 > drawingarea.connect ('expose_event', expose_event)
 > gtk.main ()

More information about the cairo mailing list