>>>>> "Gregory" == Gregory Lielens <gre...@ff...> writes:
Gregory> Hi John, Hi the list, I almost have finished updating the
Gregory> fltkAgg, to add the new subplot formatter and the blit
Gregory> method for fast animation. Everything seems to work fine,
Gregory> but I still need to do some code clean-up, and before
Gregory> commiting the changes I have to ask what you think about
Gregory> the method I used:
Hi Gregory, Great news. Ken McIvor just emailed me this morning that
he completed the same with a _wxagg module that he will post shortly.
That means we have (or will soon have) blit methods for all GUIs
except Qt and Cocoa (gentle reminder).
Gregory> Instead of writing a special method taking account of the
Gregory> bbox for bliting, similar to the agg_to_gtk_drawable, or
Gregory> a aggcanvas->str conversion like the proposed
Gregory> aggcanvas.as_rgba_str(bbox), I just changed the method
Gregory> that I introduced for normal draw in fltkAdd, that rely
Gregory> on pytho buffer objects to transfer the agg buffer to the
Gregory> fltk library.
Gregory> The old method of the Agg backend that was used to get
Gregory> the buffer object was:
....snip...
Gregory> As fltkAgg is able to use an image source with a pixel
Gregory> stride (skipping the a of the rgba quatuor) and a line
Gregory> stride (taking a part of wide w1 from an image of wide
Gregory> w2), no copy or special method is needed with the new
Gregory> version of RendererAgg::buffer_rgba...
Gregory> Now I checked and other backends (qtagg, cocoaagg) now
Gregory> use RendererAgg::buffer_rgba. If this is ok, I will
Gregory> update all the calls to pass (0,0) instead of no
Gregory> arguments, but I wonder if the new methods could not
Gregory> simplify things for those backends too...
Yep, this all looks right. I will look into this for gtkagg as well
tomorrow. It may make rendering to a bbox even faster because gtkagg
and tkagg may both make an unnecessary copy that your method avoids.
Gregory> Another remarks is that the animation code seems quite
Gregory> fragile for the moment: if one do a resize of the output
Gregory> window, the background copy is not updated correctly and
Gregory> one have a very interresting effect ;) Well, at leat I
Gregory> observe that under fltkAgg (my local version supporting
Gregory> blit) and GTKAgg. tkagg does not do this..because it is
Gregory> not possible to resize the window during animation (well,
Gregory> window is resizable but the size of the inside canvas
Gregory> does not change during this resize...).
Gregory> I think some extra steps should be performed to make
Gregory> resize and anim play nicely together (like freezing the
Gregory> anim during resize, and restarting afterwards with the
Gregory> correct background...)
What I do is connect to the new 'draw_event' which is called at the
end of Figure.draw (no GUI specific event handling required). If you
set the animated property of the Artist, and then connect your
background saver to the 'draw_event', it should handle this case.
Additionally you want to make sure your draw events are processed in
an idle queue on resizes, etc.
Check out widgets.Cursor for a complete example using draw_event
background cacheing, the guts of which are
def __init__(self, ax, useblit=False, **lineprops):
...snip
self.canvas.mpl_connect('draw_event', self.clear)
def clear(self, event):
'clear the cursor'
if self.useblit:
self.background = self.canvas.copy_from_bbox(self.ax.bbox)
Does this work for you in GTKAGG and/or FLTK? If everything looks
good, you may want to update the canonical animation examples in the
examples directory and on the wiki.
JDH