Matplotlib tkagg backend performance - python

I have a tkinter application that can plot a large number of data, where I have noticed poor pan and zoom performance when there is a large number of data on the canvas.
Looking into the tkagg_backend (as suggested by this and several other questions), the function and documentation suggests that the canvas should only be redrawn once the user is idle. However, from current and previous experience the canvas has always been updating(redrawing) mid zoom/pan. Therefore, I was looking at the specific functions that are involved and have a question regarding that.
The dynamic_update function:
def dynamic_update(self):
'update drawing area only if idle'
# legacy method; new method is canvas.draw_idle
self.canvas.draw_idle()
The canvas.draw_idle() function:
def draw_idle(self):
'update drawing area only if idle'
if self._idle is False:
return
self._idle = False
def idle_draw(*args):
try:
self.draw()
finally:
self._idle = True
self._idle_callback = self._tkcanvas.after_idle(idle_draw)
The ._idle parameter is initialized as True in the backend. This point is where I got stuck as I am unable to understand how ._idle is linked to mouse activity (I assume that it is, please correct me if that's wrong).
Interestingly enough, the canvas behaves like I would expect by commenting the self.canvas.draw_idle() line (redrawing once the mouse button is unpressed), and thus not calling the entire draw_idle function.
Therefore, my question is how is _idle set or why does it redraw my entire canvas when I am not idle?

When refering to "being idle" it is not the user or his mouse activity that is meant, but rather the GUI mainloop. The canvas should only be redrawn if the mainloop is not currently busy. Here, of course self._idle only refers to the matplotlib part of the GUI and what this structure inside draw_idle is supposed to do is to prevent the canvas from being drawn while it is being drawn.
This could easily happen when panning or zooming. The mouse moves to a new location, causing a readraw. While this redraw is happening, the mouse has already moved further and caused the next redraw. At that point in time the first redraw may not yet have finished, such that it would queue up. And so forth, such that at some point the GUI might become unresponsive. To prevent that, a new draw is only initialized once the previous one has finished and this behaviour is steered by self._idle being true or false.

Related

Is there any way to speed up the draw rate of a Scrollable Canvas in Tkinter?

So after posting another question about this issue I realised the problem was only happening with my custom resize bind. When I resize the window by the default edges of the window the Issue does not happen and the contents of the canvas is drawn accurately - however when using my custom resize bind the contents of the canvas is laggy and lags behind the true position of the window.
import tkinter as tk
def resize(self, event=None):
y=root.winfo_height()
x1 = root.winfo_pointerx()
x0 = root.winfo_rootx()
root.geometry("%sx%s" % (x1-x0,y))
root=tk.Tk()
canvas = tk.Canvas(root,bg='red')
canvas.pack(padx=20,pady=20)
inside=tk.Frame(root)
inside.pack(fill='both')
for i in range(10):
x=tk.Label(inside,text='Lsdkfjhaskfjhskljdfhaskjfhksfhkjasdhf')
x.pack()
g=tk.Button(root,text='Drag to resize')
g.bind('<B1-Motion>',resize)
g.pack()
canvas.create_window(5,5,window=inside,anchor='nw')
Original content:
Screenshot while resizing window with manual bind via button - as you can see the content is not visible while resizing the window and lags behind where the canvas is.
The issue is fixed if I call root.update() at the start of my resize function however this then causes a recursion depth error to the thousands of calls being made to update in such a small period of time.
Finally to repeat. When resizing with the default window Resize nudges at the edge of the window the canvas resizes perfectly with perfect draw rate and the content stays visible all the time. Why is my binding not acting the same way?
Geometry calculations are quite complex in general (and Tk has some fairly sophisticated solvers behind the scenes for handling them) so they only run when the GUI is considered to be "idle". Idleness occurs when the event queue would otherwise block, waiting for the next message from the external world, i.e., when there's nothing else queued up for immediate attention. Tk also postpones redrawing (another actually really expensive operation) the same way; it makes the UI seem far faster than it really is. This was critical back 20 years ago and more, and yet is still really useful now.
But it breaks if there there is a total flood of events coming in, and drags can do that, especially when there's resizing involved. That's why you can do update idletasks (in Tcl/Tk notation) and update_idletasks() (in Tkinter notation). That that does is immediately process all the pending idle event callbacks without taking further real events, running the pending resize computations and redraws. It is far less likely to trigger reentrant event processing than a full update(), which is the problem you were hitting (and why update is considered harmful under normal circumstances).
Note that it is important to let the event loop itself have some time to process still, as parts of handling widget resizes are inevitably done via real events (as they impact upon the OS/GUI). By far the best way to do that is to just make sure to return normally from your event handlers as soon as practical.
Are you using the right bind event? Try ''
g.bind('<ButtonRelease-1>',resize)

wx.lib.EVT_COLOURSELECT interfering with wx.lib.FloatCanvas.EVT_MOUSEWHEEL

The Problem:
It seems like one event is disabling another event. Before calling event B, event A works just fine. After event B fires, event A no longer works. Both events are custom ones that people made for their wxPython libraries (FloatCanvas and ColourSelect). I'd like to trace the generated events to make sure that, after event B, event A is still being fired (but perhaps not triggering the handler code?)
The Details:
I have a wxPython app where I create a wx.Panel object with two child items - a plot and a legend. This top level panel handles the majority of the events.
The plot is a wx.Panel object with a single wx.lib.FloatCanvas.FloatCanvas canvas in it.
The legend is a wx.Panel with multiple wx.StaticText and wx.lib.colourselect.ColourSelect objects in it (I'm plotting discrete data points).
On the plot, I bind FloatCanvas.EVT_MOUSEWHEEL to my zoom in/out function. On the legend, I bind the wx.lib.colourselect.EVT_COLOURSELECT event to my update_colors function which then sends the event to the parent panel via wx.PostEvent(event).
The parent panel then receives EVT_COLOURSELECT from the child and executes code that changes the colors of my plots.
Source Code:
I haven't had time to write a small sample that demo's the problem, but you can see the problem by running the source code:
https://github.com/dougthor42/wafer_map
Run the wm_app.py file and then go through the following test steps.
Testing Steps:
Scroll/zoom in and out - works just fine
Change a plot color (fires EVT_COLOURSELECT). If you're testing with the source code you do this by clicking on a legend color box and choosing a new color.
Attempt to zoom in and out again. Doesn't work!
Verify that all other events (key down, click-and-drag, mouse move) all work.
Things I've Tried:
Obviously these didn't work or else I wouldn't be here :-P
Unbinding the mousewheel event and rebinding it when the top level panel receives the event from the child.
Completely disabling the handlers for EVT_COLOURSELECT in both the parent panel and the legend.
I thought that perhaps there was something going on in my handlers. Turns out, just
triggering EVT_COLOURSELECT causes the EVT_MOUSEWHEEL to stop working.
Changing FloatCanvas.EVT_MOUSEWHEEL to some other event, for example FloatCanvas.EVT_RIGHT_DOWN
This does actually work! I loose the 'speed' info from the mouse wheel, but right-clicking
before and after step 2 both work.
Seems to indicate that it's an issue with the mouse scroll event specifically.
Has anyone had a problem where events interfere with each other?
It turns out that there wasn't any interference with events - at least not the events that I mentioned.
After a ton of fiddling and attempts at making a bare-bones version that still demonstrated the problem (which I was unsuccessful at doing), I figured out the root cause. Sadly, I did not figure out a workaround.
The Problem:
My plot wx.Panel was binding wx.MOUSE_LEFT_DOWN. It seems that this was preventing the parent wx.Frame window from giving focus to the Panel that held the plot and the legend.
The Solution:
I removed the bind to wx.EVT_LEFT_DOWN and that seems to fix the underlying issue. However, this is still not the ideal case because it causes two other problems:
I can no longer use my left mouse button on the plot. Not a big deal, I guess, because I can always bind right-mouse.
A user must first left-click on the plot area to give it focus before the mouse scroll or any keyboard shortcuts will work. This doesn't necessarily stem from not binding wx.EVT_LEFT_DOWN, but rather (I believe), from something that Windows does. See Focus-follows-mouse in wxPython? for a bit more info.
Failed Workaround:
I tried a workaround: adding code that sends the wx.EVT_LEFT_DOWN event to the parent but either that still didn't work or I was doing it wrong. Here's the left-click event handler that I tried:
def left_click(self, event):
print("left click!")
parent = wx.GetTopLevelParent()
wx.PostEvent(self.parent, event)
So anyway, this question is kinda solved. Hopefully what I've got here helps someone else out.

Using SwapBuffers() with multiple OpenGL canvases and vertical sync?

I have a GUI written using wxPython that contains two GLCanvases, a 'display' canvas and a 'preview' canvas onto which I am drawing some very simple geometry using PyOpenGL. The 'preview' and 'display' canvases display the contents of the same framebuffer:
I need both of these canvases to be updated synchronously at a consistent framerate with no tearing.
So far I have just been calling
self.SetCurrent()
# draw stuff...
self.SwapBuffers()
for both the preview and display canvases within my rendering loop. This works reasonably well provided that I disable vsync in my driver settings.
However, if I enable vsync my framerate drops to ~30Hz, half the vertical refresh rate of my display. Presumably this is because the execution of the rendering loop stalls at each SwapBuffers() call waiting for the next vertical refresh - if I comment out one of the two SwapBuffers() calls, my framerate goes back up to 60Hz, but then of course only one of my canvases gets updated.
What I'd really like is for both of my canvases to be updated synchronously, once per vsync interval. Is there any way that I can achieve this using the wx.glcanvas.GLCanvas class?
What Andon implied in his comment is, that you use {wgl|glX}SwapInterval to set the swap interval for the first, synching SwapBuffers to 1, and then right after the synching SwapBuffers returns set the swap interval to 0 and do the other SwapBuffers.

More Pyglet - Supress mouse motion events?

In the same game as last time, I've ran into a new problem. When I move the mouse, FPS increases from around 60 to over 500. I know what you're thinking; it's not because of on_draw() getting fired each event (I think), since I already did override of pyglet.app.EventLoop.idle:
class EventLoop:
def idle(self):
pyglet.clock.tick(poll=True)
return pyglet.clock.get_sleep_time(sleep_idle=True)
pyglet.app.EventLoop = EventLoop()
Also I call flip() on the window in my drawing function. All the useless mouse motion events take up a lot of CPU, which is annoying. What can I do about it?
Edit
I added window.invalid = False to my drawing function and window.invalid = True to my update function, this seems to reduce CPU usage with other mouse actions.
Edit 2
The drawing function is a typical on_draw() function.
Edit 3
After some more investigating, it seems that all those events don't take that much CPU as I thought they would. Still it would be good to know if this is the way Pyglet is supposed to act, or if it's something that should be avoided.
All you should be doing on mouse events is updating your apps' model of the input control state and using that in the next regular scheduled update and redisplay of the world model (which is presumably much more complex with physics and rendering and stuff).
ie just because mouse events come in at ~300fps doesn't mean you actually have to do all the stuff you want to do at 300fps.
It's been a while since I did any Pyglet, but the pattern I seemed to use was to subclass Pyglet's window.Window as MyGameWindow, then that registered event handlers like
#self.event
def on_mouse_motion(x,y,dx,dy):
self.mouse_position=(x,y)
self.mouse_buttons=0
self.invalid = False
(also on_mouse_drag, on_mouse_press, on_mouse_release). Hmmm... actually, I think that assignment to self.invalid might have been crucial for overriding Pyglet's default behaviour and defering any further updating/drawing until the next "clock tick".

Highlight Select Box in Python

I am trying to rebuild the functionality of the desktop's "highlight to select" feature so that I can use it in my own app. When I say "highlight to select" I mean the selection box that shows up if you click and drag on your desktop (native to all main-stream OS).
I've been working for hours trying to recreate it, and simply can't find a way. I've tried PyGTK, Xlib for python, and a couple other weird hacks. All of which have their own problems that won't allow me to move forward.
I generally don't ask for straight up example code without providing some sort of starting point, but in this project I don't even know where to start. How would you do this?
Here's the requirements:
Must draw on the root window (or a transparent layer that "appears" to be the root)
Must return the coordinates of the selection (x, y, height width)
Update: Forgot some details.
I am using Ubuntu 10.10
I have dual monitors (though, I don't think that should matter)
I don't mind downloading any extra libraries that are necessary
I don't know if this is what you're looking for, but what if you created another window in your module, and have your code show it when you release drag? You could fetch the cursor's current position, and have it draw the window there.
This should help you get the mouse position on the root window.
So, your code may look a little like this (this is untested code!) I'm only showing the relevant portions of what goes inside __ init __.
def __init__(self):
...
#Some of your code here.
...
win = gtk.Window(gtk.WINDOW_TOPLEVEL)
#Note that I am creating a popup window separately.
popwin = gtk.Window(gtk.WINDOW_POPUP)
#I am setting "decorated" to False, so it will have no titlebar or window controls.
#Be sure to compensate for this by having another means of closing it.
popwin.set_decorated(False)
def ShowPopup():
#You may need to put additional arguments in above if this is to be an event.
#For sake of example, I'm leaving this open ended.
#Get the cursor position.
rootwin = widget.get_screen().get_root_window()
curx, cury, mods = rootwin.get_pointer()
#Set the popup window position.
popwin.move(curx, cury)
popwin.show()
def HidePopup():
#This is just an example for how to hide the popup when you're done with it.
popwin.hide()
...
#More of your code here.
...
#Of course, here is the code showing your program's main window automatically.
win.show()
A very simplistic approach, but it should give the appearance of what you're wanting.

Categories

Resources