I added a console.log print here, and the function is getting called for keyboard events when using the webagg backend but not the nbagg backend. I tried the same here and it is called by both backends. The 'scroll_event' works on webagg but not nbagg.

Just to pile on, the following picker demo did not work either:

import numpy as np
x, y, c, s = np.random.rand(4, 100)
ind = 0

def onpick3(event):
    global ind
    ind = event.ind

fig, ax = plt.subplots()
col = ax.scatter(x, y, 100*s, c, picker=True)
fig.canvas.mpl_connect('pick_event', onpick3)
plt.show()

The picker demo also does not work in the webagg backend (I used print(event) in the callback).

Your going to make me learn javascript aren't you .....

I'm with you brother, I had resisted until this week...

I think somehow the IPython KeyboardManager needs to be prevented from stealing all the events for the document.

Its in IPython/html/static/notebook/js/keyboardmanager.js

Ah, I was trying to figure out what was different between webagg/nbagg that was eating keyboard events. It makes a good deal of sense that IPython is what is eating the keystrokes....

Presumably they have a way of getting around this when entering code into cells.

Yes and maybe we should ask the the IPython devs about scroll events since they seem to eat those as well.

@minrk Any advice?

The problem seems to be that the notebook keyboard handler is snarfing all of the keystrokes before they get to the <div> tag that mpl has keydown event listeners on.

attn @pelson

To avoid IPython handling keyboard events on a particular element, you will want to call

IPython.notebook.keyboard_manager.register_events($element);

Generally, you would call this once on your outermost div when you create it.

Thanks @minrk - that is the kind of knowledge that would take a good while to figure out. @blink1073 - how do you want to proceed? Do you want to take a look at this?

Not related to the fix. But if you guys are modifying mpl.js it won't hurt to fix a typo in line $43.

- this.focus_on_mousover = false;
+ this.focus_on_mouseover = false;

https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/backends/web_backend/mpl.js#L43

I'll give it a shot.

I fear I may be out of my depth. Here's what I tried (note that the keyboard_manager has moved between IPython 2.3 and 3.0):

diff --git a/lib/matplotlib/backends/web_backend/nbagg_mpl.js b/lib/matplotlib/backends/web_backend/nbagg_mpl.js
index 71c52f9..3dd7844 100644
--- a/lib/matplotlib/backends/web_backend/nbagg_mpl.js
+++ b/lib/matplotlib/backends/web_backend/nbagg_mpl.js
@@ -49,6 +49,7 @@ mpl.mpl_figure_comm = function(comm, msg) {
         return false;
     });

+    IPython.keyboard_manager.register_events($(fig.root));
 };

 mpl.figure.prototype.handle_close = function(fig, msg) {

Keyboard events are still getting eaten. I tried a few syntax variations on fig.root, but I'm just stabbing in the dark.

@blink1073 I am also whacking at this (but keep getting side tracked).

@minrk The problem seems to be that the relevant are never getting focus which I think might be a problem on our end.

Ha! I have it working, PR on the way.

The problem seems to be that the relevant are never getting focus

Ah, yes. focus is a constant source of headaches for us.

Update: the picker is working (in master), with this demo:

import matplotlib
import itertools
import numpy as np
matplotlib.use('nbagg')
import matplotlib.pyplot as plt
plt.close('all')
fig, ax = plt.subplots()
x = np.linspace(0,10,100)
y = np.sin(x)
ln, = ax.plot(x,y, 'o', picker=5)
evt = []
colors = iter(itertools.cycle(['r', 'g', 'b', 'k', 'c']))
def on_event(event):
    evt.append(event)
    ln.set_color(next(colors))
    fig.canvas.draw()
    fig.canvas.draw_idle()
fig.canvas.mpl_connect('key_press_event', on_event)
fig.canvas.mpl_connect('pick_event', on_event)
plt.show()

Woo. Thanks for working on this @blink1073 & @tacaswell. We should really get around to adding nbagg to the %matplotlib% magic in IPython.

The magic has been set up on their side already

On Tue, Jan 6, 2015, 02:05 Phil Elson notifications@github.com wrote:

Woo. Thanks for working on this @blink1073 https://github.com/blink1073
& @tacaswell https://github.com/tacaswell. We should really get around
to adding nbagg to the %matplotlib% magic in IPython.

—
Reply to this email directly or view it on GitHub
#3963 (comment)
.

This isn't technically closed until #3968 is merged.

It probably got auto-magic closed via one of my commit messages.

I think it was from the first line in your PR description.

Both are now merged and backported.

And the UAT all passes on the cherry-picked 1.4.x branch, hurray!

Well that what easy, it only took two us learning a new language. 😄

Well that what easy, it only took two us learning a new language.

And tripling the nbagg backend bus factor!