How to update an IntSlider from "inside" the function decorated by interact()?

I want to visualize a signal’s channel (signal is has dimension (n_samples, n_channels)) using interact, but before retrieving it I don’t know how many channels it will have. Is there a way I can dynamically update the channel slider max value with the n_channels information?

def foo(index, channel):
    signal = get_signal(S, index)
    n_channels = signal.shape[1]
    plt.plot(signal[:,channel])
    plt.show()
interact(foo, index=(0,n_sessions-1), channel=(1,10))

I tried defining an enclosed function like:

def foo(index):
    def foo2(channel):
         plt.plot(signal[:,channel])
         plt.show()
    signal = get_signal(S, index)
    n_channels = signal.shape[1]
    interact(foo2, channel=(0,n_channels-1))
interact(foo, index=(0,n_sessions-1))

But this way, when I change the index slider, it creates a new interact output below the old one (I tried using close() and clear_output() but it didn’t work) I’ve also tried using ipywidgets observe but couldn’t work it out.

[jasongrout]

Can you try using the newest beta of 7.0? This should work:

from ipywidgets import interact
from matplotlib import pyplot as plt
import numpy as np
signals=[np.random.rand(CHANNELS, 20) for CHANNELS in range(1,10)]

@interact(signal=(0,len(signals)-1))
def f(signal):
    s=signals[signal]
    @interact(channel=(0,s.shape[0]-1))
    def g(channel):
        plt.plot(s[channel])
        plt.show()

[jasongrout]

One final possibility, though it is a bit rough around the edges, and may break in the future if we change the order of controls returned by interactive:

from ipywidgets import interactive
from matplotlib import pyplot as plt
import numpy as np
signals=[np.random.rand(x, 20) for x in range(1,10)]

def f(signal=0, channel=0):
    s=signals[signal]
    plt.plot(s[channel])
    plt.show()
controls = interactive(f, signal=(0,len(signals)-1), channel=(0,signals[0].shape[0]-1))

signalslider = controls.children[0]
channelslider = controls.children[1]

def update_channels(change):
    signal = signals[change.new]
    channelslider.max = signal.shape[0]-1
    # reset channel value?
    # channelslider.value = 0
signalslider.observe(update_channels, 'value')

controls