matplotlib widgets in napari

🚀 Feature

Following some guidance from @story645 and @tacaswell we were able to get a matplotlib widget into napari using some instructions from the matplotlib documentation https://matplotlib.org/3.1.1/gallery/user_interfaces/embedding_in_qt_sgskip.html

If you have matplotlib installed then the following code will add a widget:

import numpy as np
import napari
from skimage import data
from matplotlib.backends.backend_qt5agg import FigureCanvas
from matplotlib.figure import Figure
    
    
with napari.gui_qt():
    blobs_raw = data.binary_blobs(length=64, n_dim=2, volume_fraction=0.1)
    viewer = napari.view_image(blobs_raw)
    
    static_canvas = FigureCanvas(Figure(figsize=(5, 3)))
    axes = static_canvas.figure.subplots()
    axes.plot(blobs_raw.mean(axis=0))
    viewer.window.add_dock_widget(static_canvas, area='bottom', name='matplotlib figure')    

We were also discussing how to make a new figure manager so that users could just do plt.plot and have it pop up in the dock widget. I was pointed to these two parts of matplotlib

https://github.com/matplotlib/matplotlib/blob/8ac3ea170fee9ab7b78767b9a7463332a5970880/lib/matplotlib/backends/backend_qt5.py#L502

and

https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/backends/backend_qt5agg.py#L98

Will be great for a plugin! But putting it here for others to see too 😃 Thanks to the matplotlib team for the help!!

[dstansby]

👋 I thought I’d drop by and mention that I’ve started work on a napari-matplotlib plugin: https://github.com/dstansby/napari-matplotlib. Currently the elevator pitch and scope are “napari-matplotlib is a bridge between napari and matplotlib, making it easy to create publication quality Matplotlib plots based on the data loaded in napari layers.”. I’d be super interested in any feedback or suggestions folks on this thread have for the plugin, including the scope I settled on for it 😄

[sofroniewn]

Would love to work on this 😃

@goanpeca that’s great! As @tlambert03 said, I’d definitely encourage reaching out to and potentially teaming up with @lukasz-migas around his 1D plotting which I believe uses vispy under the hood.

Additionally I think making it as easy and smooth to do matplotlib widgets in napari makes sense - that could actually just be a really good tutorial

finally i do think it would be good if we started trying to add some additional 1D plots to napari itself, like value histograms and profiles, maybe using the 1D stuff from either a plugin or directly in napari. it’s been a while since those last convos, and i think we’ve learnt a lot as a team on how to move work forward since then