mpl-extensions / mpl-interactions Goto Github PK

Should have the option to return the ui and contained widgets to embed however is desireable.

slope's value was never displayed here

if using many sliders it will just get really long. should probs do some nice chunking of them like boxes with no more than 3 sliders vertically and then rows of no more than three boxes. Like this but if it were drawn not poorly:

do it however ipywidgets.interact does it

They are currently a mess. Stuff in wrong order, multiple items blended together etc
https://github.com/ianhi/ipympl-interactions/blob/d423ac6b13a7f4565d2bac0f11701df5afaa3d5a/ipympl_interactions.py#L208-L225

https://thebelab.readthedocs.io/en/latest/
https://jupyter-sphinx.readthedocs.io/en/latest/#thebelab-support

ipympl-helpers?

based on the precision in the parameter. or like the differences between max and min. or maybe just always use sci notation

particularly useful for having a favicon for the docs

basically take the example https://matplotlib.org/3.1.1/gallery/subplots_axes_and_figures/zoom_inset_axes.html and add interactivity to be able to move the box and resize it. Would also be cool to have a similar ability with the inset axes

analogous to the ipywidgets fixed widget. As proposed here: jupyter-widgets/ipywidgets#2946 (comment)

now that functions can return x as well this also needs to be dealt with.

It would be cool to have interactive_[imshow, pcolormesh, hist,...]. I think these all deserve separate functions as they have different signatures and are different in matplotlib.

First list:

• [ x] hist

• [ x] scatter

• [ x] imshow

• pcolormesh

• errorbar

• seaborn distplot

each should get their own interactive_[...]_factory and interactive_[...] function

Pretty much all of the functions in pyplot could be possibly be included, but the above are probably the most useful.

• Hist

• Scatter

Originally posted by @SeanAshley5 in #59 (comment)

would be cool. Ideally this would be a part of ipywidgets (jupyter-widgets/ipywidgets#2941) but until/if that happens could probably make this work with matplotlib

via https://gist.github.com/ianhi/f54d5dc0066337ae4ad2cff4d3f30029

should add the ability to set the mask color. and think about colormaps vs straight image.

I suggest adding the Framework :: Matplotlib classifier (https://discourse.matplotlib.org/t/framework-matplotlib-trove-classifier/20927) to your setup.py and adding your package to https://github.com/matplotlib/matplotlib/blob/master/doc/thirdpartypackages/index.rst .

Example is zero contours in stark effect.

if not using this backend and you try to display a fig.canvas inside of a VBox you get fairly mystifying errors

basically to reinitalize the stretch behavior
maybe this can added to the toolbar?

def f(tau):
    blah blah
   return y
interactive_plot(f, tau = np.linspace...)

could be equivalent to plt.plot(f(tau[0])). .I think this would sacrifice the case where f returns both x and y but only a single point for both and not as numpy arrays. which like, whatever... that does not seem like a terrible thing to lose

https://gist.github.com/ianhi/b638e09fa7a00764ea539940a0519af9
https://gist.github.com/ianhi/5f514c31f13cc28fbd167849867122b7

https://github.com/ianhi/jupyter-jukebox/blob/23abca8f943e6280cb2909edaffc4bd78a4662e5/jupyter_jukebox.py#L30-L33

though I guess it also ought to get a specific if statement? any value not in those documented will result in a fixed y scale.

then use sel(..., method='nearest')

Having written the entire library I'm finding it difficult to figure out which parts need the most explanation. So the narrative documentation is a bit lacking although I think the examples are pretty good. Although even the examples do not cover every case.

%matplotlib ipympl
import numpy as np
from mpl_interactions import interactive_plot
def f(x, τ, β, type_):
    if type_ == 'sin':
        return np.sin(x*τ)*x**β
    elif type_ =='cos':
        return np.cos(x*τ)*x**β
        
fig, ax, sliders = interactive_plot(f, x=x, τ = (1, 10, 100), β = (1, 10 ), type_ = {('sin', 'cos', 'beep','boop')})

will fail with an error in the log console but not in the notebook :(

this would remove the need to wrap things in lists.

idk if its worth it though. I need to manufacture myself a user who is not me and figure out what they would like

imagining something like

λ = np.linspace(500, 600,100)
β = np.linspace(0,1,1000)
`interactive_plot(f, λ = λ, β=β

which then is also nice for adding labels to the sliders.

this line: https://github.com/ianhi/jupyter-jukebox/blob/23abca8f943e6280cb2909edaffc4bd78a4662e5/jupyter_jukebox.py#L67
should have a -1 added to the maximum

Making use of/adding to https://github.com/ianhi/mpl-interactions/blob/master/mpl_interactions/mpl_kwargs.py

• plot
• scatter
• hist
• ax{v,h}line

with a draggable line on the image following some of the examples in: https://stackoverflow.com/questions/7878398/how-to-extract-an-arbitrary-line-of-values-from-a-numpy-array

I just wrote down the following sentence when making documentation:

You can control how xlim/ylims behave using the x_scale/y_scale arguments

which demonstrates the absurdity of the current situation

With the switch to RTD and changes to .gitignore it no longer renders anyway.

Not sure what the interface ought to be. Ideas:

1. optional argument to interactive plot
2. it's own function

proposed by @kmdalton

It would be cool to have interactive_[imshow, pcolormesh, hist,...].
First list:
has a set_data method

• pcolormesh (see https://stackoverflow.com/a/31490420/835607)

• colorbar

  • this is just a pcolormesh internally

• hist (partially in #73)

  • Optimal implementation will use matplotlib/matplotlib#18275 (will give set_data)

• imshow - #107

  • needs docs (#119)

• axvline and axhline (#126)

  • #119 makes these actually useful bc they can paired with other interactive elements
  • https://github.com/matplotlib/matplotlib/blob/fc18cccbb9122ad6d7fc6156ec8f6c455e408d8c/lib/matplotlib/axes/_axes.py#L777

no set_data method

• scatter - #102
  • Has set_offsets
• seaborn distplot
• fill between
  each should get their own interactive_[...]_factory and interactive_[...] function
• errorbar
• contour(f)
  • see https://stackoverflow.com/questions/16915966/using-matplotlib-animate-to-animate-a-contour-plot-in-python
  • also see my implementation here https://discourse.matplotlib.org/t/best-way-to-animate-a-contour/22073

Pretty much all of the functions in pyplot could be possibly be included, but the above are probably the most useful.

with some sort of string format ideally

i.e. an argument like title = 'rotation: {param}' should be subjected ax.set_title(rotation.format(param=slider.value)) and/or this should apply to the figure suptitle. Or both suptitle and title default to none

basically the same as the interact decorator except that it's a special case to deal with matplotlib

Currently slices accepts: 'vertical', 'horizontal', and 'both'. This has potential to become a nightmare of special casing once arbitrary angles are supported. In particular it should probably not accept 'both' and instead you use a list to specify.

xref: #29 and #33