[Bug]: Setting alpha with an array is ignored with jupyterlab %inline backend #29595

Activity

1. ianthomas23 commented on Feb 9, 2025

   

   ianthomas23

   on Feb 9, 2025

   Member

   More actions

   Please can you show us the output of conda list.
2. jklymak commented on Feb 9, 2025

   

   jklymak

   on Feb 9, 2025

   Member

   More actions

   I can reproduce this. Worked fine with 3.8.3, and doesn't work with the inline backend and 3.10.0

   But thankfully also reproducible by just saving as a png outside of Jupyter.

   import matplotlib
   print(matplotlib.__version__)
   import numpy as np
   import matplotlib.pyplot as plt
   
   img_data = np.random.rand(256, 200)
   alpha = np.ones((256, 200))
   alpha[:, 0:100] = 0.5
   fig, ax = plt.subplots(1, 2, figsize=(10, 8))
   ax[0].imshow(img_data, alpha=alpha)
   ax[1].imshow(alpha)
   fig.savefig('Boo.png', dpi=100)
   

   

   Also occurs on main...

   This bisects to 3ceba40, and is basically caused by #28382
3. timhoffm commented on Feb 9, 2025

   

   timhoffm

   on Feb 9, 2025

   Member

   More actions

   @jklymak is #28437 the solution to this?
4. jklymak commented on Feb 9, 2025

   

   jklymak

   on Feb 9, 2025

   Member

   More actions

   Yes I think so.
5. timhoffm commented on Feb 10, 2025

   

   timhoffm

   on Feb 10, 2025

   Member

   More actions

   We need to clarify what the intended behavior for array alpha kwarg is on RGBA images. This is the blocker on #28437 (see #28437 (comment)). Unfortunately, the decision which behavior is wanted is mostly irrelevant here. As long as colormaps don't have a transparency, the produced RGBA images will have an alpha channel with constant 1, and then multiplying and replacing is the same. But we need a decision to move the PR forward.
6. ucabova commented on Apr 3, 2025

   

   ucabova

   on Apr 3, 2025

   More actions

   I also encountered this issue and noticed that the issue only appears for larger resolutions. For square images, I noticed the cutoff to be at 123 pixels.

   import numpy as np
   import matplotlib.pyplot as plt
   
   array1 = np.random.random(size=(123, 123))
   alpha1 = np.ones_like(array1) * 0.2
   alpha1[:, :64] = 0.8
   
   plt.imshow(array1, alpha=alpha1)
   

   

   Then increasing the resolution:

   array2 = np.random.random(size=(124, 124))
   alpha2 = np.ones_like(array2) * 0.2
   alpha2[:, :64] = 0.8
   
   plt.imshow(array2, alpha=alpha2)
   

   
7. timhoffm commented on Apr 3, 2025

   

   timhoffm

   on Apr 3, 2025

   Member

   More actions

   Closed by #28437. This should be fixed in 3.10.1.

   👍React with 👍1Reacted by ucabova

   🎉React with 🎉1Reacted by Martijn Wehrens
8. timhoffm

   closed this as completedon Apr 3, 2025
9. QuLogic

   added this to the v3.10.1 milestone on Apr 4, 2025