You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hexbin() has the option to display 1D marginals along the x- and y-axes. This is a bit of an oddity within Matplotlib - we don't support automatic marginals for pcolormesh, imshow etc. In addition there are lots of possible issues with the current implementation (#21349, and the list at #18875), which are a maintenance burden.
I would like to propose deprecating and removing support for automatic marginals in hexbin. Instead we can provide an example showing how to add marginals on separate axes for a range of 2D plots, and/or point to higher level libraries that build upon Matplotlib that provide automatic marginal functionality.
I think I agree with the general premise that we probably shouldn't have had these in the first place, but now that they are there, are they worth deprecating?
I think the only inconsistency is with hist2d not providing marginals. I don't think that pcolormesh or imshow have any inherent binning in them that marginals would be desired for. I am pretty sure we have a marginals example for hist2d though it is implemented as histograms, not colored strips.
If marginals is a reasonably working feature, I'd leave it as is even if we wouldn't add it nowadays.
Note that #21349 is not about marginals. So the question is: are the issues #18875 and the added code complexity worth the effort on our side to deprecate, remove and provide alternative examples and the burden on the user side to switch away from marginals?
Activity
jklymak commentedon Oct 14, 2021
I think I agree with the general premise that we probably shouldn't have had these in the first place, but now that they are there, are they worth deprecating?
I think the only inconsistency is with
hist2dnot providing marginals. I don't think thatpcolormeshorimshowhave any inherent binning in them that marginals would be desired for. I am pretty sure we have a marginals example forhist2dthough it is implemented as histograms, not colored strips.timhoffm commentedon Oct 14, 2021
If
marginalsis a reasonably working feature, I'd leave it as is even if we wouldn't add it nowadays.Note that #21349 is not about marginals. So the question is: are the issues #18875 and the added code complexity worth the effort on our side to deprecate, remove and provide alternative examples and the burden on the user side to switch away from marginals?
dstansby commentedon Dec 21, 2021
Since I'm neutral, and there are two in favour of keeping, I'm going to close this.