plot histogram with class bins

[knaaptime]

wanted to gauge interest in adding a method to Classifier that plots the histogram of y with class bins a la desktop GIS

pandas and seaborn are optional dependencies but usually available in a pysal environment. If we take the coloring logic from #211 I think the legendgram could also be a method on Classifier (without depending on palettable), which would be pretty handy. Will add a test if folks are interested

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.6%. Comparing base (88c6549) to head (61c7c25).

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #216     +/-   ##
=======================================
+ Coverage   89.4%   89.6%   +0.1%     
=======================================
  Files          8       8             
  Lines       1088    1101     +13     
=======================================
+ Hits         973     986     +13     
  Misses       115     115             

Files	Coverage Δ
mapclassify/classifiers.py	88.6% <100.0%> (+0.2%)	⬆️

[martinfleis]

Nice, I like this and the idea of moving legendgram as well.

I would probably just try to add only maptlotlib as optional dependency. You can easily use plt.hist instead of using pandas and if seaborn is needed only for a minor cosmetic change, you can easily call sns.despine() yourself.

[knaaptime]

cool will do. considered dropping the seaborn from the beginning, but despine really makes a difference in how the plot appears to me. Can drop it and add a test

[martinfleis]

I don't think we need seaborn to remove spines :).

ax.spines.right.set_visible(False)
ax.spines.top.set_visible(False)

from https://matplotlib.org/3.4.3/gallery/ticks_and_spines/spines_dropped.html

[knaaptime]

oh right :P thanks

[knaaptime]

plt.hist doesnt take or return an axes though (so the ax.vline doesn't work and isn't as composable otherwise). I think I'll keep it as pandas unless you have a different implementation?

[martinfleis]

Like this

rng = np.random.default_rng(seed=0)
normal = rng.normal(size=1000)

fig, ax = plt.subplots()

ax.hist(normal)

lim = ax.get_ylim()[1]
ax.vlines(-1, 0, lim, color="k")
ax.spines.right.set_visible(False)
ax.spines.top.set_visible(False)

[knaaptime]

ah, i see. I was doing plt.hist not ax.hist. Thanks!

[jGaboardi]

fancy!