Display both axes in sorted order for non numerical data

How to achieve the correct order for both axes:

• a-b-c instead of c-a-b
• x-y-z instead of y-z-x

import matplotlib.pyplot as plt


categories_x = ["c", "a", "b", "c", "b"]
categories_y = ["y", "z", "y", "x", "z"]

plt.scatter(categories_x, categories_y)
plt.show()

There are a lot of solutions on SO, that rely on either of two properties:

1. The data can be cast to numerical, e.g. ("1", "0", ...) -> cast to numerical
2. Only one axis has the wrong order -> sort the two arrays by this axis (the reason why this works is, that the axis-ticks are ordered by first occurrence)

But for my example neither of these solutions work.

I'm looking for a solution, of how to get this to work in matplotlib. I am aware, that there are other probably even better ways to convey the same message, or maybe other libraries that don't have this issue.

What about using [tag:pandas] and an ordered Categorical?

ax = plt.subplot()

X = pd.Categorical(categories_x, ordered=True)
Y = pd.Categorical(categories_y, ordered=True)

ax.scatter(X.codes, Y.codes)
ax.set_xticks(range(len(X.categories)), X.categories)
ax.set_yticks(range(len(Y.categories)), Y.categories)

Output:

How about simply converting everything to numerical values and playing with the x- and y-ticklabels

import matplotlib.pyplot as plt

categories_x = ["c", "a", "b", "c", "b"]
categories_y = ["y", "z", "y", "x", "z"]

def axis_to_number(values):
    # this mapping function can be customized 
    return {j:i for i, j in enumerate(sorted(set(values)))}

map_x = axis_to_number(categories_x)
map_y = axis_to_number(categories_y)

# now convert the original arrays to the 
# mapped values to keep the order
cx = [map_x[i] for i in categories_x]
cy = [map_y[i] for i in categories_y]

xticks, xticklabels = [x for x in map_x.values()], [x for x in map_x.keys()]
yticks, yticklabels = [y for y in map_y.values()], [y for y in map_y.keys()]

# plot
fig, ax = plt.subplots()
ax.plot(cx, cy, 'o')
ax.set_xticks(xticks)
ax.set_xticklabels(xticklabels)
ax.set_yticks(yticks)
ax.set_yticklabels(yticklabels)

We can use sort function to arrange them in sequence.

import matplotlib.pyplot as plt

categories_x = ["c", "a", "b", "c", "b"]
categories_y = ["y", "z", "y", "x", "z"]

plt.scatter(sorted(categories_x), sorted(categories_y))
plt.show()

You could draw a first scatter with the ordered strings, to get the ticks setup properly, then hide it (or use a transparent color) and draw the actual diagram:

import matplotlib.pyplot as plt
   
categories_x = ["c", "a", "b", "c", "b"]
categories_y = ["y", "z", "y", "x", "z"]
    
p1 = plt.scatter(sorted(categories_x),sorted(categories_y),c='#00000000')
# p1.set_visible(False)

plt.scatter(categories_x,categories_y)

plt.show()

You may want to use sorted(set(categ...)) if you have large number of points and performance becomes a concern