How to create a variable fontsize for bar plot annotations

Question

How to choose the font size for text annotations inside the bars of the bar graph with the condition:

• Text will completely cover the rectangular bar area.

Please go through the diagram and code for better clarity about the problem.

So, the requirement is only : font size should be relative to bars in the bar graphs

Code

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl

# Plot styles
mpl.style.use("ggplot")

# data
fruits = pd.Series(index = ["Apples", "Oranges", "Watermelon"], data = [324,518, 258])


# Bar graph for Fruits

# figure
plt.figure(figsize = (7,5))

# bar graph
fruits.plot(kind = "bar", color = ["red", "orange", "green"], alpha = 0.6, width = 0.5, )

# percentage of each fruit type
categories = list(fruits.index)
categories_percent = [100*(value/fruits.sum()) for value in fruits ]

# categories annotations coordinates
ax = plt.gca() # get current axes
rects = ax.patches # rectangles axes of bars in the graph

# annotations
for i in range(len(categories)):
    plt.annotate(f"{categories[i]} - {categories_percent[i] : 0.2f}%",
                 xy = (rects[i].get_x() + rects[i].get_width()/2, 
                       rects[i].get_y() + (ax.get_yticks()[1] - ax.get_yticks()[0])*.2),
                 fontsize = [20,28,12][i], # Chosen by hit and trial for adjustment
                 color = "white",
                 ha = "center",
                 rotation = 90,
                 )
    
plt.ylabel("# Counts", fontsize = 15,)
plt.title("Distribution of Fruits", fontsize = 25, fontname = "Monospace", alpha = .6)
plt.xticks([])
plt.tight_layout(rect=[0, 0, 1, 1])
plt.show()

How to deal with this line of code fontsize = [20,28,12][i], # Chosen by hit and trial for adjustment to adjust the font size dynamically with respect to bar area?

Answer 1

• Updating the existing annotation with an adjustable fontsize
  • From a logical perspective figure sizes' y acts as a scaling factor for height.
  • Think .get_height as a relative height of the figure.
  • The actual height is the y scaling factor multiplied with .get_height.
  • About including breadth, we can include relative breadth which is just .get_width (not get_width*x), however it would just act as a constant, since it's relative width.
  • We can't include actual width because the font would adjusted unproportionally for y axis.

x,y=15,15
plt.figure(figsize = (x,y))


for i in range(len(categories)):
    txt="{} - {: 0.2f} %".format(categories[i],categories_percent[i])
    plt.annotate(txt,
                 xy = (rects[i].get_x() + rects[i].get_width()/2, 
                       rects[i].get_y() + (ax.get_yticks()[1] - ax.get_yticks()[0])*.2),
                 fontsize = (rects[i].get_height())*y*.2/len(txt), # Chosen by hit and trial for adjustment
                 color = "white",
                 ha = "center",
                 rotation = 90,
                 )

• The entire code can be written more cleanly as follows

# data
fruits = pd.Series(index = ["Apples", "Oranges", "Watermelon"], data=[324,518, 258])

# calculate percent
per = fruits.div(fruits.sum()).mul(100).round(2)

# bar graph
y = 5
ax = fruits.plot(kind="bar", color=["red", "orange", "green"], alpha=0.6, width=0.5, figsize=(7, y), rot=0)

labels = [f'{fruit} - {per[fruit]}%' for fruit in fruits.index]

# annotations:
for label, p in zip(labels, ax.patches):
    left, bottom, width, height = p.get_bbox().bounds
    fs = height * y * 0.18 / len(label)
    ax.annotate(label, xy=(left+width/2, bottom+height/2), ha='center', va='center', rotation=90, fontsize=fs)

plt.ylabel("# Counts", fontsize=15,)
plt.title("Distribution of Fruits", fontsize=25, fontname="Monospace", alpha=.6)
plt.xticks([])
plt.tight_layout(rect=[0, 0, 1, 1])
plt.show()

For figsize=(15,15):

For figsize=(8,8):

For figsize=(7,5):