Matplotlib - 对数图表 - 向图表添加额外文本

Question

I am trying to create a readable layout of a logarithmic chart, based on a pandas dataframe. I have the following code:

fig, ax = plt.subplots(figsize=(15, 8))
ax.clear()
ax.barh(mydataframe['Labels'], mydataframe['Values']
plt.xscale('log')
for i, (label, value) in enumerate(zip(mydataframe['Labels'], mydataframe['Values'])):
            ax.text(value + 1, i, f'{value:,.0f}', size=8, ha='left', va='center')
            ax.text(value + 20, i, f'({int(round(extraValue1)):d} -- {extraValue2:,.2f}%)', size=8, ha='left', va='center')

As you can see in the picture, my problem is the alignment of the second ax.text print on the chart. The first line of ax.text aligns perfectly the values at the end of each bar. But the second one, with the additional information I need to add, the alignment is messed up and overwrites the previous value, the more we go up to the bigger bars. Does anyone know how I can fix this? I have a feeling it's related to the chart being set with logarithmic scale. The extravalue1 and 2 and also part of column of mydataframe, but for the simplicity of the code, I defined here the enumerate for only the main columns

Answer 1

To have a nice distance between the bar and the text, it is better to use the exact value for the x of the text and start the text with one or two spaces. So, using ax.text(value, i, " abc") instead of ax.text(value+20, i, "abc") . To have two texts, just concatenate them and place them in one go.

If you want text with multiple colors, and have them placed together with a gap adapting to the length of the first text, offset boxes can be used.

To have more space for the text, you can increase the x-margins, eg ax.margins(x=0.2) . At the same time you can decrease the y-margins, for which the defaults are a bit too large here.

import matplotlib.pyplot as plt
from matplotlib.colors import CSS4_COLORS
from matplotlib.offsetbox import AnchoredOffsetbox, TextArea, HPacker
import numpy as np

fig, ax = plt.subplots(figsize=(15, 8))
labels = ["".join(np.random.choice(list('abcdef'), 5)) for _ in range(20)]
values = np.sort(10 ** np.random.uniform(1, 5, 20))
ax.barh(labels, values, color=np.random.choice(list(CSS4_COLORS.keys()), 20))
ax.set_xscale('log')
for i, (label, value, extraValue1, extraValue2) in enumerate(
        zip(labels, values, *np.random.randint(100, 1000, (2, 20)))):
    xbox1 = TextArea(f'{value:,.0f}', textprops=dict(color="crimson", size=8, ha='left', va='baseline'))
    xbox2 = TextArea(f'({int(round(extraValue1)):d} -- {extraValue2:,.2f}%)',
                     textprops=dict(color="navy", size=8, ha='left', va='baseline'))
    xbox = HPacker(children=[xbox1, xbox2], align="baseline", pad=0, sep=5)
    anchored_xbox = AnchoredOffsetbox(loc="center left", child=xbox, pad=0.5, frameon=False, bbox_to_anchor=(value, i),
                                      borderpad=0, bbox_transform=ax.transData)
    ax.add_artist(anchored_xbox)
ax.margins(x=0.2, y=0.01)
plt.show()

Answer 2

Even for the classic (non-log) chart it was a bit messed up, but I fixed with an indent based on value (if it's three or four digits)

if value<1000:
            indent = 80
        else:
            indent = 120
ax.text(value + 20, i, f'({int(round(extraValue1)):d} -- {extraValue2:,.2f}%)', size=8, ha='left', va='center')

The problem remains with log chart

Answer 3

Just add the text you want to plot as one string not to:

for i, (label, value) in enumerate(zip(mydataframe['Labels'], mydataframe['Values'])):
    text = f'{value:,.0f}\t({int(round(extraValue1)):d} -- {extraValue2:,.2f}%)'
    ax.text(value + 1, i, text, size=8, ha='left', va='center')

Or am I missing something why you do not want to do this?

Answer 4

Since your x-axis is logarithmic, you'd want to use a multiplicative offset for your text, instead of an additive one. That is, try value * 1.2 instead of value + 20 :

fig, ax = plt.subplots(figsize=(15, 8))
ax.clear()
ax.barh(mydataframe['Labels'], mydataframe['Values']
plt.xscale('log')
for i, (label, value) in enumerate(zip(mydataframe['Labels'], mydataframe['Values'])):
            ax.text(value * 1.05, i, f'{value:,.0f}', size=8, ha='left', va='center')
            ax.text(value * 1.2, i, f'({int(round(extraValue1)):d} -- {extraValue2:,.2f}%)', size=8, ha='left', va='center')

Please tune the 1.05 and 1.2 to fit your needs. The formatting could look at bit nicer if you combined both texts into one, but that's just taste.