How can I set the x-axis tick locations for a bar plot created from a pandas DataFrame?

Question

I have a simple plot, with x labels of 1, 1.25, 1.5, 1.75, 2 etc. up to 15:

The plot was created from a pandas.DataFrame without specifying the xtick interval:

speed.plot(kind='bar',figsize=(15, 7))

Now I would like the x-interval to be in increments of 1 rather than 0.25, so the labels would read 1,2,3,4,5 etc.

I'm sure this is easy but I cannot for the life of me figure it out.

I've found plt.xticks() which seems like it's the right call but maybe it's set_xticks?

I've changed the x ticks a great amount without doing what I wanted up until this point. Any help would be greatly appreciated.

Answer 1

The way that pandas handles x-ticks for bar plots can be quite confusing if your x-labels have numeric values. Let's take this example:

import pandas as pd
import numpy as np

x = np.linspace(0, 1, 21)
y = np.random.rand(21)
s = pd.Series(y, index=x)

ax = s.plot(kind='bar', figsize=(10, 3))
ax.figure.tight_layout()

You might expect the tick locations to correspond directly to the values in x, i.e. 0, 0.05, 0.1, ..., 1.0. However, this isn't the case:

print(ax.get_xticks())
# [ 0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20]

Instead pandas sets the tick locations according to the indices of each element in x, but then sets the tick labels according to the values in x:

print(' '.join(label.get_text() for label in ax.get_xticklabels()))
# 0.0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1.0

---

Because of this, setting the tick positions directly (either by using ax.set_xticks) or passing the xticks= argument to pd.Series.plot() will not give you the effect you are expecting:

new_ticks = np.linspace(0, 1, 11)       # 0.0, 0.1, 0.2, ..., 1.0
ax.set_xticks(new_ticks)

---

Instead you would need to update the positions and the labels of your x-ticks separately:

# positions of each tick, relative to the indices of the x-values
ax.set_xticks(np.interp(new_ticks, s.index, np.arange(s.size)))

# labels
ax.set_xticklabels(new_ticks)

This behavior actually makes a lot of sense in most cases. For bar plots it is common for the x-labels to be non-numeric (e.g. strings corresponding to categories), in which case it wouldn't be possible to use the values in x to set the tick locations. Without introducing another argument to specify their locations, the most logical choice would be to use their indices instead.