Setting x-axis as dates using mdates with matplotlib

Question

I'm loading these packages:

import pandas as pd
from matplotlib import pyplot as plt
import numpy
import matplotlib.pyplot as plt
import seaborn as sns
import matplotlib
import matplotlib.dates as mdates
sns.set()
%matplotlib inline

And I have a dataframe df which looks like this

df['element_date'] = pd.to_datetime(df['element_date'])
df['mdate'] = [mdates.date2num(d) for d in df['element_date']]
df.head()

id            Tier    element     element_date           mdate
5228039     Tier B      4      2018-05-28 10:59:00  736842.457639
5232263     Tier B      3      2018-05-28 10:59:00  736842.457639
5245478     Tier B      EA     2018-05-27 13:58:00  736841.581944
4975552     Tier B      2      2018-05-30 21:01:00  736844.875694
4975563     Tier A      2      2018-05-30 21:01:00  736844.875694

I'm trying to set the x axis of a count-plot to month and day only, and I'm getting an error message. This is the code I'm running (I've removed the naming labels to save space):

fig, ax = plt.subplots(figsize=(15,10))
fig = sns.countplot(x="mdate", hue="element", data=df)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%m-%d'))
plt.show(fig)

I'm getting DateFormatter found a value of x=0, which is an illegal date. This usually occurs because you have not informed the axis that it is plotting dates, e.g., with ax.xaxis_date()

Now, of course I've tried adding ax.xaxis_date(), to no avail. I also have no x values that are equal to 0. I've dropped NA, and value counted mdate, and there is no 0 to be found.

I've looked at a bunch of different answers here, and can't seem to get to a solution. I've tried both using element_date as my date time value, as well as using "mathplotlib" dates using mdate.

Any thoughts would be much appreciated. Essentially, I'm just trying to have my x-axis be an ordered series of dates over two months, with elements being counted for each date.

Thanks!

Answer 1

Buried down on a GitHub pandas issues page, user, @pawaller, found a workaround using plt.FixedFormatter where you string format the datetime dataframe column.

ax.xaxis.set_major_formatter(plt.FixedFormatter(df['element_date'].dt.strftime("%m-%d")))

However, using above does not immediately work as value labels are out of order and not aligned properly. Hence, unique() and sort_values() are required:

x_dates = df['element_date'].dt.strftime('%m-%d').sort_values().unique()
ax.xaxis.set_major_formatter(plt.FixedFormatter(x_dates))

To demonstrate below (where mdate column is never used):

Data

from io import StringIO
...

txt = '''id            Tier    element     element_date           mdate
5228039     "Tier B"      4      "2018-05-28 10:59:00"  736842.457639
5232263     "Tier B"      3      "2018-05-28 10:59:00"  736842.457639
5245478     "Tier B"      EA     "2018-05-27 13:58:00"  736841.581944
4975552     "Tier B"      2      "2018-05-30 21:01:00"  736844.875694
4975563     "Tier A"      2      "2018-05-30 21:01:00"  736844.875694'''

df = pd.read_table(StringIO(txt), sep="\s+", parse_dates=[3])

Plot

fig, ax = plt.subplots(figsize=(13,4))

fig = sns.countplot(x="element_date", hue="element", data=df, ax=ax)

x_dates = df['element_date'].dt.strftime('%m-%d').sort_values().unique()
ax.xaxis.set_major_formatter(plt.FixedFormatter(x_dates))

plt.legend(loc='upper left')
plt.show()
plt.close()