将多级 X 轴添加到 matplotlib/Seaborn（月和年）

Question

I am having difficulty adding a multi level axis with month and then year to my plot and I have been unable to find any answers anywhere. I have a dataframe which contains the upload date as a datetime dtype and then the year and month for each row. See Below:

    Upload Date Year    Month      DocID
0   2021-03-22  2021    March      DOC146984
1   2021-12-16  2021    December   DOC173111
2   2021-12-07  2021    December   DOC115350
3   2021-10-29  2021    October    DOC150149
4   2021-03-12  2021    March      DOC125480
5   2021-06-25  2021    June       DOC101062
6   2021-05-03  2021    May        DOC155916
7   2021-11-14  2021    November   DOC198519
8   2021-03-20  2021    March      DOC159523
9   2021-07-19  2021    July       DOC169328
10  2021-04-13  2021    April      DOC182660
11  2021-10-08  2021    October    DOC176871
12  2021-09-19  2021    September  DOC185854
13  2021-05-16  2021    May        DOC192329
14  2021-06-29  2021    June       DOC142190
15  2021-11-30  2021    November   DOC140231
16  2021-11-12  2021    November   DOC145392
17  2021-11-10  2021    November   DOC178159
18  2021-11-06  2021    November   DOC160932
19  2021-06-16  2021    June       DOC131448

What I am trying to achieve is to build a bar chart which has the count for number of documents in each month and year. The graph would look something like this:

The main thing is that the x axis is split by each month and then further by each year, rather than me labelling each column with month and year (eg 'March 2021'). However I can't figure out how to achieve this. I've tried using a countplot but it only allows me to choose month or year (See Below). I have also tried groupby but the end product is always the same. Any Ideas?

This is using randomly generated data, see the code to replicate below:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.style as style
import seaborn as sns
from datetime import date, timedelta
from random import choices
np.random.seed(42)
  
# initializing dates ranges 
test_date1, test_date2 = date(2020, 1, 1), date(2021, 6, 30)
  
# initializing K
K = 2000
  
res_dates = [test_date1]
  
# loop to get each date till end date
while test_date1 != test_date2:
    test_date1 += timedelta(days=1)
    res_dates.append(test_date1)
  
# random K dates from pack
res = choices(res_dates, k=K)

# Generating dataframe
df = pd.DataFrame(res, columns=['Upload Date'])

# Generate other columns
df['Upload Date'] = pd.to_datetime(df['Upload Date'])
df['Year'] = df['Upload Date'].dt.year
df['Month'] = df['Upload Date'].dt.month_name()
df['DocID'] = np.random.randint(100000,200000, df.shape[0]).astype('str')
df['DocID'] = 'DOC' + df['DocID']

# plotting graph
sns.set_color_codes("pastel")
f, ax = plt.subplots(figsize=(20,8))
sns.countplot(x='Month', data=df)

Answer 1

A new column with year and month in numeric form can serve to indicate the x-positions, correctly ordered. The x-tick labels can be renamed to the month names. Vertical lines and manual placing of the year labels lead to the final plot:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

test_date1, test_date2 = '20200101', '20210630'

months = pd.date_range('2021-01-01', periods=12, freq='M').strftime('%B')
K = 2000
df = pd.DataFrame(np.random.choice(pd.date_range(test_date1, test_date2), K), columns=['Upload Date'])
df['Year'] = df['Upload Date'].dt.year
# df['Month'] = pd.Categorical(df['Upload Date'].dt.strftime('%B'), categories=months)
df['YearMonth'] = df['Upload Date'].dt.strftime('%Y%m').astype(int)
df['DocID'] = np.random.randint(100000, 200000, df.shape[0]).astype('str')
df['DocID'] = 'DOC' + df['DocID']

sns.set_style("white")
sns.set_color_codes("pastel")
fig, ax = plt.subplots(figsize=(20, 8))
sns.countplot(x='YearMonth', data=df, ax=ax)
sns.despine()
yearmonth_labels = [int(l.get_text()) for l in ax.get_xticklabels()]
ax.set_xticklabels([months[ym % 100 - 1] for ym in yearmonth_labels])
ax.set_xlabel('')

# calculate the positions of the borders between the years
pos = []
years = []
prev = None
for i, ym in enumerate(yearmonth_labels):
    if ym // 100 != prev:
        pos.append(i)
        prev = ym // 100
        years.append(prev)
pos.append(len(yearmonth_labels))
pos = np.array(pos) - 0.5
# vertical lines to separate the years
ax.vlines(pos, 0, -0.12, color='black', lw=0.8, clip_on=False, transform=ax.get_xaxis_transform())
# years at the center of their range
for year, pos0, pos1 in zip(years, pos[:-1], pos[1:]):
    ax.text((pos0 + pos1) / 2, -0.07, year, ha='center', clip_on=False, transform=ax.get_xaxis_transform())

ax.set_xlim(pos[0], pos[-1])
ax.set_ylim(ymin=0)
plt.tight_layout()
plt.show()