How to plot categorical variable against a date column in Python

Question

I have data that looks like this

Date         Fruit
2017-01-01   Orange
2017-01-01   Apple
2017-01-08   Orange
2017-01-09   Orange
2017-01-09   Apple

I want to plot Number of Oranges, Number of Apples by Date in a single plot. How would I do that?

I grouped them by Date and I see the result.

df.groupby(['Date','Fruit']).size()

Date         Fruit
2017-01-01   Orange  1
             Apple   1
2017-01-08   Orange  1
2017-01-09   Orange  1
             Apple   1

I tried this but it gives a bar plot having two categories but not against the dates.

sns.catplot(x="Fruit", hue="Fruit", kind="count",
            palette="pastel", edgecolor=".6",
            data=df);

How can plot a graph have Date on the x-axis and number of apples and the number of oranges for each date?

Answer 1

Framing the dataset:

df = pd.DataFrame(columns=["Date", "Fruit"], data=[['2017-01-01','Orange'], ['2017-01-01','Orange'], ['2017-01-01','Apple'], ['2017-01-08','Orange'], ['2017-01-09','Orange'], ['2017-01-09','Apple']])

By using unstack and group by a bar plot can be drawn:

(df
 .groupby(['Date', 'Fruit'])
 .size()
 .unstack()
 .plot.bar()
)

Answer 2

You can do something like this.

# dummy data
date_range = pd.date_range('2019-01-01', '2019-01-06', freq='D')
df = pd.DataFrame(['Orange', 'Apple', 'Orange', 'Orange',
                   'Apple', 'Apple', 'Apple', 'Orange', 'Orange'],
    index=[date_range[0], date_range[0], date_range[1], date_range[2],
           date_range[2], date_range[2], date_range[2], date_range[3],
           date_range[3]],
    columns=['Fruit'])
df.index.name = 'Date'

groupby as you do, then unstack, which looks like this.

>>> print(df.unstack())

Fruit       Apple  Orange
Date                     
2019-01-01    1.0     1.0
2019-01-02    NaN     1.0
2019-01-03    3.0     1.0
2019-01-04    NaN     2.0

And then plot the unstacked data.

df.unstack().plot(kind='bar')
plt.show()

(You'll have to do something about the date formatting though).