CODEWITHSUNDEEP
pandasfunctiondataframearguments
Des Grieux
Des Grieux

Reputation: 570

calling function with pandas column as argument

I wrote a function that takes a pandas data frame and two of its columns. Inside the function, I want to group the elements of the first column by the elements of the second column. The goal of the function is to generate a bar chart using matplotlib that plots the grouped counts. I do not know how to refer to the column arguments so they can be recognized by the group-by call inside the function.

I tried using df['col'] and 'col' but none of these worked. When I use df['col'], I get this error:

AttributeError: 'DataFrameGroupBy' object has no attribute 'x'

When I use 'col', I get this error:

AttributeError: 'DataFrameGroupBy' object has no attribute 'x'

Here is an example implementation, first without the function, to generate the expected result, and then with the function.

import pandas as pd

# generate dataframe
df = pd.DataFrame()
df['col_A'] = [1, 4, 3, 2, 2, 1, 1, 4, 3, 2]
df['col_B'] = ['a', 'a', 'b', 'b', 'b', 'c', 'c', 'c', 'c', 'c']

# plot counts
import matplotlib.pyplot as plt

counts = df.groupby('col_B').col_A.count()
counts = counts.sort_values(ascending=False)

fig = plt.figure(figsize=(10,8))
counts.plot.barh(ylim=0).invert_yaxis()

# plot count with function
def count_barplot(data, x, y):

    counts = data.groupby(y).x.count()
    counts = counts.sort_values(ascending=False)

    fig = plt.figure(figsize=(10,8))
    counts.plot.barh(ylim=0).invert_yaxis()

# function call
count_barplot(df, df['col_A'], df['col_B'])

How do I specify the data frame column arguments inside the function and in the function call, so that the group-by function can recognize them?

Upvotes: 1

Views: 3034

Answers (2)

Brendan
Brendan

Reputation: 4011

The problem is that your function call is providing a a dataframe and two series as its arguments, while what you want to pass is a dataframe and column names. Note that you also want to use the [] syntax for referring to the column in your groupby, and you can simplify your count method using the built-in value_counts() method.

Thus, using your syntax:

# plot count with function
def count_barplot(data, x, y):

    counts = data.groupby(y)[x].count()
    counts = counts.sort_values(ascending=False)

    fig = plt.figure(figsize=(10,8))
    counts.plot.barh(ylim=0).invert_yaxis()

count_barplot(df, 'col_A', 'col_B')

or more simply:

# plot count with function
def count_barplot(data, y):

    counts = df[y].value_counts()

    fig = plt.figure(figsize=(10,8))
    counts.plot.barh(ylim=0).invert_yaxis()

# function call
count_barplot(df, 'col_B')

or even

def count_barplot(data, x, y):
    fig = plt.figure(figsize=(10,8))
    df[y].value_counts(ascending=True).plot.barh(ylim=0)

Upvotes: 2

Hryhorii Pavlenko
Hryhorii Pavlenko

Reputation: 3910

This way it works for me:

def count_barplot(data, x, y):

    counts = data.groupby(y)[x].count()
    counts = counts.sort_values(ascending=False)

    fig = plt.figure(figsize=(10,8))
    counts.plot.barh(ylim=0).invert_yaxis()

# function call
count_barplot(df, 'col_A', 'col_B')

Upvotes: 2

Related Questions