Python functions are custom blocks of code that transform inputs into outputs. For example, the round()
function transforms an input number into a rounded version:
round(3.14)
# Output: 3
# Python function syntaxdef function_name(input_parameters):<indented code to copute output>return function_output
The code that a function executes must be indented after the function def
line. The standard indentation is to consistently use four spaces or one tab for each line of code. But the key requirement is that all lines for the function are indented the same amount.
The example function squared_difference
performs operations to compute the squared difference between two numbers. Each line of code needed for the calculation is indented consistently by four spaces.
# Function that computes the squared difference of two numbersdef squared_difference(numbers):# code blocks indented by four spacesdiff = numbers[0] - numbers[1]squared_diff = diff**2return squared_diffsquared_difference([3,1])# Output: 4
The return
statement in a Python function determines the output of the function. The output can be a value on its own or a variable storing a value.
Multiple output values can be returned by specifying each output separated by a comma:
def function(input):
<indented code>
return output1, output2
def find_min_max(numbers):min_value = np.min(numbers)max_value = np.max(numbers)return min_value, max_valueminimum, maximum = find_min_max([3,6,2,5,1])print(minimum)# Output: 1print(maximum)# Output: 6
Python functions can have multiple inputs, using a comma to separate each input inside the function parentheses.
def function(input1, input2):
# Function that computes a multivariate equationdef line(x, m, b):y = m*x + breturn y
Function inputs can have default values, to be used if the user does not provide input. Default values are assigned during definition by placing an =
sign after the input parameter name followed by the default value.
The example function line
takes in three input parameters x
, m
, and b
. When calling line
without specifying a value for b
, the function defaults to using b=0
.
# Function that computes a multivariate equation# Default value b=0def line(x, m, b=0):y = m*x + breturn yline(x=2,m=1)# Output: 2
To use/call a Python function, write the function name followed by parentheses:
name()
If the function has inputs, specify the inputs in the same order as in the function definition, or by using the input parameter name/keyword (see code snippet for examples).
# Function that computes a mathematical formuladef equation(a, b, c=0):y = 4*a + 2*b + creturn y# Calling w/ ordered argumentsequation(1,2,2)# Output: 10# Calling w/ parameter keywordsequation(b=2.c=2,a=1)# Output: 10
.apply()
Method on GroupBy ObjectsThe .apply()
method can apply custom aggregation functions to a GroupBy.
In the code snippet, we’ve written a function count_no_goals
that takes a column as input and counts the number of entries with the value 0
.
We have then applied that to the results
DataFrame grouped by the tournament
column.
This gives us a count of the number of games in each tournament with no goals.
results
year | home_team | away_team | total_goals | tournament |
---|---|---|---|---|
2009 | Czech Republic | Northern Ireland | 0 | FIFA World Cup qualification |
2012 | Egypt | Mauritania | 3 | Friendly |
2015 | Turkey | Latvia | 2 | UEFA Euro qualification |
def count_no_goals(column):return (column == 0).sum()matches_zero = results.groupby('tournament')\['total_goals'].apply(count_no_goals)
.apply()
Method Across Rows or ColumnsThe .apply()
method can apply functions across either the rows or columns of a DataFrame using the axis
keyword where
axis=1
applies the function across the rowsaxis=0
applies the function to each columnHere we apply the sum
function to df
where row_sum
is the output of the function applied across the rows and column_sum
is the output of the function applied to each column:
A | B | row_sum | ||
---|---|---|---|---|
0 | 1 | 3 | 4 | |
1 | 2 | 4 | 6 | |
column_sum | 3 | 7 |
df = pd.DataFrame({'A':[1,2], 'B':[3,4]})# Sum the values across each rowdf.apply(sum, axis=1)# Sum the values in each columndf.apply(sum, axis=0)
While applying custom functions using the .apply()
method is very flexible, it is oftentimes much slower than using built-in pandas methods, especially with bigger datasets. When building data pipelines to clean, pre-process, and model data, it is important to evaluate the advantages and disadvantages of using custom functions or built-in methods for your data task.