Making functional function in Pandas

Question

I write my own function in Python. The function is very simple and below you can see data and function:

  data_1 = {'id':['1','2','3','4','5'],
            'name': ['Company1', 'Company1', 'Company3', 'Company4', 'Company5'], 
            'employee': [10, 3, 5, 1, 0], 
            'sales': [100, 30, 50, 200, 0], 
           }
    df = pd.DataFrame(data_1, columns = ['id','name', 'employee','sales'])
    
    threshold_1=40
    threshold_2=50

And the function is written below:

  def my_function(employee,sales):
        conditions = [
        (sales == 0 ),
        (sales < threshold_1), 
        (sales >= threshold_1 & employee <= threshold_2)]
        values = [0, sales*2, sales*4]
        sales_estimation = np.select(conditions, values)    
        return (sales_estimation)

df['new_column'] = df.apply(lambda x: my_function(x.employee,x.sales), axis=1)
df

So this function works well and gives the expected result.

Now I want to make the same function but with vectorized operation across Pandas Series. I need to have this function because vectorized operation decreases the time for executing. For this reason, I wrote this function but the function is not working.

  def my_function1(
        pandas_series:pd.Series
        )-> pd.Series:
        """
        Vectorized operation across Pandas Series
        """
        conditions = [
        (sales == 0 ),
        (sales < threshold_1), 
        (sales >= threshold_1 & employee <= threshold_2)]
        values = [0, sales*2, sales*4]
        sales_estimation = np.select(conditions, values)    
        return sales_estimation
    
    df['new_column_1']=my_function1(data['employee','sales'])

Probably my error is related to the input parameters of this function. So can anybody help me how to solve this problem and make my_function1 functional?

mozway · Accepted Answer · 2022-02-22 09:11:37Z

1

You need to slightly change one condition to be able to pass Series:

(sales >= threshold_1 & employee <= threshold_2)
# equivalent to
# sales >= (threshold_1 & employee) <= threshold_2

into:

(sales >= threshold_1) & (employee <= threshold_2)

as the operator precedence was incorrect.

def my_function(employee,sales):
        conditions = [
        (sales == 0 ),
        (sales < threshold_1), 
        (sales >= threshold_1) & (employee <= threshold_2)]
        values = [0, sales*2, sales*4]
        sales_estimation = np.select(conditions, values)    
        return (sales_estimation)

df['new_column'] = my_function(df['employee'], df['sales'])

output:

  id      name  employee  sales  new_column
0  1  Company1        10    100         400
1  2  Company1         3     30          60
2  3  Company3         5     50         200
3  4  Company4         1    200         800
4  5  Company5         0      0           0

You can also pass the whole dataframe ans subset the columns there:

def my_function(df):
    employee = df['employee']
    sales = df['sales']
    conditions = [
    (sales == 0 ),
    (sales < threshold_1), 
    (sales >= threshold_1) & (employee <= threshold_2)]
    values = [0, sales*2, sales*4]
    sales_estimation = np.select(conditions, values)    
    return (sales_estimation)

df['new_column'] = my_function(df)

edited Feb 22, 2022 at 9:11

answered Feb 22, 2022 at 9:05

mozway

267k13 gold badges56 silver badges106 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

silent_hunter Over a year ago

Does this function convert the dataset into Pandas series? I am asking this because I need to have vectorized series

mozway Over a year ago

@silent_hunter this code processes the data in a vectorial way, it outputs a numpy array, but assignment to the dataframe transforms it into a Series automatically

mozway Over a year ago

@silent_hunter it should be much faster than the apply approach if this is your concern ;)

jezrael · Accepted Answer · 2022-02-22 09:12:42Z

1

Pass Series to function like and also add () for avoid ValueError: The truth value of a Series is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all(). because priority of operators:

def my_function1(employee, sales):
      conditions = [
      (sales == 0 ),
      (sales < threshold_1), 
      (sales >= threshold_1) & (employee <= threshold_2)] #<- here
      values = [0, sales*2, sales*4]
      sales_estimation = np.select(conditions, values)    
      return sales_estimation
    
df['new_column_1']= my_function1(df['employee'],df['sales'])
print (df)
  id      name  employee  sales  new_column_1
0  1  Company1        10    100           400
1  2  Company1         3     30            60
2  3  Company3         5     50           200
3  4  Company4         1    200           800
4  5  Company5         0      0             0

edited Feb 22, 2022 at 9:12

answered Feb 22, 2022 at 9:07

jezrael

868k103 gold badges1.4k silver badges1.3k bronze badges

4 Comments

silent_hunter Over a year ago

Does this function convert dataset into Pandas series? I am asking this because I need to have vectorized series

jezrael Over a year ago

@silent_hunter - If select column like df['employee'] it is Series. Same for sales

jezrael Over a year ago

@silent_hunter - you can check it print (type(df['employee']))

jezrael Over a year ago

@silent_hunter - aand in np.select are converted values to numpy.arrays (Series is built by arrays), so working fast and after assign to new column arrays is converted to new column. And it is Series if select like df['new_column_1']

Collectives™ on Stack Overflow

Making functional function in Pandas

2 Answers 2

3 Comments

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related