pandas: how to check for nulls in a float column? - pandas

Question

I am conditionally assigning a column based on whether another column is null:
df = pd.DataFrame([
{ 'stripe_subscription_id': 1, 'status': 'past_due' },
{ 'stripe_subscription_id': 2, 'status': 'active' },
{ 'stripe_subscription_id': None, 'status': 'active' },
{ 'stripe_subscription_id': None, 'status': 'active' },
])
def get_cancellation_type(row):
if row.stripe_subscription_id:
if row.status == 'past_due':
return 'failed_to_pay'
elif row.status == 'active':
return 'cancelled_by_us'
else:
return 'cancelled_by_user'
df['cancellation_type'] = df.apply(get_cancellation_type, axis=1)
df
But I don't get the results I expect:
I would expect the final two rows to read cancelled_by_user, because the stripe_subscription_id column is null.
If I amend the function:
def get_cancellation_type(row):
if row.stripe_subscription_id.isnull():
Then I get an error: AttributeError: ("'float' object has no attribute 'isnull'", 'occurred at index 0'). What am I doing wrong?

Answer 1

With pandas and numpy we barely have to write our own functions, especially since our own functions will perform slow because these are not vectorized and pandas + numpy provide a rich pool of vectorized methods for us.
In this case your are looking for np.select since you want to create a column based on multiple conditions:
conditions = [
df['stripe_subscription_id'].notna() & df['status'].eq('past_due'),
df['stripe_subscription_id'].notna() & df['status'].eq('active')
]
choices = ['failed_to_pay', 'cancelled_by_us']
df['cancellation_type'] = np.select(conditions, choices, default='cancelled_by_user')
status stripe_subscription_id cancellation_type
0 past_due 1.0 failed_to_pay
1 active 2.0 cancelled_by_us
2 active NaN cancelled_by_user
3 active NaN cancelled_by_user