[FIXED] Test for np.nan in row wise function

Issue

I have a function that I want to apply to each row of a data frame to create a new column. The value returned from the function will be different if certain columns are NaN. I have several conditions in the function (more complicated than the example below), otherwise, I would use np.where.

How do I test if the column is nan using the function below? I tried row['id'] is np.nan but that doesn't work.

# data
d = {'name': {0: 'dave', 1: 'hagen'},
     'id': {0: 123456.0, 1: np.nan},
     'position': {0: np.nan, 1: 5600.0},
     'test': {0: 'has an id', 1: 'has an id'}}

df = pd.DataFrame(d)

# function
def test_func(row):
    if row['id'] is np.nan:
        val = "missing id"
    else:
        val = 'has an id'
    return val

# apply function
df['test'] = df.apply(test_func, axis=1)

Results (I would expect the test column to say "missing an id" on the second row since id is np.nan on that row:

Solution

numpy has its own isnan() function:

the simplest way is:

val = "missing id" if np.isnan(row['id']) else 'has an id'

Answered By - Glauco