如何在Python的不均匀列上迭代函数？

Question

I know this question may not make much sense, but hopefully the following example will clarify it. I need to reference one string in column sentA and then compare it to all strings in sentB . The following example shows the dataframe I defined as questions .

sentA     sentB
str1      str1
str2      str2
          str3

The code I'm currently using can only compare even columns and looks like this:

def compare(row):
    sentA = row[0]
    return pd.Series([simalarity_funct(sentA, sentB) for sentB in questions['sentB']])

results = questions.apply(compare, axis=1).T

That code gives me 3 outputs for str1A (similarity to str1B, str2B, and str3B) and puts them in a column.

Here is another example with simplified code based on the input df numbers :

num1    num2 
   3       5    
   4       6
           7

def multiply(num1, num2):
    return num1*num2

def compare(row):
    num1 = row[0]
 # I would like to prevent this next statement from passing an "NaN" to the 
 # multiply function. The empty cells will always be at the end of the column.
    return pd.Series([multiply(num1, num2) for num2 in numbers['num2']])

results = numbers.apply(compare, axis=1).T
print(results)
15     20     NaN
18     24     NaN
21     28     NaN

The underlying problem is that my similarity function will throw an error if it is fed bad data. The easiest way I can think of fixing this is by not feeding it bad data. Is there a way I can modify the last step to prevent it from passing "NaN" to the similarity function?

Answer 1

def compare(row):
    num1 = row[0]
    pd.Series([multiply(num1, num2) for num2 in numbers[numbers.num2.notnull()].num2 ])

numbers[numbers.num1.notnull()].apply(compare, axis=1).T