How to find the longest sequence of NaNs in Dataframe?

Question

Lets say I have a dataframe like this:

Time                           A            B           C             D                                                               
2019-06-17 08:45:00     12089.89     12089.89    12087.71      12087.71      
2019-06-17 08:46:00          NaN          NaN    12087.71      12087.91      
2019-06-17 08:47:00          NaN     12088.21    12084.21      12085.21      
2019-06-17 08:48:00          NaN     12090.21         NaN           NaN      
2019-06-17 08:49:00          NaN     12090.21         NaN           NaN     
2019-06-17 08:50:00          NaN          NaN    12504.11           NaN     
2019-06-17 08:51:00          NaN          NaN    12503.11      12503.11    
2019-06-17 08:52:00     12504.11          NaN    12503.11      12503.11      
2019-06-17 08:53:00     12503.61     12503.61    12503.61      12503.61      
2019-06-17 08:54:00     12503.61     12503.61    12503.11      12503.11     

How to find the length of the longest uninteruppted NaN sequence in the whole df? ( in the example its 6 ) efficiently?

EDIT: forgot to emphasize the word "efficiently", because the df is about 1mio rows long

Answer 1

Let's try apply a user-defined function, which in turns uses cumsum() to identify blocks:

def max_na(s):
    isna = s.isna()
    blocks = (~isna).cumsum()
    return isna.groupby(blocks).sum().max()

df.apply(max_na).max()
# 6.0