Using np.busday_count with NaT values-CodePudding

I am calculating a field like this:

f['days'] = np.busday_count(pd.to_datetime(f['Start time']).values.astype('datetime64[D]'), \
            pd.to_datetime(f['Stop time']).values.astype('datetime64[D]'))

However, I have columns f['Start time'] or f['Stop time'] with NaT values, I tried 'f['Start time']is pd.NaT but I don't know how to implement this in above code

CodePudding user response：

You should read up on indexing.

Here i'm creating an index where neither start nor stop time is na (nan, nat, None etc). The .isna() method returns a series with True for na values. ~ is the bitwise not operator so we change both indexes to indicate True for valid dates. & is the bitwise and operator so we "merge" the 2 to indicate True where both dates are valid. You can then "filter" every Series your work on with this index.

valid_dates_index =  (~f['Stop time'].isna()) & (~f['Start time'].isna())

f['days'][valid_dates_index ] = np.busday_count(\
    pd.to_datetime(f['Start time'][valid_dates_index ]).values.astype('datetime64[D]'), \
    pd.to_datetime(f['Stop time'][valid_dates_index ]).values.astype('datetime64[D]'))

You could also drop the rows containing na values with .dropna(axis=0,how='any')

Btw most numpy functions have 'ignore_nan' keyword that you can set to True.