I'm trying to change the date format to a column but I cant apply my function because I have NaN cells.

# change date format
def dmy_to_dmy(d):
    return datetime.strptime(d, '%d %B %Y').strftime('%d/%m/%Y')

dates2['Dates_Reins'] = dates2['Dates_Reins'].apply(dmy_to_dmy)

My data looks like this:

46  9 September 2021
47  NaN
48  24 July 2021
49  28 September 2021
50  18 October 2021
51  8 January 2021
52  NaN

Thanks

CodePudding user response：

There are two options, ingoring the problem or detecting the problem. The first one is the easy one since you try and pass if it doesn't happen. The second one will detect the NaN

First solution

# change date format
def dmy_to_dmy(d):
    try:
        return datetime.strptime(d, '%d %B %Y').strftime('%d/%m/%Y')
    except TypeError:
        return d
    
dates2['Dates_Reins'] = dates2['Dates_Reins'].apply(dmy_to_dmy)

Second solution

# change date format
def dmy_to_dmy(d):
    if d == np.nan:
        return d
    else:
        return datetime.strptime(d, '%d %B %Y').strftime('%d/%m/%Y')
    
dates2['Dates_Reins'] = dates2['Dates_Reins'].apply(dmy_to_dmy)

Since you haven't provided the data d==np.nan you have to find for yourself

CodePudding user response：

You can detect whether the input to your function is a Nan value and return it before you start to operate with it:

def dmy_to_dmy(d):
    if d == np.nan:
      return np.nan
    return datetime.strptime(d, '%d %B %Y').strftime('%d/%m/%Y')

CodePudding user response：

You can either make your function ignore NaN:

def dmy_to_dmy(d):
    if isinstance(d, float) and np.isnan(d):
        return np.nan
        
    return datetime.strptime(d, '%d %B %Y').strftime('%d/%m/%Y')

dates2['Dates_Reins'] = dates2['Dates_Reins'].apply(dmy_to_dmy)

Or, call your function on only the non-NaN values:

notna = dates2['Dates_Reins'].notna()
dates2['Dates_Reins'] = dates2.loc[notna, 'Dates_Reins'].apply(dmy_to_dmy)

CodePudding user response：

Try:

dates2['Dates_Reins'].apply(lambda x: dmy_to_dmy(x) if x != np.nan else x)

Inspired by: https://stackoverflow.com/a/56589062/7509907

CodePudding user response：

I wouldn't use apply here:

dates2['Dates_Reins'] = (
   pd.to_datetime(dates2['Dates_Reins'], format='%d %B %Y')
     .dt.strftime('%d/%m/%Y')
)

• pd.to_datetime() can replace strptime()
• The .dt accessor allows direct access to strftime()

This is faster and deals with missing values implicitly.