Using Pandas Date Offset with Offset Aliases Directly

user_3179058 0
Viewed 1013

I'm sure I'm just not searching the correct phrase to answer this, but in Pandas, one can take any DateTime object and add or subtract from it using DateOffset like:

pd.datetime.now() + pd.DateOffset(months=2)

pd.datetime.now() + pd.DateOffset(weeks=4)

pd.datetime.now() + pd.DateOffset(days=2)

etc.

But in different context, Offset Aliases are also used: https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#offset-aliases

My question is is there a way to use DateOffset and pass an Offset Alias to determine the Offset basis?

Something like: pd.datetime.now() + pd.DateOffset(n=some_int, freq='W') for some_int weeks as an example?

If that isn't possible, where should I dig in to get something to that effect?

(The reason is because there is a function I'm utilizing the Offset Alias and I don't want to have to create a long if...else statement to convert the freq string to different instances of pd.DateOffset(weeks=n | years=n | months=n | etc.) I rather the freq string dictate the DateOffset in one line of code and still be dynamic to handle the different frequencies of time.)

Edited: to add a custom function that takes care of what I need, but it would be nice to have a solution in to_timedelta or DateOffset so that the solution is upstream and more efficient. For example, I'd love to utilize the Business versions of all the freq parameters I'm using so that the n can take in information more natural and raw to the source it came from.

def datedelta(date, n=0, freq='M'):
    from pandas import to_datetime, DateOffset
    
    if n == 0:
        date_sign = 1
    else:
        date_sign = np.abs(n)/n
        
    freq = freq.lower()
    
    if freq == 'y':
        dtOff = DateOffset(years=abs(n))
    elif freq == 'q':
        dtOff = DateOffset(quarters=abs(n))
    elif freq == 'm':
        dtOff = DateOffset(months=abs(n))
    elif freq == 'w':
        dtOff = DateOffset(weeks=abs(n))
    elif freq == 'd':
        dtOff = DateOffset(days=abs(n))
    else:
        raise ValueError("The freq parameter not one of the following: {'Y', 'Q', 'M', 'W', 'D'}")
        
    return to_datetime(date) + date_sign * dtOff
pandas datetimeoffset python-3.x datetime python
2 Answers

I think the to_timedelta function is doing what you want.

time = pd.datetime.now() + pd.to_timedelta(5, unit='W')

Update
The to_timedelta function does not support anymore month, quarter and year. There are two types of object for time offset:

  • DateOffset which respect the calendar time:
  • TimeOffset which is absolute time (the difference comes up with daylight saving time days).

More details with the documentation about it. The to_timedelta function cannot know which one you want to use.

user_118321270

Maybe this function would be helpful?

def offset(freq:str, n:int = 0):
    """
    Pandas DateOffset wrapper
    import pandas as pd
    """
    # =============================================================================
    # offsets_df = pd.read_html('https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html')[2]
    # offsets_df.drop(columns = ['Description'], inplace = True)
    # offsets_df.replace('None', np.NaN, inplace = True)
    # offsets_df.dropna(subset = ['Frequency String'], inplace = True)
    # offsets_df['Frequency String'] = offsets_df['Frequency String'].str.replace("\'", '')
    # for x in ['Date Offset', 'Frequency String']:
    #    offsets_df[x] =  offsets_df[x].str.split(' or ')
    # offsets_df['Date Offset'] = offsets_df['Date Offset'].map(lambda x: x[0])
    # offsets_df = explode_rows(offsets_df, 'Frequency String', fill_value = '')
    # offsets_df.drop_duplicates(subset = ['Frequency String'], inplace = True)
    # offsets_d = dict(zip(offsets_df['Frequency String'], offsets_df['Date Offset']))
    # =============================================================================
    offsets_d = {'B': 'BDay',
                 'C': 'CDay',
                 'W': 'Week',
                 'WOM': 'WeekOfMonth',
                 'LWOM': 'LastWeekOfMonth',
                 'M': 'MonthEnd',
                 'MS': 'MonthBegin',
                 'BM': 'BMonthEnd',
                 'BMS': 'BMonthBegin',
                 'CBM': 'CBMonthEnd',
                 'CBMS': 'CBMonthBegin',
                 'SM': 'SemiMonthEnd',
                 'SMS': 'SemiMonthBegin',
                 'Q': 'QuarterEnd',
                 'QS': 'QuarterBegin',
                 'BQ': 'BQuarterEnd',
                 'BQS': 'BQuarterBegin',
                 'REQ': 'FY5253Quarter',
                 'A': 'YearEnd',
                 'AS': 'YearBegin',
                 'BYS': 'YearBegin',
                 'BA': 'BYearEnd',
                 'BAS': 'BYearBegin',
                 'RE': 'FY5253',
                 'BH': 'BusinessHour',
                 'CBH': 'CustomBusinessHour',
                 'D': 'Day',
                 'H': 'Hour',
                 'T': 'Minute',
                 'min': 'Minute',
                 'S': 'Second',
                 'L': 'Milli',
                 'ms': 'Milli',
                 'U': 'Micro',
                 'us': 'Micro',
                 'N': 'Nano'}
    return eval(f'pd.offsets.{offsets_d[freq]}({n})')
user_35231760
Related