poyee-data-warehouse/dw_base/spark/udf/test/d2str.py

import re

pattern1 = r'(\d+)[- /\']?([A-Za-z\d]+)[- /\'\.]+(\d+ ?\d+)'
pattern2 = r'[,-;\']?(\d+)[- /\']?([A-Za-z\d]+)[- /\'\.]+(\d+ ?\d+)'
pattern3 = r'[A-Za-z]+, ([A-Za-z]+) (\d+), (\d+)'
pattern4 = r'([A-Za-z\d]+) ([A-Za-z\d]+)\.? (\d+)'
pattern5 = r'(!|\d+)[- ]+([A-Za-z]+)[ ]?(\d+)'

month_dict = {'Agsts': '08',
              'Agsutus': '08',
              'Agts': '08',
              'Agust': '08',
              'Agustus': '08',
              'Apr': '04',
              'April': '04',
              'Aprl': '04',
              'Aprll': '04',
              'Aug': '08',
              'August': '08',
              'Deaember': '12',
              'Dec': '12',
              'December': '12',
              'Des': '12',
              'Desember': '12',
              'Feb': '02',
              'Febrauri': '02',
              'Februari': '02',
              'Februaru': '02',
              'February': '02',
              'Febuari': '02',
              'JULI': '07',
              'Jan': '01',
              'Januari': '01',
              'January': '01',
              'Jul': '07',
              'Juli': '07',
              'July': '07',
              'Jun': '06',
              'June': '06',
              'Juni': '06',
              'MAy': '05',
              'Mar': '03',
              'March': '03',
              'Maret': '03',
              'Mart': '03',
              'May': '05',
              'Mei': '05',
              'Mrt': '03',
              'No': '11',
              'Nof': '11',
              'Nop': '11',
              'Nopember': '11',
              'Nov': '11',
              'November': '11',
              'Oct': '10',
              'October': '10',
              'Okober': '10',
              'Okt': '10',
              'Okt0ber': '10',
              'Oktober': '10',
              'Pebruari': '02',
              'Sep': '09',
              'Sepetember': '09',
              'Sept': '09',
              'September': '09',
              'Septembver': '09',
              'agust': '08',
              'des': '12',
              'desmb': '12',
              'juli': '07',
              'maret': '03',
              'mei': '05',
              'november': '11',
              'oct': '10',
              'oktober': '10'
              }


def get_date(text: str):
    match1 = re.match(pattern1, text)
    if match1:
        day, month, year = match1.groups()
        return year, month, day
    match2 = re.match(pattern2, text)
    if match2:
        day, month, year = match2.groups()
        return year, month, day
    match3 = re.match(pattern3, text)
    if match3:
        month, day, year = match3.groups()
        return year, month, day
    match4 = re.match(pattern4, text)
    if match4:
        day, month, year = match4.groups()
        return year, month, day
    match5 = re.match(pattern5, text)
    if match5:
        day, month, year = match5.groups()
        return year, month, day
    return None, None, None


def clean_date_indonesia(text):
    if text:
        year, month, day = get_date(text)
        year = clean_year(year)
        month = clean_month(month)
        day = clean_day(day)
        if year and month and day:
            return f'{year}-{month}-{day}'
    else:
        return None


def clean_year(year: str):
    if year:
        year = year.replace(' ', '')
        if len(year) == 1:
            return f'200{year}'
        elif len(year) == 2:
            if year < '30':
                return f'20{year}'
            else:
                return f'19{year}'
        elif len(year) == 3:
            if year[0] == '0':
                return f'2{year}'
            else:
                return f'{year[0]}0{year[1:]}'
        try:
            year_int = int(year)
            if year_int > 2024 or year_int <= 1900:
                return None
        except ValueError:
            return None
        return year
    else:
        return None


def clean_month(month: str):
    if month:
        if len(month) == 1:
            month = f'0{month}'
        elif re.match( r'^\d{2}$', month):
            month = month
        else:
            month = month_dict.get(month)
        try:
            month_int = int(month)
            if month_int < 1 or month_int > 12:
                return None
        except ValueError:
            return None
        return month
    else:
        return None


def clean_day(day: str):
    if day:
        if len(day) == 1:
            if day in ['!', 'I', '1']:
                return '01'
            else:
                return f'0{day}'
        elif len(day) == 2:
            try:
                day_int = int(day)
                if day_int < 1 or day_int > 31:
                    return None
            except ValueError:
                return None
            return day
    else:
        return None



if __name__ == '__main__':
    test_cases = [
        'Monday, November 03, 2014',
        'Tuesday, September 08, 2015',
        "28 Agustus' 09",
        "19Juli 2011",
        '25September 2027',
        'I Februari 2011',
        ',30 April 2013',
        '25 Agsts. 08',
        "4'Nov 08",
        "15 Des.08",
        "'06-Sept-10",
        "! Dec 09",
        "06- Mei 09",
        "1 Desember2009",
        "18 MAy09",
        "22-Jan013",
        "21 Okober-20 10",
        "01 Oktober 2 013",
        "19-No-13",
        "8 oct 9 ",
    ]
    for test_case in test_cases:
        print(test_case + '    ->    ', get_date(test_case))
    year_cases = [
        '00',
        '01',
        '09',
        '19',
        '20',
        '79',
        '85',
        '96',
        '99',
        '013',
        '204',
        '209',
        '210',
        '2 013',
        '20 10',
        '1028',
        '2116',
        '10209',
        '13',
    ]
    for year_case in year_cases:
        print(year_case + '    ->    ' + str(clean_year(year_case)))

    month_cases = [
        '01',
        '09',
        '11',
        '5',
        '7',
        'Agsts',
        'January',
        'Okt',
        'No',
        '17'
    ]
    for month_case in month_cases:
        print(month_case + '    ->    ' + str(clean_month(month_case)))
    day_cases = [
        '01',
        '09',
        '11',
        '5',
        '7',
        '!',
        'I',
        '31',
        '35',
        '13',
        '898']
    for day_case in day_cases:
        print(day_case + '    ->    ' + str(clean_day(day_case)))
    print('----------------------------------------------------------------|')
    for test_case in test_cases:
        print(test_case + '    ->    ', clean_date_indonesia(test_case))