justinhchae

import pandas as pd


col1 = 'event_date'
col_new = str(col1 + '_new')
col2 = 'received_date'
curr_year = 2020
past_year = 2010

# given a dataframe df with col1 and col2 as datetime columns

justinhchae

# iterate through a DataFrame with iterrows()
curr_year = 2020
past_year = 2010
# loop through index and rows of df
for idx, row in df.iterrows():
  # compare years as integers
  if row[col1].year > curr_year:
     
     # return received_date's year as integer
     new_year = row[col2].year

justinhchae

# same iterrows() solution with change log
curr_year = 2020
past_year = 2010
change_log = []
# loop through index and rows of df
for idx, row in df.iterrows():
    if row[col1].year > curr_year:
        new_year = row[col2].year
        # save reference data as tuple
        records = tuple((row['case_id']

justinhchae

# store the lambda function as an object
impute = lambda x: x[col1].replace(year=x[col2].year) if x[col1].year > curr_year \
                   else x[col1].replace(year=x[col2].year) if x[col1].year < past_year \
                   else x[col1]

# simplify the code later by calling impute
df[col_new] = df.apply(impute, axis=1)

# a new dataframe called change_log
change_log = df[(df[col1].dt.year > curr_year)]

justinhchae

# lambda with two conditional statements
df[col_new] = df.apply(lambda x: x[col1].replace(year=x[col2].year) if x[col1].year > curr_year
                        else x[col1].replace(year=x[col2].year) if x[col1].year < past_year 
                        else x[col1]
                        , axis=1)

justinhchae

# replace col1 year with col2 year on a condition
# if does not meet condition, use the original col1 value
df[col_new] = df.apply(lambda x: x[col1].replace(year=x[col2].year) 
                       if x[col1].year > curr_year else x[col1]
                       , axis=1)

# filter df where year is greater than current year
df = df[(df[col1].dt.year > curr_year)]
print(df[[col1, col_new, col2]].head(2))
# output

justinhchae

# copy col2 value into a new column
col_new = str(col1 + '_new')
df[col_new] = df.apply(lambda x: x[col2] , axis=1)
print(df[[col1, col_new, col2]].head(2))

# output

"""
event_date event_date_new received_date
2011-01-05     2011-01-31    2011-01-31

justinhchae

# replace all values with the same thing
df[col1] = df[col1].apply(lambda x: x.replace(year=2109, month=1, day=1))
print(df[col1].head(2))

# output

"""
event_date
2109-01-01
2109-01-01

justinhchae

data = {'dates':
        ['2012-05-04',
         '2012-05-04',
         '2012-06-04',
         '2012-08-08'],
        'types':
        ['a',
         'a',
         'z',
         'z',],

justinhchae

freq='M'
# or 'D' or 'Y'
df = df[['dates', 'types']].groupby([pd.Grouper(key='dates', freq=freq)]).agg('count').reset_index()

"""
       dates  count
2 2012-07-31      0
1 2012-06-30      1
3 2012-08-31      1
0 2012-05-31      2