Building Hive UDF

udf.py

#Want to try this as Hive UDF
#1. Compile using Jython (never done before)
#2. Re-write using pure Java (don't know Java)
#3. Re-write using Scala (barely done first week of Coursera FP class so far

#Hoping to return this as a single column first
#Eventually want to build a Hive UDTF, to make one row containing dates into multiple rows, one per date

from datetime import *

def burst_hour_timestamp(start_t, end_t):
    '''Function takes start_t and end_t as strings, returns starting hours between interval'''

    start_time_tuple = datetime.strptime(start_t, "%Y-%m-%d %H:%M:%S.%f")
    end_time_tuple = datetime.strptime(end_t, "%Y-%m-%d %H:%M:%S.%f")
    
    #Round start time down to beginning of hour
    start_time_tuple = datetime.combine(start_time_tuple.date(), time(start_time_tuple.hour, 0))

    hour_list = []
    while start_time_tuple <= end_time_tuple:
        hour_list.append(str(start_time_tuple))
        start_time_tuple += timedelta(hours=1)
  
    return hour_list

#Example

#burst_hour_timestamp("2014-02-14 17:56:40.17", "2014-02-14 20:00:00.1")
#>['2014-02-14 17:00:00',
# '2014-02-14 18:00:00',
# '2014-02-14 19:00:00',
# '2014-02-14 20:00:00']

#In Hive, use explode function on array returned by burst_hour_timestamp function
'''
--If values column is comma-separated string, then split value will make array
select
account_number,
values
from split_column_test
lateral view explode(split(value, ',')) a as values;

--This shows that the array format to not need split function
--["1", "2", "3"]
select
account_number,
split(value, ',') as array_result
from split_column_test;

--Apparently, exploding only works with one other column
--You explode, then match back to table
--This means we need a unique record id

select
account_number,
--start_,
--end,
start_interval_timestamp
from string_result
lateral view explode(split(array_result, ',')) a as start_interval_timestamp;
'''