问题
I'm trying to Implement Netezza AGE function in Redshift as a UDF. I can able to get the correct answer in Python (Spyder IDE - Py 3.6) but when I execute it in Redshift as UDF, it gives me incorrect output.
I've tried to execute as select AGE_UDF('1994-04-04 20:10:52','2018-09-24 11:31:05'); in Redshift.
Here is the code used in RS UDF.
CREATE OR REPLACE FUNCTION AGE_UDF (START_DATE TIMESTAMP, END_DATE TIMESTAMP)
RETURNS varchar(100)
stable
AS $$
from datetime import datetime
from dateutil import relativedelta
START_DATE = datetime.strptime(START_DATE, '%Y-%m-%d %H:%M:%S')
END_DATE = datetime.strptime(END_DATE, '%Y-%m-%d %H:%M:%S')
difference = relativedelta.relativedelta(END_DATE, START_DATE)
years = difference.years
months = difference.months
days = difference.days
hours = difference.hours
minutes = difference.minutes
seconds = difference.seconds
age=''
if years == 0:
age=''
elif years == 1:
age+=str(years)+' year '
else:
age+=str(years)+' years '
if months == 0:
age+=''
elif months == 1:
age+=str(months)+' mon '
else:
age+=str(months)+' mons '
if days == 0:
age+=''
elif days == 1:
age+=str(days)+' day '
else:
age+=str(days)+' days '
age+=str(hours)+':'+str(minutes)+':'+str(seconds)
return age
$$ language plpythonu;
Output in RS: -8809.15:20:13
Here is the Code used in Python (3.6).
from datetime import datetime
from dateutil import relativedelta
START_DATE = '1994-04-04 20:10:52'
START_DATE = datetime.strptime(START_DATE, '%Y-%m-%d %H:%M:%S')
END_DATE = '2018-09-24 11:31:05'
END_DATE = datetime.strptime(END_DATE, '%Y-%m-%d %H:%M:%S')
difference = relativedelta.relativedelta(END_DATE, START_DATE)
years = difference.years
months = difference.months
days = difference.days
hours = difference.hours
minutes = difference.minutes
seconds = difference.seconds
age=''
if years == 0:
age=''
elif years == 1:
age+=str(years)+' year '
else:
age+=str(years)+' years '
if months == 0:
age+=''
elif months == 1:
age+=str(months)+' mon '
else:
age+=str(months)+' mons '
if days == 0:
age+=''
elif days == 1:
age+=str(days)+' day '
else:
age+=str(days)+' days '
age+=str(hours)+':'+str(minutes)+':'+str(seconds)
print(age)
Output in Python: 24 years 5 mons 19 days 15:20:13
EDIT:
I found the way to achieve the Netezza functionality and I've pasted here. Still I'm Expecting an another efficient way !!! Cheers !!!
Thanks for the Support and Suggestions !!!
回答1:
No Python needed. Here's a SQL UDF that encapsulates the logic. You will need to extend it if the units plural is important to you (mons vs mon).
/*
Postgres AGE() Function
*/
CREATE OR REPLACE FUNCTION f_postgres_age(TIMESTAMP, TIMESTAMP)
RETURNS VARCHAR(64)
STABLE AS $$
-- Input: '1994-04-04 20:10:52', '2018-09-24 11:31:05'
-- Output: 24 years 5 mons 19 days 15:20:13
-- Input: '1994-10-04 20:10:52', '2019-06-12 11:31:05'
-- Output: 24 years 8 mons 7 days 15:20:13
-- Check: SELECT '1994-10-04 20:10:52'::TIMESTAMP
-- + INTERVAL '24 years' + INTERVAL '8 months' + INTERVAL '7 days'
-- + INTERVAL '15 hours' + INTERVAL '20 minutes' + INTERVAL '13 seconds';
-- Result: 2019-06-12 11:31:05
SELECT CASE WHEN DATEDIFF(year, DATE_TRUNC('year', $1)
, DATE_TRUNC('year', CASE WHEN DATEPART(month, $1) > DATEPART(month, $2)
THEN $2 - INTERVAL '1 Year' ELSE $2 END)) > 0
THEN DATEDIFF(year, DATE_TRUNC('year', $1)
, DATE_TRUNC('year', CASE WHEN DATEPART(month, $1) > DATEPART(month, $2)
THEN $2 - INTERVAL '1 Year' ELSE $2 END)) || ' years '
ELSE '' END
|| CASE WHEN ABS( DATEDIFF(month, DATE_TRUNC('month', $1), DATE_TRUNC('month', $2))
- DATEDIFF(month, DATE_TRUNC('year', $1)
, DATE_TRUNC('year', CASE WHEN DATEPART(month, $1) > DATEPART(month, $2)
THEN $2 - INTERVAL '1 Year' ELSE $2 END))) > 0
THEN DATEDIFF(month, DATE_TRUNC('month', $1), DATE_TRUNC('month', $2))
- DATEDIFF(month, DATE_TRUNC('year', $1)
, DATE_TRUNC('year', CASE WHEN DATEPART(month, $1) > DATEPART(month, $2)
THEN $2 - INTERVAL '1 Year' ELSE $2 END)) || ' mons '
ELSE '' END
|| CASE WHEN ABS( DATEDIFF(day, DATE_TRUNC('day', $1)+1, DATE_TRUNC('day', $2))
- DATEDIFF(day, DATE_TRUNC('month', $1), DATE_TRUNC('month', $2))) > 0
THEN DATEDIFF(day, DATE_TRUNC('day', $1)+1, DATE_TRUNC('day', $2))
- DATEDIFF(day, DATE_TRUNC('month', $1), DATE_TRUNC('month', $2)) || ' days '
ELSE '' END
|| TO_CHAR((TIMESTAMP 'epoch'
+ ( DATEDIFF(second, $1, DATE_TRUNC('day', $1)+1 )
+ DATEDIFF(second, DATE_TRUNC('day', $2), $2) )
* INTERVAL '1 Second '),'HH24:MI:SS') age
$$ LANGUAGE SQL
;
回答2:
I Found the way to get the output as same as Netezza ! And we need to create 4 Different UDF with different Inputs ! Here I've added the UDF for (TIMESTAMP, TIMESTAMP)
create or replace function AGE_UDF_V2 (START_DATE TIMESTAMP, END_DATE TIMESTAMP)
returns VARCHAR
stable
as $$
# -*- coding: utf-8 -*-
"""
Created on Wed Sep 26 12:59:24 2018
@author: pnataraj
"""
from dateutil import relativedelta
from dateutil.parser import parse
if (START_DATE is None or END_DATE is None):
return None
else:
START_DATE = str(START_DATE).strip()
END_DATE = str(END_DATE).strip()
START_DATE = parse(START_DATE)
END_DATE = parse(END_DATE)
difference = relativedelta.relativedelta(START_DATE, END_DATE)
years = difference.years
months = difference.months
days = difference.days
hours = difference.hours
minutes = difference.minutes
seconds = difference.seconds
age=''
if years != 0:
if years == 1 or years == -1:
age+=str(years)+' year '
else:
age+=str(years)+' years '
if months != 0:
if months == 1 or months == -1:
age+=str(months)+' mon '
else:
age+=str(months)+' mons '
if days != 0:
if days == 1 or days == -1:
age+=str(days)+' day '
else:
age+=str(days)+' days '
if (hours !=0 or minutes !=0 or seconds != 0):
if (hours < 0 or minutes < 0 or seconds < 0):
age+=str("-"+format(abs(hours),"02")+":"+format(abs(minutes),"02")+":"+format(abs(seconds),"02"))
else:
age+=str(format(hours,"02")+":"+format(minutes,"02")+":"+format(seconds,"02"))
elif(hours == 0 and minutes ==0 and seconds == 0):
if len(age)>0:
age = age
else:
age = "00:00:00"
return age.strip()
$$ language plpythonu;
Thanks for all the suggestions and Helps ! Hope It'll be helpful for those who're doing Nz to AWS RS Migration !
来源：https://stackoverflow.com/questions/52491128/netezza-in-built-age-function-as-udf-in-redshift