I have a column that holds LossDate
in the format of (yyyymmdd
). I need to verify that if the month ends in 30 days, then no loss was reported with a loss date of 31. I have millions of records. Help will be highly appreciated.
LossDate
--------
20120128
20150520
20180631
Query should return the last invalid record because 200806 ends in 30 days.
For your specific question, the following will return bad rows:
select *
from table
where substr(lossdate,5,2) in ('04','06','09','11')
and substr(lossdate,7,2) > '30'
Depending on your input interface, you may need to repeat that for months with 31 days that show values greater than 31.
select *
from table
where substr(lossdate,5,2) in ('01','03','07','08','10','12')
and substr(lossdate,7,2) > '31'
Feb is a little trickier. Start incrementally, identifying all February entries with more than 29 days, you can get rid of/fix those immediately.
select *
from table
where substr(lossdate,5,2) = '02'
and substr(lossdate,7,2) > '29'
Then you need to repeat this for the remaining rows with Feb dates greater than 28 and aren't a leap year. In the last 100 years, any year divisible by 4 is a leap year, so you can identify the remainder with this (assuming you corrected/deleted the bad Feb entries you already found):
select *
from table
where substr(lossdate,5,2) = '02'
and mod(substr(lossdate,1,4),4)) <> 0
and substr(lossdate,7,2) > '28'
You could create an ISDATE a function such below and then use it in a query such as SELECT * FROM TABLE WHERE DB_IS_DATE(LOSSDATE) = 0
to find all your invalid dates
CREATE OR REPLACE FUNCTION IS_DATE(YYYYMMDD INTEGER)
RETURNS SMALLINT
LANGUAGE SQL CONTAINS SQL DETERMINISTIC NO EXTERNAL ACTION
RETURN
CASE
WHEN YYYYMMDD/10000 BETWEEN 1 AND 9999
AND( ( MOD(YYYYMMDD/100, 100) IN (1,3,5,7,8,10,12) AND MOD(YYYYMMDD,100) BETWEEN 1 AND 31 ) -- Jan,Mar,May,Jul,Aug,Oct,Dec have 31 days
OR ( MOD(YYYYMMDD/100, 100) IN (4,6,9,11) AND MOD(YYYYMMDD,100) BETWEEN 1 AND 30 ) -- Apr,Jun,Sep,Nov have 30 days
OR ( MOD(YYYYMMDD/100, 100) = 2 AND MOD(YYYYMMDD,100) BETWEEN 1 AND 28 ) -- Feb has 28 days
OR ( MOD(YYYYMMDD/100, 100) = 2 AND MOD(YYYYMMDD,100) BETWEEN 1 AND 29 -- unless is a leap year. i.e.
AND ( ( MOD(YYYYMMDD/10000,4) = 0 AND MOD(YYYYMMDD/10000,100) <> 0) -- year is divisable by 4 but not 100
OR MOD(YYYYMMDD/10000,400) = 0 ) -- or year is divisable by 400
)
)
THEN 1
ELSE 0
END
If you are using Db2 Warehouse you could return a BOOLEAN
from the function. E.g.
CREATE OR REPLACE FUNCTION IS_DATE(YYYYMMDD INTEGER)
RETURNS BOOLEAN
LANGUAGE SQL CONTAINS SQL DETERMINISTIC NO EXTERNAL ACTION
RETURN
YYYYMMDD/10000 BETWEEN 1 AND 9999
AND( ( MOD(YYYYMMDD/100, 100) IN (1,3,5,7,8,10,12) AND MOD(YYYYMMDD,100) BETWEEN 1 AND 31 ) -- Jan,Mar,May,Jul,Aug,Oct,Dec have 31 days
OR ( MOD(YYYYMMDD/100, 100) IN (4,6,9,11) AND MOD(YYYYMMDD,100) BETWEEN 1 AND 30 ) -- Apr,Jun,Sep,Nov have 30 days
OR ( MOD(YYYYMMDD/100, 100) = 2 AND MOD(YYYYMMDD,100) BETWEEN 1 AND 28 ) -- Feb has 28 days
OR ( MOD(YYYYMMDD/100, 100) = 2 AND MOD(YYYYMMDD,100) BETWEEN 1 AND 29 -- unless is a leap year. i.e.
AND ( ( MOD(YYYYMMDD/10000,4) = 0 AND MOD(YYYYMMDD/10000,100) <> 0) -- year is divisable by 4 but not 100
OR MOD(YYYYMMDD/10000,400) = 0 ) -- or year is divisable by 400
)
)
The function can be used to convert the input to a DATE if it is a valid date. E.g.
SELECT i, IS_DATE(i) AS IS_DATE
, CASE WHEN IS_DATE(i) = 1 THEN DATE(TO_DATE(DIGITS(DECIMAL(i,8,0)),'YYYYMMDD')) END AS DATE
FROM TABLE(VALUES(-999),(0),(1),(00010101),(99991231),(20180101),(20180228),(20160229),(20180229),(20000229),(19000229)) as D(i)
which returns
I IS_DATE DATE
-------- ------- ----------
-999 0 NULL
0 0 NULL
1 0 NULL
10101 1 0001-01-01
99991231 1 9999-12-31
20180101 1 2018-01-01
20180228 1 2018-02-28
20160229 1 2016-02-29
20180229 0 NULL
20000229 1 2000-02-29
19000229 0 NULL
The function will also accept strings of "YYYYMMDD" format. Db2 will cast the string to an integer for you.
If you have a string of e.g. "YYYY-MM-DD" format, you could create a character version of the function that calls the above function such as
CREATE OR REPLACE FUNCTION IS_DATE(YYYYMMDD VARCHAR(10))
RETURNS INTEGER
LANGUAGE SQL CONTAINS SQL DETERMINISTIC NO EXTERNAL ACTION
RETURN
CASE WHEN REGEXP_LIKE(YYYYMMDD,'[0-9]{4}?-[0-9]{2}?-[0-9]{2}?') = 1
THEN IS_DATE(INTEGER(REPLACE(YYYYMMDD,'-',''))) ELSE 0 END
;