Trouble with Oracle date format

Question

Query 1: Redundant to_date usage:

SELECT 1
FROM dual
WHERE '22-APR-2018'>add_months(to_date(
  (to_date('28-02-2018' ,'dd-mm-yyyy') ) ,'dd-mm-yyyy'),60);

--

Query 2: Single to_date usage:

SELECT 1
FROM dual where '22-APR-2018'>ADD_MONTHS(TO_DATE('28-02-2018' ,'dd-mm-yyyy'),60);

In any case, these queries must not return data, since 22-APR-2018 is never greater than 28-FEB-2023.

But Query 1 returns data. I do not understand oracle functionality here.

Answer 1

There are a few issues with this code. One of them is:

to_date((to_date('28-02-2018', 'dd-mm-yyyy')), 'dd-mm-yyyy')

or without the double-bracketing:

to_date(to_date('28-02-2018', 'dd-mm-yyyy'), 'dd-mm-yyyy')

to_date() takes a string argument, so Oracle has to convert the date it got from to_date('28-02-2018', 'dd-mm-yyyy') into a string. So implicitly it's

to_date(to_char(to_date('28-02-2018', 'dd-mm-yyyy')), 'dd-mm-yyyy')

What does to_char(to_date('28-02-2018', 'dd-mm-yyyy')) give you? It will depend on your nls_date_format , which often defaults to DD-MON-RR , in which case you'll get 28-FEB-18 , and converting that back to a date with a YYYY year format gives 28th Feb 0018 (Oracle implicitly interprets FEB to match MM ), which is before 22nd April 2018.

Oracle also implicitly converts the string in the comparison to a date, again using the default date format, so although it's obviously a bad idea to compare a string to a date and hope for the best, in this case you get away with it.

In Oracle we write date literals as

date '2018-04-22'

and not

'22-APR-2018'

So the right way to write this would be

select 1
from   dual
where  date '2018-04-22' > add_months(date '2018-02-28', 60);

which gives no rows.

Answer 2

'22-APR-2018' is not a date it is a string. So both sides of the condition get implicitly converted to strings and 'APR' < 'FEB'.

Use to_date on it or (easier) an ANSI literal instead so that you are comparing date values:

SELECT 1
FROM dual where DATE '2018-04-22'>ADD_MONTHS(TO_DATE('28-02-2018' ,'dd-mm-yyyy'),60);