Oracle SELECT Query With Multiple Date Formats [duplicate] - sql

Question

I want to bring a string which contains a date to a single format date.
EX:
13-06-2012 to 13-JUN-12
13/06/2012 to 13-JUN-12
13-JUN-2012 to
13-JUN-12
13/jun-2012 to 13-JUN-12
...
I tried to delete all special characters and after that use a function to transform that string into a single format of date. My function return more exceptions, I don't know why...
The function:
CREATE OR REPLACE FUNCTION normalize_date (data_in IN VARCHAR2)
RETURN DATE
IS
tmp_month VARCHAR2 (3);
tmp_day VARCHAR2 (2);
tmp_year VARCHAR2 (4);
TMP_YEAR_NUMBER NUMBER;
result DATE;
BEGIN
tmp_day := SUBSTR (data_in, 1, 2);
tmp_year := SUBSTR (data_in, -4);
--if(REGEXP_LIKE(SUBSTR(data_in,3,2), '[:alpha:]')) then
if(SUBSTR(data_in,3,1) in ('a','j','i','f','m','s','o','n','d','A','J','I','F','M','S','O','N','D')) then
tmp_month := UPPER(SUBSTR (data_in, 3, 3));
else
tmp_month := SUBSTR (data_in, 3, 2);
end if;
DBMS_OUTPUT.put_line (tmp_year);
TMP_YEAR_NUMBER := TO_NUMBER (tmp_year);
IF (tmp_month = 'JAN')
THEN
tmp_month := '01';
END IF;
IF (tmp_month = 'FEB')
THEN
tmp_month := '02';
END IF;
IF (tmp_month = 'MAR')
THEN
tmp_month := '03';
END IF;
IF (tmp_month = 'APR')
THEN
tmp_month := '04';
END IF;
IF (tmp_month = 'MAY')
THEN
tmp_month := '05';
END IF;
IF (tmp_month = 'JUN')
THEN
tmp_month := '06';
END IF;
IF (tmp_month = 'JUL')
THEN
tmp_month := '07';
END IF;
IF (tmp_month = 'AUG')
THEN
tmp_month := '08';
END IF;
IF (tmp_month = 'SEP')
THEN
tmp_month := '09';
END IF;
IF (tmp_month = 'OCT')
THEN
tmp_month := '10';
END IF;
IF (tmp_month = 'NOV')
THEN
tmp_month := '11';
END IF;
IF (tmp_month = 'DEC')
THEN
tmp_month := '12';
END IF;
-- dbms_output.put_line(tmp_day || '~'||tmp_year || '~' ||tmp_month);
IF (LENGTH (tmp_day || tmp_year || tmp_month) <> 8)
THEN
result := TO_DATE ('31122999', 'DDMMYYYY');
RETURN result;
END IF;
-- dbms_output.put_line('before end');
result:=TO_DATE (tmp_day || tmp_month ||tmp_year , 'DDMMYYYY');
-- dbms_output.put_line('date result: '|| result);
RETURN result;
EXCEPTION
WHEN NO_DATA_FOUND
THEN
NULL;
WHEN OTHERS
THEN
result := TO_DATE ('3012299', 'DDMMYYYY');
RETURN result;
RAISE;
END normalize_date;
Usage
SELECT customer_no,
str_data_expirare,
normalize_date (str_data_expirare_trim) AS data_expirare_buletin
FROM (SELECT customer_no,
str_data_expirare,
REGEXP_REPLACE (str_data_expirare, '[^a-zA-Z0-9]+', '')
AS str_data_expirare_trim
FROM (SELECT Q1.set_act_id_1,
Q1.customer_no,
NVL (SUBSTR (set_act_id_1,
INSTR (set_act_id_1,
'+',
1,
5)
+ 1,
LENGTH (set_act_id_1)),
'NULL')
AS str_data_expirare
FROM STAGE_CORE.IFLEX_CUSTOMERS Q1
WHERE Q1.set_act_id_1 IS NOT NULL
)
);

Answer 1

If you have a sound idea of all the possible date formats it might be easier to use brute force:
create or replace function clean_date
( p_date_str in varchar2)
return date
is
l_dt_fmt_nt sys.dbms_debug_vc2coll := sys.dbms_debug_vc2coll
('DD-MON-YYYY', 'DD-MON-YY', 'DD-MM-YYYY', 'MM-DD-YYYY', 'YYYY-MM-DD'
, 'DD/MM/YYYY', 'MM/DD/YYYY', 'YYYY/MM/DD', 'DD/MM/YY', 'MM/DD/YY');
return_value date;
begin
for idx in l_dt_fmt_nt.first()..l_dt_fmt_nt.last()
loop
begin
return_value := to_date(p_date_str, l_dt_fmt_nt(idx));
exit;
exception
when others then null;
end;
end loop;
if return_value is null then
raise no_data_found;
end if;
return return_value;
exception
when no_data_found then
raise_application_error(-20000, p_date_str|| ' is unknown date format');
end clean_date;
/
Be aware that modern versions of Oracle are quite forgiving with date conversion. This function handled dates in formats which aren't in the list, with some interesting consequences:
SQL> select clean_date('20160817') from dual;
CLEAN_DAT
---------
17-AUG-16
SQL> select clean_date('160817') from dual;
CLEAN_DAT
---------
16-AUG-17
SQL>
Which demonstrates the limits of automated data cleansing in the face of lax data integrity rules. The wages of sin is corrupted data.
#AlexPoole raises the matter of using the 'RR' format. This element of the date mask was introduced as a Y2K kludge. It's rather depressing that we're still discussing it almost two decades into the new Millennium.
Anyway, the issue is this. If we cast this string '161225' to a date what century does it have? Well, 'yymmdd' will give 2016-12-15. Fair enough, but what about '991225'? How likely is that the date we really want is 2099-12-15? This is where the 'RR' format comes into play. Basically it defaults the century: numbers 00-49 default to 20, 50-99 default to 19. This window was determined by the Y2K issue: in 2000 it was more likely that '98 referred to the recent past than the near future, and similar logic applied to '02. Hence the halfway point of 1950. Note this is a fixed point not a sliding window. As we move further from the year 2000 the less useful that pivot point becomes. Find out more.
Anyway, the key point is that 'RRRR' does not play nicely with other date formats: to_date('501212', 'rrrrmmdd') hurlsora-01843: not a valid month. So, use'RR'and test for it before using'YYYY'`. So my revised function (with some tidying up) looks like this:
create or replace function clean_date
( p_date_str in varchar2)
return date
is
l_dt_fmt_nt sys.dbms_debug_vc2coll := sys.dbms_debug_vc2coll
('DD-MM-RR', 'MM-DD-RR', 'RR-MM-DD', 'RR-DD-MM'
, 'DD-MM-YYYY', 'MM-DD-YYYY', 'YYYY-MM-DD', 'YYYY-DD-MM');
return_value date;
begin
for idx in l_dt_fmt_nt.first()..l_dt_fmt_nt.last()
loop
begin
return_value := to_date(p_date_str, l_dt_fmt_nt(idx));
exit;
exception
when others then null;
end;
end loop;
if return_value is null then
raise no_data_found;
end if;
return return_value;
exception
when no_data_found then
raise_application_error(-20000, p_date_str|| ' is unknown date format');
end clean_date;
/
The key point remains: there's a limit to how smart we can make this function when it comes to interpreting dates, so make sure you lead with the best fit. If you think most of your date strings fit day-month-year put that first; you will still get some wrong casts but less that if you lead with year-month-day.

Answer 2

The String-to-Date Conversion Rules allow additional formatting rules (without any other modifiers being applied). (Also see this question) So:
MM also matches MON and MONTH;
MON also matches MONTH (and vice versa);
YY also matches YYYY;
RR also matches RRRR; and
The punctuation is optional.
Which means you can do:
CREATE OR REPLACE FUNCTION parse_Date_String(
in_string VARCHAR2
) RETURN DATE DETERMINISTIC
IS
BEGIN
BEGIN
RETURN TO_DATE( in_string, 'DD-MM-YY' );
EXCEPTION
WHEN OTHERS THEN
NULL;
END;
BEGIN
RETURN TO_DATE( in_string, 'MM-DD-YY' );
EXCEPTION
WHEN OTHERS THEN
NULL;
END;
BEGIN
RETURN TO_DATE( in_string, 'YY-MM-DD' );
EXCEPTION
WHEN OTHERS THEN
NULL;
END;
RETURN NULL;
END;
/
Query:
WITH dates ( value ) AS (
SELECT '010101' FROM DUAL UNION ALL
SELECT '02JAN01' FROM DUAL UNION ALL
SELECT '03JANUARY01' FROM DUAL UNION ALL
SELECT '04012001' FROM DUAL UNION ALL
SELECT '05JAN2001' FROM DUAL UNION ALL
SELECT '06JANUARY2001' FROM DUAL UNION ALL
SELECT 'JAN0701' FROM DUAL UNION ALL
SELECT 'JANUARY0801' FROM DUAL UNION ALL
SELECT 'JAN0901' FROM DUAL UNION ALL
SELECT 'JANUARY1001' FROM DUAL UNION ALL
SELECT '990111' FROM DUAL UNION ALL
SELECT '99JAN12' FROM DUAL UNION ALL
SELECT '99JANUARY13' FROM DUAL UNION ALL
SELECT '19990114' FROM DUAL UNION ALL
SELECT '2001-01-15' FROM DUAL UNION ALL
SELECT '2001JAN16' FROM DUAL UNION ALL
SELECT '2001JANUARY17' FROM DUAL UNION ALL
SELECT '20010118' FROM DUAL
)
SELECT value, parse_Date_String( value ) AS dt
FROM dates;
Output:
VALUE DT
------------- -------------------
010101 2001-01-01 00:00:00
02JAN01 2001-01-02 00:00:00
03JANUARY01 2001-01-03 00:00:00
04012001 2001-01-04 00:00:00
05JAN2001 2001-01-05 00:00:00
06JANUARY2001 2001-01-06 00:00:00
JAN0701 2001-01-07 00:00:00
JANUARY0801 2001-01-08 00:00:00
JAN092001 2001-01-09 00:00:00
JANUARY102001 2001-01-10 00:00:00
990111 2099-01-11 00:00:00
99JAN12 2099-01-12 00:00:00
99JANUARY13 2099-01-13 00:00:00
19990114 1999-01-14 00:00:00
2001-01-15 2001-01-15 00:00:00
2001JAN16 2001-01-16 00:00:00
2001JANUARY17 2001-01-17 00:00:00
20010118 0118-01-20 00:00:00
(Note: the date formats you are using are ambiguous, as the last example demonstrates. You can swap the order the formats are parsed in the function to get different results but if you have 010203 is it 01-FEB-2003, 02-JAN-2003, 03-FEB-2001 or even 01-FEB-0003?)
If you want it in the format DD-MON-YY (but why YY and not YYYY?) then just use:
TO_CHAR( parse_Date_String( value ), 'DD-MON-YY' )