PL SQL 'IF or CASE' using variable date - oracle

Question

I am stuck with this and could use advice/help:
Basically, trying to set the date as a variable and then run select statements, using that date variable in the 'WHERE' section of the query. Not sure if I should be using IF or CASE, or neither? If its monday, i want to run 1 set of dates (prev thur and fri) any other day (just sysdate-2 and sysdate-1) Any help is much appreciated!
Code is below:
DECLARE
today_date number;
start_date date;
end_date date;
BEGIN
today_date := to_char(sysdate, 'D');
start_date := case when today_date ='2' then 'sysdate-4' else 'sysdate-2'
end;
end_date := case when today_date ='2' then 'sysdate-3' else 'sysdate-1' end;
SELECT COLUMN A, COLUMN B, COLUMN C, COLUMN D
FROM /*csv*/REPORT_NAME
WHERE COLUMN B between trunc(start_date)+21/24 and trunc(end_date)+21/24 and
BOOK_NAME = 'xxxxxx' and SERVER = 'xxxxxx' and EX_ACTION = 'xxxxx';
end;

Answer 1

You're mixing variables/functions and strings. This should work.
DECLARE
today_date number;
start_date date;
end_date date;
BEGIN
today_date := to_char(sysdate, 'D');
start_date := case when today_date ='2' then sysdate-4 else sysdate-2
end;
end_date := case when today_date ='2' then sysdate-3 else sysdate-1 end;
/* this won't work without declaring a cursor, and returning it to the client
SELECT COLUMN A, COLUMN B, COLUMN C, COLUMN D
FROM REPORT_NAME
WHERE COLUMN B between trunc(start_date)+21/24 and trunc(end_date)+21/24 and
BOOK_NAME = 'xxxxxx' and SERVER = 'xxxxxx' and EX_ACTION = 'xxxxx';
*/
end;
Note you also have some implicit type conversion happening. today_date should probably be char(1) instead.

Answer 2

You don't really need variables for this, and you don't need PL/SQL; you can calculate the dates as part of the where clause using case expressions in that instead:
SELECT /*csv*/ COLUMN_A, COLUMN_B, COLUMN_C, COLUMN_D
FROM REPORT_NAME
WHERE COLUMN_B >= trunc(sysdate) - case to_char(sysdate, 'Dy', 'NLS_DATE_LANGUAGE=ENGLISH')
when 'Mon' then 4 else 2 end + 21/24
AND COLUMN_B < trunc(sysdate) - case to_char(sysdate, 'Dy', 'NLS_DATE_LANGUAGE=ENGLISH')
when 'Mon' then 3 else 1 end + 21/24
AND BOOK_NAME = 'xxxxxx'
AND SERVER = 'xxxxxx'
AND EX_ACTION = 'xxxxx';
I've taken the liberty of changing the logic from being based on the day number being 2 to it having a specific day abbreviation, because the value D returns is based on your NLS settings so it could vary (and give unexpected results) for someone else running your code. As day names and abbreviations are also NLS-dependent I've specified the language to use. (You can't specify how D is used in the same way, unfortunately).
I've changed the time window slightly, so it goes from 9pm on one day to up to but not including 9pm the next day. If you use between then it's inclusive at both ends, so runs on consecutive days you both pick up any data at exactly 21:00:00 from the overlapping day. That probably isn't what you want (but if it really is, just change < to <=, or revert to between if you prefer...).
If you want a PL/SQL wrapper you can do the same thing, but you either have to select the results into something like a collection, or use a ref cursor to return the result set to the caller. It isn't clear if you actually need or want to do that though.

Answer 3

have you tried selecting into your variables?
DECLARE
TODAY_DATE NUMBER;
START_DATE DATE;
END_DATE DATE;
BEGIN
TODAY_DATE := TO_CHAR (SYSDATE, 'D');
SELECT CASE WHEN TODAY_DATE = '2' THEN SYSDATE - 4 ELSE SYSDATE - 2 END,
CASE WHEN TODAY_DATE = '2' THEN SYSDATE - 3 ELSE SYSDATE - 1 END
INTO START_DATE, END_DATE
FROM DUAL;
SELECT COLUMN_A, COLUMN_B, COLUMN_C, COLUMN_D
FROM REPORT_NAME /*csv*/
WHERE COLUMN_B BETWEEN TRUNC (START_DATE) + 21 / 24
AND TRUNC (END_DATE) + 21 / 24
AND BOOK_NAME = 'xxxxxx'
AND SERVER = 'xxxxxx'
AND EX_ACTION = 'xxxxx';
END;