Comparing date stored in date and varchar columns - sql

Question

I have a simple table, in which I have to store date information both in DATE and VARCHAR2 columns:
CREATE TABLE date_test(date_in_varchar VARCHAR2(30), date_in_date DATE);
INSERT INTO date_test values ('06/21/2017 01:30:00 AM', to_date('06/21/2017 01:01:03 AM', 'MM/DD/YYYY HH:MI:SS AM'));
INSERT INTO date_test values ('06/22/2017 02:45:00 AM', to_date('06/22/2017 05:06:01 AM', 'MM/DD/YYYY HH:MI:SS AM'));
INSERT INTO date_test values ('06/23/2017 05:51:20 AM', to_date('06/23/2017 12:00:05 AM', 'MM/DD/YYYY HH:MI:SS AM'));
As you can see, format of data is 'MM/DD/YYYY HH:MI:SS AM'. I have to prepare comparisons between date passed in variable l_attr_val format 'MM/DD/YYYY' and ones stored in the table. Now my question is: What is the best and safest way, to write queries for filters: equals, like, greater, lower, to be sure, that I won't face implicit conversion? I assume, that query would be different for values stored as date and for values stored as varchar.
CASE 1: Date is stored in DATE column and it has information about day, month, year, hour, minutes, seconds. I also have a variable (of VARCHAR2 type), that has information of only day, month, year. (l_attr_val = '06/21/2017'). Now I want to make an equality filter between this variable and values stored in table, based only on date, without time (I get 21st of June 2017 in variable and I want to see all records from 21st of June 2017, no matter the time).
SELECT * FROM date_test WHERE date_in_date = TO_DATE('06/22/2017', 'MM/DD/YYYY');
It gives me 0 rows because of the times.
CASE 2, separate from CASE 1:
CASE 1: Date is stored in VARCHAR2 column and it has information about day, month, year, hour, minutes, seconds. I also have a variable (of VARCHAR2 type), that has information of only day, month, year. (l_attr_val = '06/21/2017'). Now I want to make an equality filter between this variable and values stored in table, based only on date, without time (I get 21st of June 2017 in variable and I want to see all records from 21st of June 2017, no matter the time).
SELECT * FROM date_test WHERE TO_DATE(date_in_varchar, 'MM/DD/YYYY HH:MI:SS AM' = TO_DATE('06/22/2017', 'MM/DD/YYYY');
So, what is the safest way to compare dates in above mentioned cases?

Answer 1

Case 1: Properly stored dates (date type)
You could convert your date_in_date column to drop time part but this would not take advantage of non-functional index on this colum so I would advise not to go that way. Instead, if 06/22/2017 is your VARCHAR2 type variable you could also modify it on the fly to have one more value which is incremented by 1 day. This would make use of plain index on date_in_date function and the query would look like:
SELECT * FROM date_test
WHERE date_in_date >= TO_DATE('06/22/2017', 'MM/DD/YYYY')
AND date_in_date < TO_DATE('06/23/2017', 'MM/DD/YYYY')
You could also add +1 after the second part and use the same variable to save you the time for doing math by yourself on a variable.
Case 2: Dates stored as varchars
You need to convert your date_in_varchar to date without time part (or with it, and then partially apply logic proposed for case 1) and also convert your input variable the same way. If you decide to convert it to date without time part then below query would take advantage of functional index (but not a plain one):
SELECT * FROM date_test
WHERE TO_DATE(SUBSTR(date_in_varchar, 0, 12), 'MM/DD/YYYY') = TO_DATE('06/22/2017', 'MM/DD/YYYY')
Should you decide on converting it to a date with time part query would be similar to case 1, but with proper formating. I would advise you to create a functional index on this conversion to date with trimming the varchar to characters containing only date if your queries are going to be accurate up to one day.

Answer 2

If there are indexes on these columns I would use:
Case 1:
select * from date_test
where date_in_date >= to_date('06/22/2017', 'MM/DD/YYYY')
and date_in_date < to_date('06/22/2017', 'MM/DD/YYYY') + 1;
Case 2:
select * from date_test where date_in_varchar like '06/22/2017%'

Answer 3

After some more research I used below approach:
Case 1: Properly stored dates (date type)
SELECT * FROM date_test
WHERE TRUNC(date_in_date)= to_date('06/22/2017', 'MM/DD/YYYY', 'nls_date_language = ENGLISH');
Case 2: Dates stored as varchars
SELECT * FROM date_test
WHERE TRUNC(TO_DATE(date_in_varchar, 'MM/DD/YYYY HH:MI:SS AM')) = to_date('06/22/2017', 'MM/DD/YYYY', 'nls_date_language = ENGLISH');