Get rows in which any column contains sequence using exists select 1 - sql

Question

I have a table with three columns with the following values (dbFiddle)
C1 C2 C3
----------------------------
Red Yellow Blue
null Red Green
Yellow null Violet
I'm trying to create a query that returns all the rows that contain the value "Yellow" without using IN or OR. If I execute the following query:
SELECT 1
FROM test
WHERE CONCAT(C1, C2, C3) LIKE '%Yellow%'
It correctly returns the rows specified. However, if I try to use this query inside an exists:
SELECT *
FROM test
WHERE EXISTS (SELECT 1 FROM test WHERE CONCAT(C1, C2, C3) LIKE '%Yellow%')
it returns all the rows, not just the two with the "Yellow" word. What am I doing wrong here?
Any help would be greatly appreciated.

Answer 1

Re
SELECT 1 FROM test WHERE CONCAT(C1, C2, C3) LIKE '%Yellow%'
"correctly returns the rows specified"
The select returns a single column, 1, although this is because there is a row which has one column containing Yellow somewhere in its text.
This is because EXISTS:
Returns TRUE if a subquery contains any rows.
i.e. All of the following queries also returned all rows in your test table:
SELECT * FROM test WHERE EXISTS (SELECT 1);
SELECT * FROM test WHERE EXISTS (SELECT 0);
SELECT * FROM test WHERE EXISTS (SELECT NULL);
... simply because the SELECT returns at least one row!
The usual usage of EXISTS also includes correlation of the subquery in the EXISTS back to the outer select.
Example of Correlation
In the below contrived example, we've got 4 people living in two houses. Here we're using EXISTS to figure out the names of the persons who are happy, and also have someone else who is also happy living in the same (correlated) House.
CREATE TABLE House
(
HouseId INT PRIMARY KEY,
Name VARCHAR(MAX)
);
CREATE TABLE Person
(
PersonId INT PRIMARY KEY,
HouseId INT FOREIGN KEY REFERENCES HOUSE(HouseId),
Name VARCHAR(MAX),
IsHappy BIT
);
INSERT INTO House(HouseId, Name) VALUES (1, 'House1'), (2, 'House2');
INSERT INTO Person(PersonId, HouseId, Name, IsHappy) VALUES
(1, 1, 'Joe', 0),
(2, 1, 'Jim', 1),
(3, 2, 'Fred', 1),
(4, 2, 'Mary', 1);
SELECT pOuter.Name
FROM Person pOuter
WHERE pOuter.IsHappy = 1
AND EXISTS
(SELECT 1
FROM Person pInner
WHERE pInner.HouseId = pOuter.HouseId
AND pInner.PersonId != pOuter.PersonId
AND pInner.IsHappy = 1);
Returns
Mary
Fred
(There are obviously other ways to find the same result, e.g. finding groupings of House Id where there exists 2 or more happy people, etc)

Answer 2

exists clause check specifies a subquery to test for the existence of rows.
so exists (SELECT 1 FROM test WHERE CONCAT(C1, C2, C3) LIKE '%Yellow%') will always have rows when any column contain yellow data.
if you want to use exists you need to set inner exists query CONCAT(t.C1, t.C2, t.C3) by the outer table.
SELECT *
FROM test t
where exists (SELECT 1 FROM test WHERE CONCAT(t.C1, t.C2, t.C3) LIKE '%Yellow%')
You don't need to use exists only set the condition on where
SELECT *
FROM test
where CONCAT(C1, C2, C3) LIKE '%Yellow%'
sqlfiddle

Answer 3

I would use cross apply:
SELECT 1
FROM test t CROSS APPLY
(SELECT COUNT(*) as cnt
FROM (VALLUES (C1), (C2), (C3)) V(C)
WHERE c = 'Yellow'
) v
WHERE cnt > 0;
You can readily adapt this to a subquery:
SELECT . . .
FROM test t
WHERE EXISTS (SELECT 1
FROM (VALLUES (C1), (C2), (C3)) V(C)
WHERE c = 'Yellow'
) ;
Personally, I much prefer the direct comparison of each value to 'Yellow' rather than using LIKE. For instance, this will not match "yellow-green" or any other value where "yellow" is part of the name.
And, just for the record, you can still use boolean logic, even if you don't use OR and IN:
where not (coalesce(c1, '') <> 'Yellow' and
coalesce(c2, '') <> 'Yellow' and
coalesce(c3, '') <> 'Yellow'
)
Technically, this is probably the "simplest" solution to your problem. However, I still prefer the apply method, because the intent is clearer.