Find out if column contains two distinct values for each group - sql

Question

I want to select the rows which satisfy both conditions of the same column. Below is the table schema.
Security Table
Id RoleId CompId SecurityToken Accesstype
1 1 10 abc 2
2 1 10 xyz 2
3 12 10 abc 2
4 16 12 abc 2
5 16 12 xyz 2
6 30 13 abc 2
7 1 10 efg 2
8 1 10 lmn 0
I want "All rows for each RoleID/CompID combination where accesstype = 2 and there is both a row containing securitytoken "abc" and a row containing SecurityToken "xyz" for that role/compID combination"
Output should be
Id RoleId CompId SecurityToken Accesstype
1 1 10 abc 2
2 1 10 xyz 2
4 16 12 abc 2
5 16 12 xyz 2

Answer 1

I believe the following query will produce the desired output:
SELECT *
FROM testdata
WHERE Accesstype = 2
AND SecurityToken IN ('abc', 'xyz')
AND EXISTS (
SELECT 1
FROM testdata AS tmp
WHERE RoleId = testdata.RoleId
AND CompId = testdata.CompId
AND Accesstype = testdata.AccessType
AND SecurityToken IN ('abc', 'xyz')
HAVING COUNT(DISTINCT SecurityToken) = 2
)
SQL Fiddle
In order to eliminate sets that contain extra security tokens (such as efg and lmn) change the WHERE and HAVING clause to:
WHERE RoleId = testdata.RoleId
AND CompId = testdata.CompId
AND Accesstype = testdata.AccessType
HAVING COUNT(DISTINCT SecurityToken) = 2
AND COUNT(DISTINCT SecurityToken) = COUNT(CASE WHEN SecurityToken IN ('abc', 'xyz') THEN 1 END)

Answer 2

CREATE TABLE #Table1
([Id] int, [RoleId] int, [CompId] int, [SecurityToken] varchar(3), [Accesstype] int)
;
INSERT INTO #Table1
([Id], [RoleId], [CompId], [SecurityToken], [Accesstype])
VALUES
(1, 1, 10, 'abc', 2),
(2, 1, 10, 'xyz', 2),
(3, 12, 10, 'abc', 2),
(4, 16, 12, 'abc', 2),
(5, 16, 12, 'xyz', 2),
(6, 30, 13, 'abc', 2)
;
WITH cte AS (
SELECT *,ROW_NUMBER() OVER (PARTITION BY [ROLEID],[COMPID] ORDER BY ID) AS RN FROM #TABLE1
),
COUNTED AS (
SELECT
*,
COUNT(*) OVER (PARTITION BY [ROLEID],[COMPID]) AS CNT
FROM cte
)
SELECT
[ID], [ROLEID], [COMPID], [SECURITYTOKEN], [ACCESSTYPE]
FROM COUNTED
WHERE CNT >= 2
output
ID ROLEID COMPID SECURITYTOKEN ACCESSTYPE
1 1 10 abc 2
2 1 10 xyz 2
4 16 12 abc 2
5 16 12 xyz 2
or
WITH CTE AS
(
SELECT
*,
COUNT(*) OVER (PARTITION BY [ROLEID],[COMPID]) AS CNT
FROM #TABLE1)
SELECT [ID], [ROLEID], [COMPID], [SECURITYTOKEN], [ACCESSTYPE] FROM CTE WHERE CNT>=2

Answer 3

One method uses exists:
select t.*
from t
where t.Accesstype = 2 and
t.securityToken in ('abc', 'xyz') and
exists (select 1
from t t2
where t2.RoleId = t.RoleId and
t2.CompId = t.CompId and
t2.Accesstype = t.AccessType and
t2.SecurityToken in ('abc', 'xyz') and
t2.SecrityToken <> t.SecurityToken
);
Perhaps a simpler method uses window functions:
select t.*
from (select t.*,
min(securitytoken) over (partition by roleid, compid) as min_st,
min(securitytoken) over (partition by roleid, compid) as max_st
from t
where t.Accesstype = 2 and
t.SecurityToken in ('abc', 'xyz')
) t
where minsecuritytoken = 'abc' and
maxsecuritytoken = 'xyz;