I have table like this
DocumentID | MasterStepID | StepNumber | RoleID | UserID | Status
JIEP/TT/07/000174 | Approval1 | 1 | NULL |
Try This
SELECT T1.* FROM WF_Approval T1
JOIN
(SELECT DISTINCT [Status], DocumentID, MAX(StepNumber) as [StepNumber] from WF_Approval
GROUP BY DocumentID, [Status]) T2
ON T1.DocumentID = T2.DocumentID AND T1.[Status] = T2.[Status]
AND T1.StepNumber = T2.StepNumber
ORDER BY StepNumber ASC, Status ASC
OUTPUT:
JIEP/TT/07/000174 Approval1 1 NULL 0006100022 1
JIEP/TT/07/000174 Approval1 5 18 0006100022 3
JIEP/TT/07/000174 Approval1 6 16 0006104115 6
The simple table (Student_Course) currently is like this with duplicate records assume:
Sname Sid
------------
Ravi S001
Ravi S001
Jack S002
Jack S002
Jack S002
Guru S003
Smith S004
Arun S005
Rajeev S006
Query:
SELECT S.Sid, S.Sname, Count(*) As Occurance
FROM Student_Course AS S
GROUP BY S.Sid
HAVING Count(S.Sid) > 1
Order By Count(*) Desc Limit 1
Output:
Sid Sname Occurance
-------------------------
S002 Jack 3
You're basically just missing a status comparison since you want one row per status;
SELECT *
FROM WF_Approval sr1
WHERE NOT EXISTS (
SELECT *
FROM WF_Approval sr2
WHERE sr1.DocumentID = sr2.DocumentID AND
sr1.Status = sr2.Status AND # <-- new line
sr1.StepNumber < sr2.StepNumber
) AND MasterStepID = 'Approval1'
or rewritten as a JOIN
;
SELECT *
FROM WF_Approval sr1
LEFT JOIN WF_Approval sr2
ON sr1.DocumentID = sr2.DocumentID
AND sr1.Status = sr2.Status
AND sr1.StepNumber < sr2.StepNumber
WHERE sr2.DocumentID IS NULL
AND sr1.MasterStepID = 'Approval1';
SQLfiddle with both versions of the query here.
This should be faster than a self-join because (most probably) only a single scan over the table is required.
select DocumentID,
MasterStepID,
StepNumber,
RoleID,
UserID ,
Status
from (
select wf.*
row_number() over (partition by wf.status order by wf.stepnumber desc) as rn
from WF_Approval wf
) t
where rn = 1
order by StepNumber