CTE-common table exp in t-sql - sql

Question

I have a problem with the following query:
WITH CTE_1 (stu_id
,meet_doc_id
,doc_name
,stu_name
,dob
,done
,date_meet_doc)
AS
(SELECT stu_id
,meet_doc_id
,doc_name
,stu_name
,dob
,CASE
WHEN (PATINDEX('%SMOKING%',act.VALUE)) THEN
'LMDO'
WHEN (PATINDEX('%NOT SMOKING%',act.VALUE)) THEN
'LMD1'
WHEN (ISNULL(CAST(act.VALUE as varchar(max)),'')='') THEN
'CLEAR'
ELSE
'CLEAR'
END done
,date_meet_doc
FROM
abc INNER JOIN
INNER JOIN
INNER JOIN
WHERE multiple conditions
)
SELECT * FROM CTE_1 one
WHERE date =(SELECT MAX(DATE) FROM CTE_1 two WHERE two.stu_id=one.stu_id
AND one.doc_name=two.doc_name)
ORDER BY stu_name,dob
;
The result set of the inner query(CTE_1) for three student(for eg) will be something like
stu_id meet_doc_id doc_name stu_name dob value date
101 0104 AD AM 15/06/1950 LMDO 2011-02-15
101 0105 AD AM 15/06/1950 CLEAR 2011-02-18
101 0106 AD AM 15/06/1950 CLEAR 2011-02-25
102 0107 AD AK 12/08/1987 CLEAR 2011-03-28
102 0108 AD AK 12/08/1987 LDMO 2011-04-29
103 0109 PK LMP 13/07/1970 CLEAR 2011-03-28
103 0110 PK LMP 13/07/1970 CLEAR 2011-05-12
AND when i execute the whole query my result set will be
stu_id meet_doc_id doc_name stu_name dob value date
101 0106 AD AM 15/06/1950 CLEAR 2011-02-25
102 0108 AD AK 12/19/1987 LDMO 2011-04-29
103 0110 PK LMP 13/07/1970 CLEAR 2011-05-12
What do i have to do to change the outer query to select only those value ie LDMO or LMD1 for a particular student and whose doctor is same?
Suppose if the student meets the doc multiple times and if in any of the cases if the student gets a LMDO or LMD1 then it should pick only that record irrespective of the date.
I am expecting my resultset to be something like:
stu_id meet_doc_id doc_name stu_name dob value date
101 0104 AD AM 15/06/1950 LMDO 2011-02-15
102 0107 AD AK 12/08/1987 CLEAR 2011-03-28
103 0110 PK LMP 13/07/1970 CLEAR 2011-05-12
The logic behind it is if the stu_id is same and the doc_name is same and if there exists a value either LMDO or LMD1 then show that record if not show the record which has CLEAR.
Simply i want to remove the MAX(date) and place a condition for the whole reporting period of that particular stu_id with the same doc_name.

Answer 1

Edit: To add a high-level description of what I did.
You want to get your original information (CTE_1) filtered down by 2 possible criteria. The easiest way to do this is to first establish those criteria in their own result sets. So, we have a subquery that returns a list of (Student, Doc, Max(Date)) combinations and a similar list that is filtered on the LMDO/LMD1 values.
Now, we need to LEFT JOIN to the filtered data since there may not be results for each student.
So now you have a list of Student/Doc/MaxDate and also a possible FilteredDate.
The final step is to JOIN that result set to the original data (CTE_1). Since the FilteredDate takes precedence, we check for that first through the ISNULL function and, if it isn't present, we use the MaxDate instead.
First, I'd change the original query to the following as I think you'll see some performance gains on large data-sets by eliminating the correlated subquery:
SELECT *
FROM CTE_1 one
INNER JOIN
(SELECT stu_id, doc_name, max(date) maxdate FROM CTE_1 group by stu_id,doc_name) two
ON one.stu_id = two.stu_id and one.doc_name = two.doc_name
ORDER BY stu_name,dob
Now, we can add an additional, similar, join to get the max(date) where the Value is in your desired list. We'll need to shuffle the joins around a little bit too.
SELECT realdata.*
FROM
((SELECT stu_id, doc_name, max(date) maxdate FROM CTE_1 group by stu_id,doc_name) maxdt
LEFT JOIN
(SELECT stu_id, doc_name, max(date) maxdate FROM CTE_1
WHERE value in ('LMDO', 'LMD1')
group by stu_id,doc_name) filtered
ON maxdt.stu_id = filtered.stu_id and maxdt.doc_name = filtered.doc_name)
INNER JOIN CTE_1 realdata
ON realdata.stu_id = maxdt.stu_id and realdata.doc_name = maxdt.doc_name
and realdata.date = isnull(filtered.maxdate, maxdt.maxdate)
ORDER BY realdata.stu_name,realdata.dob

Answer 2

declare #TestTable
as table
(stu_id int,
meet_doc_id char(4),
doc_name char(2),
stu_name varchar(3),
dob date,
value varchar(5),
date_meet_doc date)
insert into #TestTable
(stu_id,meet_doc_id,doc_name,stu_name,dob,value,date_meet_doc)
values
(101,'0104','AD','AM', '19500615','LDMO' ,'2011-02-15'),
(101,'0105','AD','AM', '19500615','CLEAR','2011-02-18'),
(101,'0106','AD','AM', '19500615','CLEAR','2011-02-25'),
(102,'0107','AD','AK', '19870812','CLEAR','2011-03-28'),
(102,'0108','AD','AK', '19870812','LDMO' ,'2011-04-29'),
(103,'0109','PK','LMP','19700713','CLEAR','2011-03-28'),
(103,'0110','PK','LMP','19700713','CLEAR','2011-05-12');
WITH CTE_1 (stu_id
,meet_doc_id
,doc_name
,stu_name
,dob
,done
,date_meet_doc)
AS
(SELECT stu_id
,meet_doc_id
,doc_name
,stu_name
,dob
,value
,date_meet_doc
FROM #TestTable
),
CTE_2 as(
SELECT *,row_number() over (partition by stu_id order by case when done in ('LDMO','LDM1') then 0 else 1 end, date_meet_doc desc) rn FROM CTE_1)
select stu_id
,meet_doc_id
,doc_name
,stu_name
,dob
,value
,date_meet_doc
from CTE_2 where rn=1
;
Thanks to those who tried to understand but gave up since i couldnt explain it.
Thanks Again guys :)