Crystal Report performance - subreports or sql command?

Question

Hello,
bit of quandary, I've an existing report that contains six subreports all sharing the same parameter to the main report, a 'projectID' field.
The main report approximately returns 300 records/projects and each of the 6 subreports are passed the corresponding 'projectID' field 300 times.
Now, I've succesfully incorporated the main report query and corresponding subreports into a TSQL command containing a number of subqueries in the Select clause:
SELECT   PROJECT.PROJECTID, PROJECT.TITLE, PROJECT.REFERENCE, PROJECTSTATUS.PROJECTSTATUS, PROJECT.INPUT_DATE
,(SELECT max(INPUT_DATE) FROM V_PROJECT_NOTE AS VPN
WHERE PROJECT.PROJECTID = VPN.PROJECTID) AS LastHeadline
,(SELECT
MAX(CASE
     WHEN MODIFIED_DATE IS NULL THEN INPUT_DATE
     WHEN INPUT_DATE > MODIFIED_DATE THEN INPUT_DATE
          ELSE MODIFIED_DATE
END) FROM ISSUE WHERE PROJECT.PROJECTID = ISSUE.PROJECTID) AS LastNewIssue
,(SELECT max(ISSUENOTE.INPUT_DATE) FROM ISSUE
INNER JOIN ISSUENOTE ON ISSUE.ISSUEID = ISSUENOTE.ISSUEID
WHERE PROJECT.PROJECTID = ISSUE.PROJECTID ) AS LastIssueNote 
,(SELECT max(modified_date) FROM V_PROJECT_RISK_LAST_AMMENDED AS VPR
WHERE PROJECT.PROJECTID = VPR.PROJECTID) AS LastRiskLogUpdate
,(SELECT max(INPUT_DATE) FROM PROJECT_CHECKLIST AS PC
WHERE PROJECT.PROJECTID = PC.PROJECTID) AS LastChecklistUpdate
,(SELECT max(input_date) FROM V_PROJECT_ACTIVITY_LAST_AMMENDED AS VPA
WHERE PROJECT.PROJECTID = VPA.PROJECTID) AS LastGANTTUpdate 
,V_PROJECT_KEYCONTACT.USERNAME AS KeyContact
,(SELECT     USERPROFILE.USERNAME AS Supervisor
FROM         USERPROFILE INNER JOIN
                      PROJECT_MEMBER AS SUPERVISOR ON USERPROFILE.USERID = SUPERVISOR.USERID RIGHT OUTER JOIN
                      PROJECT_MEMBER AS KEYCONTACT ON SUPERVISOR.PROJECT_MEMBERID = KEYCONTACT.PARENTID
WHERE PROJECT.PROJECTID = KEYCONTACT.PROJECTID
AND KEYCONTACT.KEY_CONTACT =1) AS Supervisor
FROM         PROJECT INNER JOIN
                      PROJECTSTATUS ON PROJECT.PROJECTSTATUSID = PROJECTSTATUS.PROJECTSTATUSID LEFT OUTER JOIN
                      V_PROJECT_KEYCONTACT ON PROJECT.PROJECTID = V_PROJECT_KEYCONTACT.PROJECTID
WHERE V_PROJECT_KEYCONTACT.USERNAME IN ('aaa','bbb','ccc')
AND (PROJECTSTATUS.PROJECTSTATUS IN ('111', '222', '333'))
AND ((PROJECT.TITLE NOT LIKE 'xxx%') AND (PROJECT.TITLE NOT LIKE 'yyy%') AND (PROJECT.TITLE NOT LIKE 'zzz%'))
ORDER BY V_PROJECT_KEYCONTACT.USERNAME, PROJECT.INPUT_DATE
Now, I've run both SQL Server Profiler and looked at the Performance Information in Crystal and the SQL command method is an order of magnitude less efficient!!
I may have to post my query on a TSQL forum but was wondering if anyone on here can possible shed any light on it?
Thanks in advance,
Dom

Answer 1

Dom,
I assume that the last edit was to remove a few curse words...
The reality is that we can give a good explanation as to why the sub-report version would execute faster that the command version w/o seeing the report (with the sub-reports included).
Looking at the SQL, I can tell that it's a fairly expensive query, firing several sub-queries for every row returned by the outer query. That said, I sounds like the same thing is taking place with the sub-report version too... So I would think that the performance would be similar.
The only thing that I can think of, and this is just a guess... With the sub-report version, each all queries are being run independently and then combined locally by CR, whereas the command version, everything is combined. The more complex the query, the harder the optimizer has to work to come up with the best execution plan... and the greater the likelihood that that the resulting plan isn't as optimized as it could be.
That said, I'd still think the command version would be faster over all. Are you seeing real differences between the two reports, in the amount of time that it takes from refresh to final render?
Either way, I think moving all of those sub-queries from the select list (where they are executing once for every row) down to the FROM area (where they only have to execute once) should increase the speed dramatically and surpass the sub-report version by a decent margin.
Give this version of your SQL a test drive and see if it yields an improvement.
SELECT  
PROJECT.PROJECTID,
PROJECT.TITLE,
PROJECT.REFERENCE,
PROJECTSTATUS.PROJECTSTATUS,
PROJECT.INPUT_DATE,
VPN.LastHeadline,
ISSUE.LastNewIssue,
ISSUE_NOTE.LastIssueNote,
VPR.LastRiskLogUpdate,
PC.LastChecklistUpdate,
VPA.LastGANTTUpdate,
V_PROJECT_KEYCONTACT.USERNAME AS KeyContact,
Supervisor.Supervisor
FROM PROJECT
INNER JOIN PROJECTSTATUS ON PROJECT.PROJECTSTATUSID = PROJECTSTATUS.PROJECTSTATUSID
LEFT OUTER JOIN V_PROJECT_KEYCONTACT ON PROJECT.PROJECTID = V_PROJECT_KEYCONTACT.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     max(INPUT_DATE) AS LastHeadline
     FROM V_PROJECT_NOTE
     GROUP BYPROJECTID) AS VPN ON PROJECT.PROJECTID = VPN.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     MAX(CASE
          WHEN MODIFIED_DATE IS NULL THEN INPUT_DATE
          WHEN INPUT_DATE > MODIFIED_DATE THEN INPUT_DATE
          ELSE MODIFIED_DATE END) AS LastNewIssue
     FROM ISSUE
     GROUP BY PROJECTID) AS ISSUE ON PROJECT.PROJECTID = ISSUE.PROJECTID
LEFT OUTER JOIN (
     SELECT ISSUE.PROJECTID,
     max(ISSUENOTE.INPUT_DATE) AS LastIssueNote
     FROM ISSUE
     INNER JOIN ISSUENOTE ON ISSUE.ISSUEID = ISSUENOTE.ISSUEID
     GROUP BY ISSUE.PROJECTID) AS ISSUE_NOTE ON PROJECT.PROJECTID = ISSUE_NOTE.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     max(modified_date) AS LastRiskLogUpdate
     FROM V_PROJECT_RISK_LAST_AMMENDED
     GROUP BY PROJECTID) AS VPR ON PROJECT.PROJECTID = VPR.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     max(INPUT_DATE) AS LastChecklistUpdate
     FROM PROJECT_CHECKLIST
     GROUP BY PROJECTID) AS PC ON PROJECT.PROJECTID = PC.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     max(input_date) AS LastGANTTUpdate
     FROM V_PROJECT_ACTIVITY_LAST_AMMENDED
     GROUP BY PROJECTID) AS VPA ON PROJECT.PROJECTID = VPA.PROJECTID
LEFT OUTER JOIN (
     SELECT PROJECTID,
     USERPROFILE.USERNAME AS Supervisor
     FROM USERPROFILE
     INNER JOIN PROJECT_MEMBER AS SUPERVISOR ON USERPROFILE.USERID = SUPERVISOR.USERID RIGHT
     OUTER JOIN PROJECT_MEMBER AS KEYCONTACT ON SUPERVISOR.PROJECT_MEMBERID = KEYCONTACT.PARENTID
     WHERE KEYCONTACT.KEY_CONTACT =1) AS Supervisor ON PROJECT.PROJECTID = Supervisor.PROJECTID
WHERE V_PROJECT_KEYCONTACT.USERNAME IN ('aaa','bbb','ccc')
AND (PROJECTSTATUS.PROJECTSTATUS IN ('111', '222', '333'))
AND ((PROJECT.TITLE NOT LIKE 'xxx%')
AND (PROJECT.TITLE NOT LIKE 'yyy%')
AND (PROJECT.TITLE NOT LIKE 'zzz%'))
ORDER BY V_PROJECT_KEYCONTACT.USERNAME, PROJECT.INPUT_DATE
HTH,
Jason