Comments on: Time Sliced Measures Part III: Making Measures

By: Dave Rathbun

Dave Rathbun — Thu, 13 Oct 2011 18:54:44 +0000

Yes, queries with only measures will be synchronized because there’s nothing to link with.

By: Sravan

Sravan — Thu, 13 Oct 2011 15:01:26 +0000

Dave, thanks for pointing me in the right direction. I think I understand now what’s causing the synchronization operation in my case. If I select just the timeslice measures (without any dimensions) in my query then the queries are synchronized which I think is expected as there are no shared dimensions across the timeslice contexts. Please correct me if I am wrong.

(also your comments in old posts from BOB forum helped me understand the ‘Sychronization’ and ‘Join’ concepts, http://www.forumtopics.com/busobj/viewtopic.php?p=264898 and http://www.forumtopics.com/busobj/viewtopic.php?p=622469
)

Thanks again.

By: Dave Rathbun

Dave Rathbun — Thu, 13 Oct 2011 02:17:02 +0000

Hi, if you see a synchronization operation it means you have some dimensions that work for some contexts but not others. If you carefully examine the SQL you should see which dimensions those are, which fact table is being used, and which time-slice route. That may help you understand where the issue is coming from.

By: Sravan

Sravan — Wed, 12 Oct 2011 14:54:31 +0000

Hi Dave,

Thanks for the detailed post. We have a similar requirement that our clients want to time slice the measures by CY, PY – MTD, YTD, QTD (based on fiscal and calendar year). With this post’s help, I was able to implement it in development for one fact table so far (within 4 hrs, database and universe) in semi-complex universe (70 tables, 287 joins, 12 contexts). I am testing the data and it looks great so far.

I have a question. When I pull CY MTD, CY YTD, PY MTD and PY YTD measures into my report as one query, the SQL viewer shows 4 sqls as ‘Synchronization’ instead of ‘Join’ like you mentioned above. What’s the different between ‘Synchronization’ and ‘Join’ here?

I have both ‘Multiple sqls for each context’ and ‘Multiple sqls for each measure’ checked in the universe.

Thanks much,
Sravan

By: edpypf

edpypf — Wed, 07 Sep 2011 16:09:18 +0000

SELECT
A_Sales_Org_Curr.Sales_Org_ID,
A_SITE_CURR_New.Site_ID,
sum(Case when Yr_Day_Trend_Range.SORT_NUM=100 then A_Daily_Site_Sales.Daily_Dept_Sales_Amt else 0 end),
sum(Case when Yr_Day_Trend_Range.SORT_NUM=200 then A_Daily_Site_Sales.Daily_Dept_Sales_Amt else 0 end),
sum(Case when Yr_Day_Trend_Range.SORT_NUM=-100 then A_Daily_Site_Sales.Daily_Dept_Sales_Amt else 0 end),
sum(Case when Yr_Day_Trend_Range.SORT_NUM=-200 then A_Daily_Site_Sales.Daily_Dept_Sales_Amt else 0 end)
FROM
A_Sales_Org_Curr INNER JOIN A_Dist_Channel_Curr ON (A_Dist_Channel_Curr.Sales_Org_ID=A_Sales_Org_Curr.Sales_Org_ID)
INNER JOIN A_SITE_CURR_New ON (A_SITE_CURR_New.Sales_Org_ID=A_Dist_Channel_Curr.Sales_Org_ID and A_SITE_CURR_New.Dist_Channel_ID=A_Dist_Channel_Curr.Dist_Channel_ID)
INNER JOIN A_Daily_Site_Sales ON (A_SITE_CURR_New.Site_ID=A_Daily_Site_Sales.Site_ID)
INNER JOIN Yr_Day_Trend_Range ON A_Daily_Site_Sales.trans_date BETWEEN Yr_Day_Trend_Range.Start_Date and Yr_Day_Trend_Range.end_Date
WHERE A_Sales_Org_Curr.Sales_Org_ID IN (’0001′)
AND Yr_Day_Trend_Range.SORT_NUM IN (100,200,-100,-200)
and ( ( Yr_Day_Trend_Range.select_date ) IN ({d ‘2011-12-28′}) )
GROUP BY
1,
2;

By: edpypf

edpypf — Tue, 30 Aug 2011 12:37:37 +0000

Thanks, learned “bit” solution,
I actually designed a date mapping table, which is similar as your customized table, but lists all the mapped date instead of Starting and Ending date column. the columns in my table are “Selected Date”, “Index Number”, “Mapped Date”(which link to fact table-Trans_dt), so with one table, I use Case When Index_Num=999 then 1 else 0 end statement to aggregate the fact in proper time slices. this reduced number of contexts. after index created and collect stats, the performance is close to the one using time partition, the performance is even better if report level filter to limit use of index number is applied. Please let me know what you thought. Thank you

By: Dave Rathbun

Dave Rathbun — Mon, 16 Aug 2010 18:34:52 +0000

Hi, I didn’t post a link to download the database or the universe. The database that I am working with is Oracle, but the same tables could be created in Access like the original Island Resorts universe.

By: Nishant

Nishant — Mon, 16 Aug 2010 15:07:59 +0000

Great post, Dave. Is there a link for the completed universe/database — sorry if I missed it? Thanks!

By: Dave Rathbun

Dave Rathbun — Wed, 04 Aug 2010 00:25:03 +0000

Hi, Avinash, and thanks for your comment. I hope I am understanding your question. You still need a reference date of some kind, as “YTD” stops on a particular date, even if it’s today’s date. Our solution was complicated by the fact that our users wanted YTD to be defined at runtime. Many (most) solutions define the “to date” in YTD as the most recent warehouse load date.

In either case, this sort of solution still works.

I am in the process of implementing this in an older universe, and it will dramatically reduce the amount of time required to do future maintenance. That’s a nice benefit as well.

By: Avinash Wagh

Avinash Wagh — Tue, 03 Aug 2010 14:58:40 +0000

Its realy good solution. But what if I dont have any refrence date. I mean if i want to see YTD for current year on all the dates available in FACT table. In this case I would need RunningSum(Revenue) on date level.
To elaborate more If I want to see YTD values on time dimension(Date, Month..),will this solution work??