Create a Hierarchy

Question

I am having a sql which has invoice date and has a index on it. I need to put parameter on this column but the user wants to enter only period name.
So if you create a hierarchy on this date column I know that performance will be really bad.
I read that if we create a dimension table and attach this to that the performance retains. Can any body explain how to create dimension folder and create a hierarchy on that, how to attach that to complex folder and make a parameter on that.
thanks in advance.
thanks
kiran

Answer 1

Hi Kiran
I have been waiting to answer this one so that I have some time to devote to the answer, not because I don't know what to do but because the answer will be time consuming.
Basically, the steps involved are:
1. Determine what your fiscal periods are
2. Determine what time components you want to report on
3. Build a table that would store those components
4. Populate the table using a script to fill in the appropriate fiscal components
5. Build a primary key on the date or some other user-defined unique key
6. Build a series of non-unique keys on the components required within your hierarchies - typically fiscal month, fiscal quarter and fiscal year
7. Import this table into Discoverer as your master time folder
8. Create lists of values on all of the indexed items
9. Copy the folder, using CTRL-C, once per transaction date, and paste it using CTRL-V into the appropriate business area
10. Move the new folder up under its designated transaction folder
11. Rename the folder and the items within it to reflect the date that it is going to manage - E.g. for a ship date the folder could be called SO Ship Date, with the main items being called Ship Date, Fiscal Shipped Month, Fiscal Shipped Quarter and Fiscal Shipped Year
12. Create a Discoverer hierarchy on Ship Date, Fiscal Shipped Month, Fiscal Shipped Quarter and Fiscal Shipped Year and name it Ship Date Hierarchy
13. Build a 1:N join from the calendar date in the copied folder to the transaction date - you probably also want the date in the underlying transaction table to be indexed for maximum performance
14. Hide the date in the transaction folder
Ok, now that the basic workflow has been outlined let's talk about some of the steps. Step 1 asks you to determine your fiscal periods. What this means is that you need to know the start and end date for your organization's fiscal year, and all of the fiscal periods in between.
If you are using Oracle E-Business Suite you can use the GL_PERIODS table to get this information. This table stores information about the accounting periods you define using the Accounting Calendar form in Apps. In other words, GL_PERIODS is the Apps table that contains your organization’s fiscal calendar. However, because you can have recursive joins between the elements of this table, report performance can be seriously impacted. Each row includes the start date and end date of the period, the period type, the fiscal year, the period number, and other information. There is also a one–to–many relationship between a row in the GL_PERIOD_SETS table and rows GL_PERIODS.
You can now write a simple script that will pull out the fiscal period information. Then, after you are certain that your algorithms are working, you should create a materialized view of your GL_PERIODS date hierarchy. After you have created the materialized view, you should place a unique index on the calendar date, with non-unique indexes on the months, quarters, and years. After you do this, inside your view you can now join your transaction date to the date in the materialized view and read off the hierarchy.
Most DBAs are rightly concerned about adding materialized views to E-Business Suite. In this case, because GL_PERIODS is only updated only once a year, you have a perfect candidate for a materialized view. You just need to make sure that you refresh the materialized view after you add the dates for the next year.
Now, GL_PERIODS only has information concerning fiscal periods, or months. You have to get the individual dates in there as well. Once again, if you are using Apps and you are using BOM (Bills of Material) you may well be using a table called BOM_CALENDAR_DATES. This table contains one row per date within your year. In order to join this table to GL_PERIODS you need to be able to determine which fiscal period a certain date falls in to. Having worked that one out you now join to the previous script adding a clause such as DATE >= GL.PERIODS.START_DATE and DATE <= GL_PERIODS.END_DATE where the PERIOD_TYPE is 'Month'.
Here is a script that might give you the basics:
SELECT DISTINCT
gl_mth.period_name month,
gl_mth.period_set_name set_name,
gl_mth.start_date mth_start_date,
gl_mth.end_date mth_end_date,
gl_mth.period_num mth_number,
gl_mth.last_update_date mth_last_update_date,
(gl_mth.end_date -
gl_mth.start_date) +1 mth_timespan,
gl_qtr.period_name quarter,
gl_qtr.start_date qtr_start_date,
gl_qtr.end_date qtr_end_date,
gl_qtr.quarter_num qtr_number,
gl_qtr.last_update_date qtr_last_update_date,
(gl_qtr.end_date -
gl_qtr.start_date) + 1 qtr_timespan,
gl_yr.period_name year,
gl_yr.start_date yr_start_date,
gl_yr.end_date yr_end_date,
(gl_yr.end_date -
gl_yr.start_date) + 1 yr_timespan,
gl_yr.last_update_date yr_last_update_date,
gl_yr.period_year period_year,
gl_qtr.entered_period_name actual_qtr
FROM
apps.gl_periods gl_mth,
apps.gl_periods gl_qtr,
apps.gl_periods gl_yr
WHERE
gl_mth.PERIOD_SET_NAME = 'your data set name'
AND gl_mth.period_type = 'Month'
AND gl_mth.start_date >= '01-OCT-1998'
AND gl_qtr.quarter_num = gl_mth.quarter_num
AND gl_qtr.period_year = gl_mth.period_year
AND gl_qtr.period_type = 'Quarter'
AND gl_yr.period_year = gl_mth.period_year
AND gl_yr.period_type = 'Year';
Does this help understand the logic behind creating fiscal date folder?
Best wishes
Michael Armstrong-Smith
URL: http://learndiscoverer.com
Blog: http://learndiscoverer.blogspot.com