Oracle OLAP adds power to your SQL applications by providing extensive analytic content and fast query response times. A SQL query interface enables any application to query cubes and dimensions without any knowledge of OLAP.
The OLAP option automatically generates a set of relational views on cubes, dimensions, and hierarchies. SQL applications query these views to display the information-rich contents of these objects to analysts and decision makers. You can also create custom views that follow the structure expected by your applications, using the system-generated views like base tables.
In this chapter, you learn the basic methods for querying dimensional objects in SQL. It contains the following topics:
See Also:
"Developing Reports and Dashboards Using SQL Tools and Application Builders" for a sample dashboard created using Oracle Application Express
"Overview of the Dimensional Data Model" for a discussion of cubes, dimensions, and hierarchies
The system-generated views are created in the same schema as the analytic workspace. Oracle OLAP provides three types of views:
Cube views
Dimension views
Hierarchy views
These views are related in the same way as fact and dimension tables in a star schema. Cube views serve the same function as fact tables, and hierarchy views and dimension views serve the same function as dimension tables. Typical queries join a cube view with either a hierarchy view or a dimension view.
Each cube has a cube view that presents the data for all the measures and calculated measures in the cube. You can use a cube view like a fact table in a star or snowflake schema. However, the cube view contains all the summary data in addition to the detail level data.
The default name for a cube view is cube_VIEW. To find the view for UNITS_CUBE in your schema, you might issue a query like this one:
SELECT view_name FROM user_views WHERE view_name LIKE 'UNITS_CUBE%'; VIEW_NAME ------------------------------ UNITS_CUBE_VIEW
The next query returns the names of all the cube views in your schema from USER_CUBE_VIEWS:
SELECT view_name FROM user_cube_views; VIEW_NAME ------------------------------ UNITS_CUBE_VIEW PRICE_CUBE_VIEW
Like a fact table, a cube view contains a column for each measure, calculated measure, and dimension in the cube. In the following example, UNITS_CUBE_VIEW has columns for the SALES, UNITS, and COST measures, for several calculated measures on SALES, and for the TIME, CUSTOMER, PRODUCT, and CHANNEL dimensions.
DESCRIBE units_cube_view Name Null? Type ----------------------------------------- -------- ---------------------------- SALES NUMBER UNITS NUMBER COST NUMBER SALES_PP NUMBER SALES_CHG_PP NUMBER SALES_PCTCHG_PP NUMBER SALES_PROD_SHARE_PARENT NUMBER SALES_PROD_SHARE_TOTAL NUMBER SALES_PROD_RANK_PARENT_PP NUMBER TIME VARCHAR2(100) CUSTOMER VARCHAR2(100) PRODUCT VARCHAR2(100) CHANNEL VARCHAR2(100)
The USER_CUBE_VIEW_COLUMNS data dictionary view describes the columns of a cube view, as shown by the following query.
SELECT column_name, column_type FROM user_cube_view_columns
WHERE view_name = 'UNITS_CUBE_VIEW';
COLUMN_NAME COLUMN_TYPE
------------------------------ --------------
SALES MEASURE
UNITS MEASURE
COST MEASURE
SALES_PP MEASURE
SALES_CHG_PP MEASURE
SALES_PCTCHG_PP MEASURE
SALES_PROD_SHARE_PARENT MEASURE
SALES_PROD_SHARE_TOTAL MEASURE
SALES_PROD_RANK_PARENT_PP MEASURE
TIME KEY
CUSTOMER KEY
PRODUCT KEY
CHANNEL KEY
13 rows selected.
You can display the contents of a cube view quickly with a query like this one. All levels of the data are contained in the cube, from the detail level to the top.
SELECT sales, units, time, customer, product, channel
FROM units_cube_view WHERE ROWNUM < 15;
SALES UNITS TIME CUSTOMER PRODUCT CHANNEL
---------- ---------- ---------- ---------- ---------- --------
1120292752 4000968 TOTAL TOTAL TOTAL TOTAL
134109248 330425 CY1999 TOTAL TOTAL TOTAL
130276514 534069 CY2003 TOTAL TOTAL TOTAL
100870877 253816 CY1998 TOTAL TOTAL TOTAL
136986572 565718 CY2005 TOTAL TOTAL TOTAL
140138317 584929 CY2006 TOTAL TOTAL TOTAL
144290686 587419 CY2004 TOTAL TOTAL TOTAL
124173522 364233 CY2000 TOTAL TOTAL TOTAL
92515295 364965 CY2002 TOTAL TOTAL TOTAL
116931722 415394 CY2001 TOTAL TOTAL TOTAL
31522409.5 88484 CY2000.Q1 TOTAL TOTAL TOTAL
27798426.6 97346 CY2001.Q2 TOTAL TOTAL TOTAL
29691668.2 105704 CY2001.Q3 TOTAL TOTAL TOTAL
32617248.6 138953 CY2005.Q3 TOTAL TOTAL TOTAL
14 rows selected.
Each dimension has one dimension view plus a hierarchy view for each hierarchy associated with the dimension. For example, a Time dimension might have these three views:
Time dimension view
Calendar hierarchy view
Fiscal hierarchy view
You can use dimension views and hierarchy views like dimension tables in a star schema.
USER_CUBE_DIM_VIEWS identifies the dimension views for all dimensions. The default name for a dimension view is dimension_VIEW.
SELECT * FROM user_cube_dim_views; DIMENSION_NAME VIEW_NAME ------------------------------ ------------------------------ PRODUCT PRODUCT_VIEW CUSTOMER CUSTOMER_VIEW CHANNEL CHANNEL_VIEW TIME TIME_VIEW
USER_CUBE_HIER_VIEWS identifies the hierarchy views for all the dimensions. For a hierarchy view, the default name is dimension_hierarchy_VIEW. The following query returns the dimension, hierarchy, and view names.
SELECT * FROM user_cube_hier_views ORDER BY dimension_name; DIMENSION_NAME HIERARCHY_NAME VIEW_NAME --------------- --------------- ------------------------------ CHANNEL PRIMARY CHANNEL_PRIMARY_VIEW CUSTOMER SEGMENT CUSTOMER_SEGMENT_VIEW CUSTOMER SHIPMENTS CUSTOMER_SHIPMENTS_VIEW PRODUCT PRIMARY PRODUCT_PRIMARY_VIEW TIME FISCAL TIME_FISCAL_VIEW TIME CALENDAR TIME_CALENDAR_VIEW
Like a dimension table, a dimension view contains a key column, level name, level keys for every level of every hierarchy associated with the dimension, and attribute columns. In the following example, TIME_VIEW has a column for the dimension keys, the level name, and the dimension attributes.
DESCRIBE time_view Name Null? Type ----------------------------------------- -------- ---------------------------- DIM_KEY VARCHAR2(100) LEVEL_NAME VARCHAR2(30) DIM_ORDER NUMBER END_DATE DATE LONG_DESCRIPTION VARCHAR2(100) SHORT_DESCRIPTION VARCHAR2(100) TIME_SPAN NUMBER
USER_CUBE_DIM_VIEW_COLUMNS describes the information in each column, as shown in this query.
SELECT column_name, column_type FROM user_cube_dim_view_columns
WHERE view_name ='TIME_VIEW';
COLUMN_NAME COLUMN_TYPE
------------------------------ --------------------
DIM_KEY KEY
LEVEL_NAME LEVEL_NAME
DIM_ORDER DIM_ORDER
END_DATE ATTRIBUTE
TIME_SPAN ATTRIBUTE
LONG_DESCRIPTION ATTRIBUTE
SHORT_DESCRIPTION ATTRIBUTE
The following query displays the level and attributes of each dimension key.
SELECT dim_key, level_name, long_description description, time_span, end_date
FROM time_view WHERE dim_key LIKE '%2005%';
DIM_KEY LEVEL_NAME DESCRIPTION TIME_SPAN END_DATE
------------ -------------------- ------------ ---------- ---------
CY2005 CALENDAR_YEAR 2005 365 31-DEC-05
CY2005.Q2 CALENDAR_QUARTER Q2.05 91 30-JUN-05
CY2005.Q4 CALENDAR_QUARTER Q4.05 92 31-DEC-05
CY2005.Q3 CALENDAR_QUARTER Q3.05 92 30-SEP-05
CY2005.Q1 CALENDAR_QUARTER Q1.05 90 31-MAR-05
2005.01 MONTH JAN-05 31 31-JAN-05
2005.05 MONTH MAY-05 31 31-MAY-05
2005.07 MONTH JUL-05 31 31-JUL-05
2005.03 MONTH MAR-05 31 31-MAR-05
2005.04 MONTH APR-05 30 30-APR-05
2005.08 MONTH AUG-05 31 31-AUG-05
2005.09 MONTH SEP-05 30 30-SEP-05
2005.02 MONTH FEB-05 28 28-FEB-05
2005.11 MONTH NOV-05 30 30-NOV-05
2005.06 MONTH JUN-05 30 30-JUN-05
2005.10 MONTH OCT-05 31 31-OCT-05
2005.12 MONTH DEC-05 31 31-DEC-05
FY2005 FISCAL_YEAR FY2005 365 30-JUN-05
FY2005.Q4 FISCAL_QUARTER Q4 FY-05 91 30-JUN-05
FY2005.Q1 FISCAL_QUARTER Q1 FY-05 92 30-SEP-04
FY2005.Q2 FISCAL_QUARTER Q2 FY-05 92 31-DEC-04
FY2005.Q3 FISCAL_QUARTER Q3 FY-05 90 31-MAR-05
22 rows selected.
Like the dimension views, the hierarchy views also contain columns for the dimension key, level name, and level keys. However, all of the rows and columns are associated with the dimension keys that belong to the hierarchy.
DESCRIBE time_calendar_view Name Null? Type ----------------------------------------- -------- ---------------------------- DIM_KEY VARCHAR2(100) LEVEL_NAME VARCHAR2(30) DIM_ORDER NUMBER HIER_ORDER NUMBER LONG_DESCRIPTION VARCHAR2(100) SHORT_DESCRIPTION VARCHAR2(100) END_DATE DATE TIME_SPAN NUMBER PARENT VARCHAR2(100) TOTAL VARCHAR2(100) CALENDAR_YEAR VARCHAR2(100) CALENDAR_QUARTER VARCHAR2(100) MONTH VARCHAR2(100)
The following query displays the dimension keys, parent key, and the full ancestry for calendar year 2005.
SELECT dim_key, long_description description, parent, calendar_year year,
calendar_quarter quarter, month FROM time_calendar_view
WHERE calendar_year='CY2005'
ORDER BY level_name, end_date;
DIM_KEY DESCRIPTION PARENT YEAR QUARTER MONTH
------------ ------------ ------------ ------------ ------------ ------------
CY2005.Q1 Q1.05 CY2005 CY2005 CY2005.Q1
CY2005.Q2 Q2.05 CY2005 CY2005 CY2005.Q2
CY2005.Q3 Q3.05 CY2005 CY2005 CY2005.Q3
CY2005.Q4 Q4.05 CY2005 CY2005 CY2005.Q4
CY2005 2005 TOTAL CY2005
2005.01 JAN-05 CY2005.Q1 CY2005 CY2005.Q1 2005.01
2005.02 FEB-05 CY2005.Q1 CY2005 CY2005.Q1 2005.02
2005.03 MAR-05 CY2005.Q1 CY2005 CY2005.Q1 2005.03
2005.04 APR-05 CY2005.Q2 CY2005 CY2005.Q2 2005.04
2005.05 MAY-05 CY2005.Q2 CY2005 CY2005.Q2 2005.05
2005.06 JUN-05 CY2005.Q2 CY2005 CY2005.Q2 2005.06
2005.07 JUL-05 CY2005.Q3 CY2005 CY2005.Q3 2005.07
2005.08 AUG-05 CY2005.Q3 CY2005 CY2005.Q3 2005.08
2005.09 SEP-05 CY2005.Q3 CY2005 CY2005.Q3 2005.09
2005.10 OCT-05 CY2005.Q4 CY2005 CY2005.Q4 2005.10
2005.11 NOV-05 CY2005.Q4 CY2005 CY2005.Q4 2005.11
2005.12 DEC-05 CY2005.Q4 CY2005 CY2005.Q4 2005.12
17 rows selected.
Querying a cube is similar to querying a star schema. In a star schema, you join a fact table to a dimension table. The fact table provides the numeric business measures, and the dimension table provides descriptive attributes that give meaning to the data. Similarly, you join a cube view with either a dimension view or a hierarchy view to provide fully identified and meaningful data to your users.
For dimensions with no hierarchies, use the dimension views in your queries. For dimensions with hierarchies, use the hierarchy views, because they contain more information than the dimension views.
When querying a cube, remember these guidelines:
Apply a filter to every dimension.
The cube contains both detail level and aggregated data. A query with an unfiltered dimension typically returns more data than users need, which negatively impacts performance.
Let the cube aggregate the data.
Because the aggregations are calculated in the cube, a typical query does not need a GROUP BY clause. Simply select the aggregations you want by using the appropriate filters on the dimension keys or attributes.
To create a level filter, you must know the names of the dimension levels. You can easily acquire them by querying the dimension or hierarchy views:
SELECT DISTINCT level_name FROM time_calendar_view; LEVEL_NAME ------------------------------ CALENDAR_YEAR CALENDAR_QUARTER MONTH TOTAL
Several data dictionary views list the names of the levels. The following example queries USER_CUBE_HIER_LEVELS.
SELECT level_name FROM user_cube_hier_levels
WHERE dimension_name = 'TIME' AND hierarchy_name ='CALENDAR';
LEVEL_NAME
--------------------
TOTAL
CALENDAR_YEAR
CALENDAR_QUARTER
MONTH
To see the importance of applying a filter to every dimension, consider the query in Example 4-1, which has no filter on the time dimension.
Example 4-1 Displaying Aggregates at All Levels of Time
/* Select key descriptions and facts */
SELECT t.long_description time,
ROUND(f.sales) sales
/* From dimension views and cube view */
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* No filter on Time */
WHERE p.level_name = 'TOTAL'
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
/* Join dimension views to cube view */
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
ORDER BY t.end_date;
Without a filter on the Time dimension, the query returns values for every level of time. This is more data than users typically want to see, and the volume of data returned can negatively impact performance.
TIME SALES
---------- ----------
JAN-98 8338545
FEB-98 7972132
Q1.98 24538588
MAR-98 8227911
APR-98 8470315
MAY-98 8160573
JUN-98 8362386
Q2.98 24993273
JUL-98 8296226
AUG-98 8377587
SEP-98 8406728
Q3.98 25080541
OCT-98 8316169
NOV-98 8984156
Q4.98 26258474
1998 100870877
.
.
.
Now consider the results when a filter restricts Time to yearly data.
Example 4-2 shows a basic query. It selects the Sales measure from UNITS_CUBE_VIEW, and joins the keys from the cube view to the hierarchy views to get descriptions of the keys.
Example 4-2 Basic Cube View Query
/* Select key descriptions and facts */
SELECT t.long_description time,
ROUND(f.sales) sales
/* From dimension views and cube view */
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* Create level filters */
WHERE t.level_name = 'CALENDAR_YEAR'
AND p.level_name = 'TOTAL'
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
/* Join dimension views to cube view */
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
ORDER BY t.end_date;
Example 4-2 selects the following rows. For CUSTOMER, PRODUCT, and CHANNEL, only one value is at the top level. TIME has a value for each calendar year.
TIME SALES -------- ---------- 1998 100870877 1999 134109248 2000 124173522 2001 116931722 2002 92515295 2003 130276514 2004 144290686 2005 136986572 2006 140138317
Dimension attributes also provide a useful way to select the data for a query. The WHERE clause in Example 4-3 uses attributes values to filter all of the dimensions.
Example 4-3 Selecting Data with Attribute Filters
/* Select key descriptions and facts */
SELECT t.long_description time,
p.long_description product,
cu.long_description customer,
ch.long_description channel,
ROUND(f.sales) sales
/* From dimension views and cube view */
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* Create attribute filters */
WHERE t.long_description in ('2005', '2006')
AND p.package = 'Laptop Value Pack'
AND cu.long_description LIKE '%Boston%'
AND ch.long_description = 'Internet'
/* Join dimension views to cube view */
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
ORDER BY time, customer;
The query selects two calendar years, the products in the Laptop Value Pack, the customers in Boston, and the Internet channel.
TIME PRODUCT CUSTOMER CHANNEL SALES ------ ------------------------------ --------------------- -------- ---------- 2005 Laptop carrying case KOSH Entrpr Boston Internet 5936 2005 56Kbps V.92 Type II Fax/Modem KOSH Entrpr Boston Internet 45285 2005 Internal 48X CD-ROM KOSH Entrpr Boston Internet 2828 2005 Standard Mouse KOSH Entrpr Boston Internet 638 2005 Envoy Standard Warren Systems Boston Internet 19359 2005 Laptop carrying case Warren Systems Boston Internet 13434 2005 Standard Mouse Warren Systems Boston Internet 130 2006 Standard Mouse KOSH Entrpr Boston Internet 555 2006 Laptop carrying case KOSH Entrpr Boston Internet 6357 2006 56Kbps V.92 Type II Fax/Modem KOSH Entrpr Boston Internet 38042 2006 Internal 48X CD-ROM KOSH Entrpr Boston Internet 3343 2006 Envoy Standard Warren Systems Boston Internet 24198 2006 Laptop carrying case Warren Systems Boston Internet 13153 2006 Standard Mouse Warren Systems Boston Internet 83 14 rows selected.
A cube contains all of the aggregate data. As shown in this chapter, a query against a cube just selects the aggregate data. It does not calculate the values.
The following is a basic query against a fact table:
/* Querying a fact table */
SELECT t.calendar_year_dsc time,
SUM(f.sales) sales
FROM time_dim t, units_fact f
WHERE t.calendar_year_dsc IN ('2005', '2006')
AND t.month_id = f.month_id
GROUP BY t.calendar_year_dsc;
The next query fetches the exact same results from a cube using filters:
/* Querying a cube */
SELECT t.long_description time, f.sales sales
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* Apply filters to every dimension */
WHERE t.long_description IN ('2005', '2006')
AND p.level_name = 'TOTAL'
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
/* Join dimension views to cube view */
AND t.dim_key = f.TIME
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
ORDER BY time;
Both queries return these results:
TIME SALES ----- ---------- 2005 136986572 2006 140138317
The query against the cube does not compute the aggregate values with a SUM operator and GROUP BY clause. Because the aggregates exist in the cube, this would re-aggregate previously aggregated data. Instead, the query selects the aggregates directly from the cube and specifies the desired aggregates by applying the appropriate filter to each dimension.
The most efficient queries allow the OLAP engine to filter the data, so that the minimum number of rows required by the query are returned to SQL.
The following are among the WHERE clause operations that are pushed into the OLAP engine for processing:
=
!=
>
!>
<
!<
IN
NOT IN
IS NULL
LIKE
NOT LIKE
The OLAP engine also processes nested character functions, including INSTR, LENGTH, NVL, LOWER, UPPER, LTRIM, RTRIM, TRIM, LPAD, RPAD, and SUBSTR.
SQL processes other operations and functions in the WHERE clause, and all operations in other parts of the SELECT syntax.
Drilling is an important capability in business analysis. In a dashboard or an application, users click a dimension key to change the selection of data. Decision makers frequently want to drill down to see the contributors to a data value, or drill up to see how a particular data value contributes to the whole. For example, the Boston regional sales manager might start at total Boston sales, drill down to see the contributions of each sales representative, then drill up to see how the Boston region contributes to the New England sales total.
The hierarchy views include a PARENT column that identifies the parent of every dimension key. This column encapsulates all of the hierarchical information of the dimension: If you know the parent of every key, then you can derive the ancestors, the children, and the descendants.
For level-based hierarchies, the LEVEL_NAME column supplements this information by providing a convenient way to identify all the keys at the same depth in the hierarchy, from the top to the base. For value-based hierarchies, the PARENT column provides all the information about the hierarchy.
See Also:
Chapter 6, "Developing Reports and Dashboards" about using bind variables to support drillingYou can use the PARENT column of a hierarchy view to select only the children of a particular value. The following WHERE clause selects the children of calendar year 2005.
/* Select children of calendar year 2005 */ WHERE t.parent = 'CY2005' AND p.dim_key = 'TOTAL' AND cu.dim_key = 'TOTAL' AND ch.dim_key = 'TOTAL'
The query drills down from Year to Quarter. The four quarters Q1-05 to Q4-05 are the children of year CY2005 in the Calendar hierarchy.
TIME SALES -------- ---------- Q1.05 31381338 Q2.05 37642741 Q3.05 32617249 Q4.05 35345244
The PARENT column of a hierarchy view identifies the parent of each dimension key. Columns of level keys identify the full heritage. The following WHERE clause selects the parent of a Time key based on its LONG_DESCRIPTION attribute.
/* Select the parent of a Time key*/ WHERE t.dim_key = (SELECT DISTINCT parent FROM time_calendar_view WHERE long_description='JAN-05') AND p.dim_key= 'TOTAL' AND cu.dim_key = 'TOTAL' AND ch.dim_key = 'TOTAL'
The query drills up from Month to Quarter. The parent of month JAN-05 is the quarter Q1-05 in the Calendar hierarchy.
TIME SALES -------- ---------- Q1.05 31381338
The following WHERE clause selects the descendants of calendar year 2005 by selecting the rows with a LEVEL_NAME of MONTH and a CALENDAR_YEAR of CY2005.
/* Select Time level and ancestor */
WHERE t.level_name = 'MONTH'
AND t.calendar_year = 'CY2005'
AND p.dim_key = 'TOTAL'
AND cu.dim_key = 'TOTAL'
AND ch.dim_key = 'TOTAL'
The query drills down two levels, from year to quarter to month. The 12 months Jan-05 to Dec-05 are the descendants of year 2005 in the Calendar hierarchy.
TIME SALES -------- ---------- JAN-05 12093518 FEB-05 10103162 MAR-05 9184658 APR-05 9185964 MAY-05 11640216 JUN-05 16816561 JUL-05 11110903 AUG-05 9475807 SEP-05 12030538 OCT-05 11135032 NOV-05 11067754 DEC-05 13142459
The hierarchy views provide the full ancestry of each dimension key, as shown in "Displaying the Contents of a Hierarchy View". The following WHERE clause uses the CALENDAR_YEAR level key column to identify the ancestor of a MONTH dimension key.
/* Select the ancestor of a Time key based on its Long Description attribute */
WHERE t.dim_key =
(SELECT calendar_year
FROM time_calendar_view
WHERE long_description = 'JAN-05')
AND p.dim_key = 'TOTAL'
AND cu.dim_key = 'TOTAL'
AND ch.dim_key = 'TOTAL'
The query drills up two levels from month to quarter to year. The ancestor of month Jan-05 is the year 2005 in the Calendar hierarchy.
TIME SALES -------- ---------- 2005 136986572
A DBA can create calculated measures in Analytic Workspace Manager, so they are available to all applications. This not only simplifies application development, but ensures that all applications use the same name for the same calculation.
Nonetheless, you may want to develop queries that include your own calculations. In this case, you can use an inner query to select aggregate data from the cube, then perform calculations in an outer query. You can select data from cubes that use any type of aggregation operators, and you can use any functions or operators in the query. You must ensure only that you select the data from the cube at the appropriate levels for the calculation, and that the combination of operators in the cube and in the query create the calculation you want.
Example 4-4 shows a query that answers the question, What was the average sales of Sentinel Standard computers to Government customers for the third quarter of fiscal year 2005. UNITS_CUBE is summed over all dimensions, so that FY2005.Q3 is a total for July, August, and September. The inner query extracts the data for these months, and the outer query uses the MIN, MAX, and AVG operator s and a GROUP BY clause to calculate the averages.
Example 4-4 Calculating Average Sales Across Customers
SELECT customer, ROUND(MIN(sales)) minimum, ROUND(MAX(sales)) maximum,
ROUND(AVG(sales)) average
FROM
(SELECT cu.long_description customer,
f.sales sales
FROM time_fiscal_view t,
product_primary_view p,
customer_segment_view cu,
channel_primary_view ch,
units_cube_view f
WHERE t.parent = 'FY2005.Q3'
AND p.dim_key = 'SENT STD'
AND cu.parent = 'GOV'
AND ch.level_name = 'TOTAL'
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
)
GROUP BY customer
ORDER BY customer;
This is the data extracted from the cube by the inner query:
CUSTOMER TIME SALES ---------------------------------------- -------- ---------- Dept. of Labor JAN-05 1553.26 Dept. of Labor MAR-05 1555.6 Ministry of Intl Trade JAN-05 1553.26 Ministry of Intl Trade FEB-05 1554.56 Ministry of Intl Trade MAR-05 1555.6 Royal Air Force JAN-05 1553.26 Royal Air Force FEB-05 6218.23 UK Environmental Department JAN-05 4659.78 UK Environmental Department FEB-05 3109.12
The outer query calculates the minimum, maximum, and average sales for each customer:
CUSTOMER MINIMUM MAXIMUM AVERAGE ------------------------------ ---------- ---------- ---------- Dept. of Labor 1553 1556 1554 Ministry of Intl Trade 1553 1556 1554 Royal Air Force 1553 6218 3886 UK Environmental Department 3109 4660 3884
An OLAP cube aggregates the data within its hierarchies, using the parent-child relationships revealed in the hierarchy views. The OLAP engine does not calculate aggregates over dimension attribute values.
Nonetheless, you may want to aggregate products over color or size, or customers by age, zip code, or population density. This is the situation when you can use a GROUP BY clause when querying a cube. Your query can extract data from the cube, then use SQL to aggregate by attribute value.
The cube must use the same aggregation operator for all dimensions, and the aggregation operator in the SELECT list of the query must match the aggregation operator of the cube. You can use a GROUP BY clause to query cubes that use these operators:
First Non-NA Value
Last Non-NA Value
Maximum
Minimum
Sum
Example 4-5 shows a query that aggregates over an attribute named Package. It returns these results:
TIME PACKAGE SALES ------ ------------------ ---------- 2005 All 1809157.64 2005 Multimedia 18083256.3 2005 Executive 19836977 2005 Laptop Value Pack 9547494.81
Units Cube uses the SUM operator for all dimensions, and the query uses the SUM operator to aggregate over Sales. The Package attribute applies only to the Item level of the Product dimension, so the query selects the Item level of Product. It also eliminates nulls for Package, so that only products that belong to a package are included in the calculation. The GROUP BY clause breaks out Total Sales by Time and Package.
Example 4-5 Aggregating Over an Attribute
SELECT t.long_description time,
p.package package,
SUM(f.sales) sales
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* Select Product by level and attribute */
WHERE p.level_name = 'ITEM'
AND p.package IS NOT NULL
AND t.long_description = '2005'
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
/* Join dimensions and cube */
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
GROUP BY t.long_description, p.package;
Before using the technique described in "Aggregating Measures Over Attributes", ensure that the calculation is meaningful. For example, the common calculation Percent Change might be defined as a calculated measure in a cube. Summing over Percent Change would produce unexpected results, because the calculation for Percent Change ((a-b)/b,) is not additive.
Consider the following rows of data. The correct Total Percent Change is .33, whereas the sum of the percent change for the first two rows is .75.
| Row | Sales | Sales Prior Period | Percent Change |
|---|---|---|---|
| 1 | 15 |
10 |
.50 |
| 2 | 25 |
20 |
.25 |
| Total | 40 |
30 |
.33 |
Example 4-6 shows a query that aggregates over the Package attribute and calculates Percent Change From Prior Period. The inner query aggregates Sales and Sales Prior Period over the attributes, and the outer query uses the results to compute the percent change. These are the results of the query, which show the expected results for PCT_CHG_PP:
TIME PACKAGE SALES PRIOR_PERIOD PCT_CHG_PP ------ ------------------ ---------- ------------ ---------- 2005 All 1809157.64 1853928.06 -.02414895 2006 All 1720399.03 1809157.64 -.04906074 2005 Executive 19836977 20603879.8 -.03722128 2006 Executive 19580638.4 19836977 -.01292226 2005 Laptop Value Pack 9547494.81 10047298.6 -.04974509 2006 Laptop Value Pack 9091450.58 9547494.81 -.04776585 2005 Multimedia 18083256.3 19607675.5 -.07774604 2006 Multimedia 18328678.7 18083256.3 .013571806 8 rows selected.
Example 4-6 Querying Over Attributes Using Calculated Measures
/* Calculate Percent Change */
SELECT TIME, package, sales, prior_period,
((sales - prior_period) / prior_period) pct_chg_pp
FROM
/* Fetch data from the cube and aggregate over Package */
(SELECT t.long_description time,
p.package package,
SUM(f.sales) sales,
SUM(f.sales_pp) prior_period
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
/* Create filters */
WHERE p.level_name = 'ITEM'
AND p.package IS NOT NULL
AND t.long_description IN ('2005', '2006')
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
/* Join dimension views to cube view */
AND t.dim_key = f.time
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
GROUP BY t.long_description, p.package
ORDER BY p.package);
You can generate and view execution plans for queries against cubes and dimensions the same as for those against relational tables.
The SQL EXPLAIN PLAN command creates a table with the content of the explain plan. The default table name is PLAN_TABLE.
The following command creates an execution plan for a basic query on a cube:
EXPLAIN PLAN FOR
SELECT t.long_description time,
p.long_description product,
cu.long_description customer,
ch.long_description channel,
f.sales sales
FROM time_calendar_view t,
product_primary_view p,
customer_shipments_view cu,
channel_primary_view ch,
units_cube_view f
WHERE t.level_name = 'CALENDAR_YEAR'
AND p.level_name = 'TOTAL'
AND cu.level_name = 'TOTAL'
AND ch.level_name = 'TOTAL'
AND t.dim_key = f.TIME
AND p.dim_key = f.product
AND cu.dim_key = f.customer
AND ch.dim_key = f.channel
ORDER BY t.end_date;
Example 4-7 shows selected columns of the execution plan. A CUBE SCAN operation is performed. The plan option is PARTIAL OUTER, which is described in "Types of Execution Plans".
Example 4-7 Selected Columns From PLAN_TABLE
SQL> SELECT operation, options, object_name FROM plan_table; OPERATION OPTIONS OBJECT_NAME -------------------- -------------------- --------------- SELECT STATEMENT SORT ORDER BY JOINED CUBE SCAN PARTIAL OUTER CUBE ACCESS UNITS_CUBE CUBE ACCESS CHANNEL CUBE ACCESS CUSTOMER CUBE ACCESS PRODUCT CUBE ACCESS TIME 8 rows selected.
The DISPLAY table function of the DBMS_XPLAN PL/SQL package formats and displays information from an execution plan, as shown in Example 4-8.
Example 4-8 Formatted Execution Plan From DBMS_XPLAN
SQL> SELECT plan_table_output FROM TABLE(dbms_xplan.display());
PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------
Plan hash value: 1667678335
----------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 100 | 104 (3)| 00:00:02 |
| 1 | SORT ORDER BY | | 1 | 100 | 104 (3)| 00:00:02 |
| 2 | JOINED CUBE SCAN PARTIAL OUTER| | | | | |
| 3 | CUBE ACCESS | UNITS_CUBE | | | | |
| 4 | CUBE ACCESS | CHANNEL | | | | |
| 5 | CUBE ACCESS | CUSTOMER | | | | |
| 6 | CUBE ACCESS | PRODUCT | | | | |
|* 7 | CUBE ACCESS | TIME | 1 | 100 | 103 (2)| 00:00:02 |
----------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
7 - filter(SYS_OP_ATG(VALUE(KOKBF$),12,13,2)='CALENDAR_YEAR' AND
SYS_OP_ATG(VALUE(KOKBF$),43,44,2)='TOTAL' AND
SYS_OP_ATG(VALUE(KOKBF$),33,34,2)='TOTAL' AND
SYS_OP_ATG(VALUE(KOKBF$),23,24,2)='TOTAL')
22 rows selected.
Table 4-1 describes the types of execution plans for cubes.
If you are developing a generic application -- that is, one where the names of the dimensional objects are not known -- then your application can retrieve this information from the data dictionary.
Among the static views of the database data dictionary are those that provide information about dimensional objects. All OLAP metadata is stored in the data dictionary. A few of the data dictionary views were introduced previously in this chapter.
Table 4-2 provides brief descriptions of the ALL views. There are corresponding DBA and USER views.
Table 4-2 Static Data Dictionary Views for OLAP
| View | Description |
|---|---|
|
Describes the visibility of the attributes for cube dimensions. |
|
|
Describes the attributes for cube dimensions. |
|
|
Describes the cube build processes and maintenance scripts. |
|
|
Describes the calculated members (keys) for cube dimensions. |
|
|
Describes the cube dimension levels. |
|
|
Describes the models for cube dimensions. |
|
|
Describes the columns of the system-generated relational views of cube dimensions. |
|
|
Describes the system-generated relational views of OLAP dimensions. |
|
|
Describes the dimension order of the OLAP cubes. |
|
|
Describes the cube dimensions. |
|
|
Describes the hierarchy levels for cube dimensions. |
|
|
Describes the columns of relational hierarchy views of cube dimensions. |
|
|
Describes the hierarchies for cube dimensions. |
|
|
Describes the OLAP dimension hierarchies. |
|
|
Describes the measures in the OLAP cubes. |
|
|
Describes the columns of the relational views of OLAP cubes. |
|
|
Describes the system-generated relational views of OLAP cubes. |
|
|
Describes the OLAP cubes. |
|
|
Describes the contents of OLAP measure folders. |
|
|
Describes the OLAP measure folders. |
See Also:
Oracle Database Reference for full descriptions of data dictionary views.