Category Archives: MicroStrategy Generated SQL

Tips & Tricks # 14: How Attribute Roles Work in MicroStrategy

Airport_Gates

Attribute Roles

The support of attribute roles is among the most requested features in the schema/modeling area. It is the result of a common modeling practice where two or more attributes are defined using the same lookup table and column. When these attributes are placed on the same report, the result may return no data. To address this issue, a unique table name alias is required.
This Tips & Tricks post demonstrates a business scenario where two possible solutions are available in MicroStrategy:
  1. Explicit logical table aliasing can be used to generate correct report Structured Query Language (SQL).
  2. MicroStrategy SQL Generation Engine automatically detects tables where the same column is used to create two different attributes.

Business Case Scenario

The Airport Example

A user defines two attributes that have the same definition but play different roles in the business model. In this example, attribute Origin Airport and Destination Airport are defined using the same Lookup Table and Column (Airport_ID).

TN5800-072-0108A.gif

Both attributes share the same forms, or information about them (Description, Location, etc.). In the fact table, however, a separate column exists for each of their roles (Origin_Airport_ID and Destination_Airport_ID).

TN5800-072-0108B.gif

If the user places both attributes on a report to obtain the number of flights originated from ‘MIA’ and arrived at ‘LGA’, an empty result set is returned. The SQL statement tries to obtain the description of an airport that is both ‘MIA’ and ‘LGA’ at the same time (Airport_ID = “MIA” AND Airport_ID = “LGA”), thus generating an empty result set as shown below:
Example

 

NOTE: The SQL for this template is similar to the following:

TN5800-072-0108D.gif

Returning no data as explained previously:

 

SOLUTIONS

1. Explicit Table Alias Definition

The recommended way to model attribute roles in MicroStrategy is using explicit table aliases. This approach gives project architects direct control over the specific attributes to be treated as different roles, as well as their mappings to logical tables. This level of control can ensure consistent Engine behavior.

  1. In MicroStrategy Desktop Schema Objects/Tables folder, right-click on the Table and select ‘Create Table Alias’.

     

  2. A logical table alias is created for LU_AIRPORT as LU_AIRPORT (1). The alias name can be modified. In this example, the alias table is named ‘LU_AIRPORT_ALIAS’.
The two attributes, Origin Airport and Destination Airport, can each have its own lookup table.
In this example, the Destination Airport attribute uses LU_AIRPORT_ALIAS as the primary lookup table. Be sure to uncheck LU_AIRPORT as a source table so no join will be made.
Form Definition for the ID

Form Definition for the DESC

TN5800-072-0108J.gif

The Origin Airport attribute’s definition remains the same, using LU_AIRPORT as the source table for both of its forms. LU_AIRPORT_ALIAS is not checked as a source table.
Form Definition for the ID

TN5800-072-0108K.gif

Form Definition for the DESC

TN5800-072-0108L.gif

Report SQL

TN5800-072-0108M.gif

2. Automatic Attribute Role Recognition

A VLDB property is available at database instance level in the Query Optimizations folder, called “Engine Attribute Role Options.” The option is disabled by default. When enabled, the engine detects columns that support more than one attribute in the same table and automatically creates a separate alias in memory only for each attribute. These automatic aliases are not saved into the metadata and cannot be edited in MicroStrategy Desktop. Thus, while the engine attribute role option makes it more convenient to use attribute roles, it also represents a loss of control over the schema representation that is ultimately used for SQL generation. As a result, some logical schema designs may function incorrectly when using automatic role recognition.
Example

 

REPORT SQL

TN5800-072-0108G.gif

Enabling/Disabling ‘Automatic Attribute Role Recognition’

Attributes are considered candidates for the unique table name alias feature if their attribute forms are defined off the same column of a lookup table. Due to the specific nature of the modeling design described above, and that users may want to create two attributes of the same definition, automatic attribute role recognition is an optional setting.

The Very Large Database (VLDB) ‘Engine Attribute Role Options’ property can be found under Project Configuration/VLDB Properties/Query Optimizations folder in the VLDB Settings.

IMPORTANT NOTES:

1. Attribute roles are intended for lookup tables. Since automatic role recognition effectively splits a logical table into several virtual logical tables, the option should not be applied when a fact table supports more than one attribute on the same column. In that case, some levels of analysis will not be available (that is, reports may return the “Fact does not exist at a level that can support the requested analysis” error). In general there should be no need to have attribute roles on fact tables. However, if it is necessary in a specific scenario, explicit table aliases should be used.
2. The automatic detection option will not work if the attributes in the roles situation are in the same hierarchy, meaning that a child attribute is shared. In the example above, the two airport attributes do not have a common child attribute.
———————————————————————————
Source: Jaime Perez, TN6197: How do Attribute Roles Work in MicroStrategy 8.x?, MicroStrategy Knowledgebase, ‎08-30-2001.

Tips & Tricks #12: How to Troubleshoot Cross Joins in SQL Reports for the SQL Generation Engine 9.x

MicroStrategy Community Banner

Readers:

In my last blog post, I blogged about the new MicroStrategy Community. Jaime Perez, VP of Worldwide Customer Services, and his crew have come up with a better way for us to engage with MicroStrategy as well as his team.

Speaking of Jaime, last June, he posted this great tip on the MicroStrategy Knowledgebase site as a TechNote. I am reblogging it since it is one of the most frequent questions I get asked and I find it an extremely useful Tip & Trick. Also, this will give you an idea of the great stuff being posted in the MicroStrategy Community.

Best Regards,

Michael

MicroStrategy and Cross Joins

In some scenarios, one may encounter cross joins in the SQL View of a standard, SQL Report in MicroStrategy.  Cross joins appear when two tables do not have any common key attributes between them in which they can inner join.  As a result, the two tables essentially combine together to create one table that has all the data from both tables, but this results in poorer performance with a common effect of increased execution times.  Sometimes these execution times, and performance hits, can be very severe.  Therefore, it is important to understand some simple steps that can be performed to resolve a cross join, as well as some steps to understand why it may be appearing in the SQL View of the report.

One common occurrence of a cross join is when a report contains at least two unrelated attributes in the grid, and no metrics are present in order to relate the unrelated attributes via a fact table.  Such a occurrence can be resolved in a few ways:

  1. Create a relationship filter, set the output level as the unrelated attributes (or the entire report level), and then relate these by a Logical Table object
  2. Create a relationship filter, set the output level as the unrelated attributes (or the entire report level), and then relate these by a Fact object
  3. Add a metric to the report that uses a fact from a table in which both attributes can inner join to

This provides a pathway from the fact table to the lookup tables in which the unrelated attributes are sourced from.  The result is an inner join between the fact table and the lookup tables, which resolves the cross join between the two unrelated lookup tables.

Options 1 and 2 provide a means in which the report template can remain as only attributes, whereas Option 3 would have a metric on the report.  Option 3 may not be desired if a metric does not want to be placed on the report.  Keep in mind that other techniques can also be employed to have the metric on the report, but formatted to be hidden from display.

More common scenarios include cross joins between a fact table and a lookup table, and are typically surprising to a developer.  These situations can be a bit more tricky to troubleshoot and resolve, but here are a few techniques that can be employed to try to resolve the issue:

  1. In SQL View look at where the cross join appears, and between which tables the cross join appears
  2. Open up those tables in the Table Editor by navigating to the Schema Objects\Tables folder, and double-clicking the tables
  3. Select the Logical View Tab of both tables to see all the logical objects mapped to the table
  4. Take note of which attributes have a key icon beside them
  5. These key attributes denote attributes at the lowest level of their hierarchy presently mapped to the table and/or attributes that are in their own hierarchy (meaning they have no parents or children)
  6. The SQL Engine will join 2 tables on common key attributes only, so if none of the key attributes on either table exist on both tables, then a cross join should appear

This means that just because a Region attribute exists on Table_A and a Region attribute exists on Table_B does not necessarily mean that the SQL Engine will join on Region.  If Region has its child attribute on the table, then that attribute should be the key as it is the lowest level attribute of its particular hierarchy mapped to the table.  If Region exists on both tables, and is also a key attribute on both tables, then an inner join should take place on Region.

This essentially means that one can find a cross join, investigate the tables in which it appears, and verify if at least 1 common key attribute exists between the tables.  If not, then that should be the first path to investigate because a cross join is correct in that scenario.

Video

You can find a detailed video on how this issue is reproduced and resolved here: Tech Note 71019 . Steps to reproduce and resolve

Note

MicroStrategy Technical Support can assist with resolving cross joins in a specific report, however caution should be taken when resolving such issues.  In some scenarios, the cross join is resolved through modifications to the schema objects, which can have a ripple effect to all other reports in an environment.  For example, if a relationship is changed in the Region attribute to resolve a cross join in one report, this change will be reflected in all other reports that use Region, and potentially the hierarchy in which Region belongs.  As a result, the SQL View of one report will have the cross join resolved, but the SQL may have changed in other reports using Region or its related attributes.  This may or may not be desired.  MicroStrategy Technical Support may not be able to fully understand the impact of such a schema change to the data model, so before a change is made to the data model the consequences of such a change should be fully understood by the developer, and any changes made to the schema should be recorded.

References

[1] Jaime Perez, TN47356: How to troubleshoot cross joins in SQL Reports for the SQL Generation Engine 9.x, MicroStrategy Community, 06/24/2014, http://community.microstrategy.com/t5/Architect/TN47356-How-to-troubleshoot-cross-joins-in-SQL-Reports-for-the/ta-p/196989.

[2] MicroStrrategy Knowledgebase, Tech Note 71019 . Steps to reproduce and resolve,

Interview Question #1: MicroStrategy Generated SQL

Readers:

Today, I am going to introduce a new blog entry related to Interview Questions. I have interviewed a lot of people for various MicroStrategy positions over the years and thought I would share with you some of the questions I have asked or have been asked on interviews. I will try to add two interview questions a month.

Best regards,

Michael

Question

Why does the SQL you created using MicroStrategy (assume it generated a single-pass SQL) produce different results than when you run it directly against a database?

Answer

There may have been steps that required the analytical engine which would only occur on the Intelligence Server and not when you directly run SQL against a database.