Refresh - Teradata Warehouse Miner

Teradata Warehouse Miner User Guide - Volume 2ADS Generation

Product
Teradata Warehouse Miner
Release Number
5.4.4
Published
July 2017
Language
English (United States)
Last Update
2018-05-03
dita:mapPath
fcf1492702067123.ditamap
dita:ditavalPath
ft:empty
dita:id
B035-2301
Product Category
Software

The Refresh analysis is provided as a means to re-execute a chain of referenced analyses with a different set of user specified parameters without modifying the original analyses. It falls under the ADS umbrella because it is designed to allow the user to refresh an analytic data set. However, in addition to ADS analyses, it may also be used to refresh Score analyses.

Creating an analytic data set can require a lot of thought and result in many steps of creating variables and reorganizing data. There can be multiple tables joined by complicated join paths and containing sophisticated arithmetic formulas, as well as the dimensioning of variables. With the use of Analysis References, that provide a means to feed the output of a previous analysis into a subsequent analysis, the result can be a complex string of analyses that make up the creation of a final analytic data set. As the source data changes over time, it might be necessary to modify the parameters used in generating the analytic data set. Apart from Refresh, there are two ways to do this.
  • The first way is to reproduce the entire set of analyses used to generate the analytic data set with the new modified parameters. This is not ideal; if it is a complicated set of analyses, it could take a significant amount of time to reproduce it when you only wanted to change a few things.
  • The second way is to actually change the original analyses with new parameters. The problem with this is that the original ADS template is now permanently changed.

With the Refresh Analysis, the original analyses can be re-executed with the modified parameters without affecting the original parameters used. If any of the parameters are not selected to be changed, then the original values are used. When Refresh is run, the analysis to be refreshed is executed (along with any analyses that it references) using the new parameters specified within Refresh. In addition, the referenced analyses will only generate the columns needed for the analysis that is being refreshed, based on column names (subject to some exceptions described below).

When Refresh limits the columns produced by a referenced analysis that has already been executed in a project (other than by a Refresh or Publish analysis), an error can occur if the analysis is referenced again later in the project by an analysis other than a Refresh or Publish analysis. This is because project execution is designed to execute each analysis only once (apart from execution by a Refresh or Publish analysis). This type of error can usually be avoided either by reordering the analyses in a project or by using the right-click option to Skip during Project Execution.

Column limiting and certain metadata gathering features are bypassed for certain special-purpose analyses such as those that represent subqueries or seed or recursive components of With (Recursive) queries. These features are also bypassed for any analyses that follow them in a chain of refreshed analyses. If, however, a bypassed analysis is also referenced by a non-bypassed analysis, a warning message is given to warn of possibly unexpected results:

An analysis being refreshed, '...', is referenced both by a bypassed analysis (such as a Subquery or With Recursive Query) and a non-bypassed analysis, possibly leading to unexpected results.

It is worth noting that the anchor table replacement feature is not supported for the same special-purpose analyses mentioned above, for the base query of a With (Recursive) analysis, or for analyses that produce a score table or do not have anchor columns selected. However, other bypassed analyses (i.e., those referenced by subquery or With (Recursive) analyses) do support anchor table replacement if they have anchor columns selected.

This is a change from releases prior to TWM 5.3.2.