IBM Cognos User Manual page 199

Results
Having reached the extent of your own expertise, you gather the following data
and prepare to contact Cognos Software Services, to have them investigate your
problem further:
v the Cognos Transformer log file from the PowerCube build, with information
about which dimensions were involved in the partitioning, and partition
consolidation details
v the populated Cognos Transformer model (.mdl or .py? file) and PowerCube
(.mdc or .mdl file), so that IBM Cognos support staff can confirm the status and
record distribution of all partitioned categories
v information about which dimensions are most queried at run-time, and how
deep those queries tend to be; that is, whether your users query mostly the
parent or child levels
v detailed information about cases when query performance was below
expectations, such as typical reports and OLAP activity, especially regarding
nesting or alternate drill-through exploration
Keeping Upgraded Models Populated if Category Codes are Not
Unique
When the category codes, or members, in Cognos Transformer models are not
unique, Cognos Transformer makes them unique by adding a numeric sequence to
each code, preceded by a tilde character (~). This blended expression cannot be
interpreted by other IBM Cognos components.
For example, suppose your cube contains more than one instance of the category
code Item, and so renames the second instance Item~1, the third instance Item~2,
and so on. These category codes are numbered as they are encountered in the
model. This means that the codes do not stay the same, in a predictable sequence,
as your organization (and model) evolves. The tilde characters are interpreted as
non-numeric and, because these codes do not appear in the data source, they are
ignored during the consolidation pass. As a result, duplicate records are
overridden and not summed, despite what may be reported in the log file.
If the category codes are not unique, and change after each cube build, reports and
drill-through parameters that contain explicit member references may no longer
run successfully. To avoid this problem, your reports should be based on level
(children of) references, which are not affected by category code changes.
Save your models with categories of this type in their fully populated form, rather
than cleaning out the categories between each cube build. Save the models as both
.mdl and .py? files, so that you have a backup version if the .py? file becomes
corrupted. This precaution is particularly important if the same category repeats
itself, or if your model contains blank categories.
To avoid such problems, use one of the following strategies to make these codes
unique:
v Ensure that all source values are unique in a dimension.
v Create a calculated column in Cognos Transformer, to make your categories
unique.
v Edit the Model Definition Language (MDL) model file to make the category
codes unique.
For more information about using MDL, see the Cognos Transformer Developer
Guide.
Chapter 8. PowerCubes in Production Environments
185