4. Key Features of Transformation Manager
4.1. Meta-data Text Repository
TM is able to connect to either a relational repository or a text -based repository. The Meta-data Text Repository, used by the majority of ETL clients, provides a much faster environment, and is able to offer better support for multi-users sharing of all aspects of models and transforms. The underlying text repository can be stored directly in a version control system, for example CVS and SourceSafeŽ. Models, transforms, user-defined functions, etc. can be checked out in read or read/write modes.
4.2. Data Access to Data of any Format
TM maps between virtually all kinds of data. In-built adapters are provided for access to RDBMS, XML, and Java classes. A generic adapter to access other data formats can be easily built by implementing Java classes that conform to TM’s published Open Interface. Multiple sources and targets, in any combination, are permitted, allowing maximum flexibility using TM’s Group Adapter.
4.3. Model Management
Relational, XML, and Java data models may be loaded using TM’s built-in loaders by simply pointing to the data source or target. Alternatively, a non-standard data model may be loaded using a generic loader conforming to TM’s published Loader Open Interface. In the case of XML, models can be loaded from DTDs, XSDs, or an XML instance.
All aspects of the models are captured and represented generically regardless of their source, e.g. entities, attribute, relationships, cardinality, attribute types, constraints, RDBMS triggers and comments, in order to provide the most comprehensive model management environment for the transform designer, as well as ensuring that all model constraints are enforced in the transforms that are generated. Model namespaces are fully supported allowing these to be viewed, added or edited.
Models are viewed as a tree structure consisting of entities, attributes , and relationships, although a graphical model viewer with print facility is available to assist the transform designer. Instance data may be used to subset the models displaying only that part of the model relating to the instance data. Instance data may also assist the designer by displaying instance values, with understanding new models. Models are easily modified, extended and exported in alternative forms (e.g. a model captured as a DTD or from a database can be generated as an XSD). A model comparator is also available to analyse two versions of a model and produces a configurable report summarising the differences.
Data changes can occur often as a result of a number of transforms which operate in sequence. TM provides a traceability feature to discover quickly the contributing source attributes and transforms that affect the results (data lineage). Soon it will also be possible to determine what target attributes and transforms depend on a data item (impact analysis).
Many organisations have used TM to implement a model drive architecture (MDA) where designs from UML tools, in the form of XML model interchange (XMI) are imported and transformed to a range of implementation forms, e.g. RDBMS, XSD, etc. implementing their specific transformation requirements.
