Data Conversion - Information Systems Reengineering, Integration and Normalization - page 174

too costly because each customized program needs to be written for each file, and
can only be used once. The interpretive transformer approach and the translator
generator approach are language dependent and also very limited in their functions.
They provide a simulator (or compiler) approach to convert from a file format to
another. The users need to learn how to use their simulation language, and even so,
the language cannot be used to serve for general database file conversion purpose.
The logical level translation approach seems to be more general. Actually, many
software utilities in the market apply this approach. However, these software tools
are proprietary.
Algorithms have been developed by converting a hierarchical database to a re-
lational database, a network database to a relational database, and a relational da-
tabase to an object-oriented database and an XML database. They all apply logical
level translation approach by using unload source database to sequential files in the
target database data structure sequence, and then upload them to the target database.
The algorithm of converting network databases to relational databases is to read
through all the network database record types from the bottom up. Each record type
accessed will be concatenated with its owner record keys. The objective is to create
the record identifier in each unload process. The foreign keys can also be unloaded
into the sequential file. They can then be uploaded into relational database.
The algorithm of converting a hierarchical database to a relational database is
similar. The objective is also to create segment identifiers from each database ac-
cess to each segment type.
The algorithm of converting a relational database to an object-oriented database
is to make use of stored OID. In other words, the superclass is stored first. Its OID
can then be used to store its subclass. Similarly, the composite class data is stored
first, and then followed by their component (or associated) class by using the stored
OID.
Data conversion must be done after schema mapping. The data conversion from
relational into an XML document is to automate the data loading according to the
translated XML schema in document type definition. For each rule of schema map-
ping for each data semantic, we can read the tuples from the relational database,
and then load them into the XML elements and their subelements according to their
translated XML schema. We can apply pair of ID and IDREF in DTD or pair of
Key and Keyref in XSD to implement a many-to-many relationship in an XML
document.