Information Systems Reengineering, Integration and Normalization - Information Systems Reengineering, Integration and Normalization - page 14

IPSB Compiler

IPSB
Source

IPSB Load Module

ESCAPE DL/1
Run-TimeInterface

DL/1 Program IDMS

IDMS
Database

Fig. 1.4 A practical database conversion approach

the disadvantage is that it may not be able to be executed efficiently (because the
program is generalized), meaning it cannot convert all the data from the source to
the target. Reconstructing data files is time-consuming, and some data files may not
be reconstructed because of drastic changes to the database semantics. Furthermore,
this approach depends on one language to describe the data structure (at both the
source and the target), and another to describe the restructuring specifications; these
languages may be cumbersome to use. With the Bridge Program Technique, some
redundant data may have to be retained in the database so that the files needed by
the existing programs can be created again.
Very often, to maximize the benefits of a database, it is better to redesign the
existing application, and the design of the new database model from scratch. In
this case, bridge programs must be written for unloading the existing database to
sequential files or serial files, and to upload them into the new database structures.
In this process, the redundancy of existing files should be removed and standards
should be strictly adhered to. Errors in current files must be detected and removed.
Also file inconsistencies must be found before the conversion, rather than later
when they may cause system malfunction.
The problem of totally automatic translation from a nonrelational DML to SQL
remains a classical problem in the area of databases. Algorithms have been devel-
oped to translate some primitive nonrelational DML to SQL, but not all DMLs can
be translated. Decompilation of lower level nonrelational DML to the higher level
SQL statements cannot therefore be used in production systems. Furthermore, the
effort of rewriting the undecompiled part of the nonrelational DML to SQL is simi-
lar to a rewrite of the whole nonrelational database program, as the time for program
analysis in both approach is about the same.