U

APPLICATION DEVELOPMENT

i PG WEEKN

APRIL 23, 1990

š z
Making the C
To help main:

— tain and enhance
the billions of
lines of existing
application code,
programmers
need powerful

|. reverse-engineer-
ing tools that can
both derive pro-
cess models and
data models: -'
from existing
code, and help
rebuild these ap-

je plications within

distributed-processing :

z

a more modern
environment.
Reverse-engineering is a software tech-
nology that has as its goal the migration
of old systems into a cleanly engineered
form that can be enhanced easily. The re-

ality is that true reverse-engineering tools
that convert existing data definitions and
procedural code into high-level specifica-
tions are not yet available. '

For instance, computer-aided software
engineering (CASE) tools are oriented
toward the specification and implemen-
tation of new applications, and thus fall
short of providing much assistance in''
the support and improvement of exist-
ing application code. zen kn '

There's a lot of confusion in the field
about what is meant by reverse-engi-
neering. To clear up the confusion, con-.
sider the three categories of tools that
support various components of reverse-
engineering:
€ Restructuring Engines. These tools
automatically convert unstructured
source code (such as COBOL) into struc-
tured code. The goal is to reduce mainte-
nanče costs, improve guality, extend the
life of old systems and enable program :/'
enhancements to be made in much less ''
time. The process automatically creates
a structured version of the original
source code without changing its func-
tionality. i ; beli
' Re-engineering Tools. These tools an:
alyze the source code to identify its
structure and detect data redundancies, |
nonstandard names and unused code, /' |
Documentation is produced that lists'
cross-reference tables, structure charts
and relationships between files, records
and fjelds. An analyst can interact with
the tool to resolve inconsistencies and
restructure process or data definitions.
The standardized definitions can then be
stored in the repository of a CASE tool
for further analysis. | pi prvi
% Reverse-Engineering. The term 're-
verse-engineering" should be reserved .:
for tools that are capable of synthesiz-
ing unstructured data definitions and
processing definitions up to the level of
pure specifications, independent of envi-
rese specifications can then

a forward-engineering process
to reimplement the application in the
same or a different environment, using .
standard CASE technigues. Extraction
of specifications from existing data defi-.
nitions and process code is an extraordi..
narily difficult task, which has been
solved, to date, only on the data side. '

Most CASE tools that claim reverse"

|

j
l

ASE

m ise |

rav
PRA

engineering capabilities actually support:
functions. They don't syn-'"

thesize low-level data and process defini:

data-structure diagram into a high-level
- data model (or conceptual schema).

€. gram are first captured from
| files and analyzed using a form of re:

— with the tool

"The result

tions up to the level of environment: ..
independent; specifications. " Ualalo:
'The figure illustrates the functions
performed by a true reverse-engineering
tool. Reverse-engineering of data func-
tions are shown on the left of the figure;
reverse-engineering of process functions
are shown on the right. ' h
As shown, conversion of data defini-
tions into specifications is done at three

"levels: capture of physical data defini-

tions in the program; development of a
schema description contained in a data-
structure diagram; and conversion of the

for Di | rue Reverse-Engineering Tools

version of data definitions into an inter: :

mediate form called a data-structure di: ;
agram. The data in the old program is
likely to be unnormalized. The data-
structure diagrams need to be normal-
ized and the normalized structure made
to conform to a data model.

'The data model is a pure specifica-
tion, typically expressed in the form of
an entity-relationship diagram.

- At this point, the data definitions
from the program have been converted
into a data model, at the specification
level. From that data model, new data
structures can be created for the same
database-management environment or a
different environment.

The Stages of Reverse and Forward Engineering

Goat Is Tb Convert Data Definitions and Program Code
Into Pure Specifications, Then Forward-Engineer to New Designs

ES

—psE Repositor
NE SEEN (stores data štid:
ote ele)

Data

fr
Models ee

Data structure
diagrams

Captured
| data |
| definitions |

Redesigned
data
definitions

Lowlevel data definitions in the pro-

ua

engineering. A programmer can interact."
to resolve inconsistencies
and produce more uniform data defini-  /.
tions and data-element formats, Howev- /.
er, if data definitions from many appl
cations are loaded into the tool, the
mes revealed may be so great that the /
jed from clean:

«4 programmer is discourag,
— ing it up. Thus, ids generally desirable to

"vrernanaii

tackle a small area at a time. "
of this analysis is the con-16:

.-—. Pa EN ERROR Sar va? 7 ramo ze mana

Ee

Process

Procedure structure
diagrams

; The reverse-engineering of the data
side has been implemented by Bachman

. Information Systems Inc. in its š
7. Bachman/Re-Engineering Product Set, -

This product set is capable of ca; turing

data definitions in IMS, IDMS, DB2 and

VSAM systems and then migrating these

designs from one database-management
ze

' system to another.

The product set can also be i
create data models fi useda

The Bachman tool set is implemented

the programmer or analyst what infor-
mation is missing to create a data-struc-
ture diagram or data model. The pro-
grammer fills in the missing pieces of in-
formation while interacting with the
system. s

Reverse-engineering the process side,
which reguires converting unstructured
program code into high-level design
specifications that can be input into the
repository of a CASE tool, is proving
very difficult. The application may be
written with spaghetti code; it needs to
be converted to structured pro-

gramming. ;

Several tools exist for automatically
restructuring COBOL programs. The
product Recoder from Language Tech-
nology Inc., for example, converts CO-
BOL programs that are messy, unstruc-
tured and badly coded into fully struc-
tured COBOL.

The resulting program is functionally
identical to the original. The restruc-
tured code can be tested, perhaps with
the original test data, to ensure that it's
functionally eguivalent to the original.

The restructuring process produces
structured source code but; not process-

-model specifications. To support, reverse-

engineering on the process side, it's nec-
essary to convert the structured code
into high-level specifications that ex-
tract the business functions performed
by the code. These specifications form
the basis for enhancing and maintaining
the application.

Action Diagrams
After the program has been converted

'to a fully structured form, it should be

synthesized into a set. of action diagrams
for a CASE tool. Action diagrams, which
define the process model for the applica-
tion, can be converted into other dia-
gram types (such as decomposition dia-
grams, structure charts or data-flow dia-
grams) and can be linked to data
models, screen designs and so on.
Although the structuring of source

- code can be done automatically, the ex-

traction of design specifications from
the structured code reguires the assis-
tance of an analyst. The analyst inter-
acts with the system and fills in missing
data, such as logical groupings of busi:
ness processes, refinement, of data flow

< and consistent naming conventions.

No CASE tool currently supports re-
verse-engineering of unstructured source
code into process models at the specifi:

Y. cation level. The introduction of such a

tool would be an invaluable addition to
the CASE market. .
Next week, I will discuss tools that
see re-engineering of existing source
e.

The concepts embodied, tn reverse-engi-
neering are described in the CASE vol-
ume in The James Martin Report Se-
ries. For more information on this vol-
ume, call (800) 242-1240. For

€. information on seminars, contact (in
or new applications, '
ted via direct

the United States and. Canada) Tech-
nology Transfer Imstitute, 741 10th St.
Santa Monica, Galif. 90402 (213) 394-

8305. In Europe, contact Savant, 2
šem se kim expert system << New St, Carnforth, Lancs, LA5 9BX
7 tance by indicating to. United Kingdom (0524) 734 505. '
| č
|
an aa LEVE Vx ; Vre Sd