ftp.umcs.maine.edu/2015-02-07.ftp.umcs.maine.edu.tar/ftp.umcs.maine.edu/pub/WISR/wisr6/proceedings/ascii/hislop2.ascii

Text File | 1993-10-19 | 12KB | 303 lines

      Using  Existing  Software  in  a  Software  Reuse  Initiative



                                           Gregory W. Hislop



                                       Working Knowledge, Inc.

                                               738Elgin Rd.

                                     Newtown Square, PA 19073

                                           Tel:  [phone]

                               Email: [e-mail]



                                                  Abstract


    Existing software is often not well-suited for reuse. At the same time, it isa natural starting
point for organizations trying to adopt a reuse strategy. Any efficient techniquesfor extracting
value from existing software would ease the transition to a reuse-based approach.
    My efforts have focused on trying to derive value for reuse from existing software.  I have
been exploring ways to apply to reuse ideas developed to detect plagiarism.  I am also quite
interested in the behavioral issues, both individual and organizational, related to starting and
sustaining a reuse strategy.


Keywords: software reuse, software similarity, plagiarism detection


Workshop Goals: My goals for the workshop are to exchange information on researchefforts
and to discuss the status and results of formal reuse initiatives in commercial and governmental
organizations.


Working Groups: Useful and collectable metrics; Reuse management, organization, andeco-
nomics; Reusable component certification.



                                                  Hislop- 1


1      Background



I approach reuse from the perspective of both research and industry.  I have recentlycompleted a
doctorate with a thesis focused on reuse. In addition, I am involved with fledgling reuse efforts at
several commercial organizations.


Having a foot on both sides of the fence keeps me quite conscious of thegap b etween state of the
art and state of the practice in software engineering.  Ihave a general interest in technology transfer
and ways to close that gap.



2      Position



Existing software contains a large amount of encoded knowledge about the organization and the
organization's  past  software  needs.  This  material  could  be  very  valuable  for  reuse  if  we  could
extract it efficiently.


Existing software is also a natural starting pointfor implementing reuse, especially for an incre-
mental implementation. Existing software provides a familiar point of reference that may reduce
uncertainty, and help create a feeling of ownership.  Finally, it may be difficultto justify a ma jor
initiative that does not start with the existing software.


At the same time, working with existing software may involve significant problems. The quality of
source code is uneven. The documentation is often inadequate. And existing software was probably
not designed for reuse.


In short,  while the idea of using existing software is app ealing,  the reality is difficult.  We need
special tools to support this work. We can extract gold from ore that looks like common rock. If
we are going to use existing software, we need todevelop more economical extraction techniques
to find the flakes of gold in our software.


Intuitively  we  might  be  inclined  to  evaluate  existing  software  by  function,  i.e.,  what  it  does.
Reusable  parts  are  generally  selected  by  function, and  functional  categories  are  the  usual  basis
for organizing parts collections.  It would be logical to analyzethe existing software by function
too.  Unfortunately, analysis  by  function  is  very  difficult  to  automate  and  so  tends  to  be  quite
expensive.


An alternative to evaluating by software function is to measure software form, i.e., what the software
looks like.  Software form includes characteristics such as size and structure.  It is well suited to
automated measurement, and there is a wealth of prior work that has studied various attributes of
form.


One way to look at software form is toconsider the broader question of software similarity. That is,
are there parts in the existing software that arevery similar to each other? A possible application of
this type of analysis would be for locating instances of informal reuse with or without mo dification.


There is an existing pocket of research that measures software form to detect similarity.  These
studies are attempts to detect plagiarized student programming assignments.  Clearly, this is quite
close to our interest. Plagiarism, after all, is simply a socially unacceptable form of reuse.


A basic problem in many plagiarism studies stems from the fact that most software metrics are



                                                         Hislop- 2


single values. For example, Non- Comment Source Statements (NCSS), is a single value. Single-
valued metrics are very convenient for reporting and analysis, but their conciseness may not provide
the discriminant power needed to detect similarity.


For example, similar programs might have similar NCSS  values.  However, it is easy to imagine
programs with similar NCSS values that are not atall similar except in size.


An alternative approach is to representattributes of software form in a less concise manner than a
single-valued metric. Whale has proposed a structure profile which is a variable length descriptor
based on control structure.[1] He reports better success at detecting plagiarism with this approach
than with methods based on single-valued metrics.


I have taken concepts from the work on plagiarism and applied them to reuse. Student assignments
are generally different from commercial software in size, complexity, and style, so it was not clear
whether the concepts would transfer to this new application. I created a prototype tool, SoftKin,
and conducted a series of case studies to test the concepts.



2.1     SoftKin Prototype



SoftKin is a prototype tool for exploring approaches to measuring software similarity.  The name
reflects the idea that SoftKin looks for related software parts.


SoftKin  consists  of  a  data  collector  and  analyzer.  The  collector  processes  existing  software  and
calculates measures of form for each module. The analyzer computes similarity measures for each
module pair. The pairs of modules can then be ranked from most similar to least similar by each
similarity measure.


SoftKin can calculate similarity based on a variety of measures of software form.  These include
single-value metrics (such as McCabe's CyclomaticComplexity), a metric composite, and also a
structure profile that is a slight variant of the profile proposed by Whale.


The case study goal was to see if any of the approaches to measuring similarity show promise for
identifying candidate parts that might be reengineered for areusable parts collection.


To evaluate the various similarity rankings, I focused on the task of locating informal reuse.  In each
of the case studies, we were ableto identify a set of instances of reuse in the existing software. This
allowed me to evaluate each measure by where the known instances of reuse fall in the similarity
ranking. A good ranking is one that places the actual reuse instances near the top of the list.


This evaluation provides a relative measure of various methods of similarity detection. As a pro-
duction tool, SoftKin would guide an analysis of existing software. The ranking by similarity would
suggest where to focus attention in looking for candidate reusable parts. If SoftKin is successful,
the best candidates for a parts collection would be clustered near the top of thelist.



2.2     Case Studies



The case studies analyzed existing software from 3large commercial organizations. The applications
were all commercial data processing systems including areas such as finance, sales, and distribution.
The analyzed software consisted of about 360 modules totaling 156,000 NCSS.



                                                         Hislop- 3


The case study results are quite interesting. The results of the single-value metrics are all fairly
poor.  In fact, a ranking by similarity of size provides as good a result as a ranking by any of the
other single-value metrics. However,the average similarity, which is a composite of thesingle-value
rankings, provides substantially improved performance.


The ranking by structure profile is the best method for locating instances of actual reuse. This is
particularly true for the top part of the ranking, which is the part that has practicalvalue.  For
example, a sizeable percentage of the known cases of reuse fall in the first 100 ranked pairs.  (The
actual results are 37, 50, and 63 percent, respectively, for the three case studies.)


We can test the statistical significance of the result based on the structure profile vs.  the other
metrics. Using Dunnett's T to control for Type I error across the set of comparisons, the structure
profile  shows  a  significant  performance  improvement  versus  the  single-value  metrics  at  the  0.05
level. The improvement over the average similarity, while substantial, is not significant at the 0.05
level.


The pattern of results is very consistent across the three case studies.  This is particularly encourag-
ing since the case study settings vary substantially including differences in programming language,
development methodology, andorganization size.


In general, concepts carried overquite well from plagiarism detection to reuse. Single-value metrics
show poor results for locating the instances of informal reuse.  On the other hand, a composite
based on these metrics performed much better. Finally, the structure profile approach provided the
best performance.



3      Comparison



A number of studies have applied measures of form to software reuse.  Selby conducted a general
analysis to determine the characteristics of software that is known to have been reused.  [2] This
work provides some general guidelines that may indicate attributes that are desirable for reusable
software.


Calderia and Basili make extensive use of form metrics in their strategy for locating reusable parts in
existing software in the Care system.[3] They propose a reusability attributes model for determining
probable reusability for a part. The major attributes are:



    fflQuality of the part.

    fflCosts of using the part.

    fflUsefulness of the part.



It is easiest to associate measures of software form with thequality and cost attributes of the model.
It is more difficult to propose automated measures to predict usefulness of the part.


For the Care system, "reuse frequency" is taken as a predictor of usability.  This value is based
on the number of static calls to a software part.  This approach may be quite effective in some
environments. However, it requires that a given part have only one name. In addition, it does not
consider cases in which there are variants of a part each with a different name.


The  software  similarity  approach  that  SoftKin  explores  could  be  used  to  complement  the  reuse
frequency measure of the Care system.


                                                         Hislop- 4


References



[1]  G. Whale, "Software Metrics and Plagiarism Detection," vol. 13, pp. 131-138, 1990.


[2]  R. W. Selby, "Quantitative Studies of Software Reuse," in Software Reusability, Vol II (T. Big-
     gerstaff and A. Perlis, eds.), pp. 213-233, ACM Press, 1989.


[3]  G.  Calderia and  V.  R.  Basili,  "Identifying  and  Qualifying  Reusable  Software  Components,"
     IEEE Computer, February 1991.



4      Biography



Greg Hislop has held technical and management positions in software organizations for almost
20 years.  He is principal of Working Knowledge, Inc., a company he founded in 1987 to provide
services and education to information systems groups.


His current work involves software reuse, downsizing mainframe applications, I/S tools and tech-
niques for distributed organizations, and practical application of software metrics. He also provides
services related to measuring and analyzing I/S workloads andcharging for I/S services.


Dr.   Hislop  holds  degrees  in  Economics  from  Georgetown University,  Computer  Science  from
Queen's University, and Information Studies from Drexel University.



                                                         Hislop- 5