The Small World of Software Reverse Engineering

1
The Small World of Software Reverse Engineering

• Ahmed E. Hassan and Richard C. Holt
• SoftWare Architecture Group (SWAG)
• University Of Waterloo

2
Publications

• We study the evolution of a field through its
  publications
• Publications give a picture of
• Collaboration
• High degree in academia in contrast to industry
• Emergence of topics
• Hot topics and their effects on collaborations

3
DBLP

• DBLP
• DataBase systems and Logic Programming
• Digital Bibliography and Library Project
• Tracks publications in several conferences and
  communities, such as
• WCRE
• Reengineering and maintenance
• Software engineering
• Records for each publication
• Title
• Authors
• Conference name and year
• Abstract
• Data available online as an XML file

4
Studying Collaboration

• Develop a social collaboration network using
  co-authorship data from DBLP
• A node exists for each author
• An edge exists between two nodes (authors), if
  they co-authored a paper together
• Size of node proportional to of pubs
• Edges have a weight proportional to of
  co-authored papers
• Use a force based algorithm to layout the network

5
Co-Authorship Graph
6
Co-Authorship Graph
7
The Largest Component over Time
8
Small World Graphs

• Large graphs with small paths connecting its
  nodes
• Stanley Milgram studied them in the 60s
• Letters were given to people in Nebraska
• Each person hands letter to someone they knew and
  whom they believe can eventually deliver the
  letter to a stockbroker in Pittsburgh
• Average chain of people between both cities is 6
  six degrees of separation
• Collaboration networks which are small world
  graphs
• Good indicator of ease of communication of
  knowledge between members of a community

9
Small World Graph

• Characteristic Path Length (L) measures on
  average how many individuals an author has to go
  through to reach other authors
• The average shortest path from any node in the
  graph to any other node in the largest component
  of the graph
• Clustering Coefficient (C) measures how
  collaborative on average are the co-authors of an
  author
• For a node, C is the ratio of edges to neighbors
  of that node to the maximum number of edges
  between these neighboring nodes
• Watts and Strogatz give a formal definition of
  small world graphs using C, L, and random graphs
• LgtLrandom and C gtgt
  Crandom

10
WCRE is a small world graph!

• Clustering coefficient is 0.76
• Characteristic path length is 4.3

Author Centrality Canfora 2.76 Koschke
2.88 Merlo 2.94 De Lucia 3.1 Holt 3.2
Towards a Standard Schema for C/C By Ferenc,
Sim, Holt, Koschke and Gyimothy 3.94 ? 4.32
11
Paper Titles Analysis for Emerging Terms
12
Bigger Small Worlds in SE

• We compare results against another two research
  communities
• Maintenance and Reengineering (MR) WCRE, IWPC,
  CSMR, ICSM
• Software Engineering MR 17 Conferences
• DBLP data is not as complete for these conferences

13
The Largest Component over Time

• Slow constant growth then rapid growth once
  researchers know each other
• Soft Eng has slower growth
• Less conferences in early days
• Incomplete DBLP data
• Wider scope
• MR and Soft Eng rapid growth since 1996
• Internet and email?
• MR and WCRE growing since late 90s
• Y2K?

14
Most Central Authors over Time
15
WCRE, MR, SE vs. other fields
16
Title Spectrograph
Joint work with J. Wu
Java
Reverse
object
Compon
orient
software
program
design
system
experi
engin
data
abstract
76 77 78 79 80 81 82 83 84 85 86 87 88 89
90 91 92 93 94 95 96 97 98 99 00 01 02 03
17
Conclusion

• A meta paper on publications and collaboration
  networks in WCRE, MR and SE communities
• Small World collaboration networks facilitate the
  exchange of ideas and results in a community
• Many of the techniques presented could be used to
  study the evolution of software systems (files or
  developers as nodes)

18
Generating Small World Graphs Using Random
Re-wiring

• Large L
• Large C

Small L Small C
Small L Large C
19
(No Transcript)
20
Percentage of Papers in a year
21
of new co-authors in a year