Software Engineering

1
Software Engineering

• Metrics - I
• Lecture 15

2
Metrics -- Overview

• As part of knowing about the quality of the
  project (and its status), we would like
  quantitative measurements
• We typically measure
• Evaluation of analysis and design
• Code complexity
• Testability

3
Metrics Overview 2

• Important to set the objectives of measurement
  before collecting the data
• Important to clearly define metrics
• Often need to tailor metrics to specific products
• Hard to find data to try-out various metrics

4
Why we dont measure

• Unclear as to the benefits
• Software processes are often poorly organized/not
  sufficiently mature to take advantage of
  measurement
• No standards for measurement limited support
  for data collection/analysis
• Often hard/impossible to measure maintainability,
  complexity, and understandability
• Hard to analyze data

5
Metric classification

• Objective vs. subjective
• Absolute vs. relative
• Explicit vs. derived
• Dynamic vs. static
• Predictive vs. explanatory

6
Problems of measurement

• Effect on people
• Peoples effect on measurements
• Difficulty of data comparisons

7
Analysis metrics

• Predicting size
• Function point metrics
• Bang metrics

8
Design metrics

• Design is often not measured!
• Architecture estimating coding/testing effort
• Card and Glass
• System complexity Structural complexity data
  complexity
• s(i) fout2(i) v(i)/(fout(i) 1)
• Henry and Kafura
• Complexity length(i) (fin(i) fout(i))2

9
Design metrics

• Architecture (continued) estimating
  coding/testing efforts
• Fenton (USAFC) Structure quality index
• Factors include modules in architecture, other
  control modules, db items, modules with single
  entry/exit, program structure, module
  independence, modules not dependent on prior
  processing, db size and compartmentalization,
  module entry/exit characteristics

10
Design metrics

• Component level design
• Cohesion
• Bierman and Ott
• Coupling
• Dhama
• Complexity
• McCabe

11
Design metrics

• Interface design

12
Code metrics

• Code complexity and thus required testing
  effort
• Halstead complexity metrics
• Live variable metrics
• Understandability (affecting the ease of the
  tester to understand the code) style metrics

13
Test metrics

• Generally focus on process of testing rather than
  the technical characteristics of testing itself
• Function/bang metrics
• High level design/Architecture
• Cyclomatic complexity
• Halstead complexity/ KLOC count
• Understandability

14
Quality metrics

• Analysis of data as part of
• Post mortems
• Defect prevention
• Looking for
• Symptoms, severity, where found, when found, how
  found, where caused, when caused, how caused,
  where fixed, when fixed, how fixed

15
Maintenance metrics

• Aim is to determine when the product becomes
  stable
• Stability metrics
• Reliability metrics
• Defects per KLOC
• Jelinski and Moranda
• Okomuto
• Musa
• Littelwood and Verall