Empirical Software Engineering. Empirical Software Engineering with Examples! is not a topic for examination. Classification.

Transcription

1 Empirical Software Engineering Empirical Software Engineering with Examples is not a topic for examination a sub-domain of software engineering focusing on experiments on software systems devise experiments on software, in collecting data from these experiments, and elaborate laws and theories from this data (wikipedia) "2 Classification Observed TRUE FALSE Software Quality? (Bugs) Predicted precision = TP/(TP + FP) recall = TP/(TP + FN) TRUE TP FP FALSE FN TN "3 "4

2 Some tests... arithmetic mean mean(2, 3, 4) = 3 mean(0, 3, 6) = 3 standard deviation R "5 "6 Predicting for Eclipse Release Files Number of Packages T. Zimmermann, R. Premraj, and A. Zeller Predicting defects for Eclipse Third International Workshop on Predictor Models in Software Engineering (PROMISE) IEEE Computer Society, , "7 "8

3 methods classes FOUT MLOC NBD PAR VG NOF NOM NSF NSM Used Metrics... Number of method calls (FAN OUT) Method Lines of Code Nested block depth Number of parameters McCabe Cyclomatic Complexity Number of fields Number of methods Number of static fields Number of static methods ACD Number of anonymous type declarations NOI Number of interfaces files NOT Number of classes TLOC Total lines of code packages NOCU Number of files FOUT MLOC NBD PAR VG NOF NOM NSF NSM Spearman correlations ACD NOI NOT TLOC (-) (-) NOCU - - 0,514 0,461 "9 "10 Classification of files Release Testing Precision Recall 2 2, Foutse Khomh, Massimiliano Di Penta, Yann-Gaël Guéhéneuc, Giuliano Antoniol An exploratory study of the impact of antipatterns on class change- and fault-proneness. Empirical Software Engineering 17(3): (2012) "11 "12

4 90 ArgoUML Eclipse Mylyn Rhino AntiSingleton List of AntiPatterns... ClassDataShouldBePrivate SpeculativeGenerality % of classes participating in antipatterns Releases Fig. 1 Percentages of classes participating in antipatterns in the releases of the four systems "13 MessageChain RefusedParentBequest Blob ComplexClass SwissArmyKnife LargeClass LongMethod LazyClass LongParameterList SpaghettiCode "14 Research Questions... Research Questions... What is the relation between anti-patterns change- & defect- proneness What is the relation between kinds change- & defect- proneness "15 "16

5 Research Questions... Research Questions... Does the presence relate to the size of these classes What kind of changes participating or not in anti-patterns "17 "18 are highlighted in gray "19 "20

6 "21 "22 Are the clients of flawed classes (also) defect prone? Working is hard... Authors: Radu & Cristina Marinescu IEEE International Working Conference on Source Code Analysis and Manipulation, 2011

7 ...especially when resources are flawed Clients of Flawed Classes vs. Does any correlation exist in Eclipse? Data Class God Class Class Feature Envy Brain Class has design flaws? Class has provider classes with design flaws has defects (pre and post release)

8 Chi-Square Odds Ratio Mann-Whitney flaws / flawed providers vs. defects Non-parametric statistical tests flawed providers correlates with defects Odds Ratio on HAS NO Odds Ratio on HAS NO HAS HAS NO - NO

9 Odds Ratio on HAS NO Odds Ratio on HAS NO HAS 2-3 HAS NO NO Odds Ratio on HAS NO Percent of classes with... NO HAS NO NO The biggest danger is using flawed classes

10 Percent of classes with... NO NO Percent of classes with NO ions onal t p e Exc xcepti ht e g i l h hig Are the Classes that use Exceptions Defect Prone? Author: Cristina Marinescu 12th International Workshop on Principles on Software Evolution 7th ERCIM Workshop on Software Evolution 2011 NO 42 40

11 WE d l u Sho MO E R CA out? b a RE Exceptions & t in s i x e n o i t ela r r o yc n a s e? s e p o i l D Ec Location, Name, Type Class Class

12 Location, Name, Type Throws, Catches Location, Name, Type Throws, Catches Class Class (Pre&Post Release) Chi-Square & Odds Ratio Non-parametric Statistical Tests No Exceptions The Structure of the Contingency Tables.

13 No Exceptions The Structure of the Contingency Tables. No No Exceptions Exceptions The Structure of the Contingency Tables. No p-value? 0.05 p-value < 0.05 Correlation exists

14 No How Much? No No Exceptions < > No Exceptions < > Exceptions > < Exceptions > < Pre-Release Post-Release Eclipse 2 2,1 3 2,4 2,03 1,98 2,98 2,45 1,69? Correlation exists

15 Improper Usages of Exceptions & Does any correlation exist? Improper Usages = Exceptions handled by Superclasses No Improper Usages Improper Usages No The Structure of the NEW Contingency Tables. Pre-Release Post-Release Eclipse 2 2,1 3 2,18 2,55 2,15 2,94 2,33 1,47

16 Improper Usages & Correlation exists Exceptions & Improper Usages & Correlation exists Nachiappan Nagappan, Brendan Murphy, Victor R. Basili The influence of organizational structure on software quality: an empirical case study. International Conference of Software Engineering, 2008 Conway s Law "63 "64

17 Conway s Law organizations that design systems are constrained to produce systems which are copies of the communication structure of these organizations. Figure 1: Example Organization Structure of Company XYZ" "65 "66 Number of Engineers (NOE) Number of Ex-Engineers (NOEE) N engineers (N*(N-1))/2 communication paths knowledge transfer "67 "68

18 Edit Frequency (EF) Depth of Master Ownership (DMO) total number times the source code was edited Figure 1: Example Organization Structure of Company XYZ" level of ownership (> 75%) ABCA (190/250) "69 "70 Percentage of Org Contributing to development (PO) Level of Organizational Code Ownership (OCO) ratio of the number of people reporting at the DMO level owner relative to the Master org size 7/30 Figure 1: Example Organization Structure of Company XYZ" "71 percents of edits from the org that contains the owner 200/250 Figure 1: Example Organization Structure of Company XYZ" "72

19 Overall Organizational Ownership(OOW) Organizational Intersection Factor(OIF) ratio of the percentage of people at the DMO level making edits relative to the total engineers editing 5/32 Figure 1: Example Organization Structure of Company XYZ" "73 Figure 1: Example Organization Structure of Company XYZ" number of different org that that contribute greater than 10% of edits lower value = better "74 Prediction based on Organizational Structure Precision = 86.2% Recall = 84.0% Ralph Peters, Andy Zaidman Evaluating the Lifespan of Code Smells using Software Repository Mining. European Conference on Software Maintenance and Reengineering, 2012 "75 "76

20 Research Questions (RQ) Engineers are aware of code smells, but are not very concerned with their impact, given the low refactoring activity. RQ1 Are some types of code smells refactored more and quicker than other smell types? "77 "78 Research Questions (RQ1) Research Questions (RQ) God Class Feature Envy Data Class Message Chain Long Parameter List Avg 49,2% 40,39% 55,69% 47,04% 49,41% Sd 12,1% 18,08% 20,79% 18,76% 12,81% RQ2 Are relatively more code smells being refactored at an early or later stage of a system s life cycle? "79 "80

21 Research Questions (RQ) Research Questions (RQ) RQ3 Do some developers refactor more code smells than others and to what extend? RQ4 What refactoring rationales for code smells can be identified? "81 "82 Research Questions (RQ4) Cleaning up dead or obsolete code. Dedicated refactoring. Maintenance activities. Venera Arnaoudova, Massimiliano Di Penta, Giuliano Antoniol, Yann-Gaël Guéhéneuc A New Family of Software Anti-Patterns: Linguistic Anti-Patterns. European Conference on Software Maintenance and Reengineering, 2013 "83 "84

22 Linguistic Antipatterns LAs in software systems are recurring poor practices in the naming, documentation, and choice of identifiers in the implementation of an entity, thus possibly impairing program understanding. A. Does more than it says B. Says more than it does C. Does the opposite D. Contains more than it says E. Says more than it contains F. Contains the opposite "85 "86 A. Get - more than an accessor A. Is - returns more than a Boolean "87 "88

23 A. Set - method returns A. Expecting but not getting a single instance "89 "90 B. Not implemented condition B. Validation method does not confirm "91 "92

24 B. Get method does not return B. Not answered question "93 "94 B. Transform method does not return B. Expecting but not getting a collection "95 "96

25 C. Method name and return type are opposite C. Method signature and comment are opposite "97 "98 D. Contains more than it says E. Says more than it contains 1. Says one but contains many Vector _target; 2. Name suggests Boolean but type does not int[] isreached; private static boolean _stats = true; "99 "100

26 F. Contains the opposite 1. Attribute name and type are opposite MAssociationEnd start = null; 2. Attribute signature and comment are opposite "101 "102