Development by the numbers
•
16 likes•9,198 views
This document discusses software complexity metrics like cyclomatic complexity and N-path complexity. It provides examples of calculating these metrics for code samples and discusses how they can be used to measure complexity at the method, class, and project level. Tools are introduced that can automatically calculate these metrics for codebases, like PHPLOC and PDepend, which can help identify parts of code that are complex or hard to test. Maintaining low complexity and high test coverage over time can help improve code quality and reliability.
Report
Share
Development by the numbers
- 6. We Are Going To Measure Complexity
- 7. Why Should We Care About Complexity?
- 8. - Geer et al. "The Central Enemy Of Reliability is Complexity"
- 9. Complexity And Quality Are Strongly Related
- 11. Basic Metrics
- 13. Cyclomatic Complexity Number Of "Decision Points" In A Routine
- 14. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 15. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 16. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; } 4
- 17. Simple Right?
- 19. Cyclomatic Complexity (Single Method) 1 - 4: Low Complexity 5 - 7: Moderate Complexity 8 - 10: High Complexity 11+: Very High Complexity
- 20. Cyclomatic Complexity (Average Per Method) 1 - 2: Low Complexity 2 - 4: Moderate Complexity 4 - 6: High Complexity 6+: Very High Complexity
- 21. Compare: Average CC per Method Wordpress: 6.28 Drupal 7: 3.02 Drupal 8: 2.10 Symfony 2: 1.81 Zend Framework 2: 2.62 Laravel: 1.79
- 22. Cyclomatic Complexity (Average Per Line Of Code) .01 - .05: Low Complexity .05 - .10: Moderate Complexity .10 - .15: High Complexity .15+: Very High Complexity
- 23. Compare: Average CC per LOC Wordpress: 0.20 Drupal 7: 0.04 Drupal 8: 0.07 Symfony 2: 0.06 Zend Framework 2: 0.10 Laravel: 0.07
- 26. N-Path Complexity Number Of "Unique Paths" In A Routine
- 27. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 28. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 29. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 30. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 31. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; }
- 32. function foo($a, $b) { $c = 0; if ($a) { $c = $a; } elseif ($b) { $c = $b; } if ($a && $b) { $c = $a + $b; } return $c; } 4
- 34. Not Generally!
- 35. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: NPath:
- 36. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: NPath:
- 37. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 38. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 39. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 40. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 41. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 42. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 43. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 44. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 45. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath:
- 46. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath: 8
- 47. function foo2($a, $b, $c) { $d = 0; if ($a) { $d += $a; } if ($b) { $d += $b; } if ($c) { $d += $c; } return $d; } CC: 4 NPath: 8 2^(CC-1)
- 48. N-Path Complexity <16: Low Complexity 17-128: Moderate Complexity 129-1024: High Complexity 1025+: Very High Complexity
- 50. N-Path Complexity Minimum Number Of Tests Required To Completely Test A Routine
- 55. Cyclomatic Complexity 1 - 4: Low Complexity 5 - 7: Moderate Complexity 8 - 10: High Complexity 11+: Very High Complexity
- 64. Cyclomatic Complexity 1 - 4: Low Complexity 5 - 7: Moderate Complexity 8 - 10: High Complexity 11+: Very High Complexity
- 65. N-Path Complexity drupal_http_request() CC: 41 N-Path: 25,303,344,960
- 67. To Completely Test drupal_http_request() At 1 Line Of Code Per Test Would Require 2 Terabytes Worth Of Tests
- 68. To Completely Test drupal_http_request() At 1 Line Of Code Per Test Would Require 412 DVD's Worth Of Tests
- 69. To Completely Test drupal_http_request() At 1 Line Of Code Per Test Would Require 670k Drupals Worth Of Tests
- 70. And That's Not The Worst One!
- 74. N-Path Complexity _date_repeat_rrule_process() CC: 81 N-Path: 19,781,719,256
- 75. N-Path Complexity _date_repeat_rrule_process() CC: 81 N-Path: 19,781,719,256 ,250,000,000,000
- 76. N-Path Complexity _date_repeat_rrule_process() CC: 81 N-Path: 19,781,719,256 ,250,000,000,000 ,000,000,000
- 77. To Completely Test _date_repeat_rrule_process() At 1 Line Of Code Per Test Would Require 336T 2009's Worth Of Tests
- 78. To Completely Test _date_repeat_rrule_process() At 1 Line Of Code Per Test Would Require 1 Greenland Ice Cap of microSD cards Worth Of Tests
- 80. CRAP
- 82. CRAP (Change Risk Analysis Predictions)
- 83. CC = Cyclomatic Complexity (method) COV = Test Coverage (percent) CRAP = CC + (CC^2 * (1 - COV)^3)
- 85. CRAP Relates Complexity And Test Coverage
- 86. CRAP Increasing Test Coverage Lowers CRAP Decreasing Complexity Lowers CRAP
- 87. CRAP A Low Complexity Method With No Tests Is Good
- 88. CRAP A Low Complexity Method With Good Tests Is Great
- 89. CRAP A Moderate Complexity Method With Good Tests Is OK
- 90. CRAP A Moderate Complexity Method With No Tests Is CRAP
- 91. CRAP < 5: GREAT Code 5 - 15: Acceptable Code 15-30: Eih... Code 30+: CRAPpy Code
- 93. How Do We Apply These Metrics?
- 97. Sebastian Bergmann PHPUnit DbUnit PHPLOC PHPCPD PHPCOV hphpa www.phpqatools.org www.jenkins-php.org
- 98. PHPLOC
- 100. PHPLOC By Sebastian Bergmann Command Line Tool
- 101. PHPLOC By Sebastian Bergmann Command Line Tool Summarizes An Entire Codebase
- 102. $ phploc path/to/Drupal7/ Directories: 73 Files: 180 Lines of Code (LOC): 63347 Cyclomatic Complexity / Lines of Code: 0.04 Comment Lines of Code (CLOC): 19321 Non-Comment Lines of Code (NCLOC): 44026
- 103. Namespaces: 0 Interfaces: 1 Traits: 0 Classes: 38 Abstract: 2 (5.26%) Concrete: 36 (94.74%) Average Class Length (NCLOC): 197
- 104. Methods: 433 Scope: Non-Static: 378 (87.30%) Static: 55 (12.70%) Visibility: Public: 255 (58.89%) Non-Public: 178 (41.11%) Average Method Length (NCLOC): 17 Cyclomatic Complexity / Number of Methods: 3.02 Anonymous Functions: 0 Functions: 521 Constants: 22 Global constants: 15 Class constants: 7
- 105. PDepend
- 106. PDepend By Manuel Pichler (Also German)
- 107. PDepend By Manuel Pichler (Also German) Like PHPLOC, But Granular
- 108. PDepend By Manuel Pichler (Also German) Like PHPLOC, But Granular Lower Level Analysis
- 110. Fanout: Describes Outward Dependencies - Describes Dependence on Other Classes ANDC: Average Num of Derived Classes - Describes How Much Inheritance Is Used AHH: Average Hiearchy Height - Describes How Deep Of Inheritance Is Used
- 113. PHPMD By Manuel Pichler (German) Finds "Messy" Parts Of Code
- 114. PHPMD By Manuel Pichler (German) Finds "Messy" Parts Of Code Finds Rule Violations
- 115. PHPMD Rules CodeSize - (CC, NPath, Number of Methods, Size of Methods, etc) Design - (Eval, Goto, Exit(), Inheritance Depth) Naming - (Short names, Inconsistent Names) Unused Code Controversial - (Superglobal Access, Naming Conventions)
- 117. Prevent Complex Code From Even Getting In!
- 118. By Themselves Useful
- 119. Over Time
- 120. Over Time Invaluable
- 121. Drupal 8.x Branch Non-Comment Lines Of Code
- 122. Drupal 8.x Branch Number Of Classes
- 123. Drupal 8.x Branch Cyclomatic Complexity Per Method
- 124. Drupal 8.x Branch Cyclomatic Complexity Per Line
- 125. Drupal 8.x Branch
- 129. One More Thing To Keep In Mind