Code metrics in PHP

Code metrics in PHP
From lines to code semantic

● Julio Martinez
● Developing PHP since 2001
● 1.5 years working at Ulabox
● Find me: @liopic
Code metrics in PHP: 0. Introduction
Who am I?

● El antisúper!
● 6-years-old startup
● 11 developers
● monolith & new services
● #rigor
● We are hiring!
Code metrics in PHP: 0. Introduction
What is Ulabox?

● Evaluate quality!
● We need objective, reproducible and quantifiable metrics
Could you tell me some examples of metrics?
Code metrics in PHP: 1. Looking for quality
Why do we need software metrics?

● number of bugfixes per month
● lines of code
● test coverage
● number of user stories covered
● follows clean code’s rules
● documentation lines / total of code lines
● etc
Some examples?

● Testing first!
● Code “surface”
● Lines grouping
● Code semantic
Disclaimer: I’ll discuss locally-executable tools (non SaaS)
● SaaS: Insight, Code climate, Scrutinizer, SonarQube...
Let’s start our knowledge journey...

● User stories: behat
● General testing: phpunit
○ Code coverage
○ Mutant testing (=test your tests): humbug
● Unit/spec testing: phpspec
Code metrics in PHP. Testing
Testing First!

● Code sniffer (code style, PSR2): phpcs
● Copy+paste detector (DRY): phpcpd
● Clean code & common smells:
○ phpmd (“mess detector”, a bit old)
○ Exakat (it’s SaaS but has trial download)
Code metrics in PHP: 3. Code surface
Code “surface”

● Counting lines: phploc
● Getting ratios: pdepend
Code metrics in PHP: 4. Lines grouping
Lines grouping

1. Cyclomatic complexity (paths of execution)
2. Coupling metrics (relations between “modules”)
3. Lack of cohesion, LCOM (relations between methods)
4. Halstead’s metrics (operands and operators)
5. Maintainability Index
Code metrics in PHP: 5. Code semantic
Code semantic analysis

Cyclomatic complexity
abstract class BaseBird
{
private $eggs = 0;
public function makeEgg()
{
$this->eggs++;
}
public function crackEgg()
{
if ($this->eggs <=0) {
throw new NoEggsException();
}
$this->eggs--;
}
public function fly()
{
return 'flap-flap';
}
abstract public function sound();
}
class Duck extends BaseBird
{
public function sound()
{
return 'quack';
}
public function swim()
{
return 'splash';
}
}
class Parrot extends BaseBird
{
{
$friend = new Duck();
if($friend->sound()){
return 'Err '.$friend->sound();
}
return '';
}
}

Cyclomatic complexity
{
private $eggs = 0;
{
$this->eggs++;
}
{
}
$this->eggs--;
}
{
return 'flap-flap';
}
}
class Duck extends BaseBird
{
{
return 'quack';
}
public function swim()
{
return 'splash';
}
}
class Parrot extends BaseBird
{
{
$friend = new Duck();
if($friend->sound()){
return 'Err '.$friend->sound();
}
return '';
}
}
2
2
1

● Afferent couplings (Ca) - “they use you”
● Efferent couplings (Ce) - “you use them”
● Instability, resilience to change (I): I = Ce / (Ce + Ca)
○ I=0 is a completely stable package
○ I=1 is a completely unstable package
Coupling metrics

● Abstractness (A): ratio of abstract classes
○ A=0 is a completely concrete package
○ A=1 is a completely abstract package
● Examples:
○ BaseBird: Ce=1 (uses NoEggsException), Ca=0; A=1, I=1
○ Parrot, Ce=1 (uses Duck), Ca=0; A=0, I=1
○ Duck, Ce=0, Ca=0; A=0, I=0
Coupling metrics

● Distance from the main sequence (D): D = |A+I-1|
○ Balance between abstractness and stability
○ Ideal packages: (I=1, A=0), (I=0, A=1)
Coupling metrics: main sequence

● Groups of methods
● LCOM = 1 is ideal
Lack of cohesion of methods (LCOM)
{
private $eggs = 0;
{
$this->eggs++;
}
{
}
$this->eggs--;
}
{
return 'flap-flap';
}
}

● Groups of methods
● LCOM = 1 is ideal
● BaseBird LCOM = 3
Lack of cohesion of methods (LCOM)
{
private $eggs = 0;
{
$this->eggs++;
}
{
}
$this->eggs--;
}
{
return 'flap-flap';
}
}

● η1
= number of distinct operators
● η2
= number of distinct operands
● N1
= the total number of operators
● N2
= the total number of operands
● Program vocabulary: η = η1
+ η2
● Program length: N = N1
+ N2
Halstead complexity measures (1977)

● Volume: V = N × log2
η (linearly with length, log with vocabulary)
● Difficulty : D = η1
/2 × N2
/η2
(half distinct operators, scarcity of operands)
● Effort: E = D × V
● Time required to program: T = E/18 seconds
● Delivered bugs:
Halstead complexity measures

● University of Idaho, Oman and Hagemeister
● Volume (V), Cyclomatic Complexity (G), Lines of Code (LOC)
● Original MI = 171 - 5.2*ln(V) - 0.23*G - 16.2*ln(LOC)
Maintainability Index (1991)

● phpmetrics
○ Let’s see all those indexes!
Coupling + LCOM + Halstead + MI
Maintainability.................... 65.83 / 100
Accessibility for new developers... 41.03 / 100
Simplicity of algorithms........... 42.57 / 100
Volume............................. 64.90 / 100
Reducing bug's probability......... 65.57 / 100

● MUST!
○ phpcs - code sniffer - No extra effort need; always
○ phpunit - integration & unit tests - Write tests; always
● Adopt
○ phpspec - unit/specs - Write tests; always
○ behat - user stories (functional tests) - Write user stories; always
○ phpcpd - copy & paste - No effort; always
○ phpmetrics - No effort; main indexes weekly, others when refactor
● Give a try
○ humbug - test your tests - No effort; often
○ SaaS options - Customization effort; always
Code metrics in PHP: 6. Summing up!
RADAR of tools (how? when?)

Questions?

Thank you!

Code metrics in PHP

Recommended

More Related Content

Similar to Code metrics in PHP (20)

More from Julio Martinez (7)

Recently uploaded (20)

Code metrics in PHP