Test driven development - Zombie proof your code

Post on 10-May-2015

400 views 1 download

Tags:

description

Changing Legacy code can cause it to come back to life and bite you in the APP! Learn the symptoms and techniques to write testable code

Transcript of Test driven development - Zombie proof your code

Test Driven DevelopmentZombie proof your code

Pascal Larocque

● TrustCharge Team● Behat guy● Testing guy● SOLID guy● Pattern guy● Father of 3● Star Wars Geek @pascallarocque

Why do we write code?

● Add New Features● Maintaining old features

What are the challenges?● I can’t understand what the code does● Only the original programmer understands what it does… He’s not

with us anymore● Fixing a Bug creates a NEW BUG● There is no documentation● It’s Only working on my local● We can’t upgrade because all are code is dependent on this

version● The only way to test is with the real data on production● I have to test this manually

What are we really saying?

“I’m Scared...When I change something, some other feature might stop working”

YOU’VE CREATED ZOMBIE CODE

How to survive the Zombie apocalypse

1. Write TESTS!2. Write TESTABLE CODE

The Power of Testable code● Forces Simplification of the code● Simplifies Design● Documentation of the code● Less time debugging● New code doesn’t break old code● Refactoring becomes easier● Interfaces are better designed● Easier to do code reviews● Fearless programming ● Faster than writing code without

tests

How to write testable codePHPUnitDatabase TestingLegacy code testingPHPspec

Why aren’t we testing?● Testing main flow is enough● The code is legacy, impossible to test● Don’t have time to test● Fix bug first then write a test● To be on the safe side manual testing is mandatory● No one else is writing tests● Testing and maintaining tests will cost more time● After changing code, I have to change a bunch of test

Sorry to say this but...

DON’T BLAME TESTING!MISSING PROFESSIONALISM

WRONG DESIGN

LOSING DISCIPLINE NO-TEST DEVELOPMENT

INEXPERIENCED DEVELOPERS

UNREVIEWED CODE

“THE SECRET TO TESTING IS WRITING TESTABLE CODE”

“If the answer is not obvious, or if the tests looks like the tests would be ugly or had to write, then take that as a warning signal.

Your design probably needs to modified; change things around until the code is easy to test, and your design will end up being better for the effort”

HOW TO GET STARTED?

How can you tell if your code isn’t going to come back from the SVN and BYTE you in the APP?

Look for the following SYMPTOMS.

SYMPTOMS - CONSTRUCTOR DOES TOO MUCH

● Using the NEW keyword in the constructor● STATIC METHOD calls● Anything more than field assignment● Object not fully initialized after constructor● CONTROL FLOW (conditions or loops) in constructor● Constructor build complex Collections instead of using

FACTORY or BUILDER● There is an INITIALIZE block● Not asking for Objects, looking for Objects

(DEPENDENCIES in constructor)

SYMPTOMS - Digging into dependencies

● Object passed are used to access other objects● LAW OF DEMETER VIOLATION: Method call chain

with more than one ->● SUSPICIOUS NAMES: Manager, Context, Registry,

Container● Creating mocks that return MOCKS● Deceitful API (Real Dependencies are unclear)● Too many “Middle Men” objects● Needle in a haystack (due to breaking demeter law)

SYMPTOMS - Singletons & Global state (the dark side)

● Using SINGLETONS● Accessing GLOBAL STATE STATICALLY● Using STATIC FIELDS or STATIC METHODS● Using a STATIC INITIALIZATION block● Using a REGISTRIES● NOT (or MIS-) using DEPENDENCY INJECTION● Hard Coded Dependencies● You have to read EVERY LINE OF CODE to

understand the potential side effects

CURE

USE DEPENDENCY INJECTION

Replace your static methodWrap static method from third party libraries

If you can’t wrap them, write adapters

SYMPTOMS - Object does too much

● When you have to use the word “AND” to describe what the Object does

● Object is hard to read for team members● Object have fields that are only used in some methods● Object has static methods that only operate on

parameters● MANY collaborators that you need to reach into for

more collaborators● HIDDEN INTERACTIONS behind public methods

CURESOLID

Single ResponsibilityOpen/Close Principle

Liskov Substitution PrincipleInterface Segregation

Dependency Inversion

● Run Without○ Network connection○ Production Data○ External API

● Run Fast○ ~milliseconds

Tests should...

● Must○ be Easy to maintain○ run before every

commit

Rules for TDD

1. Devise possible tests2. Implement 1 FAILING

test3. Implement just enough

code to PASS the test4. Refactor code and clean

up design5. Repeat until ALL tests

are passed

THINK DIFFERENTLY

IMPLEMENTDESIGN

TEST

TEST

How to write testable codePHPUnitDatabase TestingLegacy code testingPHPspec

PHPUnit - Installation

● Pear● Phar● Ubuntu package● Composer

Optional Packages● Selenium● DbUnit● SkeletonGenerator

PHPUnit - Bootstrap

● Common Initialization of testing Environment

$baseDir = realpath(__DIR__.'/..');

require_once $baseDir. '/tc_platform/Include/minimalBootstrap.php';

require_once $baseDir . '/TrustChargeLibrairies/vendor/autoload.php';

require_once $baseDir . '/PHPLibs/vendor/autoload.php';

PHPUnit - Writing Test● Use the skeleton generator (should be in Eclipse)

phpunit-skelgen --bootstrap bootstrap.php --test -- Class_Something Lib/Class/Something.php Class_SomethingTest Tests/Lib/Class/SomethingTest.php

● Write it yourselfclass Class_SomethingTest extends PHPUnit_Framework_TestCase

PHPUnit - Setup● setUpBeforeClass() - Executed before class is instantiated

○ Setup DB○ Setup Files

● setUp() - Executed before every test○ Reset DB○ Load Extra fixtures

● tearDown() - Executed after Test○ Clean up DB○ Remove Files

Depends@depends - For tests that depend on output from other tests

DataProvider@dataProvider - To test the same function with multiple values

Stubs and MocksStubs - Test double that allows you to control the flow of data

Mock - Test double that allows you to control the flow of data and that assert that the object was called in a specific way

Test DoublesDoubles - When you cannot replace a static dependency you can preload a test doubles instead

Test DoublesDoubles - When you cannot replace a static dependency you can preload a test doubles instead

Preload Test Doubles

Run in own process

How to write testable codePHPUnitDatabase TestingLegacy code testingPHPspec

Database testing

● Must be fast● Need to control the data in the database● Must be ran in TESTING environment● Clean up after yourself

1. Setup - DBConnection

Make sure you’re using TEST DB

Use MYISAM to bypass foreign key issues

Create the tables

2. Setup - FixturesLoad Data

Export Data

3. Tests

Hard coded dependencies

Hard coded DB name will cause you to truncate production DB instead of TEST DB

CURECreate new test double interface

How to write testable codePHPUnitDatabase TestingLegacy code testingPHPspec

"Behavior is the most important thing about software. It is what users depends on. Users like it when we add behavior, but if we change or remove behavior they depend on, they stop trusting us."

Legacy Code Change Algorithm

1. Identify the change points2. Find test points3. Break Dependencies4. Write Tests5. Makes changes and refactor

Identify test point

Find the code you need to change

Find test points

Bridge Class● Class that bridges the gap between legacy code and

new architecture● Allows you to test in isolation● Helps refactor out dependencies

Extract Method

Extract 535 Lines of code

Bridge

How to write testable codePHPUnitDatabase TestingLegacy code testingPHPspec

PHPSpec● http://www.phpspec.net

While PHPUnit focuses on testing the functions and methods of the classes used to develop features, specBDD focuses on testing the behaviors of these classes.

“ describing the code before you actually write it is a fear management technique. You don’t have to write all the code, just the spec of the next thing you want to work on. ” -- Kent Beck

Demo

It’s not worth writing test unless you have CONTINUOUS INTEGRATION

I ZOMBIES

Don't leave "broken windows" (bad designs, wrong decisions, or poor code) unrepaired. Fix each one as soon as it is discovered. Take some action to prevent further damage and to show that you're on top of the situation.

We've seen clean, functional systems deteriorate pretty quickly once builds start breaking. There are other factors that can contribute to software rot, but neglect accelerates the rot faster than any other factor.

The broken window Theory