Modern software development challenges

Context of modern software development

You can update web application instantly

You rarely have full control of all code you use

Development of a system is never done

Modern tools have complex multi-layer structure

Can you fix mistakes after product "release"?

Facebook, modern start-ups and others

Do multiple "releases" per day

Monitoring to avoid big failures

Small issues can be corrected in minutes

Do you control all the code your system runs?

How complex is your infrastructure?

Modern JavaScript
web development

eXtreme Programming (2000s)

The Joel Test: 12 Steps to Better Code (2000)

"If you get these 12 things right, you’ll have a disciplined team that can consistently deliver" (selection)

Continuous integration

Ensuring your project is always in a good state

Your master branch builds reproducibly

Everything also builds on a clean machine

All your tests run automatically and pass

Continuous integration tools help with this!

Anatomy of a modern JavaScript project

Anatomy of a modern JavaScript project

What really matters from the previous slide

Code and tests - writing and running tests

Dependency management - getting versions right

Build configuration - build and run with one command

Continuous integration - ensure reproducibility

Dependency management

Package dependencies

• Public repository such as NPM
• Version 1.0.0 major, minor, revision
• Want exact version? Latest revision?

Why is choosing versions hard

• Packages depend on other packages
• Transitive dependencies can change
• Packages can have conflicting requirements

Solving version hell with lock files and the Yarn for JavaScript

Solving version hell with lock files and the Yarn for JavaScript

Build configuration

Build with one command

• Typically pre-processing, compilation, post-processing
• Also dev and testing
• Makefile in UNIX world

Building in JavaScript world

• Specified in the scripts section in package.json
• Invoked using yarn build, yarn test etc.

Rapid feedback during development

Forms of feedback when coding

• Syntax and type errors (milliseconds)
• Reload when code changes (seconds)
• Run tests when code changes (minutes)

How and why of unit testing

Tests for ensuring quality

• Unit tests to prevent regressions
• Write tests after you write code
• Continuous integration to run tests

 1: 
 2: 
 3: 
 4: 
 5: 
 6: 
 7: 
 8: 
 9: 
10: 
11: 
12: 
13: 
14: 
15: 

var assert = require('assert');

describe('Arithmetic', function () {
  it('should add', function () {
    var actual = 40 + 2
    var expected = 42
    assert.equal(actual, expected)
  });

  it('should multiply', function () {
    var actual = 7 * 6
    var expected = 42
    assert.equal(actual, expected)
  });
});

Writing JavaScript tests with Mocha

You describe logic with tests written using it notation

Run tests from terminal using
the mocha tool

Beyond simple tests

Unit testing and general testing concerns

Unit tests are just one, easy to automate kind

How to structure code to make testing easy?

How to test not just functions, but user interfaces

Property testing to specify more than one case

Continuous integration

What you need to use CI

• Automatic reproducible build
• Package management, tests
• Service to run your builds!

Continuous integration support on GitHub

Runs build and tests every time you git push

Runs build and tests for every pull request

Shows warning when
PR build fails

Beyond simple CI

Continuous development process

Changing context of software development
Release frequency and system complexity
Extreme programming: feedback and continuous process

Modern (JavaScript) project infrastructure
One command build to run, test and build
Dependency management, versioning and lock files

Testing and continuous integration tools
Tests for ensuring quality and tests for feedback
Running build automatically with CI systems

CO886: Continuous development process

What you should remember from this lecture

• What challenges is continuous process solving
• Reproducible builds and dependency management
• Why use tests and continuous integration tools

Tomas Petricek
t.petricek@kent.ac.uk | @tomaspetricek

Papers and articles

• Joel Spolsky (2000). The Joel Test: 12 Steps to Better Code
• Chris Williams (2016). How one developer just broke Node, Babel and thousands of projects in 11 lines of JavaScript

Libraries and tools

• JavaScript tools: Webpack, Mocha, yarn.
• Continuous integration: Travis