Intro

To run the samples in this demo you will first need to run npm install from inside the repo root. All demos have npm scripts to simplify execution.

What is TAP?

• TAP = Test Anything Protocol

  • Basic protocol created for testing Perl scripts.
  • Implementations in multiple languages

• tap JS library written by Isaac Z. Schlueter (@izs, founder of npm)

  • All of npm's test are built on tap, long term support.
  • Pros:
    • Can detect a test timing out.
    • Each test runs in own child process, less risk of environment corruption
  • Cons:
    • Cannot do browser testing

• tape JS library written by James Halliday (@substack, created Browserify)

  • Now has multiple maintainers and is used in hundreds of projects.
  • Pros:
    • Lighter-weight, faster.
    • Can be used in the browser.
  • Cons:
    • Most browser test centered around Browserify
    • No examples/tools for using with RequireJS

• Both libraries consist of two parts:

  1. Testing Framework (require('tape').test)
  • Outputs TAP to stdout
  • All tests assumed to be asynchronous
  • test('name', function (t) {})
    • Nest test() calls for subtests, instead of define() and it()
  • t contains assertions and .end()
  • Meshes well with sinon
    • t.pass(spy.calledOnce)
  • Can also use chai assertions.
    • Anything thrown within the test() function gets registered as a failed assertion, just like in mocha.
    • Wrap chai assertions in a t.shouldNotThrow() function to prevent undone tests.
  2. Testing Harness (shell command)
  • Receives TAP from tests.
  • tap formats into clean output, add -C to get raw TAP
  • tape aggregates results into a single output, must be piped to formatter.

• Both libraries can run each others tests.

  • For node testing there is a benefit to building tests with tape and running them with tap.

• Tests can be parallelized using Prove, a Perl TAP harness.

  • Dunno why nobody has made a node based parallelized test runner.

Why use Tape over Mocha?

Example 1

• Show index.js

  • Something wrong with this file, but don't look for it yet.

• Show mocha/test.js

  • Walk through test structure, point out assertions.

• npm run mocha

  • Show that test passes, everything looks good right?

• Show tape/test.js

  • Same test structure, just different way of calling assertions.

• npm run tape

  • Test is failing, why?
  • plan != count

• node tape/test.js to get raw output

  • Tape tests are normal node processes that output TAP to stdout
  • Each assertion is enumerated with ok and not ok, indents for sub-tests.
  • t.plan(4) call identifies expected four total assertions
  • Count off assertions, only three ran.
  • Note that close connection is missing
  • Identify missing parens on closeConnection call in index.js
  • Add parens, re-run test. Ta da.

• Show mocha/test-working.js

  • Mocha can be made to catch missing executions
  • Mock functions wrapped in sinon spies
  • Assertions that spies were called at end of test
  • Requires more effort, greater risk, complicates tests

Example 2

• Show index2.js

  • Basic unit, semi-contrived example. Returns collection of data.

• Show mocha/test2.js

  • Asserting that all three pieces of data pass.

• npm run mocha2

  • Test fails at first assertion.

• Show tape/test2.js

  • Exact same test.

• npm run tape2

  • Outputs all three failing assertions.
  • Because tape doesn't rely on assertions, it doesn't fall down on the first failure, giving more useful output.

Coverage

• npm run cover

  • Show coverage folder.

• npm run cover-mocha

  • Show coverage.

• Both are created the same way through istanbul.

  • istanbul cover tape tape/*
  • istanbul cover _mocha mocha/*
    • Must use _mocha due to test runner isolation in mocha
    • tap runner has same problem, solves by bundling istanbul into itself