ginkgo

README

Jump to the docs to learn more. To start rolling your Ginkgo tests now keep reading!

To discuss Ginkgo and get updates, join the google group.

Feature List

• Ginkgo uses Go's testing package and can live alongside your existing testing tests. It's easy to bootstrap and start writing your first tests

• Structure your BDD-style tests expressively:

  • Nestable Describe and Context container blocks
  • BeforeEach and AfterEach blocks for setup and teardown
  • It blocks that hold your assertions
  • JustBeforeEach blocks that separate creation from configuration (also known as the subject action pattern).
  • BeforeSuite and AfterSuite blocks to prep for and cleanup after a suite.

• A comprehensive test runner that lets you:

  • Mark specs as pending
  • Focus individual specs, and groups of specs, either programmatically or on the command line
  • Run your tests in random order, and then reuse random seeds to replicate the same order.
  • Break up your test suite into parallel processes for straightforward test parallelization

• Built-in support for testing asynchronicity

• Built-in support for benchmarking your code. Control the number of benchmark samples as you gather runtimes and other, arbitrary, bits of numerical information about your code.

• ginkgo: a command line interface with plenty of handy command line arguments for running your tests and generating test files. Here are a few choice examples:

  • ginkgo -nodes=N runs your tests in N parallel processes and print out coherent output in realtime
  • ginkgo -cover runs your tests using Golang's code coverage tool
  • ginkgo convert converts an XUnit-style testing package to a Ginkgo-style package
  • ginkgo -focus="REGEXP" and ginkgo -skip="REGEXP" allow you to specify a subset of tests to run via regular expression
  • ginkgo -r runs all tests suites under the current directory
  • ginkgo -v prints out identifying information for each tests just before it runs

  And much more: run ginkgo help for details!

  The ginkgo CLI is convenient, but purely optional -- Ginkgo works just fine with go test

• ginkgo watch watches packages and their dependencies for changes, then reruns tests. Run tests immediately as you develop!

• Straightforward support for third-party testing libraries such as Gomock and Testify. Check out the docs for details.

• A modular architecture that lets you easily:

  • Write custom reporters (for example, Ginkgo comes with a JUnit XML reporter and a TeamCity reporter).
  • Adapt an existing matcher library (or write your own!) to work with Ginkgo

Gomega: Ginkgo's Preferred Matcher Library

Learn more about Gomega here

Set Me Up!

You'll need Golang v1.2+ (Ubuntu users: you probably have Golang v1.0 -- you'll need to upgrade!)

go get github.com/onsi/ginkgo/ginkgo  # installs the ginkgo CLI
go get github.com/onsi/gomega         # fetches the matcher library

cd path/to/package/you/want/to/test

ginkgo bootstrap # set up a new ginkgo suite
ginkgo generate  # will create a sample test file.  edit this file and add your tests then...

go test # to run your tests

ginkgo  # also runs your tests

I'm new to Go: What are my testing options?

Of course, I heartily recommend Ginkgo and Gomega. Both packages are seeing heavy, daily, production use on a number of projects and boast a mature and comprehensive feature-set.

With that said, it's great to know what your options are :)

What Golang gives you out of the box

Testing is a first class citizen in Golang, however Go's built-in testing primitives are somewhat limited: The testing package provides basic XUnit style tests and no assertion library.

Matcher libraries for Golang's XUnit style tests

A number of matcher libraries have been written to augment Go's built-in XUnit style tests. Here are two that have gained traction:

• testify
• gocheck

You can also use Ginkgo's matcher library Gomega in XUnit style tests

BDD style testing frameworks

There are a handful of BDD-style testing frameworks written for Golang. Here are a few:

• Ginkgo ;)
• GoConvey
• Goblin
• Mao
• Zen

Finally, @shageman has put together a comprehensive comparison of golang testing libraries.

Go explore!

License

Ginkgo is MIT-Licensed

Documentation

Overview

Ginkgo is a BDD-style testing framework for Golang

The godoc documentation describes Ginkgo's API. More comprehensive documentation (with examples!) is available at http://onsi.github.io/ginkgo/

Ginkgo's preferred matcher library is [Gomega](http://github.com/onsi/gomega)

Ginkgo on Github: http://github.com/onsi/ginkgo

Ginkgo is MIT-Licensed

Index

• Constants
• Variables
• func AfterEach(body interface{}, timeout ...float64) bool
• func AfterSuite(body interface{}, timeout ...float64) bool
• func BeforeEach(body interface{}, timeout ...float64) bool
• func BeforeSuite(body interface{}, timeout ...float64) bool
• func Context(text string, body func()) bool
• func Describe(text string, body func()) bool
• func FContext(text string, body func()) bool
• func FDescribe(text string, body func()) bool
• func FIt(text string, body interface{}, timeout ...float64) bool
• func FMeasure(text string, body interface{}, samples int) bool
• func Fail(message string, callerSkip ...int)
• func GinkgoParallelNode() int
• func GinkgoRecover()
• func It(text string, body interface{}, timeout ...float64) bool
• func JustBeforeEach(body interface{}, timeout ...float64) bool
• func Measure(text string, body interface{}, samples int) bool
• func PContext(text string, body func()) bool
• func PDescribe(text string, body func()) bool
• func PIt(text string, _ ...interface{}) bool
• func PMeasure(text string, _ ...interface{}) bool
• func RunSpecs(t GinkgoTestingT, description string) bool
• func RunSpecsWithCustomReporters(t GinkgoTestingT, description string, specReporters []Reporter) bool
• func RunSpecsWithDefaultAndCustomReporters(t GinkgoTestingT, description string, specReporters []Reporter) bool
• func SynchronizedAfterSuite(allNodesBody interface{}, node1Body interface{}, timeout ...float64) bool
• func SynchronizedBeforeSuite(node1Body interface{}, allNodesBody interface{}, timeout ...float64) bool
• func XContext(text string, body func()) bool
• func XDescribe(text string, body func()) bool
• func XIt(text string, _ ...interface{}) bool
• func XMeasure(text string, _ ...interface{}) bool
• type Benchmarker
• type Done
• type GinkgoTInterface
  • func GinkgoT(optionalOffset ...int) GinkgoTInterface
• type GinkgoTestDescription
  • func CurrentGinkgoTestDescription() GinkgoTestDescription
• type GinkgoTestingT
• type Reporter

Constants

const GINKGO_PANIC = `` /* 332-byte string literal not displayed */

const GINKGO_VERSION = config.VERSION

Variables

var GinkgoWriter io.Writer

GinkgoWriter implements an io.Writer When running in verbose mode any writes to GinkgoWriter will be immediately printed to stdout. Otherwise, GinkgoWriter will buffer any writes produced during the current test and flush them to screen only if the current test fails.

Functions

func AfterEach

func AfterEach(body interface{}, timeout ...float64) bool

AfterEach blocks are run after It blocks. When multiple AfterEach blocks are defined in nested Describe and Context blocks the innermost AfterEach blocks are run first.

Like It blocks, AfterEach blocks can be made asynchronous by providing a body function that accepts a Done channel

func AfterSuite

func AfterSuite(body interface{}, timeout ...float64) bool

AfterSuite blocks are *always* run after all the specs regardless of whether specs have passed or failed. Moreover, if Ginkgo receives an interrupt signal (^C) it will attempt to run the AfterSuite before exiting.

When running in parallel, each parallel node process will call AfterSuite.

AfterSuite blocks can be made asynchronous by providing a body function that accepts a Done channel

You may only register *one* AfterSuite handler per test suite. You typically do so in your bootstrap file at the top level.

func BeforeEach

func BeforeEach(body interface{}, timeout ...float64) bool

BeforeEach blocks are run before It blocks. When multiple BeforeEach blocks are defined in nested Describe and Context blocks the outermost BeforeEach blocks are run first.

Like It blocks, BeforeEach blocks can be made asynchronous by providing a body function that accepts a Done channel

func BeforeSuite

func BeforeSuite(body interface{}, timeout ...float64) bool

BeforeSuite blocks are run just once before any specs are run. When running in parallel, each parallel node process will call BeforeSuite.

BeforeSuite blocks can be made asynchronous by providing a body function that accepts a Done channel

You may only register *one* BeforeSuite handler per test suite. You typically do so in your bootstrap file at the top level.

func Context

func Context(text string, body func()) bool

Context blocks allow you to organize your specs. A Context block can contain any number of BeforeEach, AfterEach, JustBeforeEach, It, and Measurement blocks.

In addition you can nest Describe and Context blocks. Describe and Context blocks are functionally equivalent. The difference is purely semantic -- you typical Describe the behavior of an object or method and, within that Describe, outline a number of Contexts.

func Describe

func Describe(text string, body func()) bool

Describe blocks allow you to organize your specs. A Describe block can contain any number of BeforeEach, AfterEach, JustBeforeEach, It, and Measurement blocks.

In addition you can nest Describe and Context blocks. Describe and Context blocks are functionally equivalent. The difference is purely semantic -- you typical Describe the behavior of an object or method and, within that Describe, outline a number of Contexts.

func FContext

func FContext(text string, body func()) bool

You can focus the tests within a describe block using FContext

func FDescribe

func FDescribe(text string, body func()) bool

You can focus the tests within a describe block using FDescribe

func FIt

func FIt(text string, body interface{}, timeout ...float64) bool

You can focus individual Its using FIt

func FMeasure

func FMeasure(text string, body interface{}, samples int) bool

You can focus individual Measures using FMeasure

func Fail

func Fail(message string, callerSkip ...int)

Fail notifies Ginkgo that the current spec has failed. (Gomega will call Fail for you automatically when an assertion fails.)

func GinkgoParallelNode

func GinkgoParallelNode() int

GinkgoParallelNode returns the parallel node number for the current ginkgo process The node number is 1-indexed

func GinkgoRecover

func GinkgoRecover()

GinkgoRecover should be deferred at the top of any spawned goroutine that (may) call `Fail` Since Gomega assertions call fail, you should throw a `defer GinkgoRecover()` at the top of any goroutine that calls out to Gomega

Here's why: Ginkgo's `Fail` method records the failure and then panics to prevent further assertions from running. This panic must be recovered. Ginkgo does this for you if the panic originates in a Ginkgo node (an It, BeforeEach, etc...)

Unfortunately, if a panic originates on a goroutine *launched* from one of these nodes there's no way for Ginkgo to rescue the panic. To do this, you must remember to `defer GinkgoRecover()` at the top of such a goroutine.

func It

func It(text string, body interface{}, timeout ...float64) bool

It blocks contain your test code and assertions. You cannot nest any other Ginkgo blocks within an It block.

Ginkgo will normally run It blocks synchronously. To perform asynchronous tests, pass a function that accepts a Done channel. When you do this, you can also provide an optional timeout.

func JustBeforeEach

func JustBeforeEach(body interface{}, timeout ...float64) bool

JustBeforeEach blocks are run before It blocks but *after* all BeforeEach blocks. For more details, read the [documentation](http://onsi.github.io/ginkgo/#separating_creation_and_configuration_)

Like It blocks, BeforeEach blocks can be made asynchronous by providing a body function that accepts a Done channel

func Measure

func Measure(text string, body interface{}, samples int) bool

Measure blocks run the passed in body function repeatedly (determined by the samples argument) and accumulate metrics provided to the Benchmarker by the body function.

The body function must have the signature:

func(b Benchmarker)

func PContext

func PContext(text string, body func()) bool

You can mark the tests within a describe block as pending using PContext

func PDescribe

func PDescribe(text string, body func()) bool

You can mark the tests within a describe block as pending using PDescribe

func PIt

func PIt(text string, _ ...interface{}) bool

You can mark Its as pending using PIt

func PMeasure

func PMeasure(text string, _ ...interface{}) bool

You can mark Maeasurements as pending using PMeasure

func RunSpecs

func RunSpecs(t GinkgoTestingT, description string) bool

RunSpecs is the entry point for the Ginkgo test runner. You must call this within a Golang testing TestX(t *testing.T) function.

To bootstrap a test suite you can use the Ginkgo CLI:

ginkgo bootstrap

func RunSpecsWithCustomReporters

func RunSpecsWithCustomReporters(t GinkgoTestingT, description string, specReporters []Reporter) bool

To run your tests with your custom reporter(s) (and *not* Ginkgo's default reporter), replace RunSpecs() with this method. Note that parallel tests will not work correctly without the default reporter

func RunSpecsWithDefaultAndCustomReporters

func RunSpecsWithDefaultAndCustomReporters(t GinkgoTestingT, description string, specReporters []Reporter) bool

To run your tests with Ginkgo's default reporter and your custom reporter(s), replace RunSpecs() with this method.

func SynchronizedAfterSuite

func SynchronizedAfterSuite(allNodesBody interface{}, node1Body interface{}, timeout ...float64) bool

SynchronizedAfterSuite blocks complement the SynchronizedBeforeSuite blocks in solving the problem of setting up external singleton resources shared across nodes when running tests in parallel.

SynchronizedAfterSuite accomplishes this by taking *two* function arguments. The first runs on all nodes. The second runs only on parallel node #1 and *only* after all other nodes have finished and exited. This ensures that node 1, and any resources it is running, remain alive until all other nodes are finished.

Both functions have the same signature: either func() or func(done Done) to run asynchronously.

Here's a pseudo-code example that complements that given in SynchronizedBeforeSuite. Here, SynchronizedAfterSuite is used to tear down the shared database only after all nodes have finished:

var _ = SynchronizedAfterSuite(func() {
	dbClient.Cleanup()
}, func() {
	dbRunner.Stop()
})

func SynchronizedBeforeSuite

func SynchronizedBeforeSuite(node1Body interface{}, allNodesBody interface{}, timeout ...float64) bool

SynchronizedBeforeSuite blocks are primarily meant to solve the problem of setting up singleton external resources shared across nodes when running tests in parallel. For example, say you have a shared database that you can only start one instance of that must be used in your tests. When running in parallel, only one node should set up the database and all other nodes should wait until that node is done before running.

SynchronizedBeforeSuite accomplishes this by taking *two* function arguments. The first is only run on parallel node #1. The second is run on all nodes, but *only* after the first function completes succesfully. Ginkgo also makes it possible to send data from the first function (on Node 1) to the second function (on all the other nodes).

The functions have the following signatures. The first function (which only runs on node 1) has the signature:

func() []byte

or, to run asynchronously:

func(done Done) []byte

The byte array returned by the first function is then passed to the second function, which has the signature:

func(data []byte)

or, to run asynchronously:

func(data []byte, done Done)

Here's a simple pseudo-code example that starts a shared database on Node 1 and shares the database's address with the other nodes:

var dbClient db.Client
var dbRunner db.Runner

var _ = SynchronizedBeforeSuite(func() []byte {
	dbRunner = db.NewRunner()
	err := dbRunner.Start()
	Ω(err).ShouldNot(HaveOccurred())
	return []byte(dbRunner.URL)
}, func(data []byte) {
	dbClient = db.NewClient()
	err := dbClient.Connect(string(data))
	Ω(err).ShouldNot(HaveOccurred())
})

func XContext

func XContext(text string, body func()) bool

You can mark the tests within a describe block as pending using XContext

func XDescribe

func XDescribe(text string, body func()) bool

You can mark the tests within a describe block as pending using XDescribe

func XIt

func XIt(text string, _ ...interface{}) bool

You can mark Its as pending using XIt

func XMeasure

func XMeasure(text string, _ ...interface{}) bool

You can mark Maeasurements as pending using XMeasure

Types

type Benchmarker

type Benchmarker interface {
	Time(name string, body func(), info ...interface{}) (elapsedTime time.Duration)
	RecordValue(name string, value float64, info ...interface{})
}

Measurement tests receive a Benchmarker.

You use the Time() function to time how long the passed in body function takes to run You use the RecordValue() function to track arbitrary numerical measurements. The optional info argument is passed to the test reporter and can be used to provide the measurement data to a custom reporter with context.

See http://onsi.github.io/ginkgo/#benchmark_tests for more details

type Done

type Done chan<- interface{}

Asynchronous specs are given a channel of the Done type. You must close or write to the channel to tell Ginkgo that your async test is done.

type GinkgoTInterface

type GinkgoTInterface interface {
	Fail()
	Error(args ...interface{})
	Errorf(format string, args ...interface{})
	FailNow()
	Fatal(args ...interface{})
	Fatalf(format string, args ...interface{})
	Log(args ...interface{})
	Logf(format string, args ...interface{})
	Failed() bool
	Parallel()
	Skip(args ...interface{})
	Skipf(format string, args ...interface{})
	SkipNow()
	Skipped() bool
}

The interface returned by GinkgoT(). This covers most of the methods in the testing package's T.

func GinkgoT

func GinkgoT(optionalOffset ...int) GinkgoTInterface

Some matcher libraries or legacy codebases require a *testing.T GinkgoT implements an interface analogous to *testing.T and can be used if the library in question accepts *testing.T through an interface

For example, with testify: assert.Equal(GinkgoT(), 123, 123, "they should be equal")

Or with gomock: gomock.NewController(GinkgoT())

GinkgoT() takes an optional offset argument that can be used to get the correct line number associated with the failure.

type GinkgoTestDescription

type GinkgoTestDescription struct {
	FullTestText   string
	ComponentTexts []string
	TestText       string

	IsMeasurement bool

	FileName   string
	LineNumber int
}

GinkgoTestDescription represents the information about the current running test returned by CurrentGinkgoTestDescription

FullTestText: a concatenation of ComponentTexts and the TestText
ComponentTexts: a list of all texts for the Describes & Contexts leading up to the current test
TestText: the text in the actual It or Measure node
IsMeasurement: true if the current test is a measurement
FileName: the name of the file containing the current test
LineNumber: the line number for the current test

func CurrentGinkgoTestDescription

func CurrentGinkgoTestDescription() GinkgoTestDescription

CurrentGinkgoTestDescripton returns information about the current running test.

type GinkgoTestingT

type GinkgoTestingT interface {
	Fail()
}

The interface by which Ginkgo receives *testing.T

type Reporter

type Reporter reporters.Reporter

Custom Ginkgo test reporters must implement the Reporter interface.

The custom reporter is passed in a SuiteSummary when the suite begins and ends, and a SpecSummary just before a spec begins and just after a spec ends