Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Test Anything Protocol tools for node
JavaScript Shell
Latest commit f45630b @isaacs v2.3.1


A TAP test framework for Node.js.

It includes a command line test runner for consuming TAP-generating test scripts, and a JavaScript framework for writing such scripts.

Coverage Status


Write your tests in JavaScript

var tap = require('tap')

// you can test stuff just using the top level object.
// no suites or subtests required.

tap.equal(1, 1, 'check if numbers still work')
tap.notEqual(1, 2, '1 should not equal 2')

// also you can group things into sub-tests.
// Sub-tests will be run in sequential order always,
// so they're great for async things.

tap.test('first stuff', function (t) {
  t.ok(true, 'true is ok')
  t.similar({a: [1,2,3]}, {a: [1,2,3]})
  // call t.end() when you're done

// If you have a bunch of setup stuff that MUST work or else
// the rest of the tests are not worth running, then you can
// pass `{ bail: true }` to make it bail out on failure.

tap.test('must succeed or all is lost', { bail: true }, function (t) {
  db = new DataBorscht()
  t.ok(db, 'borscht setup must succeed')

// You can also bail out based on specific conditions, or with a
// different error message of your choosing.
tap.test('must mostly succeed or all is lost', function (t) {
  db = new DataBorscht()

  t.ok(db, 'borscht setup')
  if (!db) {
    t.bailout('the borscht is lost.  I cannot continue.')

  t.ok(db.connection, 'db must have connection')
  t.ok(db.username, 'db must have username')
  t.equal(db.color, 'red', 'borscht should be red')
  if (!t.passing())
    t.bailout('something weird with the data borscht.')


// you can specify a 'plan' if you know how many
// tests there will be in advance. Handy when
// order is irrelevant and things happen in parallel.

// Note that the function name is used if no name is provided!
tap.test(function planned (t) {
  setTimeout(function () {
    t.ok(true, 'a timeout')
  setTimeout(function () {
    t.ok(true, 'b timeout')

// you can do `var test = require('tap').test` if you like
// it's pre-bound to the root tap object.

var test = require('tap').test

// subtests can have subtests
test(function parent (t) {
  t.test(function child (tt) {
    tt.throws(function () {
      throw new Error('fooblz')
    }, {
      message: 'fooblz'
    }, 'throw a fooblz')

    // throws also uses function name if no name provided
    tt.throws(function throw_whatever () { throw 1 })



// thrown errors just fail the current test, so you can
// also use your own assert library if you like.
// Of course, this means it won't be able to print out the
// number of passing asserts, since passes will be silent.

test('my favorite assert lib', function (t) {
  var assert = require('assert')
  assert.ok(true, 'true is ok')
  assert.equal(1, 1, 'math works')

  // Since it can't read the plan, using a custom assert lib
  // means that you MUST use t.end()

// You can mark tests as 'todo' either using a conf object,
// or simply by omitting the callback.
test('solve halting problem')
test('prove p=np', { todo: true }, function (t) {
  // i guess stuff goes here'traveling salesmen must pack their own bags')

// Prefer mocha/rspec/lab style global objects?
// Got you covered.  This is a little experimental,
// patches definitely welcome.
describe('suite ride bro', function () {
  it('should have a wheel', function () {
    assert.ok(thingie.wheel, 'wheel')
  it('can happen async', function (done) {
    setTimeout(function () {

// Read on for a complete list of asserts, methods, etc.

You can run tests using the tap executable. Put this in your package.json file:

  "scripts": {
    "test": "tap test/*.js"

and then you can run npm test to run your test scripts.

Command line behavior and flags:

$ tap -h
  tap [options] <files>

Executes all the files and interprets their output as TAP
formatted test result data.

To parse TAP data from stdin, specify "-" as a filename.

Short options are parsed gnu-style, so for example '-bCRspec' would be
equivalent to '--bail --no-color --reporter=spec'


  -c --color                  Use colors (Default for TTY)

  -C --no-color               Do not use colors (Default for non-TTY)

  -b --bail                   Bail out on first failure

  -B --no-bail                Do not bail out on first failure (Default)

  -R<type> --reporter=<type>  Use the specified reporter.  Defaults to
                              'classic' when colors are in use, or 'tap'
                              when colors are disabled.

                              Available reporters:
                              classic doc dot dump html htmlcov json
                              jsoncov jsonstream landing list markdown
                              min nyan progress silent spec tap xunit

  -s<file> --save=<file>      If <file> exists, then it should be a line-
                              delimited list of test files to run.  If
                              <file> is not present, then all command-line
                              positional arguments are run.

                              After the set of test files are run, any
                              failed test files are written back to the
                              save file.

                              This way, repeated runs with -s<file> will
                              re-run failures until all the failures are
                              passing, and then once again run all tests.

                              It's a good idea to .gitignore the file
                              used for this purpose, as it will churn a

  --coverage --cov            Capture coverage information using 'nyc'

                              If a COVERALLS_REPO_TOKEN environment
                              variable is set, then coverage is
                              captured by default and sent to the
                     service. If a CODECOV_TOKEN
                              environment variable is set, then coverage is
                              captured by default and sent to the

  --no-coverage --no-cov      Do not capture coverage information.
                              Note that if nyc is already loaded, then
                              the coverage info will still be captured.

  --coverage-report=<type>    Output coverage information using the
                              specified istanbul/nyc reporter type.

                              Default is 'text' when running on the
                              command line, or 'text-lcov' when piping
                              to coveralls or codecov.

                              If 'lcov' is used, then the report will
                              be opened in a web browser after running.

                              This can be run on its own at any time
                              after a test run that included coverage.

  -t<n> --timeout=<n>         Time out test files after <n> seconds.
                              Defaults to 30, or the value of the
                              TAP_TIMEOUT environment variable.

  -h --help                   print this thing you're looking at

  -v --version                show the version of this program

  -gc --expose-gc             Expose the gc() function to Node tests

  --debug                     Run JavaScript tests with node --debug

  --debug-brk                 Run JavaScript tests with node --debug-brk

  --harmony                   Enable all Harmony flags in JavaScript tests

  --strict                    Run JS tests in 'use strict' mode

  --                          Stop parsing flags, and treat any additional
                              command line arguments as filenames.


This module uses nyc to track code coverage, even across subprocess boundaries. It is included by default, and there's nothing you need to do but enable it. Adding coverage will make your tests run slightly slower, but that's to be expected.

To generate coverage information, run your tests with the --cov argument.

To specify a report format, you can use --coverage-report=<type>. The default type is text, which produces a pretty text-only table on the terminal. If you specify --coverage-report=lcov, then tap will attempt to open a web browser to view the report after the test run.

If you use this a lot, you may want to add coverage and .nyc_output to your .gitignore and/or .npmignore files.

Travis-CI and Integration

You can very easily take advantage of continuous test coverage reports by using Travis-CI and Coveralls.

  1. Enable Travis-CI by signing up, enabling tests on your repo, and adding a .travis.yml file to your repo. You can use this module's .travis.yml file as an example
  2. Enable or by signing up, and adding the repo. Note the repo API token.
  3. Back at Travis-CI, add a private environment variable. The name of the environment variable is COVERALLS_REPO_TOKEN for Coveralls, or CODECOV_TOKEN for, and the value is the token you got from Coveralls or CodeCov.
  4. When that token is set in the environment variable, tap will automatically generate coverage information and send it to the appropriate place.


tap = require('tap')

The root tap object is an instance of the Test class with a few slight modifications.

  1. The teardown(), plan(), and test() methods are pre-bound onto the root object, so that you don't have to call them as methods.
  2. By default, it pipes to stdout, so running a test directly just dumps the TAP data for inspection. (You can of course tap.unpipe(process.stdout) if you want to direct it elsewhere.)
  3. Various other things are hung onto it for convenience, since it is the main package export.
  4. The test ends automatically when process.on('exit') fires, so there is no need to call tap.end() explicitly.
  5. Adding a teardown function triggers autoend behavior. Otherwise, the end would potentially never arrive, if for example teardown is used to close a server or cancel some long-running process, because process.on('exit') would never fire of its own accord.


A list of all of the canonical assert methods and their synonyms.


Method that injects describe() and it() into the global environment for mocha-like BDD style test definition.

This feature is incomplete, experimental, and may change drastically in the future. Feedback is welcome.


The Test class is the main thing you'll be touching when you use this module.

The most common way to instantiate a Test object by calling the test method on the root or any other Test object. The callback passed to test(name, fn) will receive a child Test object as its argument.

A Test object is a Readable Stream. Child tests automatically send their data to their parent, and the root require('tap') object pipes to stdout by default. However, you can instantiate a Test object and then pipe it wherever you like. The only limit is your imagination.

t.test([name], [options], [function])

Create a subtest.

If the function is omitted, then it will be marked as a "todo" or "pending" test.

If the function has a name, and no name is provided, then the function name will be used as the test name. If no test name is provided, then the name will be (unnamed test).

The function gets a Test object as its only argument. From there, you can call the t.end() method on that object to end the test, or use the t.plan() method to specify how many child tests or asserts the test will have.

If the function returns a Promise object (that is, an object with a then method), then when the promise is rejected or fulfilled, the test will be either ended or failed.

The options object is the same as would be passed to any assert, with two additional fields that are only relevant for child tests:

  • timeout: The number of ms to allow the test to run.
  • bail: Set to true to bail out on the first test failure.
  • autoend: Automatically end() the test on the next turn of the event loop after its internal queue is drained.


Run the supplied function when t.end() is called, or when the plan is met.

Note that when called on the root tap export, this also triggers autoend behavior.


Automatically end the test as soon as there is nothing pending in its queue.

The automatic end is deferred with a setTimeout, and any new action will cancel and re-schedule the timer. Nonetheless, calling this method means that any slow asynchronous behavior may be lost, if it comes after the end() is auto-triggered.

This behavior is triggered on the root tap object when tap.tearDown() is called.


Specify that a given number of tests are going to be run.

This may only be called before running any asserts or child tests.


Call when tests are done running. This is not necessary if t.plan() was used.


Pull the proverbial ejector seat.

Use this when things are severely broken, and cannot be reasonably handled. Immediately terminates the entire test run.


Return true if everything so far is ok.

Note that all assert methods also return true if they pass.


Print the supplied message as a TAP comment.

Note that you can always use console.error() for debugging (or console.log() as long as the message doesn't look like TAP formatted data)., extra)

Emit a failing test point. This method, and pass(), are the basic building blocks of all fancier assertions.


Emit a passing test point. This method, and fail(), are the basic building blocks of all fancier assertions.


Sets a pragma switch for a set of boolean keys in the argument.

The only pragma currently supported by the TAP parser is strict, which tells the parser to treat non-TAP output as a failure.


var t = require('tap')
console.log('this non-TAP output is ok')
t.pragma({ strict: true })
console.log('but this will cause a failure')


The Test object has a collection of assertion methods, many of which are given several synonyms for compatibility with other test runners and the vagaries of human expectations and spelling. When a synonym is multi-word in camelCase the corresponding lower case and snake_case versions are also created as synonyms.

All assertion methods take optional message and extra arguments as the last two params. The message is the name of the test. The extra argument can contain any arbitrary data about the test, but the following fields are "special".

  • todo Set to boolean true or a String to mark this as pending
  • skip Set to boolean true or a String to mark this as skipped
  • at Generated by the framework. The location where the assertion was called. Do not set this field.
  • stack Generated by the framework. The stack trace to the point where the assertion was called. Do not set this field.

t.ok(obj, message, extra)

Verifies that the object is truthy.

Synonyms: t.true, t.assert

t.notOk(obj, message, extra)

Verifies that the object is not truthy.

Synonyms: t.false, t.assertNot

t.error(obj, message, extra)

If the object is an error, then the assertion fails.

Note: if an error is encountered unexpectedly, it's often better to simply throw it. The Test object will handle this as a failure.

Synonyms: t.ifErr, t.ifError

t.throws(fn, [expectedError], message, extra)

Expect the function to throw an error. If an expected error is provided, then also verify that the thrown error matches the expected error.

If the function has a name, and the message is not provided, then the function name will be used as the message.

Caveat: if you pass a extra object to t.throws, then you MUST also pass in an expected error, or else it will read the diag object as the expected error, and get upset when your thrown error doesn't match {skip:true} or whatever.

For example, this will not work as expected:

t.throws(function() {throw new Error('x')}, { skip: true })

But this is fine:

// note the empty 'expected error' object.
// since it has no fields, it'll only verify that the thrown thing is
// an object, not the value of any properties
t.throws(function() {throw new Error('x')}, {}, { skip: true })

The expected error is tested against the throw error using t.match, so regular expressions and the like are fine. If the expected error is an Error object, then the stack field is ignored, since that will generally never match.

Synonyms: t.throw

t.doesNotThrow(fn, message, extra)

Verify that the provided function does not throw.

If the function has a name, and the message is not provided, then the function name will be used as the message.

Note: if an error is encountered unexpectedly, it's often better to simply throw it. The Test object will handle this as a failure.

Synonyms: t.notThrow

t.equal(found, wanted, message, extra)

Verify that the object found is exactly the same (that is, ===) to the object that is wanted.

Synonyms: t.equals, t.isEqual,, t.strictEqual, t.strictEquals, t.strictIs, t.isStrict, t.isStrictly

t.notEqual(found, notWanted, message, extra)

Inverse of t.equal().

Verify that the object found is not exactly the same (that is, !==) as the object that is wanted.

Synonyms: t.inequal, t.notEqual, t.notEquals, t.notStrictEqual, t.notStrictEquals, t.isNotEqual, t.isNot, t.doesNotEqual, t.isInequal

t.same(found, wanted, message, extra)

Verify that the found object is deeply equivalent to the wanted object. Use non-strict equality for scalars (ie, ==).

Synonyms: t.equivalent, t.looseEqual, t.looseEquals, t.deepEqual, t.deepEquals, t.isLoose, t.looseIs

t.notSame(found, notWanted, message, extra)

Inverse of t.same().

Verify that the found object is not deeply equivalent to the unwanted object. Uses non-strict inequality (ie, !=) for scalars.

Synonyms: t.inequivalent, t.looseInequal, t.notDeep, t.deepInequal, t.notLoose, t.looseNot

t.strictSame(found, wanted, message, extra)

Strict version of t.same().

Verify that the found object is deeply equivalent to the wanted object. Use strict equality for scalars (ie, ===).

Synonyms: t.strictEquivalent, t.strictDeepEqual, t.sameStrict, t.deepIs, t.isDeeply, t.isDeep, t.strictDeepEquals

t.strictNotSame(found, notWanted, message, extra)

Inverse of t.strictSame().

Verify that the found object is not deeply equivalent to the unwanted object. Use strict equality for scalars (ie, ===).

Synonyms: t.strictInequivalent, t.strictDeepInequal, t.notSameStrict, t.deepNot, t.notDeeply, t.strictDeepInequals, t.notStrictSame

t.match(found, pattern, message, extra)

Verify that the found object matches the pattern provided.

If pattern is a regular expression, and found is a string, then verify that the string matches the pattern.

If the pattern is a string, and found is a string, then verify that the pattern occurs within the string somewhere.

If pattern is an object, then verify that all of the (enumerable) fields in the pattern match the corresponding fields in the object using this same algorithm. For example, the pattern {x:/a[sdf]{3}/} would successfully match {x:'asdf',y:'z'}.

This is useful when you want to verify that an object has a certain set of required fields, but additional fields are ok.

Synonyms: t.has, t.hasFields, t.matches, t.similar,, t.isLike, t.includes, t.include, t.contains

t.notMatch(found, pattern, message, extra)

Interse of match()

Verify that the found object does not match the pattern provided.

Synonyms: t.dissimilar, t.unsimilar, t.notSimilar, t.unlike, t.isUnlike, t.notLike, t.isNotLike, t.doesNotHave, t.isNotSimilar, t.isDissimilar

t.type(object, type, message, extra)

Verify that the object is of the type provided.

Type can be a string that matches the typeof value of the object, or the string name of any constructor in the object's prototype chain, or a constructor function in the object's prototype chain.

For example, all the following will pass:

t.type(new Date(), 'object')
t.type(new Date(), 'Date')
t.type(new Date(), Date)

Synonyms: t.isa, t.isA

Advanced Usage

These methods are primarily for internal use, but can be handy in some unusual situations. If you find yourself using them frequently, you may be Doing It Wrong. However, if you find them useful, you should feel perfectly comfortable using them.

Please let us know if you frequently encounter situations requiring advanced usage, because this may indicate a shortcoming in the "non-advanced" API.


Parse standard input as if it was a child test named /dev/stdin.

This is primarily for use in the test runner, so that you can do some-tap-emitting-program | tap other-file.js - -Rnyan.

t.spawn(command, arguments, [options], [name], [extra])

Sometimes, instead of running a child test directly inline, you might want to run a TAP producting test as a child process, and treat its standard output as the TAP stream.

That's what this method does.

It is primarily used by the executable runner, to run all of the filename arguments provided on the command line.

The options object is passed to child_process.spawn, and can contain stuff like stdio directives and environment vars.

The extra arg is the same that would be passed to any assertion or child test, with the addition of the following fields:

  • bail: Set to true to bail out on the first failure. This is done by checking the output and then forcibly killing the process, but also sets the TAP_BAIL environment variable, which node-tap uses to set this field internally as well.
  • timeout: The number of ms to allow the child process to continue. If it goes beyond this time, the child process will be forcibly killed.

t.addAssert(name, length, fn)

This is used for creating assertion methods on the Test class.

It's a little bit advanced, but it's also super handy sometimes. All of the assert methods below are created using this function, and it can be used to create application-specific assertions in your tests.

The name is the method name that will be created. The length is the number of arguments the assertion operates on. (The message and extra arguments will alwasy be appended to the end.)

For example, you could have a file at test/setup.js that does the following:

var tap = require('tap')

// convenience
if (module === require.main) {

// Add an assertion that a string is in Title Case
// It takes one argument (the string to be tested)
tap.Test.prototype.addAssert('titleCase', 1, function (str, message, extra) {
  message = message || 'should be in Title Case'
  // the string in Title Case
  // A fancier implementation would avoid capitalizing little words
  // to get `Silence of the Lambs` instead of `Silence Of The Lambs`
  // But whatever, it's just an example.
  var tc = str.toLowerCase().replace(/\b./, function (match) {
    return match.toUpperCase()

  // should always return another assert call, or
  // this.pass(message) or, extra)
  return this.equal(str, tc, message, extra)

Then in your individual tests, you'd do this:

require('./setup.js') // adds the assert
var tap = require('tap')
tap.titleCase('This Passes')
tap.titleCase('however, tHis tOTaLLy faILS')


Call the end() method on all child tests, and then on this one.


Return the currently active test.

Something went wrong with that request. Please try again.