About | Usage | CLI | API | Goals

Noiseless testing framework

• ~1100 LOC including the CLI runner¹
• terser and faster test code than with mocha, jasmine or tape
• test code reads like bullet points
• assertion code follows SVO structure in present tense for terseness and readability
• supports:
  • test grouping
  • assertions
  • spies
  • equals, notEquals, deepEquals and notDeepEquals assertion types
  • before/after/beforeEach/afterEach hooks
  • test exclusivity (i.e. .only)
  • async tests and hooks
• explicitly regulates test-space configuration to encourage focus on testing, and to provide uniform test suites across projects

Note: ¹ ospec is currently in the process of changing some of its API surface. The legacy and updated APIs are both implemented right now to ease the transition, once legacy code has been removed we'll clock around 800 LOC.

Both tests and assertions are declared via the o function. Tests should have a description and a body function. A test may have one or more assertions. Assertions should appear inside a test's body function and compare two values.

var o = require("ospec")

o("addition", function() {
    o(1 + 1).equals(2)
})
o("subtraction", function() {
    o(1 - 1).notEquals(2)
})

Assertions may have descriptions:

o("addition", function() {
    o(1 + 1).equals(2)("addition should work")

    /* in ES6, the following syntax is also possible
    o(1 + 1).equals(2) `addition should work`
    */
})
/* for a failing test, an assertion with a description outputs this:

addition should work

1 should equal 2

Error
  at stacktrace/goes/here.js:1:1
*/

Tests may be organized into logical groups using o.spec

o.spec("math", function() {
    o("addition", function() {
        o(1 + 1).equals(2)
    })
    o("subtraction", function() {
        o(1 - 1).notEquals(2)
    })
})

Group names appear as a breadcrumb trail in test descriptions: math > addition: 2 should equal 2

Groups can be nested to further organize test groups. Note that tests cannot be nested inside other tests.

o.spec("math", function() {
    o.spec("arithmetics", function() {
        o("addition", function() {
            o(1 + 1).equals(2)
        })
        o("subtraction", function() {
            o(1 - 1).notEquals(2)
        })
    })
})

The o.spy() method can be used to create a stub function that keeps track of its call count and received parameters

//code to be tested
function call(cb, arg) {cb(arg)}

//test suite
var o = require("ospec")

o.spec("call()", function() {
    o("works", function() {
        var spy = o.spy()
        call(spy, 1)

        o(spy.callCount).equals(1)
        o(spy.args[0]).equals(1)
        o(spy.calls[0]).deepEquals([1])
    })
})

A spy can also wrap other functions, like a decorator:

//code to be tested
var count = 0
function inc() {
    count++
}

//test suite
var o = require("ospec")

o.spec("call()", function() {
    o("works", function() {
        var spy = o.spy(inc)
        spy()

        o(count).equals(1)
    })
})

If a test body function declares a named argument, the test is assumed to be asynchronous, and the argument is a function that must be called exactly one time to signal that the test has completed. As a matter of convention, this argument is typically named done.

o("setTimeout calls callback", function(done) {
    setTimeout(done, 10)
})

Alternativly you can return a promise or even use an async function in tests:

o("promise test", function() {
    return new Promise(function(resolve) {
        setTimeout(resolve, 10)
    })
})

o("promise test", async function() {
    await someOtherAsyncFunction()
})

By default, asynchronous tests time out after 200ms. You can change that default for the current test suite and its children by using the o.specTimeout(delay) function.

o.spec("a spec that must timeout quickly", function() {
    // wait 20ms before bailing out of the tests of this suite and
    // its descendants
    o.specTimeout(20)
    o("some test", function(done) {
        setTimeout(done, 10) // this will pass
    })

    o.spec("a child suite where the delay also applies", function () {
        o("some test", function(done) {
            setTimeout(done, 30) // this will time out.
        })
    })
})
o.spec("a spec that uses the default delay", function() {
    // ...
})

This can also be changed on a per-test basis using the o.timeout(delay) function from within a test:

o("setTimeout calls callback", function(done) {
    o.timeout(500) //wait 500ms before bailing out of the test

    setTimeout(done, 300)
})

Note that the o.timeout function call must be the first statement in its test. It also works with Promise-returning tests:

o("promise test", function() {
    o.timeout(1000)
    return someOtherAsyncFunctionThatTakes900ms()
})

o("promise test", async function() {
    o.timeout(1000)
    await someOtherAsyncFunctionThatTakes900ms()
})

Asynchronous tests generate an assertion that succeeds upon calling done or fails on timeout with the error message async test timed out.

These hooks can be declared when it's necessary to setup and clean up state for a test or group of tests. The before and after hooks run once each per test group, whereas the beforeEach and afterEach hooks run for every test.

o.spec("math", function() {
    var acc
    o.beforeEach(function() {
        acc = 0
    })

    o("addition", function() {
        acc += 1

        o(acc).equals(1)
    })
    o("subtraction", function() {
        acc -= 1

        o(acc).equals(-1)
    })
})

It's strongly recommended to ensure that beforeEach hooks always overwrite all shared variables, and avoid if/else logic, memoization, undo routines inside beforeEach hooks.

Like tests, hooks can also be asynchronous. Tests that are affected by asynchronous hooks will wait for the hooks to complete before running.

o.spec("math", function() {
    var acc
    o.beforeEach(function(done) {
        setTimeout(function() {
            acc = 0
            done()
        })
    })

    //tests only run after async hooks complete
    o("addition", function() {
        acc += 1

        o(acc).equals(1)
    })
    o("subtraction", function() {
        acc -= 1

        o(acc).equals(-1)
    })
})

One or more tests can be temporarily made to run exclusively by calling o.only() instead of o. This is useful when troubleshooting regressions, to zero-in on a failing test, and to avoid saturating console log w/ irrelevant debug information.

o.spec("math", function() {
    // will not run
    o("addition", function() {
        o(1 + 1).equals(2)
    })

    // this test will be run, regardless of how many groups there are
    o.only("subtraction", function() {
        o(1 - 1).notEquals(2)
    })

    // will not run
    o("multiplication", function() {
        o(2 * 2).equals(4)
    })

    // this test will be run, regardless of how many groups there are
    o.only("division", function() {
        o(6 / 2).notEquals(2)
    })
})

//define a test
o("addition", function() {
    o(1 + 1).equals(2)
})

//run the suite
o.run()

The o.new() method can be used to create new instances of ospec, which can be run in parallel. Note that each instance will report independently, and there's no aggregation of results.

var _o = o.new('optional name')
_o("a test", function() {
    _o(1).equals(1)
})
_o.run()

Create a script in your package.json:

    "scripts": {
        "test": "ospec",
        ...
    }

...and run it from the command line:

npm test

NOTE: o.run() is automatically called by the cli - no need to call it in your test code.

Running ospec without arguments is equivalent to running ospec '**/tests/**/*.js'. In english, this tells ospec to evaluate all *.js files in any sub-folder named tests/ (the node_modules folder is always excluded).

If you wish to change this behavior, just provide one or more glob match patterns:

ospec '**/spec/**/*.js' '**/*.spec.js'

You can also provide ignore patterns (note: always add --ignore AFTER match patterns):

ospec --ignore 'folder1/**' 'folder2/**'

Finally, you may choose to load files or modules before any tests run (note: always add --preload AFTER match patterns):

ospec --preload esm

Here's an example of mixing them all together:

ospec '**/*.test.js' --ignore 'folder1/**' --preload esm ./my-file.js

For Node.js versions >= 13.2, ospec supports both ES6 modules and CommonJS packages out of the box. --preload esm is thus not needed in that case.

ospec comes with an executable named ospec. npm auto-installs local binaries to ./node_modules/.bin/. You can run ospec by running ./node_modules/.bin/ospec from your project root, but there are more convenient methods to do so that we will soon describe.

ospec doesn't work when installed globally (npm install -g). Using global scripts is generally a bad idea since you can end up with different, incompatible versions of the same package installed locally and globally.

Here are different ways of running ospec from the command line. This knowledge applies to not just ospec, but any locally installed npm binary.

If you're using a recent version of npm (v5+), you can use run npx ospec from your project folder.

If you're using a recent version of npm (v5+), you can use run npx ospec from your project folder.

Otherwise, to work around this limitation, you can use npm-run which enables one to run the binaries of locally installed packages.

npm install npm-run -g

Then, from a project that has ospec installed as a (dev) dependency:

npm-run ospec

If you understand how your system's PATH works (e.g. for OSX), then you can add the following to your PATH...

export PATH=./node_modules/.bin:$PATH

...and you'll be able to run ospec without npx, npm, etc. This one-time setup will also work with other binaries across all your node projects, as long as you run binaries from the root of your projects.

Square brackets denote optional arguments

Defines a group of tests. Groups are optional

Defines a test.

If an argument is defined for the assertions function, the test is deemed to be asynchronous, and the argument is required to be called exactly one time.

Starts an assertion. There are six types of assertion: equals, notEquals, deepEquals, notDeepEquals, throws, notThrows.

Assertions have this form:

o(actualValue).equals(expectedValue)

As a matter of convention, the actual value should be the first argument and the expected value should be the second argument in an assertion.

Assertions can also accept an optional description curried parameter:

o(actualValue).equals(expectedValue)("this is a description for this assertion")

Assertion descriptions can be simplified using ES6 tagged template string syntax:

o(actualValue).equals(expectedValue) `this is a description for this assertion`

Asserts that two values are strictly equal (===)

Asserts that two values are strictly not equal (!==)

Asserts that two values are recursively equal

Asserts that two values are not recursively equal

Asserts that a function throws an instance of the provided constructo

Asserts that a function throws an Error with the provided message

Asserts that a function does not throw an instance of the provided constructor

Asserts that a function does not throw an Error with the provided message

Defines code to be run at the beginning of a test group

If an argument is defined for the setup function, this hook is deemed to be asynchronous, and the argument is required to be called exactly one time.

Defines code to be run at the end of a test group

If an argument is defined for the teardown function, this hook is deemed to be asynchronous, and the argument is required to be called exactly one time.

Defines code to be run before each test in a group

If an argument is defined for the setup function, this hook is deemed to be asynchronous, and the argument is required to be called exactly one time.

Defines code to be run after each test in a group

If an argument is defined for the teardown function, this hook is deemed to be asynchronous, and the argument is required to be called exactly one time.

Declares that only a single test should be run, instead of all of them

Returns a function that records the number of times it gets called, and its arguments

The number of times the function has been called

The arguments that were passed to the function in the last time it was called

Runs the test suite. By default passing test results are printed using console.log and failing test results are printed using console.error.

If you have custom continuous integration needs then you can use a reporter to process test result data yourself.

If running in Node.js, ospec will call process.exit after reporting results by default. If you specify a reporter, ospec will not do this and allow your reporter to respond to results in its own way.

The default reporter used by o.run() when none are provided. Returns the number of failures, doesn't exit Node.js by itself. It expects an array of test result data as argument.

Returns a new instance of ospec. Useful if you want to run more than one test suite concurrently

var $o = o.new()
$o("a test", function() {
    $o(1).equals(1)
})
$o.run()

When an error is thrown some tests may be skipped. See the "run time model" for a detailed description of the bailout mechanism.

Test results are available by reference for integration purposes. You can use custom reporters in o.run() to process these results.

o.run(function(results) {
    // results is an array

    results.forEach(function(result) {
        // ...
    })
})

• true if the assertion passed.
• false if the assertion failed.
• null if the assertion was incomplete (o("partial assertion) // and that's it).

The Error object explaining the reason behind a failure. If the assertion failed, the stack will point to the actuall error. If the assertion did pass or was incomplete, this field is identical to result.testError.

An Error object whose stack points to the test definition that wraps the assertion. Useful as a fallback because in some async cases the main may not point to test code.

If an exception was thrown inside the corresponding test, this will equal that Error's message. Otherwise, this will be a preformatted message in SVO form. More specifically, ${subject}\n${verb}\n${object}.

As an example, the following test's result message will be "false\nshould equal\ntrue".

o.spec("message", function() {
    o(false).equals(true)
})

If you specify an assertion description, that description will appear two lines above the subject.

o.spec("message", function() {
    o(false).equals(true)("Candyland") // result.message === "Candyland\n\nfalse\nshould equal\ntrue"
})

A >-separated string showing the structure of the test specification. In the below example, result.context would be testing > rocks.

o.spec("testing", function() {
    o.spec("rocks", function() {
        o(false).equals(true)
    })
})

• A test is the function passed to o("description", function test() {}).
• A hook is a function passed to o.before(), o.after(). o.beforeEach() and o.afterEach().
• A task designates either a test or a hook.
• A given test and its associated beforeEach and afterEach hooks form a streak. The beforeEach hooks run outermost first, the afterEach run outermost last. The hooks are optional, and are tied at test-definition time in the o.spec() calls that enclose the test.
• A spec is a collection of streaks, specs, one before hook and one after hook. Each component is optional. Specs are defined with the o.spec("spec name", function specDef() {}) calls.

For a given instance, an ospec run goes through three phases:

1. tests definition
2. tests execution and results accumulation
3. results presentation

This phase is synchronous. o.spec("spec name", function specDef() {}), o("test name", function test() {}) and hooks calls generate a tree of specs and tests.

At test execution time, for each spec, the before hook is called if present, then nested specs the streak of each test, in definition order, then the after hook, if present.

Test and hooks may contain assertions, which will populate the results array.

Once all tests have run or timed out, the results are presented.

While some testing libraries consider error thrown as assertions failure, ospec treats them as super-failures. Throwing will cause the current spec to be aborted, avoiding what can otherwise end up as pages of errors. What this means depends on when the error is thrown. Specifically:

• A syntax error in a file causes the file to be ignored by the runner.
• At test-definition time:
  • An error thrown at the root of a file will cause subsequent tests and specs to be ignored.
  • An error thrown in a spec definition will cause the spec to be ignored.
• At test-execution time:
  • An error thrown in the before hook will cause the streaks and nested specs to be ignored. The after hook will run.
  • An error thrown in a task...
    • ...prevents further streaks and nested specs in the current spec from running. The after hook of the spec will run.
    • ...if thrown in a beforeEach hook of a streak, causes the streak to be hollowed out. Hooks defined in nested scopes and the actual test will not run. However, the afterEach hook corresponding to the one that crashed will run, as will those defined in outer scopes.

For every error thrown, a "bail out" failure is reported.

Ospec started as a bare bones test runner optimized for Leo Horie to write Mithril v1 with as little hasle as possible. It has since grown in capabilities and polish, and while we tried to keep some of the original spirit, the current incarnation is not as radically minimalist as the original. The state of the art in testing has also moved with the dominance of Jest over Jasmine and Mocha, and now Vitest coming up the horizon.

• Do the most common things that the mocha/chai/sinon triad does without having to install 3 different libraries and several dozen dependencies
• Limit configuration in test-space:
  • Disallow ability to pick between API styles (BDD/TDD/Qunit, assert/should/expect, etc)
  • No "magic" plugin system with global reach. Custom assertions need to be imported or defined lexically (e.g. in o(value)._(matches(refence)), matches can be resolved in file).
  • Provide a default simple reporter
• Make assertion code terse, readable and self-descriptive
• Have as few assertion types as possible for a workable usage pattern

These restrictions have a few benefits:

• tests always look the same, even across different projects and teams
• single source of documentation for entire testing API
• no need to hunt down plugins to figure out what they do, especially if they replace common javascript idioms with fuzzy spoken language constructs (e.g. what does .is() do?)
• no need to pollute project-space with ad-hoc configuration code
• discourages side-tracking and yak-shaving