jsverify

Property-based testing for JavaScript.

JSVerify

Property based checking. Like QuickCheck.

Install the module with: npm install jsverify

var jsc = require("jsverify");
 
// forall (f : bool -> bool) (b : bool), f (f (f b)) = f(b). 
var boolFnAppliedThrice =
  jsc.forall("bool -> bool", "bool", function (fb) {
    return f(f(f(b))) === f(b);
  });
 
jsc.assert(boolFnAppliedThrice);
// OK, passed 100 tests 

Usage with mocha

Using jsverify with mocha is easy, just define the properties and use jsverify.assert.

Starting from version 0.4.3 you can write your specs without any boilerplate:

describe("sort", function () {
  jsc.property("idempotent", "array nat", function (arr) {
    return _.isEqual(sort(sort(arr)), sort(arr));
  });
});

You can also provide --jsverifyRngState state command line argument, to run tests with particular random generator state.

$ mocha examples/nat.js
 
1) natural numbers are less than 90:
 Error: Failed after 49 tests and 1 shrinks. rngState: 074e9b5f037a8c21d6; Counterexample: 90;
 
$ mocha examples/nat.js --grep 'are less than' --jsverifyRngState 074e9b5f037a8c21d6
 
1) natural numbers are less than 90:
   Error: Failed after 1 tests and 1 shrinks. rngState: 074e9b5f037a8c21d6; Counterexample: 90;

Errorneous case is found with first try.

Usage with jasmine

Check jasmineHelpers.js and jasmineHelpers2.js for jasmine 1.3 and 2.0 respectively.

Testing shows the presence, not the absence of bugs.

Edsger W. Dijkstra

To show that propositions hold, we need to construct proofs. There are two extremes: proof by example (unit tests) and formal (machine-checked) proof. Property-based testing is somewhere in between. We formulate propositions, invariants or other properties we believe to hold, but only test it to hold for numerous (randomly generated) values.

Types and function signatures are written in Coq/Haskell influented style: C# -style List<T> filter(List<T> v, Func<T, bool> predicate) is represented by filter (v : array T) (predicate : T -> bool) : array T in our style.

jsverify can operate with both synchronous and asynchronous-promise properties. Generally every property can be wrapped inside functor, for now in either identity or promise functor, for synchronous and promise properties respectively.

  • forall(arbs: arbitrary a ..., userenv: (map arbitrary)?, prop : a -> property): property

    Property constructor

  • check (prop: property, opts: checkoptions?): result

    Run random checks for given prop. If prop is promise based, result is also wrapped in promise.

    Options:

    • opts.tests - test count to run, default 100

    • opts.size - maximum size of generated values, default 5

    • opts.quiet - do not console.log

    • opts.rngState - state string for the rng

  • assert(prop: property, opts: checkoptions?) : void

    Same as check, but throw exception if property doesn't hold.

  • property(name: string, ...)

    Assuming there is globally defined it, the same as:

    it(name, function () {
      jsc.assert(jsc.forall(...));
    }

    You can use property to write facts too:

    jsc.property("+0 === -0", function () {
      return +0 === -0;
    });
  • compile(desc: string, env: typeEnv?): arbitrary a

    Compile the type describiption in provided type environment, or default one.

  • sampler(arb: arbitrary a, genSize: nat = 10): (sampleSize: nat?) -> a

    Create a sampler for a given arbitrary with an optional size. Handy when used in a REPL:

    > jsc = require('jsverify') // or require('./lib/jsverify') w/in the project
    ...
    > jsonSampler = jsc.sampler(jsc.json, 4)
    [Function]
    > jsonSampler()
    0.08467432763427496
    > jsonSampler()
    [ [ [] ] ]
    > jsonSampler()
    ''
    > sampledJson(2)
    [-0.4199344692751765, false]
  • generator a is a function (size: nat) -> a.
  • show is a function a -> string.
  • shrink is a function a -> [a], returning smaller values.
  • arbitrary a is a triple of generator, shrink and show functions.
    • { generator: nat -> a, shrink : a -> array a, show: a -> string }

There is a small DSL to help with forall. For example the two definitions below are equivalent:

var bool_fn_applied_thrice = jsc.forall("bool -> bool", "bool", check);
var bool_fn_applied_thrice = jsc.forall(jsc.fn(jsc.bool()), jsc.bool(), check);

The DSL is based on a subset of language recognized by typify-parser:

  • identifiers are fetched from the predefined environment.
  • applications are applied as one could expect: "array bool" is evaluated to jsc.array(jsc.bool).
  • functions are supported: "bool -> bool" is evaluated to jsc.fn(jsc.bool()).
  • square brackets are treated as a shorthand for the array type: "[nat]" is evaluated to jsc.array(jsc.nat).
  • .smap(f: a -> b, g: b -> a, newShow: (b -> string)?): arbitrary b

    Transform arbitrary a into arbitrary b. For example:

    g should be a right inverse of f.

    positiveIntegersArb = nat.smap(
      function (x) { return x + 1; },
      function (x) { return x - 1; });
  • bless(arb: {...}): arbitrary a

    Bless almost arbitrary structure to be proper arbitrary. Note: this function mutates argument.

    Example:

    var arbTokens = jsc.bless({
      generatorfunction () {
        switch (jsc.random(0, 2)) {
          case 0: return "foo";
          case 1: return "bar";
          case 2: return "quux";
        }
      }
    });
  • integer: arbitrary integer

  • integer(maxsize: nat): arbitrary integer

  • integer(minsize: integer, maxsize: integer): arbitrary integer

    Integers, ℤ

  • nat: arbitrary nat

  • nat(maxsize: nat): arbitrary nat

    Natural numbers, ℕ (0, 1, 2...)

  • number: arbitrary number

  • number(maxsize: number): arbitrary number

  • number(min: number, max: number): arbitrary number

    JavaScript numbers, "doubles", ℝ. NaN and Infinity are not included.

  • uint8: arbitrary nat

  • uint16: arbitrary nat

  • uint32: arbitrary nat

  • int8: arbitrary integer

  • int16: arbitrary integer

  • int32: arbitrary integer

  • bool: arbitrary bool

    Booleans, true or false.

  • datetime: arbitrary datetime

    Random datetime

  • elements(args: array a): arbitrary a

    Random element of args array.

  • falsy: arbitrary *

    Generates falsy values: false, null, undefined, "", 0, and NaN.

  • constant(x: a): arbitrary a

    Returns an unshrinkable arbitrary that yields the given object.

  • nonshrink(arb: arbitrary a): arbitrary a

    Non shrinkable version of arbitrary arb.

  • unit: arbitrary ()

  • either(arbA: arbitrary a, arbB : arbitrary b): arbitrary (either a b)

  • pair(arbA: arbitrary a, arbB : arbitrary b): arbitrary (pair a b)

    If not specified a and b are equal to value().

  • tuple(arbs: (arbitrary a, arbitrary b...)): arbitrary (a, b...)

  • dict(arb: arbitrary a): arbitrary (dict a)

    Generates a JavaScript object with properties of type A.

  • array(arb: arbitrary a): arbitrary (array a)

  • nearray(arb: arbitrary a): arbitrary (array a)

  • json: arbitrary json

    JavaScript Objects: boolean, number, string, array of json values or object with json values.

  • oneof(gs : array (arbitrary a)...) : arbitrary a

    Randomly uses one of the given arbitraries.

  • record(spec: { key: arbitrary a... }): arbitrary { key: a... }

    Generates a javascript object with given record spec.

  • fn(arb: arbitrary a): arbitrary (b -> a)

  • fun(arb: arbitrary a): arbitrary (b -> a)

A generator function, generator a, is a function (size: nat) -> a, which generates a value of given size.

Generator combinators are auto-curried:

var xs = generator.array(shrink.nat, 1); // ≡ 
var ys = generator.array(shrink.nat)(1);

In purely functional approach generator a would be explicitly stateful computation: (size: nat, rng: randomstate) -> (a, randomstate). JSVerify uses an implicit random number generator state, but the value generation is deterministic (tests reproduceable), if the primitives from random module are used.

  • generator.bless(f: nat -> a): generator a

    Bless function with .map and .flatmap properties.

  • .map(f: a -> b): generator b

    Map generator a into generator b. For example:

    positiveIntegersGenerator = nat.generator.map(
      function (x) { return x + 1; });
  • .flatmap(f: a -> generator b): generator b

    Monadic bind for generators.

  • generator.constant(x: a): generator a

  • generator.combine(gen: generator a..., f: a... -> b): generator b

  • generator.oneof(gens: list (generator a)): generator a

  • generator.recursive(genZ: generator a, genS: generator a -> generator a): generator a

  • generator.pair(genA: generator a, genB: generator b): generator (a, b)

  • generator.either(genA: generator a, genB: generator b): generator (either a b)

  • `generator.unit: generator ()

    unit is an empty tuple, i.e. empty array in JavaScript representation. This is useful as a building block.

  • generator.tuple(gens: (generator a, generator b...): generator (a, b...)

  • generator.array(gen: generator a): generator (array a)

  • generator.nearray(gen: generator a): generator (array a)

  • generator.dict(gen: generator a): generator (dict a)

A shrink function, shrink a, is a function a -> [a], returning an array of smaller values.

Shrink combinators are auto-curried:

var xs = shrink.array(shrink.nat, [1]); // ≡ 
var ys = shrink.array(shrink.nat)([1]);
  • shrink.bless(f: a -> [a]): shrink a

    Bless function with .smap property.

  • .smap(f: a -> b, g: b -> a): shrink b

    Transform shrink a into shrink b. For example:

    positiveIntegersShrink = nat.shrink.smap(
      function (x) { return x + 1; },
      function (x) { return x - 1; });
  • shrink.noop: shrink a

  • shrink.pair(shrA: shrink a, shrB: shrink b): shrink (a, b)

  • shrink.either(shrA: shrink a, shrB: shrink b): shrink (either a b)

  • shrink.tuple(shrs: (shrink a, shrink b...)): shrink (a, b...)

  • shrink.array(shr: shrink a): shrink (array a)

  • shrink.nearray(shr: shrink a): shrink (nearray a)

  • shrink.record(shrs: { key: shrink a... }): shrink { key: a... }

  • show.def(x : a): string

    Currently implemented as JSON.stringify.

  • show.pair(showA: a -> string, showB: b -> string, x: (a, b)): string

  • show.either(showA: a -> string, showB: b -> string, e: either a b): string

  • show.tuple(shrinks: (a -> string, b -> string...), x: (a, b...)): string

  • show.array(shrink: a -> string, x: array a): string

  • random(min: int, max: int): int

    Returns random int from [min, max] range inclusively.

    getRandomInt(2, 3) // either 2 or 3 
  • random.number(min: number, max: number): number

    Returns random number from [min, max) range.

  • either.left(value: a): either a b

  • either.right(value: b): either a b

  • either.either(l: a -> x, r: b -> x): x

  • `either.isEqual(other: either a b): bool

    TODO: add eq optional parameter

  • either.bimap(f: a -> c, g: b -> d): either c d

    either.bimap(compose(f, g), compose(h, i)) ≡ either.bimap(g, i).bimap(f, h);
  • either.first(f: a -> c): either c b

    either.first(f) ≡ either.bimap(f, utils.identity)
  • either.second(g: b -> d): either a d

    either.second(g) === either.bimap(utils.identity, g)

Utility functions are exposed (and documented) only to make contributions to jsverify more easy. The changes here don't follow semver, i.e. there might be backward-incompatible changes even in patch releases.

Use underscore.js, lodash, ramda, lazy.js or some other utility belt.

  • utils.isEqual(x: json, y: json): bool

    Equality test for json objects.

  • utils.force(x: a | () -> a) : a

    Evaluate x as nullary function, if it is one.

  • utils.merge(x... : obj): obj

    Merge two objects, a bit like _.extend({}, x, y).

  • README.md is generated from the source with ljs
  • jsverify.standalone.js is also generated by the build process
  • Before creating a pull request run make test, yet travis will do it for you.
  • 0.6.0-alpha.42015-04-26 — Fix issue #87

    • jsc.property didn't fail with asynchronous properties
    • thanks @Ezku for reporting
  • 0.6.0-alpha.32015-04-24 — promise shrink fixed

  • 0.6.0-alpha.22015-04-24 — jsc.bless

    • Added jsc.bless
  • 0.6.0-alpha.12015-04-22 — Preview

    • Using lazy sequences for shrink results
    • Breaking changes:
      • jsc.map renamed to jsc.dict
      • jsc.value removed, use jsc.json
      • jsc.string() removed, use jsc.string
      • shrink.isomap renamed to shrink.smap
  • 0.5.32015-04-21 — More algebra

    • unit and either arbitraries
    • arbitrary.smap to help creating compound data
  • 0.5.22015-04-10show.def -change

  • 0.5.12015-02-19 — Dependencies bump

    • We also work on 0.12 and iojs!
  • 0.5.02014-12-24 — Merry Chrismas 2014!

    • Documention cleanup
  • 0.5.0-beta.22014-12-21 — Beta 2!

    • Pair & tuple related code cleanup
    • Update CONTRIBUTING.md
    • Small documentation type fixes
    • Bless jsc.elements shrink
  • 0.5.0-beta.12014-12-20 — Beta!

    • bless don't close over (uses this)
    • Cleanup generator module
    • Other code cleanup here and there
  • 0.4.62014-11-30 — better shrinks & recursive

    • Implemented shrinks: #51
    • jsc.generator.recursive: #37
    • array, nearray & map generators return a bit smaller results (log2 of size)
  • 0.4.52014-11-22 — stuff

    • generator.combine & .flatmap
    • nat, integer, number & and string act as objects too
  • 0.4.42014-11-22 — new generators

    • New generators: nearray, nestring
    • generator.constant
    • zero-ary jsc.property (it ∘ assert)
    • jsc.sampler
  • 0.4.32014-11-08 — jsc.property

    • Now you can write your bdd specs without any boilerplate
    • support for nat-litearls in dsl #36
      describe("Math.abs", function () {
        jsc.property("result is non-negative", "integer 100", function (x) {
          return Math.abs(x) >= 0;
        });
      });
    • Falsy generator #42
  • 0.4.22014-11-03 — User environments for DSL

    • User environments for DSL
    • Generator prototype map, and shrink prototype isomap
    • JSON generator works with larger sizes
  • 0.4.1 Move to own organization in GitHub

  • 0.4.02014-10-27 — typify-dsl & more arbitraries. Changes from 0.3.6:

    • DSL for forall and suchthat
    • new primitive arbitraries
    • oneof behaves as in QuickCheck (BREAKING CHANGE)
    • elements is new name of old oneof
    • Other smaller stuff under the hood
  • 0.4.0-beta.4 generator.oneof

  • 0.4.0-beta.3 Expose shrink and show modules

  • 0.4.0-beta.2 Move everything around

    • Better looking README.md!
  • 0.4.0-beta.1 Beta!

    • Dev Dependencies update
  • 0.4.0-alpha8 oneof & record -dsl support

    • also jsc.compile
    • record is shrinkable!
  • 0.4.0-alpha7 oneof & record

    • oneof and record generator combinators (@fson)
    • Fixed uint* generators
    • Default test size increased to 10
    • Numeric generators with size specified are independent of test size (#20)
  • 0.4.0-alpha6 more primitives

    • int8, int16, int32, uint8, uint16, uint32
    • char, asciichar and asciistring
    • value → json
    • use eslint
  • 0.4.0-alpha5 move david to be devDependency

  • 0.4.0-alpha4 more typify

    • suchchat supports typify dsl
    • oneofelements to be in line with QuickCheck
    • Added versions of examples using typify dsl
  • 0.4.0-alpha3 David, npm-freeze and jscs

  • 0.4.0-alpha2 Fix typo in readme

  • 0.4.0-alpha1 typify

    • DSL for forall

      var bool_fn_applied_thrice = jsc.forall("bool -> bool", "bool", check);
    • generator arguments, which are functions are evaluated. One can now write:

      jsc.forall(jsc.nat, check) // previously had to be jsc.nat() 
  • 0.3.6 map generator

  • 0.3.5 Fix forgotten rngState in console output

  • 0.3.4 Dependencies update

  • 0.3.3 Dependencies update

  • 0.3.2 funfn

  • 0.3.1 Documentation typo fixes

  • 0.3.0 Major changes

    • random generate state handling
    • --jsverifyRngState parameter value used when run on node
    • karma tests
    • use make
    • dependencies update
  • 0.2.0 Use browserify

  • 0.1.4 Mocha test suite

    • major cleanup
  • 0.1.3 gen.show and exception catching

  • 0.1.2 Added jsc.assert

  • 0.1.1 Use grunt-literate

  • 0.1.0 Usable library

  • 0.0.2 Documented preview

  • 0.0.1 Initial preview

The MIT License (MIT)

Copyright (c) 2013, 2014 Oleg Grenrus

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.