Implementation of the consumer driven contract library Pact for Javascript.

From the Pact website:

The Pact family of frameworks provide support for Consumer Driven Contracts testing.

A Contract is a collection of agreements between a client (Consumer) and an API (Provider) that describes the interactions that can take place between them.

Consumer Driven Contracts is a pattern that drives the development of the Provider from its Consumers point of view.

Pact is a testing tool that guarantees those Contracts are satisfied.

Read Getting started with Pact for more information for beginners.

• Pact JS
  • Installation
    • Do Not Track
  • Which Library/Package should I use?
  • Using Pact JS
  • HTTP API Testing
    • Consumer Side Testing
      • API
      • Example
    • Provider API Testing
      • Verification Options
      • API with Provider States
      • Before and After Hooks
      • Pending Pacts
      • WIP Pacts
      • Verifying multiple contracts with the same tag (e.g. for Mobile use cases)
      • Modify Requests Prior to Verification (Request Filters)
      • Lifecycle of a provider verification
    • Publishing Pacts to a Broker
      • Pact publishing options
      • Publishing Verification Results to a Pact Broker
  • Asynchronous API Testing
    • Consumer
    • Provider (Producer)
    • Pact Broker Integration
  • Matching
    • Match common formats
    • Match based on type
    • Match based on arrays
    • Match by regular expression
  • GraphQL API
  • Tutorial (60 minutes)
  • Examples
    • HTTP APIs
    • Asynchronous APIs
  • Using Pact in non-Node environments
    • Using Pact with Karma
    • Using Pact with RequireJS
  • Pact JS V3
    • Using the V3 matching rules
    • Using Pact with XML
    • Verifying providers with VerifierV3
      • Request Filters
      • Provider state callbacks
  • Troubleshooting / FAQs
    • Alpine Docker
    • Parallel tests
    • Splitting tests across multiple files
    • Test fails when it should pass
    • Test intermittent failures
    • Re-run specific verification failures
    • Timeout
    • Usage with Jest
    • Usage with Angular
    • Debugging
  • Contributing
  • Contact

npm i -S @pact-foundation/pact@latest

In order to get better statistics as to who is using Pact, we have an anonymous tracking event that triggers when Pact installs for the first time. The only things we track are your type of OS, and the version information for the package being installed. No PII data is sent as part of this request. To respect your privacy, you can disable tracking by simply adding a 'do not track' flag within your package.json file or setting the environment variable PACT_DO_NOT_TRACK=1:

{
	"name": "some-project",
	...
	"config": {
		"pact_do_not_track": true
	},
	...
}

See the Changelog for versions and their history.

TL;DR - you almost always want Pact JS.

* The "I need to run it in the browser" question comes up occasionally. The question is this - for your JS code to be able to make a call to another API, is this dependent on browser-specific code? In most cases, people use tools like React/Angular which have libraries that work on the server and client side, in which case, these tests don't need to run in a browser and could instead be executed in a Node.js environment.

Pact supports synchronous request-response style HTTP interactions and asynchronous interactions with JSON-formatted payloads.

To use the library on your tests, add the pact dependency:

const { Pact } = require("@pact-foundation/pact")

The Pact class provides the following high-level APIs, they are listed in the order in which they typically get called in the lifecycle of testing a consumer:

The first step is to create a test for your API Consumer. The example below uses Mocha, and demonstrates the basic approach:

1. Create the Pact object
2. Start the Mock Provider that will stand in for your actual Provider
3. Add the interactions you expect your consumer code to make when executing the tests
4. Write your tests - the important thing here is that you test the outbound collaborating function which calls the Provider, and not just issue raw http requests to the Provider. This ensures you are testing your actual running code, just like you would in any other unit test, and that the tests will always remain up to date with what your consumer is doing.
5. Validate the expected interactions were made between your consumer and the Mock Service
6. Generate the pact(s)

Check out the examples folder for examples with Karma Jasmine, Mocha and Jest. The example below is taken from the integration spec.

const path = require("path")
const chai = require("chai")
const { Pact } = require("@pact-foundation/pact")
const chaiAsPromised = require("chai-as-promised")
const expect = chai.expect

chai.use(chaiAsPromised)

describe("Pact", () => {
  // (1) Create the Pact object to represent your provider
  const provider = new Pact({
    consumer: "TodoApp",
    provider: "TodoService",
    port: 1234,
    log: path.resolve(process.cwd(), "logs", "pact.log"),
    dir: path.resolve(process.cwd(), "pacts"),
    logLevel: "INFO",
  })

  // this is the response you expect from your Provider
  const EXPECTED_BODY = [
    {
      id: 1,
      name: "Project 1",
      due: "2016-02-11T09:46:56.023Z",
      tasks: [
        { id: 1, name: "Do the laundry", done: true },
        { id: 2, name: "Do the dishes", done: false },
        { id: 3, name: "Do the backyard", done: false },
        { id: 4, name: "Do nothing", done: false },
      ],
    },
  ]

  const todoApp = new TodoApp()

  context("when there are a list of projects", () => {
    describe("and there is a valid user session", () => {
      before(() =>
        provider
          // (2) Start the mock server
          .setup()
          // (3) add interactions to the Mock Server, as many as required
          .then(() =>
            provider.addInteraction({
              // The 'state' field specifies a "Provider State"
              state: "i have a list of projects",
              uponReceiving: "a request for projects",
              withRequest: {
                method: "GET",
                path: "/projects",
                headers: { Accept: "application/json" },
              },
              willRespondWith: {
                status: 200,
                headers: { "Content-Type": "application/json" },
                body: EXPECTED_BODY,
              },
            })
          )
      )
    })

    // (4) write your test(s)
    it("generates a list of TODOs for the main screen", async () => {
      const projects = await todoApp.getProjects() // <- this method would make the remote http call
      expect(projects).to.be.a("array")
      expect(projects).to.have.deep.property("projects[0].id", 1)
    })

    // (5) validate the interactions you've registered and expected occurred
    // this will throw an error if it fails telling you what went wrong
    // This should be performed once per interaction test
    afterEach(() => provider.verify())
  })

  // (6) write the pact file for this consumer-provider pair,
  // and shutdown the associated mock server.
  // You should do this only _once_ per Provider you are testing,
  // and after _all_ tests have run for that suite
  after(() => provider.finalize())
})

1. Start your local Provider service.
2. Optionally, instrument your API with ability to configure provider states
3. Then run the Provider side verification step

const { Verifier } = require('@pact-foundation/pact');
let opts = {
  ...
};

new Verifier(opts).verifyProvider().then(function () {
	// do something
});

To dynamically retrieve pacts from a Pact Broker for a provider, provide the broker URL, the name of the provider, and the consumer version tags that you want to verify:

let opts = {
  pactBroker: "http://my-broker",
  provider: "Animal Profile Service",
  consumerVersionTag: ["master", "prod"],
}

To verify a pact at a specific URL (eg. when running a pact verification triggered by a 'contract content changed' webhook, or when verifying a pact from your local machine, or a network location that's not the Pact Broker, set just the pactUrls, eg:

let opts = {
  pactUrls: [process.env.PACT_URL],
}

To publish the verification results back to the Pact Broker, you need to enable the 'publish' flag, set the provider version and optional provider version tags:

let opts = {
  publishVerificationResult: true, //generally you'd do something like `process.env.CI === 'true'`
  providerVersion: "version", //recommended to be the git sha
  providerVersionTag: "tag", //optional, recommended to be the git branch
}

If your broker has a self signed certificate, set the environment variable SSL_CERT_FILE (or SSL_CERT_DIR) pointing to a copy of your certificate.

Read more about Verifying Pacts.

If you have defined any states in your consumer tests, the Verifier can put the provider into the right state prior to sending the request. For example, the provider can use the state to mock away certain database queries. To support this, set up a handler for each state using hooks on the stateHandlers property. Here is an example from our e2e suite:

const opts = {
  ...
  stateHandlers: {
    [null]: () => {
      // This is the "default" state handler, when no state is given
    }
    "Has no animals": () => {
      animalRepository.clear()
      return Promise.resolve(`Animals removed from the db`)
    },
    "Has some animals": () => {
      importData()
      return Promise.resolve(`Animals added to the db`)
    },
    "Has an animal with ID 1": () => {
      importData()
      return Promise.resolve(`Animals added to the db`)
    }
  }
}

return new Verifier(opts).verifyProvider().then(...)

As you can see, for each state ("Has no animals", ...), we configure the local datastore differently. If this option is not configured, the Verifier will ignore the provider states defined in the pact and log a warning.

Read more about Provider States.

Sometimes, it's useful to be able to do things before or after a test has run, such as reset a database, log a metric etc. A beforeEach hook runs on each verification before any other part of the Pact test lifecycle, and a afterEach hook runs as the last step before returning the verification result back to the test.

You can add them to your verification options as follows:

const opts = {
  ...
  beforeEach: () => {
    console.log('I run before everything else')
  },

  afterEach: () => {
    console.log('I run after everything else has finished')
  }
}

If the hook errors, the test will fail. See the lifecycle of an interaction below.

NOTE: This feature is available on Pactflow by default, and requires configuration if using a self-hosted broker.

Pending pacts is a feature that allows consumers to publish new contracts or changes to existing contracts without breaking Provider's builds. It does so by flagging the contract as "unverified" in the Pact Broker the first time a contract is published. A Provider can then enable a behaviour (via enablePending: true) that will still perform a verification (and thus share the results back to the broker) but not fail the verification step itself.

This enables safe introduction of new contracts into the system, without breaking Provider builds, whilst still providing feedback to Consumers as per before.

See the docs and this article for more background.

NOTE: This feature is available on Pactflow by default, and requires configuration if using a self-hosted broker.

WIP Pacts builds upon pending pacts, enabling provider tests to pull in any contracts applicable to the provider regardless of the tag it was given. This is useful, because often times consumers won't follow the exact same tagging convention and so their workflow would be interrupted. This feature enables any pacts determined to be "work in progress" to be verified by the Provider, without causing a build failure. You can enable this behaviour by specifying a valid timestamp for includeWipPactsSince. This sets the start window for which new WIP pacts will be pulled down for verification, regardless of the tag.

See the docs and this article for more background.

Tags may be used to indicate a particular version of an application has been deployed to an environment - e.g. prod, and are critical in configuring can-i-deploy checks for CI/CD pipelines. In the majority of cases, only one version of an application is deployed to an environment at a time. For example, an API and a Website are usually deployed in replacement of an existing system, and any transition period is quite short lived.

Mobile is an exception to this rule - it is common to have multiple versions of an application that are in "production" simultaneously. To support this workflow, we have a feature known as consumer version selectors. Using selectors, we can verify that all pacts with a given tag should be verified. The following selectors ask the broker to "find all pacts with tag 'prod' and the latest pact for 'master'":

consumerVersionSelectors: [
  {
    tag: "prod",
    all: true,
  },
  {
    tag: "master",
    latest: true,
  },
]

NOTE: Using the all flag requires you to ensure you delete any tags associated with application versions that are no longer in production (e.g. if decommissioned from the app store)

Sometimes you may need to add things to the requests that can't be persisted in a pact file. Examples of these are authentication tokens with a small life span. e.g. an OAuth bearer token: Authorization: Bearer 0b79bab50daca910b000d4f1a2b675d604257e42.

For these cases, we have two facilities that should be carefully used during verification:

1. the ability to specify custom headers to be sent during provider verification. The flag to achieve this is customProviderHeaders.
2. the ability to modify a request/response and modify the payload. The flag to achieve this is requestFilter.

Example API with Authorization

For example, to have an Authorization bearer token header sent as part of the verification request, set the verifyProvider options as per below:

let token
let opts = {
  provider: 'Animal Profile Service',
  ...
  stateHandlers: {
    "is authenticated": () => {
      token = "1234"
      Promise.resolve(`Valid bearer token generated`)
    },
    "is not authenticated": () => {
      token = ""
      Promise.resolve(`Expired bearer token generated`)
    }
  },

  // this middleware is executed for each request, allowing `token` to change between invocations
  // it is common to pair this with `stateHandlers` as per above, that can set/expire the token
  // for different test cases
  requestFilter: (req, res, next) => {
    req.headers["Authorization"] = `Bearer: ${token}`
    next()
  },

  // This header will always be sent for each and every request, and can't be dynamic
  // (i.e. passing a variable instead of the bearer token)
  customProviderHeaders: ["Authorization: Bearer 1234"]
}

return new Verifier(opts).verifyProvider().then(...)

As you can see, this is your opportunity to modify\add to headers being sent to the Provider API, for example to create a valid time-bound token.

Important Note: You should only use this feature for things that can not be persisted in the pact file. By modifying the request, you are potentially modifying the contract from the consumer tests!

For each interaction in a pact file, the order of execution is as follows:

BeforeEach -> State Handler -> Request Filter (request phase) -> Execute Provider Test -> Request Filter (response phase) -> AfterEach

If any of the middleware or hooks fail, the tests will also fail.

Sharing is caring - to simplify sharing Pacts between Consumers and Providers, we have created the Pact Broker.

The Broker:

• versions your contracts
• tells you which versions of your applications can be deployed safely together
• allows you to deploy your services independently
• provides API documentation of your applications that is guaranteed to be up-to date
• visualises the relationships between your services
• integrates with other systems, such as Slack or your CI server, via webhooks
• ...and much much more.

Host your own, or signup for a free hosted Pact Broker.

const { Publisher } = require("@pact-foundation/pact")
const opts = {
   ...
};

new Publisher(opts)
  .publishPacts()
  .then(() => {
    // ...
  })

If your broker has a self signed certificate, set the environment variable SSL_CERT_FILE (or SSL_CERT_DIR) pointing to a copy of your certificate.

If you're using a Pact Broker (e.g. a hosted one at https://pactflow.io), you can publish your verification results so that consumers can query if they are safe to release.

It looks like this:

To publish the verification results back to the Pact Broker, you need to enable the 'publish' flag, set the provider version and optional provider version tags:

let opts = {
  publishVerificationResult: true, //recommended to only publish from CI by setting the value to `process.env.CI === 'true'`
  providerVersion: "version", //recommended to be the git sha eg. process.env.MY_CI_COMMIT
  providerVersionTag: "tag", //optional, recommended to be the git branch eg. process.env.MY_CI_BRANCH
}

Since version v6.0.0 or later

Modern distributed architectures are increasingly integrated in a decoupled, asynchronous fashion. Message queues such as ActiveMQ, RabbitMQ, SQS, Kafka and Kinesis are common, often integrated via small and frequent numbers of microservices (e.g. lambda.).

Furthermore, the web has things like WebSockets which involve bidirectional messaging.

Pact supports these use cases, by abstracting away the protocol and focussing on the messages passing between them.

For further reading and introduction into this topic, see this article and our asynchronous examples for a more detailed overview of these concepts.

A Consumer is the system that will be reading a message from a queue or some other intermediary - like a DynamoDB table or S3 bucket - and be able to handle it.

From a Pact testing point of view, Pact takes the place of the intermediary (MQ/broker etc.) and confirms whether or not the consumer is able to handle a request.

The following test creates a contract for a Dog API handler:

const {
  MessageConsumerPact,
  synchronousBodyHandler,
} = require("@pact-foundation/pact")

// 1 Dog API Handler
const dogApiHandler = function(dog) {
  if (!dog.id && !dog.name && !dog.type) {
    throw new Error("missing fields")
  }

  // do some other things to dog...
  // e.g. dogRepository.save(dog)
  return
}

// 2 Pact Message Consumer
const messagePact = new MessageConsumerPact({
  consumer: "MyJSMessageConsumer",
  dir: path.resolve(process.cwd(), "pacts"),
  pactfileWriteMode: "update",
  provider: "MyJSMessageProvider",
})

describe("receive dog event", () => {
  it("accepts a valid dog", () => {
    // 3 Consumer expectations
    return (
      messagePact
        .given("some state")
        .expectsToReceive("a request for a dog")
        .withContent({
          id: like(1),
          name: like("rover"),
          type: term({ generate: "bulldog", matcher: "^(bulldog|sheepdog)$" }),
        })
        .withMetadata({
          "content-type": "application/json",
        })

        // 4 Verify consumers' ability to handle messages
        .verify(synchronousBodyHandler(dogApiHandler))
    )
  })
})

Explanation:

1. The Dog API - a contrived API handler example. Expects a dog object and throws an Error if it can't handle it.
   • In most applications, some form of transactionality exists and communication with a MQ/broker happens.
   • It's important we separate out the protocol bits from the message handling bits, so that we can test that in isolation.
2. Creates the MessageConsumer class
3. Setup the expectations for the consumer - here we expect a dog object with three fields
4. Pact will send the message to your message handler. If the handler returns a successful promise, the message is saved, otherwise the test fails. There are a few key things to consider:
   • The actual request body that Pact will send, will be contained within a Message object along with other context, so the body must be retrieved via content attribute.
   • All handlers to be tested must be of the shape (m: Message) => Promise<any> - that is, they must accept a Message and return a Promise. This is how we get around all of the various protocols, and will often require a lightweight adapter function to convert it.
   • In this case, we wrap the actual dogApiHandler with a convenience function synchronousBodyHandler provided by Pact, which Promisifies the handler and extracts the contents.

A Provider (Producer in messaging parlance) is the system that will be putting a message onto the queue.

As per the Consumer case, Pact takes the position of the intermediary (MQ/broker) and checks to see whether or not the Provider sends a message that matches the Consumer's expectations.

const { MessageProviderPact } = require("@pact-foundation/pact")

// 1 Messaging integration client
const dogApiClient = {
  createDog: () => {
    return new Promise((resolve, reject) => {
      resolve({
        id: 1,
        name: "fido",
        type: "bulldog",
      })
    })
  },
}

describe("Message provider tests", () => {
  // 2 Pact setup
  const p = new MessageProviderPact({
    messageProviders: {
      "a request for a dog": () => dogApiClient.createDog(),
    },
    provider: "MyJSMessageProvider",
    providerVersion: "1.0.0",
    pactUrls: [
      path.resolve(
        process.cwd(),
        "pacts",
        "myjsmessageconsumer-myjsmessageprovider.json"
      ),
    ],
  })

  // 3 Verify the interactions
  describe("Dog API Client", () => {
    it("sends some dogs", () => {
      return p.verify()
    })
  })
})

Explanation:

1. Our API client contains a single function createDog which is responsible for generating the message that will be sent to the consumer via some message queue
2. We configure Pact to stand-in for the queue. The most important bit here is the messageProviders block
   • Similar to the Consumer tests, we map the various interactions that are going to be verified as denoted by their description field. In this case, a request for a dog, maps to the createDog handler. Notice how this matches the original Consumer test.
3. We can now run the verification process. Pact will read all of the interactions specified by its consumer, and invoke each function that is responsible for generating that message.

As per HTTP APIs, you can publish contracts and verification results to a Broker.

Matching makes your tests more expressive making your tests less brittle.

Rather than use hard-coded values which must then be present on the Provider side, you can use regular expressions and type matches on objects and arrays to validate the structure of your APIs.

NOTE: Make sure to start the mock service via the Pact declaration with the option specification: 2 to get access to these features.

Often times, you find yourself having to re-write regular expressions for common formats. We've created a number of them for you to save you the time:

const { like, string } = Matchers

provider.addInteraction({
  state: "Has some animals",
  uponReceiving: "a request for an animal",
  withRequest: {
    method: "GET",
    path: "/animals/1",
  },
  willRespondWith: {
    status: 200,
    headers: {
      "Content-Type": "application/json; charset=utf-8",
    },
    body: {
      id: 1,
      name: string("Billy"),
      address: like({
        street: "123 Smith St",
        suburb: "Smithsville",
        postcode: 7777,
      }),
    },
  },
})

Note that you can wrap a like around a single value or an object. When wrapped around an object, all values and child object values will be matched according to types, unless overridden by something more specific like a term.

Matching provides the ability to specify flexible length arrays. For example:

pact.eachLike(obj, { min: 3 })

Where obj can be any javascript object, value or Pact.Match. It takes optional argument ({ min: 3 }) where min is greater than 0 and defaults to 1 if not provided.

Below is an example that uses all of the Pact Matchers.

const { somethingLike: like, term, eachLike } = pact

const animalBodyExpectation = {
  id: 1,
  first_name: "Billy",
  last_name: "Goat",
  animal: "goat",
  age: 21,
  gender: term({
    matcher: "F|M",
    generate: "M",
  }),
  location: {
    description: "Melbourne Zoo",
    country: "Australia",
    post_code: 3000,
  },
  eligibility: {
    available: true,
    previously_married: false,
  },
  children: eachLike({ name: "Sally", age: 2 }),
}

// Define animal list payload, reusing existing object matcher
// Note that using eachLike ensure that all values are matched by type
const animalListExpectation = eachLike(animalBodyExpectation, {
  min: MIN_ANIMALS,
})

provider.addInteraction({
  state: "Has some animals",
  uponReceiving: "a request for all animals",
  withRequest: {
    method: "GET",
    path: "/animals/available",
  },
  willRespondWith: {
    status: 200,
    headers: {
      "Content-Type": "application/json; charset=utf-8",
    },
    body: animalListExpectation,
  },
})

If none of the above matchers or formats work, you can write your own regex matcher.

The underlying mock service is written in Ruby, so the regular expression must be in a Ruby format, not a Javascript format.

const { term } = pact

provider.addInteraction({
  state: "Has some animals",
  uponReceiving: "a request for an animal",
  withRequest: {
    method: "GET",
    path: "/animals/1",
  },
  willRespondWith: {
    status: 200,
    headers: {
      "Content-Type": "application/json; charset=utf-8",
    },
    body: {
      id: 100,
      name: "billy",
      gender: term({
        matcher: "F|M",
        generate: "F",
      }),
    },
  },
})

GraphQL is simply an abstraction over HTTP and may be tested via Pact. There are two wrapper APIs available for GraphQL specific testing: GraphQLInteraction and ApolloGraphQLInteraction.

These are both lightweight wrappers over the standard DSL in order to make GraphQL testing a bit nicer.

See the history, and below for an example.

Learn everything in Pact JS in 60 minutes: https://github.com/pact-foundation/pact-workshop-js.

The workshop takes you through all of the key concepts using a React consumer and an Express API.

• Complete Example (Node env)
• Pact with AVA (Node env)
• Pact with Jest (Node env)
• Pact with TypeScript Mocha
• Pact with Mocha
• Pact with GraphQL
• Pact with Karma Jasmine
• Pact with Karma Mocha
• Pact with React Jest

• Asynchronous messages
• Serverless

Pact requires a Node runtime to be able to start and stop Mock servers, write logs and other things.

However, when used within browser or non-Node based environments - such as with Karma or ng-test - this is not possible.

To address this challenge, we have released a separate 'web' based module for this purpose - pact-web. Whilst it still provides a testing DSL, it cannot start and stop mock servers as per the pact package, so you will need to coordinate this yourself prior to and after executing any tests.

To get started, install pact-web and Pact Node:

npm install --save-dev @pact-foundation/pact-web @pact-foundation/pact-node

If you're not using Karma, you can start and stop the mock server using Pact Node or something like Grunt Pact.

We have create a plugin for Karma, which will automatically start and stop any Mock Server for your Pact tests.

Modify your karma.conf.js file as per below to get started:

    // Load pact framework - this will start/stop mock server automatically
    frameworks: ['pact'],

    // Load the pact and default karma plugins
    plugins: [
      'karma-*',
      '@pact-foundation/karma-pact'
    ],

    // load pact web module
    files: [
      'node_modules/@pact-foundation/pact-web/pact-web.js',
      ...
    ],

    // Configure the mock service
    pact: [{
      port: 1234,
      consumer: 'KarmaMochaConsumer',
      provider: 'KarmaMochaProvider',
      logLevel: 'DEBUG',
      log: path.resolve(process.cwd(), 'logs', 'pact.log'),
      dir: path.resolve(process.cwd(), 'pacts')
    }],

Check out the Examples for how to use the Karma interface.

The module name should be "Pact" - not "pact-js". An example config with a karma test might look like the following:

In client-spec.js change the define to:

define(['client', 'Pact'], function (example, Pact) {

In test-main.js:

require.config({
  baseUrl: "/base",
  paths: {
    Pact: "node_modules/pact-web/pact-web",
    client: "js/client",
  },
  deps: allTestFiles,
  callback: window.__karma__.start,
})

See this Stack Overflow question for background, and this gist with a working example.

An initial alpha version of Pact-JS with support for V3 specification features and XML matching has been released. Current support is for Node 10, 12 and 14. Thanks to the folks at Align Tech for sponsoring this work.

To install it:

npm i @pact-foundation/pact@beta

For examples on how to use it, see examples/v3/e2e and examples/v3/todo-consumer in the v3.0.0 branch.

NOTE: This implementation is not ready for production use yet, as it DOES NOT yet support the following features:

• Verifying a pact by direct URL
• consumerVersionTags to specify which pacts to verify - only the latest pact will be verified.
• Any features that make use of the "Pacts for Verification" API which include:
  • Support for consumerVersionSelectors
  • Pending pacts
  • WIP pacts

There are a number of new matchers that can be used, like integer and timestamp. There are defined in the MatchersV3 class that needs to be used with PactV3 DSL.

For example:

const { PactV3, MatchersV3 } = require("@pact-foundation/pact/v3")
const {
  eachLike,
  atLeastLike,
  integer,
  timestamp,
  boolean,
  string,
  regex,
  like,
} = MatchersV3

const animalBodyExpectation = {
  id: integer(1),
  available_from: timestamp("yyyy-MM-dd'T'HH:mm:ss.SSSX"),
  first_name: string("Billy"),
  last_name: string("Goat"),
  animal: string("goat"),
  age: integer(21),
  gender: regex("F|M", "M"),
  location: {
    description: string("Melbourne Zoo"),
    country: string("Australia"),
    post_code: integer(3000),
  },
  eligibility: {
    available: boolean(true),
    previously_married: boolean(false),
  },
  interests: eachLike("walks in the garden/meadow"),
}

You can write both consumer and provider verification tests with XML requests or responses. For an example, see examples/v3/todo-consumer/test/consumer.spec.js. There is an XmlBuilder class that provides a DSL to help construct XML bodies with matching rules and generators (NOTE that generators are not supported for XML at this time).

for example:

body: new XmlBuilder("1.0", "UTF-8", "ns1:projects").build(el => {
  el.setAttributes({
    id: "1234",
    "xmlns:ns1": "http://some.namespace/and/more/stuff",
  })
  el.eachLike(
    "ns1:project",
    {
      id: integer(1),
      type: "activity",
      name: string("Project 1"),
      due: timestamp("yyyy-MM-dd'T'HH:mm:ss.SZ", "2016-02-11T09:46:56.023Z"),
    },
    project => {
      project.appendElement("ns1:tasks", {}, task => {
        task.eachLike(
          "ns1:task",
          {
            id: integer(1),
            name: string("Task 1"),
            done: boolean(true),
          },
          null,
          { examples: 5 }
        )
      })
    },
    { examples: 2 }
  )
})

The VerifierV3 class can verify your provider in a similar way to the existing one.

Request filters now take a request object as a parameter, and need to return the mutated one.

requestFilter: req => {
    req.headers["MY_SPECIAL_HEADER"] = "my special value"

    // e.g. ADD Bearer token
    req.headers["authorization"] = `Bearer ${token}`

    // Need to return the request back again
    return req
},

Provider state callbacks have been updated to support parameters and return values. The first parameter is a boolean indicating whether it is a setup call (run before the verification) or a tear down call (run afterwards). The second optional parameter is a key-value map of any parameters defined in the pact file. Provider state callbacks can also return a map of key-value values. These are used with provider-state injected values, but support for that will be added in a later release.

stateHandlers: {
  "Has no animals": setup => {
    if (setup) {
      animalRepository.clear()
      return Promise.resolve({ description: `Animals removed to the db` })
    }
  },
  "Has some animals": setup => {
    if (setup) {
      importData()
      return Promise.resolve({
        description: `Animals added to the db`,
        count: animalRepository.count(),
      })
    }
  },
  "Has an animal with ID": (setup, parameters) => {
    if (setup) {
      importData()
      animalRepository.first().id = parameters.id
      return Promise.resolve({
        description: `Animal with ID ${parameters.id} added to the db`,
        id: parameters.id,
      })
    }
  },

You can change the log levels using the LOG_LEVEL environment variable.

If you are having issues, a good place to start is setting logLevel: 'debug' when configuring the new Pact({...}) object. Setting it to trace will give you detailed in/out requests as far as Pact sees it.

See https://docs.pact.io/docker/.

Pact tests are inherently stateful, as we need to keep track of the interactions on a per-test basis, to ensure each contract is validated in isolation from others. However, in larger test suites, this can result in slower test execution.

Modern testing frameworks like Ava and Jest support parallel execution out-of-the-box, which

The good news is, parallel test execution is possible, you need to ensure that:

1. Before any test run invocation, you remove any existing pact files, to prevent invalid / stale interactions being left over from previous test runs
2. Each test is fully self-contained, with its own mock server on its own port
3. You set the option pactfileWriteMode to "merge", instructing Pact to merge any pact documents with the same consumer and provider pairing at the end of all test runs.

When all of your tests have completed, the result is the union of the all of the interactions from each test case in the generated pact file.

See the following examples for working parallel tests:

• Pact with AVA (Node env)
• Pact with Mocha

Pact tests tend to be quite long, due to the need to be specific about request/response payloads. Often times it is nicer to be able to split your tests across multiple files for manageability.

You have a number of options to achieve this feat:

1. Consider implementing the Parallel tests guidelines.

2. Create a Pact test helper to orchestrate the setup and teardown of the mock service for multiple tests.

   In larger test bases, this can significantly reduce test suite time and the amount of code you have to manage.

   See this example and this issue for more.

3. Set pactfileWriteMode to merge in the Pact() constructor

   This will allow you to have multiple independent tests for a given Consumer-Provider pair, without it clobbering previous interactions, thereby allowing you to incrementally build up or modify your pact files.

   This feature addresses the use case of "my pact suite takes bloody ages to run, so I just want to replace the interactions that have been run in this test execution" and requires careful management

   NOTE: If using this approach, you must be careful to clear out existing pact files (e.g. rm ./pacts/*.json) before you run tests to ensure you don't have left over requests that are no longer relevant.

   See this PR for background.

TL;DR - you almost certainly have not properly handled (returned) a Promise.

We see this sort of thing all of the time:

it("returns a successful thing", () => {
  executeApiCallThatIsAPromise()
    .then((response) => {
      expect(response.data).to.eq({...})
    })
    .then(() => {
      provider.verify()
    })
  })

There are several problems with this:

1. in the "returns a successful thing", the call to executeApiCallThatIsAPromise() is a function that returns a Promise, but is not returned by the function (it block) - this leaves a dangling, unhandled Promise. In your case it fails, but by the time it does the it block has already completed without problems - and returns a green result ✅.
2. In the then block, the call to provider.verify() is also not returned, and will suffer the same fate as (1)

Side note: Jasmine and other test frameworks may detect an unhandled promise rejection and report on it.

The correct code for the above is:

it("returns a successful thing", () => {
  return executeApiCallThatIsAPromise() // <- explicit return here, you could also use the "async/await" syntax here
    .then((response) => {
      expect(response.data).to.eq({...})
    })
    .then(() => provider.verify()) // provider.verify() also returned
  })

See above - you probably have not returned a Promise when you should have.

If you prefix your test command (e.g. npm t) with the following two environment variables, you can selectively run a specific interaction during provider verification.

For the e2e example, let's assume we have the following failure:

3 interactions, 2 failures

Failed interactions:

* A request for all animals given Has some animals

* A request for an animal with id 1 given Has an animal with ID 1

If we wanted to target the second failure, we can extract the description and state as the bits before and after the word "given":

PACT_DESCRIPTION="a request for an animal with ID 1" PACT_PROVIDER_STATE="Has an animal with ID 1" npm t

Also note that PACT_DESCRIPTION is the failing description and PACT_PROVIDER_STATE is the corresponding providerState from the pact file itself.

Under the hood, Pact JS spins up a Ruby Mock Service. On some systems, this may take more than a few seconds to start. It is recommended to review your unit testing timeout to ensure it has sufficient time to start the server.

See here for more details.

Jest uses JSDOM under the hood which may cause issues with libraries making HTTP request.

You'll need to add the following snippet to your package.json to ensure it uses the proper Node environment:

"jest": {
  "testEnvironment": "node"
}

Also, from Jest 20, you can add the environment to the top of the test file as a comment. This will allow your pact test to run along side the rest of your JSDOM env tests.

/**
 * @jest-environment node
 */

Jest also runs tests in parallel by default, which can be problematic with Pact which is stateful. See parallel tests to see how to make it run in parallel, or run Jest with the --runInBand option to run them sequentially.

See this issue for background, and the Jest example for a working example.

You way want to consider using this starter schematic: https://github.com/niklas-wortmann/ngx-pact

Angular's HttpClient filters out many headers from the response object, this may cause issues when validating a response in tests.

You'll need to add the additional header Access-Control-Expose-Headers, this will allow specified headers to be passed to the response object. This can be done by declaring the header in the willRespondWith section of your interaction:

"willRespondWith": {
  "headers": {
    "Access-Control-Expose-Headers": like("My-Header"),
    "My-Header": "..."
  },
  ...
}

See this issue for background.

If your standard tricks don't get you anywhere, setting the logLevel to trace and increasing the timeout doesn't help and you don't know where else to look, it could be that the binaries we use to do much of the Pact magic aren't starting as expected.

Try starting the mock service manually and seeing if it comes up. When submitting a bug report, it would be worth running these commands before hand as it will greatly help us:

./node_modules/.bin/pact-mock-service

...and also the verifier (it will whinge about missing params, but that means it works):

./node_modules/.bin/pact-provider-verifier

1. Fork it
2. Create your feature branch from the relevant tree (e.g. [v5] or [v6]) (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request

Join us on Slack

or chat to us at

• Twitter: @pact_up
• Stack Overflow: https://stackoverflow.com/questions/tagged/pact