CQRS framework in go

Join the chat at https://gitter.im/andrewwebber/cqrs GoDoc Build Status

#Project Summary The package provides a framework for quickly implementing a CQRS style application. The framework attempts to provides helpful functions to facilitate: - Event Sourcing - Command issuing and processing - Event publishing - Read model generation from published events


Example code

Example test scenario (inmemory)

Example test scenario (couchbase, rabbitmq)

Example CQRS scaleout/concurrent test

Test Scenario

The example test scenario is of a simple bank account that seeks to track, using event sourcing, a customers balance and login password

The are two main areas of concern at the application level, the Write model and Read model. The read model is aimed to facilitate fast reads (read model projections) The write model is where the business logic get executed and asynchronously notifies the read models

Write model - Using Event Sourcing

Account

type Account struct {
  cqrs.EventSourceBased

  FirstName    string
  LastName     string
  EmailAddress string
  PasswordHash []byte
  Balance      float64
}

To compensate for golang’s lack of inheritance, a combination of type embedding and a call convention pattern are utilized.

func NewAccount(firstName string, lastName string, emailAddress string, passwordHash []byte, initialBalance float64) *Account {
  account := new(Account)
  account.EventSourceBased = cqrs.NewEventSourceBased(account)

  event := AccountCreatedEvent{firstName, lastName, emailAddress, passwordHash, initialBalance}
  account.Update(event)
  return account
}

The ‘attached’ Update function being called above will now provide the infrastructure for routing events to event handlers. A function prefixed with ‘Handle’ and named with the name of the event expected with be called by the infrastructure.

func (account *Account) HandleAccountCreatedEvent(event AccountCreatedEvent) {
  account.EmailAddress = event.EmailAddress
  account.FirstName = event.FirstName
  account.LastName = event.LastName
  account.PasswordHash = event.PasswordHash
}

The above code results in an account object being created with one single pending event namely AccountCreatedEvent. Events will then be persisted once saved to an event sourcing repository. If a repository is created with an event publisher then events saved for the purposes of event sourcing will also be published

persistance := cqrs.NewInMemoryEventStreamRepository()
bus := cqrs.NewInMemoryEventBus()
repository := cqrs.NewRepositoryWithPublisher(persistance, bus)
...
repository.Save(account)

Account Events

type AccountCreatedEvent struct {
  FirstName      string
  LastName       string
  EmailAddress   string
  PasswordHash   []byte
  InitialBalance float64
}

type EmailAddressChangedEvent struct {
  PreviousEmailAddress string
  NewEmailAddress      string
}

type PasswordChangedEvent struct {
  NewPasswordHash []byte
}

type AccountCreditedEvent struct {
  Amount float64
}

type AccountDebitedEvent struct {
  Amount float64
}

Events souring events are raised using the embedded Update function. These events will eventually be published to the read models indirectly via an event bus

func (account *Account) ChangePassword(newPassword string) error {
  if len(newPassword) < 1 {
    return errors.New("Invalid newPassword length")
  }

  hashedPassword, err := GetHashForPassword(newPassword)
  if err != nil {
    panic(err)
  }

  account.Update(PasswordChangedEvent{hashedPassword})

  return nil
}

func (account *Account) HandlePasswordChangedEvent(event PasswordChangedEvent) {
  account.PasswordHash = event.NewPasswordHash
}

Again the calling convention routes our PasswordChangedEvent to the corresponding HandlePasswordChangedEvent instance function

Read Model

Accounts projection

type ReadModelAccounts struct {
  Accounts map[string]*AccountReadModel
}

type AccountReadModel struct {
  ID           string
  FirstName    string
  LastName     string
  EmailAddress string
  Balance      float64
}

Users projection

type UsersModel struct {
  Users    map[string]*User
}

type User struct {
  ID           string
  FirstName    string
  LastName     string
  EmailAddress string
  PasswordHash []byte
}

Infrastructure

There are a number of key elements to the CQRS infrastructure. - Event sourcing repository (a repository for event sourcing based business objects) - Event publisher (publishes new events to an event bus) - Event handler (dispatches received events to call handlers) - Command publisher (publishes new commands to a command bus) - Command handler (dispatches received commands to call handlers)

Event sourcing and integration events

Nested packages within this repository show example implementations using Couchbase Server and RabbitMQ. The core library includes in-memory implementations for testing and quick prototyping

persistance := cqrs.NewInMemoryEventStreamRepository()
bus := cqrs.NewInMemoryEventBus()
repository := cqrs.NewRepositoryWithPublisher(persistance, bus)

With the infrastructure implementations instantiated a stock event dispatcher is provided to route received events to call handlers

readModel := NewReadModelAccounts()
usersModel := NewUsersModel()

eventDispatcher := cqrs.NewVersionedEventDispatchManager(bus)
eventDispatcher.RegisterEventHandler(AccountCreatedEvent{}, func(event cqrs.VersionedEvent) error {
  readModel.UpdateViewModel([]cqrs.VersionedEvent{event})
  usersModel.UpdateViewModel([]cqrs.VersionedEvent{event})
  return nil
})

We can also register a global handler to be called for all events. This becomes useful when logging system wide events and when our read models are smart enough to filter out irrelevant events

integrationEventsLog := cqrs.NewInMemoryEventStreamRepository()
eventDispatcher.RegisterGlobalHandler(func(event cqrs.VersionedEvent) error {
  integrationEventsLog.SaveIntegrationEvent(event)
  readModel.UpdateViewModel([]cqrs.VersionedEvent{event})
  usersModel.UpdateViewModel([]cqrs.VersionedEvent{event})
  return nil
})

Within your read models the idea is that you implement the updating of your pre-pared read model based upon the incoming event notifications

Commands

Commands are processed by command handlers similar to event handlers. We can make direct changes to our write model and indirect changes to our read models by correctly processing commands and then raising integration events upon command completion.

commandBus := cqrs.NewInMemoryCommandBus()
commandDispatcher := cqrs.NewCommandDispatchManager(commandBus)
RegisterCommandHandlers(commandDispatcher, repository)

Commands can be issued using a command bus. Typically a command is a simple struct. The application layer command struct is then wrapped within a cqrs.Command using the cqrs.CreateCommand helper function

changePasswordCommand := cqrs.CreateCommand(
  ChangePasswordCommand{accountID, "$ThisIsANOTHERPassword"})
commandBus.PublishCommands([]cqrs.Command{changePasswordCommand})

The corresponding command handler for the ChangePassword command plays the role of a DDD aggregate root; responsible for the consistency and lifetime of aggregates and entities within the system)

commandDispatcher.RegisterCommandHandler(ChangePasswordCommand{}, func(command cqrs.Command) error {
  changePasswordCommand := command.Body.(ChangePasswordCommand)
  // Load account from storage
  account, err := NewAccountFromHistory(changePasswordCommand.AccountID, repository)
  if err != nil {
    return err
  }

  account.ChangePassword(changePasswordCommand.NewPassword)

  // Persist new events
  repository.Save(account)  
  return nil
})

As the read models become consistant, within the tests, we check at the end of the test if everything is in sync

if account.EmailAddress != lastEmailAddress {
  t.Fatal("Expected emailaddress to be ", lastEmailAddress)
}

if account.Balance != readModel.Accounts[accountID].Balance {
  t.Fatal("Expected readmodel to be synced with write model")
}

Related Repositories

cqrs-journey

cqrs-journey

Microsoft patterns & pratices CQRS Journey sample application ...

cqrs-starter-kit

cqrs-starter-kit

A starter kit for working with CQRS and intentful testing. ...

cqrs-server

cqrs-server

An opinionated Clojure CQRS/ES implementation using Onyx, Datomic, DynamoDB, Kafka and Zookeeper. ...

fun-cqrs

fun-cqrs

Fun.CQRS is a Scala CQRS/ES framework. It provides the basic blocks to build event driven aggregates with Event Sourcing. ...

node-cqrs-domain

node-cqrs-domain

Node-cqrs-domain is a node.js module based on nodeEventStore that. It can be very useful as domain component if you work with (d)ddd, cqrs, eventdenormalizer, host, etc. ...