eGo is a minimal library that help build event-sourcing and CQRS application through a simple interface, and it allows developers to describe their commands, events and states **are defined using google protocol buffers **. Under the hood, ego leverages Go-Akt to scale out and guarantee performant, reliable persistence.
The EventSourcedBehavior
is crucial for maintaining data consistency, especially in distributed
systems. It defines how to handle the various commands (requests to perform actions) that are always directed at the
event sourced entity.
In eGo commands sent to the EventSourcedBehavior
are processed in order. When a command is processed,
it may result in the generation of events, which are then stored in an event store. Every event persisted has a revision
number
and timestamp that can help track it. The EventSourcedBehavior
in eGo is responsible for defining how
to handle events that are the result of command handlers.
The end result of events handling is to build the new state of the event sourced entity. When running in cluster mode,
aggregate root are sharded.
Commands handler
: The command handlers define how to handle each incoming command, which validations must be applied, and finally, which events will be persisted if any. When there is no event to be persisted a nil can be returned as a no-op. Command handlers are the meat of the event sourced actor. They encode the business rules of your event sourced actor and act as a guardian of the event sourced entity consistency. The command handler must first validate that the incoming command can be applied to the current model state. Any decision should be solely based on the data passed in the commands and the state of the Behavior. In case of successful validation, one or more events expressing the mutations are persisted. Once the events are persisted, they are applied to the state producing a new valid state.Events handler
: The event handlers are used to mutate the state of the event sourced entity by applying the events to it. Event handlers must be pure functions as they will be used when instantiating the event sourced entity and replaying the event store.
To define an event sourced entity, one needs to:
- define the state of the event sourced entity using google protocol buffers message
- define the various commands that will be handled by the event sourced entity
- define the various events that are result of the command handlers and that will be handled by the event sourced entity to return the new state of the event sourced entity
- define and make sure the
events store
is properly implemented. - implement the
EventSourcedBehavior
interface. - call the
Entity
method of eGo engine
Every event handled by event sourced entity are pushed to an events stream. That enables real-time processing of events
without having to interact with the events store.
Just use Subscribe
method of Engine and start iterating through the messages and cast every message to
the Event.
One can add a projection to the eGo engine to help build a read model. Projections in eGo rely on an offset store to track how far they have consumed events persisted by the write model. The offset used in eGo is a timestamp-based offset. One can also:
- remove a given projection: this will stop the projection and remove it from the system
- check the status of a given projection
One can implement a custom events store. See EventsStore. eGo comes packaged with two events store:
One can implement a custom offsets store. See OffsetStore. eGo comes packaged with two offset store:
The DurableStateBehavior
represents a type of Actor that persists its full state after processing
each command instead of using event sourcing.
This type of Actor keeps its current state in memory during command handling and based upon the command response
persists its full state into a durable store. The store can be a SQL or NoSQL database.
The whole concept is given the current state of the actor and a command produce a new state with a higher version as
shown in this diagram: (State, Command) => State
DurableStateBehavior
reacts to commands which result in a new version of the actor state. Only the
latest version of the actor state is persisted to the durable store.
There is no concept of history regarding the actor state since this is not an event sourced actor.
However, one can rely on the version number of the actor state and exactly know how the actor state has evolved
overtime.
DurableStateBehavior
version number are numerically incremented by the command handler which means it
is imperative that the newer version of the state is greater than the current version by one.
DurableStateBehavior
will attempt to recover its state whenever available from the durable state.
During a normal shutdown process, it will persist its current state to the durable store prior to shutting down. This
behavior help maintain some consistency across the actor state evolution.
One can implement a custom state store. See StateStore. eGo comes packaged with two state stores:
To define a durable state entity, one needs to:
- define the state of the entity using google protocol buffers message
- define the various commands that will be handled by the entity
- define and make sure the
durable state store
is properly implemented. - implements the
DurableStateBehavior
interface - start eGo engine with the option durable store using
WithStateStore
- call the
DurableStateEntity
method of eGo engine
DurableStateBehavior
full state is pushed to an events stream.
That enables real-time processing of state without having to interact with the state store.
Just use Subscribe
method of Engine and start iterating through the messages and cast every message to
the DurableState.
The cluster mode heavily relies on Go-Akt clustering.
eGo ships in some mocks
Check the examples
go get github.com/tochemey/ego
package main
import (
"context"
"errors"
"log"
"os"
"os/signal"
"syscall"
"time"
"github.com/google/uuid"
"google.golang.org/protobuf/proto"
"github.com/tochemey/ego/v3"
"github.com/tochemey/ego/v3/plugins/eventstore/memory"
samplepb "github.com/tochemey/ego/v3/example/pbs/sample/pb/v1"
)
func main() {
// create the go context
ctx := context.Background()
// create the event store
eventStore := memory.NewEventsStore()
// connect the event store
_ = eventStore.Connect(ctx)
// create the ego engine
engine := ego.NewEngine("Sample", eventStore)
// start ego engine
_ = engine.Start(ctx)
// create a persistence id
entityID := uuid.NewString()
// create an entity behavior with a given id
behavior := NewAccountBehavior(entityID)
// create an entity
_ = engine.Entity(ctx, behavior)
// send some commands to the pid
var command proto.Message
// create an account
command = &samplepb.CreateAccount{
AccountId: entityID,
AccountBalance: 500.00,
}
// send the command to the actor. Please don't ignore the error in production grid code
reply, _, _ := engine.SendCommand(ctx, entityID, command, time.Minute)
account := reply.(*samplepb.Account)
log.Printf("current balance: %v", account.GetAccountBalance())
// send another command to credit the balance
command = &samplepb.CreditAccount{
AccountId: entityID,
Balance: 250,
}
reply, _, _ = engine.SendCommand(ctx, entityID, command, time.Minute)
account = reply.(*samplepb.Account)
log.Printf("current balance: %v", account.GetAccountBalance())
// capture ctrl+c
interruptSignal := make(chan os.Signal, 1)
signal.Notify(interruptSignal, os.Interrupt, syscall.SIGINT, syscall.SIGTERM)
<-interruptSignal
// disconnect the event store
_ = eventStore.Disconnect(ctx)
// stop the actor system
_ = engine.Stop(ctx)
os.Exit(0)
}
// AccountBehavior implements EntityBehavior
type AccountBehavior struct {
id string
}
// make sure that AccountBehavior is a true persistence behavior
var _ ego.EntityBehavior = (*AccountBehavior)(nil)
// NewAccountBehavior creates an instance of AccountBehavior
func NewAccountBehavior(id string) *AccountBehavior {
return &AccountBehavior{id: id}
}
// ID returns the id
func (a *AccountBehavior) ID() string {
return a.id
}
// InitialState returns the initial state
func (a *AccountBehavior) InitialState() ego.State {
return ego.State(new(samplepb.Account))
}
// HandleCommand handles every command that is sent to the persistent behavior
func (a *AccountBehavior) HandleCommand(_ context.Context, command ego.Command, _ ego.State) (events []ego.Event, err error) {
switch cmd := command.(type) {
case *samplepb.CreateAccount:
// TODO in production grid app validate the command using the prior state
return []ego.Event{
&samplepb.AccountCreated{
AccountId: cmd.GetAccountId(),
AccountBalance: cmd.GetAccountBalance(),
},
}, nil
case *samplepb.CreditAccount:
// TODO in production grid app validate the command using the prior state
return []ego.Event{
&samplepb.AccountCredited{
AccountId: cmd.GetAccountId(),
AccountBalance: cmd.GetBalance(),
},
}, nil
default:
return nil, errors.New("unhandled command")
}
}
// HandleEvent handles every event emitted
func (a *AccountBehavior) HandleEvent(_ context.Context, event ego.Event, priorState ego.State) (state ego.State, err error) {
switch evt := event.(type) {
case *samplepb.AccountCreated:
return &samplepb.Account{
AccountId: evt.GetAccountId(),
AccountBalance: evt.GetAccountBalance(),
}, nil
case *samplepb.AccountCredited:
account := priorState.(*samplepb.Account)
bal := account.GetAccountBalance() + evt.GetAccountBalance()
return &samplepb.Account{
AccountId: evt.GetAccountId(),
AccountBalance: bal,
}, nil
default:
return nil, errors.New("unhandled event")
}
}
The version system adopted in eGo deviates a bit from the standard semantic versioning system. The version format is as follows:
- The
MAJOR
part of the version will stay atv3
for the meantime. - The
MINOR
part of the version will cater for any new features, breaking changes with a note on the breaking changes. - The
PATCH
part of the version will cater for dependencies upgrades, bug fixes, security patches and co.
The versioning will remain like v3.x.x
until further notice.
Contributions are welcome! The project adheres to Semantic Versioning and Conventional Commits. This repo uses Earthly.
There are two ways you can become a contributor:
- Request to become a collaborator, and then you can just open pull requests against the repository without forking it.
- Follow these steps
- Fork the repository
- Create a feature branch
- Set your docker credentials on your fork using the following secret names:
DOCKER_USER
andDOCKER_PASS
- Submit a pull request
Prior to submitting a pull request, please run:
earthly +test