Provides a eventsourcing high level API. https://github.com/YoEight/eventsource-api#readme
|Latest on Hackage:||1.0.0|
This package is not currently in any snapshots. If you're interested in using it, we recommend adding it to Stackage Nightly. Doing so will make builds more reliable, and allow stackage.org to host generated Haddocks.
This project provides what we think be a lean eventsourcing API. The goal is to set a common ground for eventsourcing-based application, yet doesn't force you to be trapped under a restrictive framework.
This library main abstraction is the
-- | Main event store abstraction. It exposes essential features expected from -- an event store. class Store store where -- | Appends a batch of events at the end of a stream. appendEvents :: (EncodeEvent a, MonadIO m) => store -> StreamName -> ExpectedVersion -> [a] -> m (Async EventNumber) -- | Appends a batch of events at the end of a stream. readBatch :: MonadIO m => store -> StreamName -> Batch -> m (Async (ReadStatus Slice)) -- | Subscribes to given stream. subscribe :: MonadIO m => store -> StreamName -> m Subscription
The idea is to use any backend of your liking as long as it's able to derive that typeclass with its related implicit constraints.
appendEventsMUST add events at the end of a stream without doing any other alteration on the stream. The order of the event's array MUST be respected within the store. Most important, the implementation MUST comply to the given
ExpectedVersionpassed by the user.
readBatchSHOULD retrieve a set of event given a starting position and a batch size (represented by the
Batchparameter). The returned array MUST have its events ordered by their
subscribeSHOULD register a subscription to the given stream. An implementation MUST allow to subscribe to a stream that doesn't exist yet. A subscription allows the user to be notified of any event added to the stream. There is no mandatory timing to meet regarding how fast the subscription is notified by a change. We only suggest to dispatch new events as soon as possible.
A word on
ExpectedVersion. It's a mean to preserve data consistency in a concurrent setting.
-- | The purpose of 'ExpectedVersion' is to make sure a certain stream state is -- at an expected point in order to carry out a write. data ExpectedVersion = AnyVersion -- Stream is a any given state. | NoStream -- Stream shouldn't exist. | StreamExists -- Stream should exist. | ExactVersion EventNumber -- Stream should be at givent event number.
On a write, if the
ExpectedVersion condition given by the user is not met within the store, the implementation should raise an exception. At the moment, the API doesn't capture this situation in the type system.