Skip to content

ymotton/NEventStore

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NEventStore

Overview

NEventStore is a persistence library used to abstract different storage implementations when using event sourcing as storage mechanism. Event sourcing is most closely associated with a concept known as CQRS.

Need Help? Have a Question?

Ask your question on Stack Overflow and tag your question with the CQRS tag and the word "NEventStore" in the title.

Purpose and Theory

The purpose of an event store is to represent a series of events as a stream. Furthermore, it provides hooks whereby any events committed to the stream can be dispatched to interested parties.

Guided by a number strategic design decisions based upon the needs of applications using event sourcing, NEventStore is able to liberate applications from the stringent requirements often imposed by infrastructure components. Specifically, most CQRS-style applications read from a message queue and perform some processing. When processing is complete, the application then commits the work to storage and publishes the completed work. In almost all cases, this requires a two-phase commit managed by a distributed transaction coordinator (MSDTC in .NET) along with various security settings and firewall ports opened and avaiable whereby such components can communicate, not to mention a ubiquitous requirement for Microsoft Windows on all machines in .NET environments.

When using two-phase commit in .NET, there are very few database drivers that support this scenario and even fewer message queues that support it. In essence, if you want to implement a typical CQRS-style application, you're stuck with MSMQ and SQL Server using MSDTC. Granted, there are other choices, but the constraints imposed by a two-phase commit are burdensome. This also creates additional issues when utilizing shared hosting or running on Mono as support in frameworks and drivers is either poor, buggy, or unavailable.

NEventStore liberates application developers from this level of infrastructure awareness and concern by committing all work within a separate isolated atomic unit--all without using transactions. Furthermore, it does this outside of any ambient transaction from a message queue or other persistence mechanisms. In other words, application developers are free to use virtually any messaging queuing infrastructure, message bus (if at all), and storage engine. Each will perform its own specific task in an isolated manner with full transactional integrity all without enlisting any resources (other than a message queue) in some form of transaction.

Interestingly enough, even without the presence of distributed transactions across the various resources involved, such as a message queue and persistent storage, the EventStore is able to ensure a fully transactional experience. This is achieved by breaking apart a distributed transaction into smaller pieces and performing each one individually. This is one of the primary goals and motivations in the underlying model found in the EventStore. Thus each message delivered by the queuing infrastructure is made to be idempotent, even though the message may be delivered multiple times, as per message queue "at-least-once" guarantees. Following this, the EventStore is able to ensure that all events committed are always dispatched to any messaging infrastructure.

New in v3.0: I knew you guys couldn't live without it, so for those storage engines and message systems which support and participate two-phase commits, there now exists the ability to specify a TransactionScopeOption of 'Required'. Simply indicate this using the following wireup overload:

var store = Wireup.Init()
	.UsingSqlPersistence("connection-name-here") // also works with UsingRavenPersistence()
		.EnlistInAmbientTransaction()

Supported Storage Engines

Relational Databases

[Complete] Microsoft SQL Server 2005 (or later)
[Complete] MySQL 5.0 (or later)

  • [Complete] InnoDB
  • [Complete] NDB/MySQL Cluster
  • [Complete] Drizzle
  • [Complete] MariaDB
  • [Complete] XtraDB
  • [Complete] PBXT
  • [Complete] Xeround
  • [Complete] Galera
  • [Complete] Percona
  • [Complete] OurDelta
  • [Untested] MyISAM
  • [Untested] BerkleyDB

[Complete] PostgreSQL 8.0 (or later)
[Complete] Firebird 2.0 (or later)
[In progress] Sybase (ASE)
[In progress] Sybase (SQL Anywhere)
[Planned] Oracle 8.0 (or later)
[TBA] IBM DB2
[TBA] Informix

Embedded Relational Databases

[Complete] SQLite 3.0 (or later)
[Complete] Microsoft SQL Server Compact Edition 3.5 (or later)
[Complete] Microsoft Access 2000 (or later)

Cloud-based Databases (relational or otherwise)

[Complete] Microsoft SQL Azure
[Complete] Amazon RDS (MySQL)
[Planned] Amazon RDS (Oracle)
[In progress] Azure Tables/Blobs
[In progress] Amazon SimpleDB/S3

Document Databases

[Complete] RavenDB r322 (or later)
[Complete] MongoDB 1.6 (or later)
[Planned] CouchDB 1.0 (or later)
[TBA] OrientDB

File System

[Planned] .NET Managed System.IO APIs

Dynamo Clones

[Planned] Cassandra
[Planned] Riak
[TBA] Voldemort
[TBA] Dynomite

KV Stores / NoSQL

[Planned] Redis
[Planned] Memcached (Membase, Gear6, etc.)
[Planned] HBase
[TBA] HyperTable
[TBA] Tokyo Cabinet
[TBA] Microsoft Velocity
[TBA] SharedCache
[TBA] Hibari
[TBA] Scalaris
[TBA] Keyspace
[TBA] OrientKV
[TBA] VoltDB
[TBA] BerkleyDB
[TBA] Hazelcast
[TBA] HampsterDB

Project Goals

  • Mono 2.4 support
  • Medium-trust support
  • Support more storage engines than any other event storage implementation
  • Easily support virtually any storage engine (NoSQL, etc.)
  • Avoid dependence upon TransactionScope or Transactions while maintaining full data integrity
  • Full test coverage of storage implementations
  • Easily hook into any bus implementation (NServiceBus, MassTransit, etc.)
  • Synchronous and asynchronous dispatching of events
  • Extreme performance
  • Multi-thread safe
  • Fluent builder

Building

To build the solution and execute all tests, run Build.RunTask.bat from the command line.

If you would also like to package the solution for distribution, pass in the 'package' command by running Build.RunTask.bat package from the command line. Once built, the files will be placed in the "publish-net40" subdirectory.

Note: You may need to run the build with Administrator rights the first time. This is because NEventStore will attempt to create several performance counters and that action requires admin rights. If you would prefer to manually create the performance counters yourself, refer to NEventStore.Diagnostics.PerformanceCounters.

Using NEventStore

var store = Wireup.Init()
	.UsingSqlPersistence("Name Of EventStore ConnectionString In Config File")
		.InitializeStorageEngine()
		.UsingJsonSerialization()
			.Compress()
			.EncryptWith(EncryptionKey)
	.HookIntoPipelineUsing(new[] { new AuthorizationPipelineHook() })
	.UsingAsynchronousDispatchScheduler()
		// Example of NServiceBus dispatcher: https://gist.github.com/1311195
		.DispatchTo(new My_NServiceBus_Or_MassTransit_OrEven_WCF_Adapter_Code())
	.Build();		

/* NOTE: This following is merely *example* code. */

using (store)
{
	// some business code here
	using (var stream = store.CreateStream(myMessage.CustomerId))
	{
		stream.Add(new EventMessage { Body = myMessage });
		stream.CommitChanges(myMessage.MessageId);
	}
	
	using (var stream = store.OpenStream(myMessage.CustomerId, 0, int.MaxValue))
	{
		foreach (var @event in stream.CommittedEvents)
		{
			// business processing...
		}
	}
}

For a more complete example, please see NEventStore.Example project in the doc subdirectory.

Running the Example

The NEventStore.Example project is configured by default to use a SQL event store. To run the example program, either change the SQL connection string in the app.config file to connect to a existing SQL database or change WireupEventStore() to call UsingInMemoryPersistence() rather than UsingSqlPersistence().

About

A persistence agnostic Event Store for .NET

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C# 89.5%
  • PowerShell 8.7%
  • XSLT 1.7%
  • Shell 0.1%