Using a DynamicDataStore instance correctly

I’ve recently seen a couple of examples of code where the Dynamic Data Store (being released as part of CMS 6) has been used in a multithreaded environment (web app for example) in a singleton pattern, i.e. a single instance of a DynamicDataStore is shared between all requests (read threads) in the application.

THIS IS BAD!

The default implementation (EPiServerDynamicDataStore) of the DynamicDataStore is not thread safe.

Let me just say that again

DYNAMIC DATA STORE IS NOT THREAD SAFE.

Think of it like an ADO.NET connection object:

- Obtain – Use – Throw Away

on the stack in a method.

The internals of the Dynamic Data Store are thread safe, that is all instances access shared data with thread synchronization. The cost of obtaining a DDS instance using the DynamicDataStoreFactory GetStore method is low due to all store metadata being cached in memory so there really is no need to hold DDS instances as member variables in a class.

An example of the recommended use pattern can be found below:

// Create and populate my object
Person p = new Person();
p.FirstName = "Micky";
p.LastName = "Mouse";

// Obtain a store instance
EPiServer.Data.Dynamic.DynamicDataStore store = GetStore(typeof(Person));

// save the object
store.Save(p);

static EPiServer.Data.Dynamic.DynamicDataStore GetStore(Type t)
{
  // GetStore will only return null the first time this method is called for a Type
  // In that case the ?? C# operator will call CreateStore
  return EPiServer.Data.Dynamic.DynamicDataStoreFactory.Instance.GetStore(t) ??
      EPiServer.Data.Dynamic.DynamicDataStoreFactory.Instance.CreateStore(t);
}