Stacking Code

It's true, I have a general aversion to EF, partly from the negative reviews of EF1 and partly because that problem space has been solved for me for a long time with NHibernate. As a result I haven't checked out the latest incarnation of EF as yet, which made the recent series by Dino Esposito on the functional differences between EF4 and NHibernate a valuable read.

• Part 1, Multiple Database Support
• Part 2, Lazy Loading
• Part 3, Fetch Plans
• Part 4, Second-Level Caching
• Part 5, Self-Tracking Entities
• Part 6, Queries

I was surprised to find out EF4 has a respectable story for working with POCOs. Dino came to the conclusion that for the majority of scenarios EF4 is quite adequate. So what exactly is stopping me from giving it a go? Here is my shortlist of excuses reasons.

• NHibernate's breadth of fetching strategies. NHibernate lets me load a lazy collection a number of different ways, and I can specify the fetching strategy at the time of of the query or in the mapping file.

• EF's lack of a second-level cache. Without a second-level cache, my mediocre designs would be hitting the database a lot more often than they do right now. It really is a wonderful feature.

• Lazy properties, new in NHibernate 3.0. This is where a property like a byte array can be defined as lazy even though it's not a mapped entity. It should eliminate some workarounds I have in a few of my projects.

• The choice of DBMS. This is still an attractive feature of NHibernate. SQL Server CE looks great, but there are more than a few very usable database platforms out there. The option to deploy an application on any DBMS is a nice-to-have, for OSS projects in particular.

When it comes down to it, as Dino says, NHibernate is a mature framework with most of its quirks worked out. The feature set is impressive and deep enough to cover off the complex scenarios whenever they arise. In addition, the community continues to come up with easier ways to use it with initiatives like FluentNHibernate and QueryOver. Overcoming all that is a tough ask for a new ORM, but perhaps the next version of EF will be a bit more compelling.