This post is a little late coming , as I first starting using Entity Framework 4 when it was on CTP 1 at the start of the year. I used it up to CTP2 in March/April 2010. Forgive me if there have been changes since and please do visit the MSDN site here and do read up on it (I read Julia Lermans book and her blog here)
First let me start by saying EF4 is a great product. It has the potential to not only generate code and classes for communication with database but also now has the ability to generate the database schema based on your code. This is great for developers who are not confident with databases or T-SQL, or do not have a lot of time to create a working system.
My background is primarily as a sql dba and developer, perhaps this is why I had a few problems with EF.
I have always believed that a good database layer is the best starting point for a good application design. I realise there are times when this is not always possible …
However, I like using stored procedure when writing real applications.
- They have the ability to protect the application from schema changes making it easier to change code without redeploying your front end.
- They can protect the database from unscrupulous client developers who like writing badly performing inline sql (you know who you are …!)
- They generally have the capability to perform well (of course they have to also be written well!)
- They can be used to protect from direct table access and permissions can be applied both externally and programatically within the procedure.
- They can enforce atomic updates to your data across multiple tables.
- Most importantly for me. They can remove a lot of complexity from the application and can allow you to create a normalised data schema, but protect the developers from the complexity of multi-tabled updates.
So why is this a problem I hear your ask. Entity framework can be used with stored procedures can’t they?
Well yes and no …
Of course you can map your insert/update/delete portion of the entity to your stored procedures. But they have to be written in a certain way and they must return key values in a select statement (no output parameters allowed here).
No out of the box support for the Get data process being a stored procedure, this needs to be a table or a view, unless you like editing the xml files. I found this side of the entity framework to be incredibly painful, in terms of setting an exact entity that could use a view for the Get and stored procedures for the rest. Let alone when you want to add a column and have to start all over again.
However what I did find useful in the entity framework was the ability to knock up a working demo system in an incredibly short amount of time. I had a WPF/WCF/SQL 2008 system working with a single table process in a matter of days. But what I found was that I couldn’t take it any further without wasting a lot of time remapping schema changes to the entities. I also decided to take a look at some of the generated SQL statements from the entity sql and linq sql. It wasn’t the worst I’ve seen but boy was it verbose !
My conclusion is … use the EF if you want create quick proof of concepts or prototypes but when it comes to writing real database applications stick to the stored procedures and real T-SQL, your DBA will thank you for it when he has to take it over and support it. Microsoft seem to be approaching the whole code generation tools from the front end. What I really need is the reverse, I want a really good database design, with standard stored procedures for data updates and tool to help me connect it to a front end without me writing all the code!
Please try it for yourself and do read the material at the start of the post – make up your own mind!