Review: DDD6 at Microsoft UK Campus, Building 3, Thames Valley Park, Reading, Berkshire. RG6 1WG. 08:45 to 17:00

I'm more of a web developer than a 'Microsoft developer', but I use a lot of the MS stuff at work and I really enjoyed WebDD last February so I thought I'd give the more .Net focussed parent event a try.

Why IronRuby? (Presentation) Dave Verwer - I'd been to a couple of Dave's talks at the WebDD event and enjoyed them tremendously, so I went in to this one with high expectations. The talk started with some of the benefits of Ruby - a dynamic language with elegant syntax designed for simplicity and productivity, before going on to discuss how IronRuby and the DLR are implemented in .Net. The DLR is basically a layer for dynamic languages on top of the standard CLR, it is an abstraction of the what was developed in the course of getting IronPython up and running. At this point Dave admitted that IronRuby wasn't yet ready for use, so the rest of his talk would be about (C based) Ruby 1.8 - which IronRuby is targeting for compatibility. This was a bit of a disappointment for me, because really I was after a more practical demonstration of IronRuby than another presentation on Ruby - but it's possible I was in the minority given the audience. Check out the IronRuby website if you want to learn more.

Entity Framework (Presentation) James Winters - I had no idea what this was before the talk, but the title sounded more interesting than the others available in this slot, and it saved me leaving the room when I knew I wanted to see the talks either side. In the event, I was glad I stayed. To give a simple sound bite - Entity Framework is like an ORM on steroids, which will allow transparent mapping to non-relational stores, such as LDAP, as well as the more usual SQL stuff. The plan is for all MS server products to move to Entity Framework for their data access, though at the time of the talk the only backend available was SQL Server and all the examples were basic ORM stuff. The logical model of Entity Framework is made up of three levels:

  • O-SPACE - the object model, eg. LINQ
  • C-SPACE - the conceptual model, eg. Entity Client
  • S-SPACE - the store - eg. ADO.Net.SqlProvider

All levels are scoped by an entity container, which defines a namespace and an object model, and the goal of the MS Visual Studio team is to be able to generate the lower levels using visual tools.

Entity Framework supports multiple querying methods:

  • ESQL - very similar to HQL in Hibernate
  • Object Query - a generic object interface with caching and lazy loading, which can either be read only or mergeable
  • Use Entities directly
  • LINQ

There are some limitiation, there is no locking by default and, while it's going to be possible to map stored procedures the visual tools aren't going to support it for some time so you have to write up the XML manually. Generally, though it has a slight whiff of the architecture astronaut, it looks like an interesting addition to the .Net toolbox (Slides available here).

Why do I need an Inversion of Control Container? (Presentation) Mike Hadlow - I'd heard of Inversion of Control (IoC), but hadn't really understood what it was for. The presenter, Mike, had been in a similar position until he started studying it seriously, and now he was here to share his epiphany. The main problems IoC is trying to address is that the relationship between component size and complexity is not linear - double the size of your components and you quadruple the complexity - but breaking components into smaller pieces tends to lead to increased dependencies. Inversion of control reduces the complexity of large components and manages the dependencies between them. The principals of inversion of control (paraphrased from Robert Martin) are:

  • High level modules should not depend on low level modules
  • Abstractions should not depend on details
  • Details should not depend on abstractions
Typical problems of OO code without IoC are:

Inversion of control helps with all these by allowing constructor injection and property injection. If these can be injected into components at runtime, rather than hard-wired into the components, then it becomes possible to modify the behaviour of components without having to modify the components thus satisfying the open/closed principal and reducing the coupling (property injection) and also to easily mock objects for unit testing (constructor injection). Mike then moved on to some detailed demonstrations using the Windsor component from the Castle project, before concluding that IoC is really just the thorough application of design patterns and other OOP best practices. Therefore you should only start considering inversion of control once you're already familiar with these and have already implemented things like unit testing (Slides available here).

Dynamic Languages on .NET (Presentation) Michael Foord - My first session of the afternoon and by this point I was flagging a bit after having to get up early and make the trip out to Reading. Michael is one of the authors of Resolver, a spreadsheet based application environment, aimed at the financial markets and written in IronPython. The app is approximately 30000 lines of code, thus refuting the myth that it's impossible to write large applications in a dynamic language, and it's spreadsheet model is implemented in a Python hash table, making it very easy to extend Resolver with new types and operators. After a discussion of what constitutes a 'dynamic language' and the overall benefits of Python and IronPython in particular, Michael moved on to some practical examples based on his IronPython Web IDE (Slides available here)

Testing Your Applications With MbUnit Gallio (Presentation) Ben Hall - unit testing is a popular topic on programmer blogs these days, and this was reflected in a fairly full room for one of the 'minor' talks (ie. we weren't in one of the big rooms). Ben covered some best practices of automating unit testing on .Net with MbUnit and Gallio, starting at the database layer (use transactions and rollbacks or COM+ transactions), the business object layer (mocking and inversion of control) and the user interface (don't try and automate it!). An interesting if slightly hurried talk which I was unfortunately too tired to pay full attention too (Slides available here)

Overall a good day, I definitely learned more than one new thing per talk and the Microsoft conference room facility is a good venue for these sorts of things, so 4 out of 5, and I'll almost certainly try and get to the next one.

Technorati tags for this review: