A prevalent theme in the Volta documentation is the concept of the three “R”s. These three principles are at the heart of what Volta attempts to provide for developers.  They are retargeting, remodulating, and refactoring.

Retargeting

To me this is the coolest aspect of Volta and the feature that makes it a very interesting technology. At its core Volta is a compiler or more accurately a recompiler.  Instead of recompiling the source code Volta uses a technique called MSIL rewriting. This is what enables the developer to write client-side browser code in C#. Since all CLR languages compile to MSIL, Volta can use that as the common denominator. Volta is capable of MSIL to MSIL, and MSIL to JavaScript conversion.

Remodulating

Another goal of the Volta toolset is seamless cross-browser support. As the documentation states:

 Volta hides as many browser-specific differences as possible, but still allows developers to leverage the unique capabilities of particular browsers. Instead of targeting solely the intersection of browser capabilities, Volta targets the entire union, but makes the intersection browser-agnostic. This is browser remodulating.

Remodulating also deals with the debugging experience. Regardless of the browser platform developers will be stepping through their application code within Visual Studio. That's assuming of course that your target is either IE or FireFox which are the only browsers supported in the current release.

Refactoring

Volta refactoring is about enabling developers to create their applications while deferring decisions about what tier a particular component will run in. In my opinion this will be the most controversial feature of the toolset.  Larry O’Brien has some reservations about it:

This sounds like a bad idea to me. You can't refactor away the difference between an in-memory method call and an Internet message: one happens in nanoseconds and the other in milliseconds

I have some questions of my own about this. The docs claim that:

During development, all code runs in the client for ease of testing and debugging.

What are the implications of this? Not every .NET class is meant to run in a browser context. Does Volta offer any automatic guidance with regard to this? The notion of clicking a “Split Tiers” check box and decorating a class with a [RunAtOrigin] does seem implicitly powerful, I’m just curious to know where the model breaks down. I will try to answer these questions and others as I dig deeper into the framework.

It's kind of ironic.  This morning I was seriously considering taking a look at the Google Web Toolkit.  This is despite the fact that GWT is targeted at Java developers, which I am not.  It's just that GWT seems like such a useful idea - I wanted to play with it a bit.

Luckily Microsoft has, apparently, decided to put out a framework that allows developers to build rich internet applications without having to do a ton of JavaScript coding.

It's called Volta and at first glance it seems pretty cool.

When I was doing a lot of work with ASP.NET AJAX last year, the idea of coding in a more productive language and environment than JavaScript kept coming up again and again.  There was Script# at the time, but I didn't get a chance to look into it as much as I would have liked.

This certainly seems like a great development.  Maybe this even brings .NET developers one step closer to what Joel Spolsky described a few month ago.

Now…on to the samples, tutorials, and docs!

One of my favorite sections in SICP so far has been section 1.3 Formulating Abstractions with Higher-Order Procedures.  Higher-Order Procedures are procedures that deal with other procedures.  They can take a procedure as an argument and/or return a procedure as a result.

Though it is somewhat academic, I really like the example that Abelson and Sussman use in section 1.3.1 to explain the “procedures as arguments” scenario.  To illustrate this concept they implement summation using Scheme.  They implement a higher-order procedure called “sum” that takes two procedures and a range of numbers as arguments.  One procedure is used for computing the value that is added to the sum and the other is used to determine the next value to compute.  The Scheme code can be seen here.

It’s not as useful as map and reduce, but those are not discussed until Chapter 2.

As some of you may know C# 3.0 adds a whole slew of functional/dynamic features.  Whether this is an effort to keep up with the cool kids or merely a necessity for making LINQ work is not important to me.  What I’m curious to know is if the features make it easier to implement the concepts in SICP using C# 3.0.

Implementing “Sum” in C# 3.0

Following along with the example from SICP - let’s say that you have defined the following methods.

I decided to break with the book here and use iteration instead of recursion to make the code a little more digestible.

Common to each method is the idea of iterating over a range of numbers and adding the result of each computation to the total.  If we create a higher-order procedure called “Sum” to abstract out this behavior we can then redefine each method.

The “=>” and “Func” constructs come from the new Lambda expressions feature in C#.  Lambda expressions are a great new aspect of the language and they make implementing higher-order procedures a lot easier.  In this case the "term" and "next" arguments can be passed as Lambda expressions.

I had to cheat and do some casting shenanigans because the division operation in "PiSum()" brings floating point numbers into the mix, but I feel like this code does capture the essence of the original Scheme implementation.  Implementing "Sum" as a generic method might be a way to make the code more elegant.

If subverting the type system does not appeal to you, the following JavaScript implementation might be of interest.

JavaScript is more similar to Scheme (dynamic typing/first class functions) so it makes sense that the translation is smoother.

It turns out that there are other people interested in implementing SICP in different languages.  Chris Rathman has translated parts of SICP into 19 other languages.  One disturbing thing to note about Chris’s C# translation is that it doesn’t go beyond chapter 1.  Does that mean that it is not possible to implement the code from the rest of the book in C#?