Wtf? Gecko!

Maybe I don't get outside enough...

(click for big)

Apart from that interesting diversion (40km bike ride) nothing else has happened. Oh, flight was bumpy as hell on the way in to Schiphol though

Step 1

Decide the feature you need to implement. In this case I have decided that my DefaultNewsProvider class needs a method for returning a single news item when passed its Id.

Step 2

Open your test file that is responsible for testing the DefaultNewsProvider class (in this case it is in my Keima.Website.Data.Test/NewsProviderTests).

Step 3

Write the test with the expected result. I know that in my database their exists a news item with an Id of "tag:keima.co.uk,2008-06-26:Composer 2.0 Launch", therefore I construct a a test that calls the new GetNewsItem method with the Id, then compares the result to check we have been given the correct item back.

Step 4

We have hit a problem, this method does not actually exist despite having the test for it. So, bring up the code helper and generate the method stub.

Step 5

Test the code. Go on. Watch it fail.

Step 6

Make it pass. In this case its easy, a sort of 'fill in the blanks'. So back to DefaultNewsProvider it is and the implementation is finished off.

Step 7

Re-run the test, watch it pass. All is well, the code has been added, a test has been created, and if in the future the method breaks due to other code changes you will know immediatly.

List of reasons why test driven development can help you:

+ Allows your code to be maintainable, shows the exact point of failure of a program, none of this nasty debugger related stack tracing and breakpointing anymore
+ Forces you to focus upon the expected usage of a function, not the specifics of its implementation. Once you get in to the TDD mindset this allows you to be more concerned with the abstract ideas of how the project should be designed and structured with maintainability in mind as opposed to language specific implementation details (e.g. "I need a function that returns a file handle given a file name or null on an invalid file" rather than "What calls need to be made inside C# to open a file?"
+ At any one time, your code WORKS. Ensuring that code is correctly covered by tests allows you to guarantee that at any one time your project can be built and shipped to a customer. 
Your tests are more accurate (bold assumption I know, but here me out). Designing the code by considering how it will be used, and ensuring that tests are constructed at the beginning is much more preferable to designing tests once a project is finished, where code may be missed, meanings of code blocks forgotten about or motivation does not allow for proper testing.

As ever address any personal experiences, observations, suggestions or death threats through the comments.

C# 3.0 is god like. Undeniably. The ability to reduce your code is amazing. Now, I understand the code zealots out there will say "Yeah, but X language has been able to do this for years". Well I don't really care about X language truth be told. The .NET platform is so widely deployed and supported now I don't really see why in 9/10 cases I should use language X.

Consider the following basic example, of calculating an Inverse Distance Weighted plot of signal strength of a given location given a set (plots) of locations and their corresponding signal measure.

The formula is simple, for a given point x, the strength is the sum of wk(x) (one over the distance between the two points to the power of a fixed parameter p) times uk (the value at k) over the sum of wk(x):

Traditionally, this would yeild the following code:

double sumWkUk = 0;
double sumWk = 0; 
 
foreach(Location loc in plots)
{
sumWkUk += wk(x, loc) * loc.Value;
sumWk += wk(x, loc);
}
 
x.Value = sumWkUk / sumWk;

Not too much code, but this can be reduced a lot. Consider using the C# 3.0 sum operator:

x.Value = plots.Sum(xk => wk(x, xk) * xk.Value)
/ plots.Sum(xk => wk(x, xk));

Wait, what, did we just reduce that to ONE LINE of code? Shit.

Are you listening Java? No, thought not, no wonder you are still losing the great programming race (or maybe I am just bitter).

Summary:

- Less code to maintain is good.

- Gives us less chances for errors

- makes it easier to read and understand (in this case at least)

Here we are, nearly a year since the last update.

Wait, where is the last update?

Well I am not entirely sure where it went, but my Wordpress fell apart, so here we are, fresh install (must say I am rather impressed with the whole Wordpress 2.5 thing, its certainly a setup up from 2.2).

Anyway, moving swiftly onwards...

I've got to say, things have been fucking boring since the last update. However, my interest in taking my camera around and taking copious amounts of photos has somewhat increased (although the corresponding increase in quality of the photos is somewhat debatable).

My latest addition to the set of camera accessories is the much-loved Sigma 10-20mm, which I picked up from Jessops on Monday (OK, so I can get it a lot cheaper online, but after various experiences with online shopping I don't really trust this anymore, but I'll see when we get to the next planned (my first piece of L glass, the 70-200mm f/4, but not until I get paid). For my first HSM motored lense, I must say the focus is much quicker and quieter than my other Sigma glass (the 24-70 f/2.8, not that it is that slow, but it is fucking loud).

So, without further ado, heres a few shots I took down the bay with OJ on Monday.

Thats all for now really.