Finding a Balance: Code 'Complexity'

"If you want to create a masterpiece - if you want to paint the Sistine Chapel - you can do that in your own time. Here, it's magnolia. All magnolia, in every room"

These words of guidance were spoken to me by a technical lead when establishing the level of code 'complexity' for a project a while back. I'm putting 'complexity' in scare-quotes here because 'complexity' wasn't really referring to something like the number of moving parts, it was referring to coding style. Without looking to bias anyone before I get to the discussion, I think it's accurate to say we were actually talking about code modernity. Allow me to elaborate. Or don't - close the tab, y'know - you don't have to do what I say Smile

A quick aside - I'm about to talk about 'average programmers', and I'm aware that by implication I'm placing myself in the 'above-average' group. I'm additionally aware there could be an arrogance (perceived or real) about doing this, but… hey. That's the impression I've got from working in various teams at several companies on different projects. And it's my blog Smile

So. *ahem* Dear reader - it's probably fair to say that if you are a programmer, you're an above-average one. I feel some confidence in saying this because research has shown and experience has attested that the average programmer does not read programming blogs, and yet here you are, reading one. The average programmer 'learns how to code' when they're starting out and expands their horizons and improves only a little over the course of their career. This means that most programmers get used to the style and practices of coding that were most popular when they were starting out, and anything introduced and established later looks foreign and strange, and - depending on how set-in-their ways they are - even convoluted and over-complex. Let's look at some concrete examples.

  • LINQ has been around since the .NET framework 3.5 in late 2007. That's [at the time of writing] more than seven years.
  • F# has been around since mid 2005. Using F# as a symbol for the use of functional programming in .NET, that's more than nine years.

So .NET programmers have had nearly a decade to stumble across functional programming and long enough to get itchy in a marriage to play around with LINQ. It has been my experience however that most programmers aren't particularly familiar or comfortable with either. In the name of fairness I should point out that when exposed to either, most of 'most programmers' are interested and want to learn how it all fits together, but the point remains that it at first just looks like 'complexity'.

Some examples. These two code blocks both print the numbers 1 to 10, grouped by the remainder when you divide each by 2. See what you think:

var numbersByRemainder = new SortedList<int, List<string>>();

for (var i = 1; i <= 10; i++)
{
    var remainderFromDivisionByTwo = i % 2;

    if (!numbersByRemainder.ContainsKey(remainderFromDivisionByTwo))
    {
        numbersByRemainder[remainderFromDivisionByTwo] = new List<string>();
    }

    var name = i.ToString(CultureInfo.InvariantCulture);

    numbersByRemainder[remainderFromDivisionByTwo].Add(name);
}

foreach (var remainder in numbersByRemainder.Keys)
{
    var numbers = string.Join(", ", numbersByRemainder[remainder]);

    Console.WriteLine("Remainder: {0}, numbers: {1}", remainder, numbers);
}

…and:

Enumerable
    .Range(1, 10)
    .Select(i => new
    {
        Name = i.ToString(CultureInfo.InvariantCulture),
        RemainderFromDivisionByTwo = i % 2
    })
    .GroupBy(i => i.RemainderFromDivisionByTwo)
    .Select(grp => new
    {
        RemainderFromDivisionByTwo = grp.Key,
        Numbers = string.Join(", ", grp.Select(i => i.Name))
    })
    .OrderBy(i => i.RemainderFromDivisionByTwo)
.ToList
() .ForEach(i => Console.WriteLine( "Remainder: {0}, numbers: {1}", i.RemainderFromDivisionByTwo, i.Numbers));

You can see the first one in action here, and the second one here. Those two links by the way both go to DotNetFiddle, which for my money (of which I'm offering none) is one of the best things on the internet, especially for programmers Smile

So, both code blocks accomplish the same task - which do you find easier to read? Personally, it's the second one. I start with a sequence of numbers and run each one through a series of declared operations to get the result - it reads to me like a series of English commands, and my brain wraps around that more easily. The grouping and ordering of the results is explicitly stated instead of being implied by the use of the SortedList. On blogs and in books I read, code like this is more and more common. You have to admit, though - it's not Magnolia.

And so to the point I want to address. On any project, there's a tension between writing code everyone finds easily accessible, and using more [relatively] modern techniques which can make code more terse, expressive and perhaps more attractive to future 'above-average'-programmer recruits. This tension should be resolved with a team discussion and an agreed approach - the earlier the better. I think it does everyone involved a service to aim higher than the most junior or inexperienced team members - keeping in mind that they will need coaching at least to begin with - and by doing so raise the skill level of the average team member and therefore the capabilities of the team. For this to be viable, anyone identifying themselves as 'above average' has a responsibility to try and educate the team they're on and provide the coaching where necessary. For work life to be tolerable however, that same group has to keep their mouths shut and paint everything magnolia if that's the decision that's taken.

You can always paint the Sistine Chapel at home Smile

Print | posted @ Thursday, January 1, 2015 3:42 PM

Comments on this entry:

Gravatar # re: Finding a Balance: Code 'Complexity'
by Tom at 1/2/2015 12:24 PM

Hi Steve! Nice article, I would be very happy to read more on this topic.

If you are fond of functional programming in C#, check out my hobby project called Endless:

https://github.com/tompazourek/Endless

I would be grateful for any feedback.
Post A Comment
Title:
Name:
Email:
Comment:
Verification: