I’m currently reading Steve McConnell’s Code Complete (for the first time – yes, I know that’s somewhat worrying) and there was one section was disturbed me a little. For those of you with a copy to hand, it’s in section 4.3, discussing the difference between programming in a language and programming into a language:
Programmers who program “in” a language limit their thoughts to constructs that the language directly supports. If the language tools are primitive, the programmer’s thoughts will also be primitive.
Programmers who program “into” a language first decide what thoughts they want to express, and then they determine how to express those thoughts using the tools provided by their specific language.
Now don’t get me wrong – I can see where he’s coming from, and the example he then provides (Visual Basic – keeping the forms simple and separating them from business logic) is fine, but he only seems to give one side of the coin. Here’s a different – and equally one-sided – way of expressing the same terms:
Programmers who program “in” a language understand that language’s conventions and idioms. They write code which integrates well with other libraries, and which can be easily understood and maintained by other developers who are familiar with the language. They benefit from tools which have been specifically designed to aid coding in the supported idioms.
Programmers who program “into” a language will use the same ideas regardless of their target language. If their style does not mesh well with the language, they will find themselves fighting against it every step of the way. It will be harder to find libraries supporting their way of working, and tools may well prove annoying. Other developers who come onto the project later and who have experience in the language but not the codebase will find it hard to navigate and may well accidentally break the code when changing it.
There is a happy medium to be achieved, clearly. You certainly shouldn’t restrict your thinking to techniques which are entirely idiomatic, but if you find yourself wanting to code in a radically different style to that encouraged by the language, consider changing language if possible!
If I were attacking the same problem in C# 1 and C# 3, I could easily end up with radically different solutions. Some data extraction using LINQ in a fairly functional way in C# 3 would probably be better solved in C# 1 by losing some of the functional goodness than by trying to back-port LINQ and then use it without the benefit of lambda expressions or even anonymous methods.
Accents and Conventions
That’s just between different versions of the same language. Between different actual languages, it can get much worse. If you’ve ever seen Java code written in a C++ style or vice versa, you’ll know what I mean. I’ve previously referred to this in terms of speaking a language with an accent – you can speak C# with a Java accent just as you can speak French with an English accent. Neither is pleasant.
At the lowest level, this is likely to be about conventions – and I’m pretty sure that when Steve writes “Invent your own coding conventions, standards, class libraries, and other augmentations” he doesn’t actually mean us to do it in a gratuitous fashion. It can be worth deviating from the “platform favoured” conventions sometimes, particularly if those differences are invisible to clients, but it should always be done with careful consideration. In a Java project I worked on a few years ago, we took the .NET naming conventions for interfaces (an I prefix) and constants (CamelCasing instead of SHOUTY_CAPS). Both of these made the codebase feel slightly odd, particularly where Java constants were used near our constants – but I personally found the benefits to be worth the costs. Importantly, the whole team discussed it before making any decisions.
At a slightly higher level, many design patterns are just supported much, much better by some languages than others. The iterator pattern is a classic example. Compare the support for it from Java 6 and C# 2. On the “client” side, both languages have specific syntax: the enhanced
for loop in Java and the
foreach loop in C#. However, there is one important difference: if the iterator returned by
IDisposable (which the generic form demands, in fact) C# will call
Dispose at the end of the loop, no matter how that occurs (reaching the end of the sequence, breaking early, an exception being thrown, etc). Java has no equivalent of this. Imagine that you want to write a class to iterate over the lines in a file. In Java, there’s just no safe way of representing it: you can make your iterator implement
Closeable but then callers can’t (safely) use the enhanced for loop. You can make your code close the file handle when it reaches the end, but there’s no guarantee that will happen.
Then consider the “server” side of the iterator – the code actually providing the data. Java is like C# 1 – there’s no specific support for implementing an iterator. In C# 2 and above, iterator blocks (i.e. methods with
yield statements) make life much, much easier. Writing iterators by hand can be a real pain. Reading a file line by line isn’t too bad, leaving aside the resource lifetime issue – but the complexity can balloon very quickly. Off by one errors are really easy to introduce.
So, if I were tackling a project which required reading text files line by line in various places, what would I do? In Java, I would take the reasonably small hit of a
while loop in each place I needed it. In C# I’d write a
LineReader class (if I didn’t already have one!) and use a more readable
foreach loop. The contortions involved in introducing that idea into Java just wouldn’t be worth the effort.
At a much higher level, we get into whole programming styles and paradigms. If your natural inclination is to write imperative code, you’re likely to create a mess (or get very frustrated) in a functional language. If the problem really does call for a functional language, find someone else to help you think in a more functional way. If the problem suits imperative programming just as well as it does functional programming, see if you can change the environment to something more familiar.
I’m not suggesting that Steve’s point isn’t valid – but he’s done his readers a disservice by only presenting one side of the matter. Fortunately, the rest of the book (so far) is excellent and humbling – to such a degree that this minor quibble stuck out like a sore thumb. In a book which had more problems, I would probably barely have even noticed this one.
There’s another possibility, of course – I could be competely wrong; maybe I’ve been approaching problems from a restrictive viewpoint all this time. How about you?