Shahzad Bhatti

March 25, 2009

When in Rome, code like how Romans code

Filed under: Languages — admin @ 12:11 pm

I have been programming for over twenty years and I have learned a number of programming languages over the years. One of recurring behavior I have seen in a lot of programmers is that they take a lot of programming habbits (good or bad) from old language(s) to the new language. This could be how you design the application, style of coding, naming conventions, etc. I remember when I switched from C to C++, I was used to procedural thinking and had to learn how to break the problem into classes and how to assign responsibility to different classes. Similarly, when I starting using Java back in 95-95, I had to learn about Java’s peculiar style. For example, I used to declare public methods in C++ at top and all private methods including attributes at the bottom. I also tended to use underscores to prefix member attributes. I slowly learned Java’s style of declaring class attributes at top, using all uppercase for constants, camel case, etc.

In early 2000s, I learned Ruby from PicAxe Ruby book that taught me Ruby from object oriented style and I missed all its functional or meta-programming features. I slowly learned more functional style of programming and meta programming. I even had to switched back to underscores as opposed to camel case. I read Ruby code of other programmers to learn how they code and what conventions they use. I did similar excercises when I learned Python, Erlang, Scala, Objective-C, etc, i.e., I tried to learn not only language itself, its core and third party libraries but how people write the code, package applications or create libraries. Though, I think it helps if there are examples of good usage or style for that language. For example, I have seen plenty of abuses of Javascript that misunderstood its prototype or functional roots and used it as either procedural or class oriented language.

At my work, we use code reviews before any code checkin and I see conventions and styles of other languages mixed in all the time. I think learning different styles of programming makes you a better programmer. For example, I learned from functional programming how immutability can make sure programs safer and I tend to use it more in other object oriented or multi-paradigm langauges that don’t enforce immutability. Though, in other cases it’s hard to force yourself to use features from one language to another when that feature isn’t available inherently. For example, I like mixins feature of Ruby or traits of Scala but I can’t really use them in langauges that support only single inheritance such as Java. So instead of jumping over hoops to use features from other language, I try to use the style suitable for that specific langauge such as using multiple interfaces. I have been learning iphone development and been reading iphone SDK book by Jonathan Zdziarski. One of peculiar thing about his coding examples is that he does not use Interface Builder and creates all UI components from the code. Though, such style is acceptable in many situations but I would prefer to use Interface Builder and follow path of least resistence.

In practice, you will often find multiple styles or approaches of doing a thing in a single language. For example, Ruby encourage multiple ways to do things that can be quite confusing. Though, I like Python’s philolsophy of only one way to do things, but there are plenty of divergent opinions in that language as well. Another somewhat related topic is how to pick a language as languages vary in their core areas of strength. For example, Java was originally marketed as language for Web platform but these days I tend to use Ruby or Python for web development and Java for system development. Also, I tend to use Erlang for network oriented or concurrent applications and use C/C++ where performance is critical. Last year, there was big hoopla over Erlang’s aweful performance for search engine that sparked WideFinder benchmarks but it missed the point that Erlang’s core strength is distributed/concurrent applications and not text searching. So in nutshell, I think it helps to pick a language based on the problem and take advantage of its strengths. Finally, stick to general style of coding and conventions of the language especially when working with large codebase or large number of programmers.

Powered by WordPress