Why Python?

{.alignleft .floatleft}A friend who is a doctor is considering learning Python as his first programming language, to do some processing on some research data. He asked me to give him the 30 second elevator pitch for Python, to evaluate whether it's a wise choice. I enjoyed constructing the reply so much that I decided to post it here, just in case it helps anyone else in a similar situation.

Why Python?

Python is very accessible and intuitive. You should be able to produce simple, useful programs in your first day of experimentation. The syntax is clean and concise, without too much cryptic punctuation (Perl, I'm looking at you), redundancy or unnecessary verbosity.

This accessibility isn't just a superficial convenience. Because of it, writing a program in Python will take noticeably less time than many other programming languages. The resulting program will be shorter and more comprehensible, and will be easy to modify or extend in the future.

The simplicity of Python is not because it is in any way cut-down or incapable. In fact, it is one of the most limber languages available, including a carefully chosen cross-section of advanced language design features, which enable it to adapt gracefully to many different situations and programming styles. Its beauty lies in its ability to provide the aforementioned simplicity regardless of the complexity of the task to which you choose to put it.

Of those language design features, a couple are worthy of special mention.

Python is one of a number of dynamic languages, which are in vogue at the moment. Proponents would say that the entire history of programming has been a gradual migration towards progressively more dynamic languages. Dynamic languages, amongst other things, allow you to write programs that modify themselves when they run. Instead of simply writing a function yourself, you can instead write a function which creates a second function, and then call this second function, which will do the thing you want done. This, and other sorts of brain-bending meta-programming, seem a little abstract at first, but sometimes allow some tremendous conceptual ju-jitsu, allowing very small amounts of code to achieve enormous things.

Secondly, Python's dynamism facilitates a programming style known as test driven development, of which I am big fan. The idea is that for every bit of code you write, you also write a test, which verifies that your code is doing the right thing. It isn't immediately obvious that this is necessarily a very useful thing to do, but in practice it reaps tremendous benefits. I evangelise about it often, because I feel it is the single most important thing that most programmers could do in order to be more productive and write better code.

As well as the language itself, Python comes bundled with a comprehensive set of pragmatic built-in standard libraries, which your program can lean on to help you get things done with a minimum of hassle. These libraries are augmented by a vibrant community of authors producing third-party modules you can download and use as well.

As any good language should be, Python is cross-platform, so with a minimum of tweaking, most Python programs should run on Windows or Macs or Linux.

Why not Python?

A notable alternative to Python is Ruby, which looks like a delightful environment and community to be in. As a general-purpose tool, Ruby is just as good as Python, and it excels in certain areas such as website development. But Ruby is not compellingly better than Python. They are more similar than they are different, and form healthy rivals.

There are other languages that are better than Python at particular things, but none, in my opinion, are better than it for most things.

Something like C++ is better for sheer speed of program execution, or for addressing the low-level bits and bytes that make up the electronics of your computer. But it takes years to master C++. It's a hard-core programmers language. I spent seven years living and breathing it, and feel qualified to say that its practitioners can be slightly masochistic about its inaccessible superiority. Even once mastered, it is still a lot of work to write C++ programs.

Java and C# are both very popular indeed - orders of magnitude more so than Python, and are ubiquitous in conservative corporate enterprise consulting shops. Both are slightly frowned-upon by computer science academics (C#, for example, for being ostensibly tied to Windows), but nevertheless, these languages are not bad choices for many people.

Programs written in Python are usually slower than any other mainstream programming language. This could be an issue if you intend to intensively crunch large amounts of data in CPU intensive ways, for example running a finite element analysis.

There are many Python libraries you can call which are, under the covers, written in C. A prominent example is NumPy, for doing numerical processing. Libraries like this might circumvent the performance issue if one of them happens to handle your particular problem.

Even if there is no appropriate library available, slow performance isn't as serious a drawback as it sounds. 99% of programs don't need to do these sort of CPU intensive tasks, so Python's slowness makes no discernible difference. Even in cases where performance is a factor, Python makes it easy to modify and optimise your code to make it run faster, which often alleviates the problem entirely.

Python uses indentation to define blocks of code instead of 'begin/end' or '{}' delimiters like other languages. This caused no small amount of controversy when it was introduced, with many veteran programmers recoiling in horror, imagining nightmare scenarios in which simply changing the whitespace in a program (eg adding more spaces or tab characters) would unexpectedly change a program's behaviour. In practice, however, this does not ever cause problems, and actually eliminates an entire class of errors, wherein a programs appear to behave strangely because the programmer has failed to keep the indentation (which is useful to human readers of the code) in sync with the delimiters (which are used by the computer.)

Multi-threading is an advanced technique in which a program casts off new versions of itself, all running around simultaneously helping each other out, sorcerers apprentice style. Python does not handle this well, only utilising a single CPU on dual or quad core machines, and often requiring careful crafting of finicky constructs to get it working reliably. However, this is equally awkward in almost every other language, and has had programmers tearing their hair out for decades, no matter what language they use. There are exciting new approaches to this in the language Erlang, but this is still too fringe to recommend as a first language.

Python lacks some of the delightful academic brilliance of hardcore functional languages such as Lisp and its derivatives, which are based on the mathematics of the lambda calculus. In the right hands, these tools can be devastatingly elegant and highly productive. However, many of them lack a degree of day-to-day practicality in terms of available libraries, and most people feel that they are initially unintuitive to learn. Such languages will no doubt remain highly influential in computer science circles, and are having something of a renaissance these days, but they are sufficiently unorthodox for me not to recommend as someone's first (and possibly only) programming language.