Saturday, January 12, 2013

Don't Learn a New Programming Language Every Year. Write one!

Don't learn a new language every year

Programmers often recommend learning a new programming language every year. I don't agree. I've studied many languages by reading their specs and dabbling with some code until I feel like I've got the basics, but I don't think I really learned them. My path to really "learning" Ruby showed me why.

Ruby was not a challenging language to learn. I picked up the Pragmatic Programmers book and read it end to end (an endorsement to both the language and the book). When the time came to start Imikimi, I chose Rails to do the web-side. However, I discovered that I hadn't really learned to *think* in the language until after several years of continuous professional use.

I was a C++ programmer before I was a Ruby programmer. A lot of my early Ruby code looks and feels a lot like C++ code. There were constructs in Ruby that I wasn't comfortable with, so I simply didn't use them. Over time I re-discovered language features I "learned" in my first reading of the Pragmatic book. Over time I learned how to program in Ruby-style from the community. By studying the Rails source and other gems I learned how to better use the language and how to write clean, readable code.

Learning a language by reading a book and writing a few small sample apps is equivalent to learning the grammar and a few phrases of spoken language. It's a shallow understanding which only broadens your perspective in shallow ways. I don't think you can claim to really know a language, spoken or programming, until you are conversant in it. It's a long, slow, but ultimately rewarding processes.

Here's my recommendation: To get started, do learn a few extra languages, shallowly. The second and third language you learn will broaden your horizons even if you don't dive too deep into them. After that, though, you need to focus on deepening your understanding either of your current language of choice, or a new one, through hours and hours of work on real projects, OR through my second recommendation:

Write a Toy language every few years

I'm writing a programming language. I've spent more than a decade thinking about programming languages and PL design. From 2001 until 2006 I worked on new language ideas full-time. Unfortunately none of that work saw the light of day, so I'm circling around again. I've learned a lot since then, primarily from Ruby and the Ruby community.

My current project, the River programming language, is an attempt to build the simplest Prototype-Based core language that can support a Ruby-like class-system on-top, written in the language itself. I've learned more about Ruby in the past few months than I have in the past few years.

You learn a lot about programming when you write a language implementation. It forces you to examine all the edge cases you normally ignore. And it happens fast. You are immediately forced into uncomfortable areas of programming languages you were unaware of. There are trade-offs to be made at every level. There are boot-strapping issues to resolve, and there are a host of subtle semantics in your host language that you never thought about.

How to get started

If you decide to follow this path, here are a few tips.
  • Use a flexible, agile host language. You will make many major refactors along the way, and you don't want to get bogged down in irrelevant details. This is true for any project, but especially true when working on something as meta as writing one language inside another. I'd recommend Ruby, but Python or even Javascript would work well.
  • Use a well-designed parser-generator (resource: my Babel-Bridge parser-generator gem)
  • Don't worry about performance at first, even if that is your primary goal.
  • Target Turing-complete functionality before all else. This will give you a solid base-camp to build out from as you explore the rest of your language. (resource: my How to write a Turing-Complete Programming Language in 40 minutes video)

You can follow my design progress on Twitter: #riverlang / @shanebdavis

Happy PL design!

3 comments:

  1. Thank you and looking for more posts. I am really satisfied with this posting that you have given us. This is really a stupendous work done by you.languageprestige.com

    ReplyDelete
  2. Wow, awesome!
    Recently I made a card game algorithm. And I thought exactly same thing you said.
    Make my own language is widen your perspective and force you to deal with the detail in the huge system.

    ReplyDelete