Friday, February 27, 2009

Tinyrb Interview

After reading about tinyrb I wanted to ask it's developer, Marc-André Cournoyer (@macournoyer), a few questions about Ruby and what he's doing with it. Here's what we talked about.


Poking at your blog, it looks like you've done a lot of work building ruby implementations, or parts of them. Why? What's the value in this kind of hacking?

Marc-André VM implementation has been one of my interest for about a year, right after releasing Thin I think. I bumped into tinypy and thought it was the coolest idea ever. I tried porting it to Ruby with tinyrb and my first attempt (last summer) was a total disaster. But I learned a lot about Ruby's internal and YARV bytecode. It's very enlighting to understand how things work inside, that your if is compiled to a branchunless instruction. It's not just magic anymore.

Speaking of magic, when implementing a language there's this kind of magical moment when you run your first chunk of code. You've created all the parts but seeing them work together is a nice feeling. You're hooked once you've been thought that and know you'll be spending a lot of time on it. It's the perfect mix of art and science. You're creating a way to express yourself but at the same time reading all those crazy research papers.

I wish more people would stop redoing all those 20 lines Rails plugins and find the courage to learn something new again. Software is an amazing domain to be in. There's no limit to what you can do and the only thing you have to invest to push your limits is time. 2 years ago I was doing ASP.NET and didn't know the difference between a GET and POST request. A year ago I didn't know very much about C. I'm sad when I ear someone say they have no side project or that they just do Rails stuff. That's like eating the same meal every day.

What have you learned about Ruby from working with/looking at implementations of other languages?

Marc-André I've learned that Ruby is not that dynamic. There are much more powerful languages out there. Io for example, allows full introspection of the message chain. Meaning you can, amongst other things, control the evaluation of method arguments. Also it's prototype based like JavaScript and unlike Ruby, which is class based like Java. It's another way to structure your code. Learning new languages and programming paradigms helped me think outside of the box when I go back to Ruby.

Ruby is probably the best combination of simplicity, speed and power. But, if you think Ruby is the most powerful and extensible language, like I did, it's time for you to look at other languages.

Other than Ruby and C, what languages are you using or investigating, and why?

Marc-André I felt in love with Io's simplicity and power. It's an amazing language to play with. I don't know if it's usable for larger projects, but we sure can learn a lot from it. All language constructs are implemented in Io itself. You can add operators at runtime. And there's no parser, just a lexer that creates a chain of messages. Also, I had trouble with using space instead of dot for message separator at first, but after spending more time with it, I find it makes the code lighter to the eye. But because of the way it lets you evaluate arguments lazily, it's impossible to compile it to bytecode, which makes it bloody slow.

Lua is becoming famous for it's small and fast VM in the language community and it already is in the gaming industry. They did a couple things differently, like using a register based VM. The code is relatively simple and well structured. There are a couple great papers about Lua on the Internet which makes it a great starting point if you want to study a VM.

Potion was one of the main inspiration for tinyrb internal design. Although I'm not sure about the language syntax, I like the way it is implemented. In fact, I started coding on my own programming language, called Min, about the same time _why started Potion and we shared some of the same concepts, like using an open extensible object model, from a paper (pdf) by Ian Piumarta. A couple parts of tinyrb are directly derived (stolen) from Potion. So I owe _why a big thanks for this.

How serious is your TinyRb project? How far do you plan on taking it?

Marc-André If by serious you mean stable, then yes I hope to bring it to a stable state someday and make it usable for some limited "real world" usage. But I had the feeling all the Ruby implementations right now are a bit too serious. Each are supported by a company. When there's money involved, there's less freedom. I have no problem saying that tinyrb is the less serious Ruby implementation of them all. I want it to be the Ruby VM people use to learn and play with so they can write better code and maybe contribute to other implementations.

As for my specific goals with tinyrb. I'd like keep low memory footprint and be fast and complete enough to run small web/desktop apps. You know, when you don't need the full thing. By being small, it would enable you to use Ruby for small daemons that need to use as little memory as possible on servers and this kind of stuff. Also, I'd like to add features such as Sandboxing so you could run tinyrb inside another VM when you need to eval unsafe code.

How much do you interact with the other Ruby implementation projects?

Marc-André I've been a passive follower of Rubinius for a while. Evan and Brian noticed tinyrb and answered my questions about VM design. I hope someday to contribute something back as tinyrb is helping me understand more of Rubinius internals.

I have lots of admiration for all the people working full-time on a Ruby implementation. It requires constant learning about very complex things but at the same time answering questions of people with various knowledge level. It's very demanding, I'm sure.

What kinds of contributions would be most welcome for tinyrb?

Marc-André Anyone that wants to help can take a look at the TODO file in the project. The main goal right now is to run RubySpecs. So anything that can push it in that direction would be awesome. Here's a short list of things I need help with: write a better grammar (which I totally suck at), implement more core libs (IO, Dir), help find out what is missing to run RubySpecs or simply compile and run tests on your machine and report any error or warning.

But if you want to hack on something else, that's cool too. As long as it doesn't involve a kilt and 2 potatoes I'm OK with it.

Since you've managed to get your hands pretty dirty dealing with Ruby, what do you wish the language did differently?

Marc-André I wish Ruby had a simpler parser and trimmed down the syntax to remove useless stuff. I'm not sure what makes it that complex but having implemented my own in tinyrb I'm sure there's a way to keep what useful and simplify it.

I wish they'd remove everything that does not comply to the principle of least surprise. For example, the difference between proc and Proc.new, namespace lookup weirdness. Maybe if MRI team would use RubySpec a bit more we'd see less of this. I'm just guessing, I'm not following 1.9 development that much. But looks like they are breaking a couple RubySpecs on each release.

I wish MRI/YARV code had consistent indentation.

I wish there was a way to implement macros in Ruby, much like in Io with lazy arguments evaluation. This way we could implement all language constructs in Ruby (if, while, def, etc.) but that would kill the speed we've all been waiting for.

I wish Matz would stop wars, hunger, poverty and bring peace around the world, but I guess he's too busy saving us Rubyists first.

Click here to Tweet this article

1 comment:

ekse-0x said...

great interview. I really liked the part where Marc-André said he didn't know much about C a year ago but this hasn't stopped him from using it in projects. I'm currently studying computer science and I often refrain myself to start coding stuff because I feel I don't master the langage, while I should jump right in it and learn.