Is there a more modern, OO version of "Let's Build a Compiler"? [closed]

Question

Is there a more modern, maybe object-oriented, equivalent to Jack Crenshaw's "Let's Build a Compiler" series?

A while back I stumbled across "Let's Build a Compiler" and could just not resist writing some code. I wrote a recursive-descent C compiler in C# that output .NET CIL. "Write once, leak everywhere" was my slogan.

Too bad I did not realize until too late that parsing C is a nightmare.

I am now interested in writing a Java compiler in Java that outputs .NET CIL or assemblies with the goal of being self-bootstrapping. I was hoping there might some newer tutorials kicking around.

As an aside, would you spend more time with up-front design or would you just write a ton of tests to support the ability to mercilessly refactor. Thinking back, I am leaning towards the latter. The compiler worked but the code was really awful.

Answer 1

It sounds like you completely missed the point of Crenshaw's tutorials. LBC isn't about writing pretty, clean, or efficient code. It's all about bringing something that's steeped in formal theory down to a level where the casual coder can easily and rapidly hack out a rudimentary (but working!) compiler.

When I read through LBC years back, I rewrote the examples in C#. I'm sure the class layout isn't the best, or tasks segregated properly, but it's comparable to his Pascal. I'd be happy to share the code with you if you like-- let me know and I can post it online and share the link.

In my spare time I've been hacking out some writing with the aim of unifying the philosophies of LBC and Basics of Compiler Design together-- walkling away with practical, working code at the end of each unit/chapter, with also discuss some theoretical stuff after exploring the ideas so the reader understands why things are the way they are. But it took Crenshaw years to write his incomplete series, so mine my be a pipe dream... and I use C (exactly because it's not C++ or Java).

Answer 2

Take a look at Terence Parr's "Language Implementation Patterns". He wrote ANTLR - a parser generator for Java - so knows his stuff. It explains the principles of compiler design really well and builds up gradually.

Martin Fowler's "Domain Specific Languages" is also good. It has a slightly different agenda than being a pure compilers course, but is a good reference on the key concepts of language design.

Answer 3

I'm a fan of "MiniJava" and associated work based on the "Modern Compiler Implementation in Java" family of books. This doesn't quite meet all the requirements you mention as a MiniJava implementation will, generally, generate native code - but the backend can easily be changed to emit MSIL or whatever.