Learning to write a compiler [closed]

Answer 1

I think Modern Compiler Implementation in ML is the best introductory compiler writing text. There's a Java version and a C version too, either of which might be more accessible given your languages background. The book packs a lot of useful basic material (scanning and parsing, semantic analysis, activation records, instruction selection, RISC and x86 native code generation) and various "advanced" topics (compiling OO and functional languages, polymorphism, garbage collection, optimization and single static assignment form) into relatively little space (~500 pages).

I prefer Modern Compiler Implementation to the Dragon book because Modern Compiler implementation surveys less of the field--instead it has really solid coverage of all the topics you would need to write a serious, decent compiler. After you work through this book you'll be ready to tackle research papers directly for more depth if you need it.

I must confess I have a serious soft spot for Niklaus Wirth's Compiler Construction. It is available online as a PDF. I find Wirth's programming aesthetic simply beautiful, however some people find his style too minimal (for example Wirth favors recursive descent parsers, but most CS courses focus on parser generator tools; Wirth's language designs are fairly conservative.) Compiler Construction is a very succinct distillation of Wirth's basic ideas, so whether you like his style or not or not, I highly recommend reading this book.

Answer 2

Take a look at the book below. The author is the creator of ANTLR.

Language Implementation Patterns: Create Your Own Domain-Specific and General Programming Languages.

Answer 3

One book not yet suggested but very important is "Linkers and Loaders" by John Levine. If you're not using an external assembler, you'll need a way to output a object file that can be linked into your final program. Even if you're using an external assembler, you'll probably need to understand relocations and how the whole program loading process works to make a working tool. This book collects a lot of the random lore around this process for various systems, including Win32 and Linux.

Answer 4

I found the Dragon book much too hard to read with too much focus on language theory that is not really required to write a compiler in practice.

I would add the Oberon book which contains the full source of an amazingly fast and simple Oberon compiler Project Oberon.

Answer 5

You might want to look into Lex/Yacc (or Flex/Bison, whatever you want to call them). Flex is a lexical analyzer, which will parse and identify the semantic components ("tokens") of your language, and Bison will be used to define what happens when each token is parsed. This could be, but is definitely not limited to, printing out C code, for a compiler that would compile to C, or dynamically running the instructions.

This FAQ should help you, and this tutorial looks quite useful.

Answer 6

If you're willing to use LLVM, check this out: http://llvm.org/docs/tutorial/. It teaches you how to write a compiler from scratch using LLVM's framework, and doesn't assume you have any knowledge about the subject.

The tutorial suggest you write your own parser and lexer etc, but I advise you to look into bison and flex once you get the idea. They make life so much easier.