Book Image

Build Your Own Programming Language

By : Clinton L. Jeffery
Book Image

Build Your Own Programming Language

By: Clinton L. Jeffery

Overview of this book

The need for different types of computer languages is growing rapidly and developers prefer creating domain-specific languages for solving specific application domain problems. Building your own programming language has its advantages. It can be your antidote to the ever-increasing size and complexity of software. In this book, you’ll start with implementing the frontend of a compiler for your language, including a lexical analyzer and parser. The book covers a series of traversals of syntax trees, culminating with code generation for a bytecode virtual machine. Moving ahead, you’ll learn how domain-specific language features are often best represented by operators and functions that are built into the language, rather than library functions. We’ll conclude with how to implement garbage collection, including reference counting and mark-and-sweep garbage collection. Throughout the book, Dr. Jeffery weaves in his experience of building the Unicon programming language to give better context to the concepts where relevant examples are provided in both Unicon and Java so that you can follow the code of your choice of either a very high-level language with advanced features, or a mainstream language. By the end of this book, you’ll be able to build and deploy your own domain-specific languages, capable of compiling and running programs.
Table of Contents (25 chapters)
1
Section 1: Programming Language Frontends
7
Section 2: Syntax Tree Traversals
13
Section 3: Code Generation and Runtime Systems
21
Section 4: Appendix

Type representation in the compiler

Frequently, our compiler will need to do things such as compare the types of two variables to see whether they are compatible. Program source code represents types with string data, which is incorporated in our syntax tree. In some languages, it might be possible to use little syntax subtrees to represent the types that are used in type checking, but in general, type information does not exactly correspond to a subtree within our syntax tree. This is because part of the type information is pulled in from elsewhere, such as another type. For this reason, we need a new data type just to represent the type information associated with any given value that is declared or computed in the program.

It would be nice if we could just represent types with a single atomic value such as an integer code or a string type name. For example, we could use 1 for an integer, 2 for a real number, or 3 for a string. If a language had only a small, fixed set of built...