Generz NPM | npm.io

generz

A programming (?) language that generates code for building compilers.

Preface

Compilers have various architectures. The most common one is building a pipeline, consisting in some phases, where output of one phase is the input for the next phase. Compilers are commonly separated in front-end (source language dependent) and back-end (source language independent):

Front-end
- Lexical analysis
- Syntax analysis
- Semantic analysis
Back-end
- Intermediate code generation
- Code optimization
- Target code generation

The de facto standard for the back-end implementation is the LLVM project.

Meanwhile, as the front-end is source language dependant, it's implementation is generally coded (manually written and/or partially generated) by the compiler's developers.

Description

The purpose of generz is to generate highly-optimized code*, that implements specific lexical and syntax analysis phases, AKA code that accepts input source (sequence of characters) and emits AST (Abstract Syntax Tree), leaving to the developers of a compiler to implement just the semantic analysis and interact with the intermediate code generation.

The behavior of the generated phases is described with the Generz Language (.erz source files).

*As highly-optimized code is intended that kind of code that sacrifices readability in favor of performances, reducing everything to a low-level of abstraction and builds an API that should be considered as a magical black box, that transform the input in the proper output.

Project status

TLDR: The generz compiler is capable of generating code (in TypeScript) that validates sources in a custom syntax. AST is not yet built from the generated code.

Generated code for TypeScript, why? Because generz itself is written in TypeScript and one of the first things it should do is generating code for itself. Obviously the next targets are C++ and Rust.

The following steps would be:

`v0.9` - Language capable of building AST

Design the language so that is becomes capable of building AST from the derivation-tree, using user-defined data structures of the target programming language.

`v0.10` - Generz builds its own phases

At this point, with some minor feature implementations, generz will be capable of producing code for its own use, and therefore run on code produced by itself.