lotuslang/lotus

A toy language/compiler written from scratch in C#

lotus

A toy language for personally exploring compilers and language design.

You can find some (somewhat outdated) notes and drafts about lotus in the wiki, although a large amount of ideas are just floating around in my personal notes. In either case, lotus is an ever-changing thing, so most samples and document are subject to documentation-rot. For example, some features (such as enums or generic types) are currently severely limited or even not supported at all. Once I'm satisfied with the parser and front-end for lotus (e.g. good CLI, or the ability to process a project instead of a single file), I'll most likely write a sample project to test things on; but until then test.lts is the most up-to-date example of lotus code.

Usage

Although lotus is nowhere-near usable, it is distributed with a very basic CLI, currently built with System.CommandLine. It supports checking an input file for parsing errors, as well as output the parse tree in Graphviz/DOT format (optionally with const-coloring). It can also try to reconstruct the original source file solely from the parse tree. For more info, run dotnet run --project ./src/Lotus.CLI/ -- -h.

lotus's Issues

Pre-process text to get rid/replace illegal characters or sequences

For example, replacing \r\n with \n

Logger doesn't enforce/add consistent punctuation and casing

For example, it'd be nice for the logger to automatically add periods at the end of messages that don't have any, or to make sure that there's no uppercase letter in the middle of a sentence.

Migrate `lotus-samples` notes to this repo's wiki

Write a README.md file

How should GraphMakers be organized ?

Should we put all GraphMakers into the same (partial) class but separate each part (tl/stmt/expr) into different files, each in the right folder ?

Make `LotusError.ExtraNotes` able to use (small) code samples and format them from the logger

We should make it possible for ExtraNotes to contain code sections with e.g. pointed locations or diff-style lines. They'd be formatted by the logger (or some logger helper class) obviously, but probably shouldn't have the same style as the user's invalid code.

Escape sequences in strings and char

Should visitors that accept only values also accept `StatementExpressionNode` ?

Should simple semantic checks be done during parsing, or during semantic analysis ?

Checks like IsName() are pretty simple and don't require any other context; as such, they can be done during parsing.

However, this might make them harder to extend or maintain, and having a weird separation like that may come back to bite us later.

It also might be better to do all the checking at analysis-time, because right now some nodes are valid while others not.