MJr

MJr-compiler implementation notes

These notes explain some of the design and implementation details of the MJr-compiler project, which may be of interest to potential contributors, people developing similar projects, or anyone who is curious.

Pipeline

The compiler works in several stages, using multiple intermediate representations:

• The tokenizer converts raw MJr source code into a sequence of tokens.
• The parser converts this sequence of tokens into an abstract syntax tree (AST).
• The resolver converts the AST into an abstract semantic graph (ASG) and performs static checks.
• The compiler itself converts the ASG into a high-level intermediate representation (IR) for the output code.
• The final output is then generated by the code generator for the target language.

Next steps

Several features of the MJr programming language are not yet supported:

• convchain, map and path statements
• convolution statements and sum expressions
• load expressions
• field and observe rules
• Search
• Periodic grids

Additionally, there some features of the MarkovJunior programming language which are not yet included:

• 3D grids
• Wave Function Collapse statements

Some of these features need some more thought to the language design before they are implemented. See also the project’s GitHub issue tracker.

This site is open source. Improve this page.