marnix / zigmmverify

Note that this branch is built using nightly 'master' zig. The more actively developed branch is zig-0.9.x. Changes are periodically merged from there to this branch.

A Metamath proof verifier in Zig

At least, the beginnings of one.

This is partly to learn Zig, partly to have a fast verifier that I can hack (Python is way too slow, Rust is too weird for me still, Nim seems less clean, C is out), so that I can perhaps one day try to do experiments around parsing (where parse tree of X... is proof of |- TOPLEVEL X... ), abbreviations, and modules.

How to build and run

Build using Zig master, then just run the resulting binary, passing a single .mm file on the command line.

For example, use set.mm, which you can get from metamath.org, or download the most recent version directly from GitHub: https://github.com/metamath/set.mm/raw/develop/set.mm.

The version of the Metamath specification that was implemented, is the one from the 2nd edition Metamath book. (Norman D. Megill, David A. Wheeler, "Metamath: A Computer Language for Mathematical Proofs". Lulu Press, 2019. http://us.metamath.org/downloads/metamath.pdf .)

Next tasks

Verifier completeness:

• Verify that 'normal' tokens don't start with $, label tokens use only allowed characters, etc.

• Support ? in proofs.

Clean up:

• Clean-up / refactor RuleIterator + ScopeDiff: Add methods, move functionality between these structs. Also encapsulate some parts. Also try to avoid duplication in statement handling.

• Work towards a library that can also be used for $a/$p/$e parse trees.

Functionality:

• Generate a parse tree for every $a/$p/$e statement. If possible/feasible, check that it is the only possible parse tree. Try to also support 'conditional syntax rules, so e.g. ${ $e -. B = 0 $. $a class ( A / B ) $. $} which expresses that A / B is only a syntactically valid expression if B is not zero.

Verifier performance:

• Optimize performance by reducing memory use: A token is currently represented by a Zig slice (= pointer and a length), and this could be replaced by a single small(ish) integer by 'interning' all tokens.

• Optimize performance by parallelizing using Zig async.

• Optimize performance by reducing memory use: I suspect ArrayList is not the right data structure for some of the lists.

Verifier tests:

• Run the test suite from https://github.com/david-a-wheeler/metamath-test probably by checking it out as a Git submodule and using the zig-mm-verify binary as the approved-or-not judge.

• (Separate project.) Extend that test suite, to capture as many as possible deviations from the specification as possible.

Verifier usability:

• Identify the location (line/column) of at least the first error.

• Don't use 'error union' for Metamath verification errors.

Language dialects:

• Optional modes where $c/$v is allowed to be duplicated (useful to create set-parsed.mm which declares stuff before $[ set.mm $]); where $f implicitly does $c/$v; and perhaps more.