Rust-LaTeX-Parser

LaTeX math parser for Rust. Takes a string, gives you an AST. No dependencies.

Install

[dependencies]
rust-latex-parser = { git = "https://github.com/William-Selna/Rust-LaTeX-Parser" }

Usage

use rust_latex_parser::{parse_equation, EqNode};

let tree = parse_equation("\\frac{-b \\pm \\sqrt{b^2 - 4ac}}{2a}");

match tree {
    EqNode::Frac(numerator, denominator) => {
        // do whatever you want with it
    }
    _ => {}
}

You get back an EqNode tree. Render it, convert it, inspect it — that part's on you.

What it handles

Fractions (\frac{a}{b}, a/b), roots (\sqrt{x}), super/subscripts, matrices, cases, big operators with limits (\sum, \int, \prod), trig and log functions, accents (\hat, \vec, \bar), \left...\right delimiters, math fonts (\mathbb, \mathcal, etc.), binomials, over/underbraces, stacked relations, and 130+ symbols (Greek, arrows, set theory, relations, the usual).

You can also skip backslashes for common stuff — pi works, sqrt(x) works, int_0^1 works. Parens after barewords act as grouping, so sqrt(x+1) grabs the whole x+1.

The tree

19 variants:

EqNode::Text(String)                    // "x", "42", "α"
EqNode::Space(f32)                      // inserted around operators
EqNode::Seq(Vec<EqNode>)               // horizontal sequence
EqNode::Sup(base, superscript)
EqNode::Sub(base, subscript)
EqNode::SupSub(base, sup, sub)
EqNode::Frac(numerator, denominator)
EqNode::Sqrt(contents)
EqNode::BigOp { symbol, lower, upper }
EqNode::Accent(contents, kind)
EqNode::Limit { name, lower }
EqNode::TextBlock(String)               // \text{...}
EqNode::MathFont { kind, content }
EqNode::Delimited { left, right, content }
EqNode::Matrix { kind, rows }
EqNode::Cases { rows }
EqNode::Binom(top, bottom)
EqNode::Brace { content, label, over }
EqNode::StackRel { base, annotation, over }

There's also EqMetrics (width/ascent/descent) if you need it for layout.

Symbol lookup

latex_to_unicode works standalone if you just need that:

use rust_latex_parser::latex_to_unicode;

latex_to_unicode("alpha");      // Some("α")
latex_to_unicode("rightarrow"); // Some("→")
latex_to_unicode("garbage");    // None

Error handling

It doesn't error. Bad input gives you a best-effort tree — unmatched braces get skipped, unknown commands turn into text nodes. I needed this for a live editor where the input is always half-finished.

Tests

cargo test

95 tests — parser correctness, adversarial input (deep nesting, emoji, 10k tokens), real-world expressions (quadratic formula, Gaussian integral, Schrodinger, Maxwell), Greek alphabet coverage.

License

MIT