#tree-traversal #generics #traversal #macro

uniplate

Simple, boilerplate-free operations on tree-shaped data types

3 releases

0.1.2 Sep 27, 2024
0.1.1 Sep 26, 2024
0.1.0 Sep 26, 2024

#350 in Rust patterns

Download history 483/week @ 2024-09-26 182/week @ 2024-10-03 353/week @ 2024-10-10 246/week @ 2024-10-17 188/week @ 2024-10-24

1,027 downloads per month

MPL-2.0 license

36KB
587 lines

Uniplate

Uniplate helps you write simple, boilerplate-free operations on tree shaped data types.

It is a port of the Haskell library Uniplate into Rust.


Quick links:

A simple example

Adapted from (Mitchell and Runciman 2007)

Uniplate makes the traversal and querying of tree shaped data easy and boilerplate free. A good use case of Uniplate is the manipulation of abstract syntax trees.

Consider the AST for a simple calculator language:

enum Expr {
    Add(Box<Expr>, Box<Expr>),
    Sub(Box<Expr>, Box<Expr>),
    Mul(Box<Expr>, Box<Expr>),
    Div(Box<Expr>, Box<Expr>),
    Val(i32),
    Var(String),
    Neg(Box<Expr>),
}

Say we want to list all the variable names used inside a given expression:

fn vars_names(expr: &Expr) -> Vec<String>{
    match expr {
        Add(a,b) => {
            [vars(a),vars(b)].concat()
        },
        Sub(a,b) => {
            [vars(a),vars(b)].concat()
        },
        Mul(a,b) => {
            [vars(a),vars(b)].concat()
        },
        Div(a,b) => {
            [vars(a),vars(b)].concat()
        },
        Val(a) => {
            Vec::new()
        },
        Var(a) => {
            vec![a.clone()]
        },
        Neg(a) =>{
            vars(a)
        }
    }
}

Functions like these are annoying to write: the first 4 constructors are basically identical, adding a new expression type requires a new line to be added to all match statement, and this code cannot be shared with similar functions (e.g. one that change all the variable names).

With Uniplate, this boilerplate can be eliminated:

use uniplate::{Uniplate,Biplate};
use uniplate::derive::Uniplate;
#[derive(Clone,PartialEq,Eq,Debug,Uniplate)]
#[uniplate()]
#[biplate(to=String)]
enum Expr {
    Add(Box<Expr>, Box<Expr>),
    Sub(Box<Expr>, Box<Expr>),
    Mul(Box<Expr>, Box<Expr>),
    Div(Box<Expr>, Box<Expr>),
    Val(i32),
    Var(String),
    Neg(Box<Expr>),
}

fn vars_names(expr: &Expr) -> Vec<String>{
    <Expr as Biplate<String>>::universe_bi(expr).into_iter().collect()
}

Uniplate also supports trees with multiple recursive types. Lets extend our calculator language to include statements as well as expressions:

enum Expr {
    Add(Box<Expr>, Box<Expr>),
    Sub(Box<Expr>, Box<Expr>),
    Mul(Box<Expr>, Box<Expr>),
    Div(Box<Expr>, Box<Expr>),
    Val(i32),
    Var(String),
    Neg(Box<Expr>),
}

enum Stmt {
    Assign(String, Expr),
    Sequence(Vec<Stmt>),
    If(Expr, Box<Stmt>, Box<Stmt>),
    While(Expr, Box<Stmt>),
}

When looking for variable names in a given statement, we want to identify not only the variable names directly used inside the statement, but also any variable names used by child expressions.

use uniplate::{Uniplate,Biplate};
use uniplate::derive::Uniplate;
#[derive(Clone,PartialEq,Eq,Debug,Uniplate)]
// look for strings inside expressions as well as statements 
#[biplate(to=String,walk_into=[Expr])]
#[biplate(to=Expr)]
#[uniplate()]
enum Stmt {
    Assign(String, Expr),
    Sequence(Vec<Stmt>),
    If(Expr, Box<Stmt>, Box<Stmt>),
    While(Expr, Box<Stmt>),
}

#[derive(Clone,PartialEq,Eq,Debug,Uniplate)]
#[biplate(to=String])]
#[uniplate()]
enum Expr {
    Add(Box<Expr>, Box<Expr>),
    Sub(Box<Expr>, Box<Expr>),
    Mul(Box<Expr>, Box<Expr>),
    Div(Box<Expr>, Box<Expr>),
    Val(i32),
    Var(String),
    Neg(Box<Expr>),
}

fn vars_names(stmt: &Stmt) -> Vec<String>{
    <Stmt as Biplate<String>>::universe_bi(stmt).into_iter().collect()
}

Despite having to recursively look through multiple types, this operation is no harder to write!

Acknowledgements

This library is inspired by Neil Mitchell's Haskell library Uniplate and its accompanying paper: Neil Mitchell and Colin Runciman. 2007. Uniform boilerplate and list processing.

Dependencies

~5MB
~101K SLoC