Rust - Differens

The Rust extractor maps tree-sitter Rust CST nodes to canonical concepts. It detects structural changes at the item level: functions, structs, enums, traits, and their members.

Detected concepts

tree-sitter node type	Canonical concept
`function_item`	Function
`struct_item`	Struct
`enum_item`	Enum
`trait_item`	Trait
`impl_item`	Impl
`function_item` inside `impl`	Function
`use_declaration`	Import
`mod_item`	Module
`macro_definition`	Macro
`let_declaration`	Variable
`const_item`	Constant
`type_item`	TypeDef

Example

Two versions of a Rust module. One renames a function and adds a struct field:

old.rs

use std::num::ParseIntError;

pub fn parse_config(input: &str) -> Result<Config, ParseIntError> {
    Ok(Config { workers: input.parse()? })
}

pub struct Config {
    pub workers: u32,
}

new.rs

use std::num::ParseIntError;

pub fn load_config(input: &str) -> Result<Config, ParseIntError> {
    Ok(Config { workers: input.parse()?, timeout_ms: 1000 })
}

pub struct Config {
    pub workers: u32,
    pub timeout_ms: u64,
}

Running differens old.rs new.rs:

  ~ renamed function `parse_config` to `load_config`
  ~ modified struct `Config` (1 addition)
  ~ modified function `load_config` (1 addition)

Known limitations

Macro expansion is not followed. The extractor works on the surface AST, so code generated by macros (derive, proc macros) is invisible to the diff.
impl blocks use the Impl concept, and their function items remain Function nodes. The matcher uses syntax and context; it does not resolve trait implementations or apply Rust orphan rules.
Attributes have Attribute or InnerAttribute nodes. JSON keeps their edits; human reports may group them under a declaration.