# Rust Abstraction > Use when deciding whether a long Rust file in the koan repo has an extractable seam — and what *shape* the extraction should take. File-level companion to `modgraph` (complexity scoring) and `doc-abstraction` (cross-doc concept surfacing); seams proposed here can be scored with `modgraph item`. Reach for it when asked "can this be simplified?", "are there seams here?", "what should come out of this file?". - Skill: `tomevault-io/rust-abstraction` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/rust-abstraction` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/rust-abstraction/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/rust-abstraction --- # rust-abstraction Decide whether a long Rust file has seams worth extracting, and if there are, how to extract each. ## Check the cheap fixes first Before looking for abstractions, see if a mechanical fix handles it. - **Tests inline?** Measure prod vs. test lines. If tests are >20% of the file and live in `#[cfg(test)] mod ...` blocks, lift them to `foo/tests.rs` per the project convention. That alone often resolves "this file is too long." - **Hand-written `Clone`/`Debug` that could `derive`?** Replace and stop. - **Dead code from a half-done refactor?** Delete and stop. ## Seam smells Look for these if there are no cheap fixes. One strong signal is enough; three weak ones is not. - **Inline `struct`/`enum` in a function body.** The author already factored the *concept* but didn't lift the *boundary*. Promote it. - **Repeated loop or match shape with varying bodies.** N near-identical `for/match` walks differing only in body or predicate → consolidate. - **Read-only narrow access to a wide type.** A method that touches one or two fields of a large struct and never the rest. That slice *is* the boundary. - **Load-bearing invariant living as a docstring.** A rule that must be preserved (cache-safety, ordering, "don't recurse here") is one rename away from being lost — promote it to a type name. ## When picking a shape Your goal is to make the code easier to read and safer to edit. Try to encapsulate data and invariants. - **No empty wrappers.** If `Foo::new(thing).do_x()` just renames `Thing::do_x(thing)`, reject it. The test: *can you state in one sentence what the new type guarantees?* If not, it's a wrapper. ## Workflow 1. Read the file end to end (not skim). 2. Run it by the **cheap-fixes** list, then the **seam smells** list. Give honest judgment - weak signals are not seams. 3. If you find one or more seams, report the name of the proposed abstraction and (briefly) how it improves the code. 4. f you don't find a seam: report "no seams found" and move on. --- > Source: [jackhall/koan](https://github.com/jackhall/koan) — distributed by [TomeVault](https://tomevault.io).