rlsp-fmt

A reusable Wadler-Lindig pretty-printing engine for the rlsp language server project.

Overview

rlsp-fmt provides a document IR (Doc) and a width-aware printer. You describe your output as a tree of Doc nodes — text, groups, indentation, soft/hard line breaks — and the printer decides where to break lines to fit within a configurable width.

Usage

use rlsp_fmt::{Doc, FormatOptions, format, group, concat, indent, line, text};

let doc = group(concat(vec![
    text("["),
    indent(concat(vec![line(), text("a"), text(","), line(), text("b")])),
    line(),
    text("]"),
]));

// Fits on one line at width 80
assert_eq!(format(&doc, &FormatOptions::default()), "[ a, b ]");

// Breaks across lines at width 10
let narrow = FormatOptions { print_width: 10, ..Default::default() };
assert_eq!(format(&doc, &narrow), "[\n  a,\n  b\n]");

Doc Nodes

Node	Description
`Text(String)`	Literal text (must not contain newlines)
`HardLine`	Mandatory line break regardless of mode
`Line`	Space in flat mode, newline + indent in break mode
`Indent(Doc)`	Increases indentation level for the child
`Group(Doc)`	Tries flat mode first; breaks if content exceeds width
`Concat(Vec<Doc>)`	Sequential composition of documents
`FlatAlt { flat, break_ }`	Different content depending on flat/break mode

Builder Functions

text(s) — create a Text node
hard_line() — create a HardLine node
line() — create a Line node (soft break)
indent(doc) — wrap in Indent
group(doc) — wrap in Group
concat(docs) — combine into Concat
flat_alt(flat, break_) — mode-dependent content
join(separator, docs) — intersperse a separator between documents

FormatOptions

Option	Default	Description
`print_width`	80	Maximum line width before breaking groups
`tab_width`	2	Spaces per indentation level (ignored when `use_tabs` is true)
`use_tabs`	false	Use tab characters instead of spaces

Documentation

Feature Log — implemented capabilities and design decisions

Building

cargo build -p rlsp-fmt
cargo test -p rlsp-fmt

License

MIT — Christoph Dalski

AI Note

Every line of source in this crate was authored, reviewed, and committed by AI agents working through a multi-agent pipeline (planning, implementation, independent review, and test/security advisors for high-risk tasks). The human role is designing the architecture, rules, and review process; agents execute them. Conformance against the YAML Test Suite is a measured acceptance criterion — not an aspiration — and any change touching parser behaviour or untrusted input passes through formal test and security advisor review before being merged.

rlsp-fmt 0.1.9