Crate snapper_fmt 

snapper

Semantic line break formatter for prose documents. Reformats text so each sentence occupies its own line, producing minimal git diffs when collaborating on papers and documentation.

The crate is published as snapper-fmt on crates.io. Installers ship two CLI names for the same program: snapper and snapper-fmt (the latter avoids colliding with openSUSE’s Btrfs snapshot tool of the same name).

Supported formats

• Org-mode: drawers, tables, keywords preserved; #+BEGIN_SRC is Region::Code (comment reflow via [code.<lang>], optional formatters)
• LaTeX: preamble and math preserved; minted / lstlisting are code regions
• Markdown: front matter and headings preserved; fenced blocks are code regions
• RST: directives and literals preserved; .. code-block:: is a code region
• Plaintext: everything treated as prose

Library usage

use snapper_fmt::{format_text, FormatConfig};
use snapper_fmt::format::Format;

let input = "Hello world. This is a test. Another sentence.";
let config = FormatConfig {
    format: Format::Plaintext,
    ..Default::default()
};
let output = format_text(input, &config).unwrap();
assert_eq!(output, "Hello world.\nThis is a test.\nAnother sentence.");

Modules

abbreviations

check

Line-level --check diagnostics: fused, wrap, and long.

cli

code_block

Comment-aware reflow inside a Region::Code.

config

diff

files

format

Document format detection and representation.

git_diff

init

lsp

mcp

MCP (Model Context Protocol) server for snapper.

oracle

Format-local render oracle used as a fail-closed backstop.

output

parser

reflow

sdiff

sentence

watch

Structs

FormatConfig

Configuration for the formatting pipeline.

InvalidUtf8Error

Typed error for invalid UTF-8 input. Branch with error.downcast_ref.

PandocCannotSplice

Pandoc’s AST has no source offsets, so it cannot splice into original bytes. format_text refuses rather than reconstruct.

Functions

build_splitter

Build the appropriate sentence splitter from config.

format_bytes

Format raw bytes. Invalid UTF-8 is a hard error (InvalidUtf8Error).

format_range

Format only lines within a range (1-indexed, inclusive). Lines outside the range pass through unchanged.

format_text

Format text with semantic line breaks.

format_text_with_splitter

Format text using a pre-constructed splitter (avoids reloading models per file).

run_fixpoint

Run step until the output is a byte fixpoint or the cap is hit.