FAQ¶

Why not just wrap at 80 columns?¶

Column wrapping causes paragraph-wide reflows when you change a single word. Reviewers see the entire paragraph in the diff, obscuring the actual edit. Sentence-per-line formatting means a one-word change produces a one-line diff.

Does this affect how the document renders?¶

No. LaTeX, Org-mode, Markdown, RST, and most prose formats treat single newlines as whitespace. Two consecutive newlines start a new paragraph. Snapper preserves paragraph breaks (blank lines) and only changes where single newlines occur within a paragraph.

My collaborators don’t use snapper.¶

Will this cause problems?

No. The formatting affects only whitespace within paragraphs. Collaborators can edit normally and their changes will merge cleanly. For fully transparent integration, use a git smudge/clean filter (see the how-to guide).

Why does snapper handle “Dr. Smith” correctly?¶

Snapper maintains curated abbreviation lists for English (80+), German, French, Icelandic, and Polish. After Unicode sentence boundary detection, it merges any splits that occurred at known abbreviation periods. Select a language with --lang de (or set lang = "de" in .snapperrc.toml). You can also add project-specific abbreviations via extra_abbreviations in the config file.

What about non-English text?¶

Snapper supports non-English text in two ways:

Rule-based abbreviation sets for German (de), French (fr), Icelandic (is), and Polish (pl).

Select with --lang de. The Unicode UAX #29 sentence boundary algorithm works across all languages.

Neural sentence detection via nnsplit for 9 languages (en, de, fr, no, sv, zh, tr, ru, uk).

Enable with --neural --lang de. Models download on first use (~4MB, cached).

The rule-based approach handles most academic text well and runs 18x faster. Use neural for languages where abbreviation rules fall short (Chinese, Turkish, etc.).

Can I use snapper as a library?¶

Yes. The crate (snapper-fmt) exposes format_text() and all internal modules:

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

let config = FormatConfig {
    format: Format::Org,
    max_width: 0,
    use_neural: false,
    neural_lang: "en".to_string(),
    neural_model_path: None,
    extra_abbreviations: vec![],
};
let output = format_text(input, &config).unwrap();

How fast is it?¶

Snapper processes typical academic papers (10-50 pages) in under 10ms with the rule-based splitter. Neural mode adds ~200ms per file (model load + inference via tract, pure Rust ONNX). No external runtime dependencies.

What about CRLF line endings?¶

Snapper detects the input’s line ending convention (LF or CRLF) and preserves it in the output. Internally, processing normalizes to LF and converts back before writing.

Can I skip formatting for specific sections?¶

Yes. Use snapper:off and snapper:on pragmas in comments:

Org: # snapper:off / # snapper:on
LaTeX: % snapper:off / % snapper:on
Markdown:  / 
RST/Plaintext: snapper:off / snapper:on on their own line

Text between these markers passes through unchanged. Useful for poetry, aligned text, or ASCII art.

What formats does snapper support?¶

Org-mode, LaTeX, Markdown, reStructuredText, and plaintext. Format auto-detects from the file extension. Override with --format rst (or org, latex, markdown, plaintext).

How do I compare versions at the sentence level?¶

Use snapper sdiff to diff two files at sentence granularity:

snapper sdiff paper_v1.org paper_v2.org

Or diff against a git ref:

snapper git-diff HEAD~1 paper.org

Whitespace reflow produces zero diff. Only actual content changes appear.

How do I report a bug or request a feature?¶

File an issue at github.com/TurtleTech-ehf/snapper/issues.