Matching

Snippet application is deterministic, left-to-right, and single-pass.

At each position in the input:

1. Test snippets in configuration order.
2. Test each snippet's matchers in when order.
3. Use the first matcher that succeeds at the current position.
4. Append its snippet's body to the output and advance past the matched input.
5. If nothing matches, append the next input character unchanged and advance one character.

First match wins, not longest match

An earlier matcher always wins. The library never prefers the longer match.

If a snippet matching a appears before a snippet matching ab, the input ab matches a first, and the output is that snippet's body followed by a literal b.

Order your configuration accordingly: put more specific matchers before the ones they overlap with.

Literal matchers

• Match case-insensitively, using ECMAScript Unicode semantics.
• Match their complete value at the current position.
• Treat regex metacharacters as ordinary characters.
• Add no word boundaries, so they can match inside a word.

Regex matchers

• Must match beginning at the current position; they never search ahead.
• See the complete input as context, so anchors and lookaround assertions behave as expected.
• Are case-sensitive unless the i flag is present.
• Are ignored when they produce a zero-length result, and testing continues with the remaining matchers.

Bodies

The body is inserted verbatim. Regex captures such as $1 are not expanded.

Inserted bodies are never scanned again, so snippets cannot recurse or chain into each other. The cursor stays on the original input throughout.

Input and output are neither trimmed nor Unicode-normalized.

Regex standard

Regex syntax and behaviour follow ECMAScript 2018 regular expressions in Unicode mode. Unicode mode is always enabled, whether or not you pass u.

Supported flags:

The flags d, g, v, and y are rejected. The library controls global scanning and cursor positioning itself.

The TypeScript library uses the platform's native RegExp. The Rust library uses regress, a native Rust ECMAScript-compatible engine — it does not embed JavaScript, and the TypeScript library does not call into Rust through WebAssembly.

Cross-language parity

A shared conformance suite runs the same cases against both implementations, covering configuration validation, ordering, literals, regexes, flags, single-pass behaviour, whitespace, and Unicode. Every case must produce the same validation result and the exact same output in both libraries, and regex engine or Unicode data upgrades must pass it before release.