Optimized for Large Language Models and High-Throughput Streams.
Type-safe, structure-aware, and 60% smaller than JSON.
JSON is verbose. SLD/MLD eliminates structural overhead like quotes, braces, and whitespace without losing data integrity.
{
"name": "Alice",
"age": 30,
"tags": ["admin", "user"],
"verified": true
}
name[Alice;age[30;tags{admin~user};verified[^1
Authoritative specification, evolving RFC draft, quick references and source code. Use these links to implement, validate or extend the formats.
Work-in-progress formalization of SLD/MLD for future standardization and community review.
Canonical single-line data rules: delimiters, header metadata, inline types, canonicalization profile.
Line-oriented variant optimized for grep/awk/sed, streaming ingestion and incremental processing.
Real-world samples plus syntax guides to learn formatting choices and edge cases quickly.
Reference implementations, open issue tracker, roadmap and contribution guidelines.
Looking for Spanish docs? Check Referencia Rápida SLD y MLD.
Convert between JSON and SLD/MLD directly in your browser. This demo supports core v2.0 features: fields, arrays, booleans (^1/^0), null (^_), and escapes. Advanced inline types and headers are omitted for simplicity.
Ready.
The encoder walks object keys and emits key[value pairs separated by ;. Arrays produce key{item~item}. Booleans become ^1/^0; null becomes ^_. Special characters ; ~ [ { } ^ inside values are escaped with a caret.
The decoder splits on record separators (~ for SLD/MLD, \n for MLD), then fields by ; (ignoring empty tail). Each field is parsed as either property (name[Alice) or array (tags{admin~user}). Escapes are resolved by scanning values for caret-prefixed tokens.
This demo is intentionally minimal—no canonicalization, inline type tags, headers, or nested objects. Treat it as an educational preview, not a production parser.
Uses rare characters (; [ {) as delimiters to minimize token splitting in LLM tokenizers.
MLD (Multi Line Data) is the newline-separated variant. Perfect for logs, Unix pipes, and streaming APIs.
Native support for Strings, Numbers, Booleans (^1), Nulls (^_), and nested Arrays.
No quoting required for standard alphanumeric strings. Only escape what is necessary.
Optional first record advertises version, schema, timestamp, source, and enabled features via reserved ! keys (e.g., !v, !schema, !ts, !source, !features{types~canon}).
Explicit typing on keys with !i !f !b !s !n !d !t !ts before [ or {, enabling schema validation and typed nulls (!n[).
Production-ready libraries available for major languages.