TOON to TONL Converter

Example: TOON input and real TONL output

The output below was generated by the same conversion code the converter uses, run on this sample when the page was built.

employees[4]{id,name,dept,salary,remote}:
  1,Alice Chen,Engineering,128000,true
  2,Bob Okafor,Sales,86000,false
  3,Carla Diaz,Engineering,119000,true
  4,Dan Weiss,Support,64000,false

#version 1.0
root:
  employees[4]{dept,id,name,remote,salary}:
    Engineering,1,Alice Chen,true,128000
    Sales,2,Bob Okafor,false,86000
    Engineering,3,Carla Diaz,true,119000
    Support,4,Dan Weiss,false,64000

TONL output is 16% larger than the TOON input. The same data is 324 characters as minified JSON and 532 as pretty-printed JSON. Sizes are character counts, not tokens: token counts depend on the model's tokenizer (see the tokenizer comparison). Across larger datasets, TOON averages about 43% fewer tokens than JSON (42.6%) according to the official TOON benchmark.

TOON to TONL: what changes

• Both formats encode uniform arrays as a header plus rows, so a tabular TOON block maps directly onto a TONL block. TONL reorders the columns alphabetically and adds the #version header.
• Reading TOON: TOON is strict about structure: the [N] in an array header must match the number of rows, and each row needs exactly as many values as the {field} list. Hand-edited TOON that breaks either rule fails to decode.
• Writing TONL: TONL output starts with #version 1.0 and a root key, lists field names in braces, and sorts keys alphabetically. On small payloads the header and per-object field lists make it larger than TOON; TONL earns its place through its query, schema, and streaming tooling.

Why Convert TOON to TONL?

Use TONL when you want a compact text format plus tooling around it: querying, schema validation, or streaming. For plain prompt compression, TOON is usually the simpler choice.