TOON to CSV

From TOON tables to CSV

TOON (Token-Oriented Object Notation) is closer to CSV than any other JSON encoding. Its signature feature, the tabular array, declares a length and a header once — products[3]{sku,name,price}: — and then lists one comma-separated row per line. That is why TOON is so token-efficient for LLM prompts and responses. But a TOON table is not a CSV file: it is indented, nested inside a document, and can use quotes, tabs or pipes according to its own rules. This converter decodes the TOON with the official library, finds the arrays inside it and writes a clean, RFC 4180-style CSV you can open in Excel, Google Sheets, LibreOffice or import into a database.

A typical workflow is asking a model to extract records — leads, invoices, products, test cases — in TOON to save tokens, then exporting the result here for analysis or bulk upload.

Choosing the table and handling nesting

A TOON document may hold several arrays, such as friends and hikes in the sample. The Table list shows every array by its dotted path; the first one is selected by default. Rows that are objects contribute their keys as columns; nested objects are flattened into dotted names like owner.email, while nested arrays are kept as JSON text in one cell so no data is lost. When rows have different keys, the header is the union of all of them and missing cells stay empty. A list of plain values becomes a single value column.

Delimiters and spreadsheet tips

Comma is the standard, but Excel in many European locales expects a semicolon, so switch the delimiter if columns end up merged. Tab and pipe are also available. Fields containing the delimiter, quotes or line breaks are quoted and escaped automatically. Keep Strict mode on to make sure the row count matches the declared [N] before you trust a model's export. For the reverse direction or a JSON view, see CSV to JSON and TOON to JSON; other targets are TOON to YAML and TOON to XML. Everything runs locally in your browser.