Introduction

CSV Table Schema is a pragmatic approach for defining CSV schemas. CSV Table Schema provides a CSV schema description via JSON.

Main features:

Easy schema definiton via JSON which can be validated against the CSV Table Meta Schema.
Dedicated to the CSV format.
Supports the following data types: string, enum, enum-set, boolean, integer, numeric, date, time, date-time, json and xml.
Supports CSV tables with ordered and unordered columns.
Supports CSV tables with or without header row.
Supports CSV dictionaries with key/value rows.
Supports polimorphic CSV tables with different row types, distinguished by a discriminator in the first (or another) column.
Multilingual support

Examples

The following examples are available:

Why yet another CSV schema definition?

Unlike JSON or XML there is no well established schema definition language for CSV. But we are of course not the first ones with a proposal:

The CSV Schema Language is a language for defining and validating CSV data. It uses its own domain language for describing the schema.
The Table Schema is a format to declare a schema for tabular data (CSV is just one example). The schema is designed to be expressible in JSON.
The Metadata Vocabulary for Tabular Data is a standard of the World Wide Web Consortium (W3C) for declaring a schema for tabular data records on the web. The schema is designed to be expressible in JSON.

All three approaches are well thought out, but unfortunately do not meet our needs (However, the Metadata Vocabulary for Tabular Data comes very close).

CSV Table Schema is an alternative schema description which is heavily inspired by those existing schema languages.

What's missing?

A data schema has two main purposes:

Semantical documentation of a data structure
The possibility to validate real data against the data schema.

The documentation part is done. What's still missing is a reference validator.