Warning

This are just some random thoughts for a new data language. It will probably never be implemented.

Pure

The pure data language.

Spiritual successor of YAML, JSON and TOML. One file format to rule them all.

Goals

• Simple specification
• Only one correct way to format a file (highly whitespace sensitive)
• Maximum space efficiency while keeping readability
• Ability to reproduce large share of commonly used data & config files (>95%)
• Streaming friendly (no lookback necessary => state machine)
• Typed

Specification

• a-zA-z - String
• & - Boolean
• -x - List of datatype x (if x in a-zA-z type = String)
• = - Dictionary
• % - Number (arbitrary precision floating point, no scientific)
• # - Comment
• @ - Datetime
• - Multi line continuation
• | - Multi line continuation with newline
• \ - Escape first character

Reserved chars: +*^}{][/:;().,?'"

null is conveyed by a single _ as the value

Example

name Pure
version 0.1.0
description The pure data language
&private false
-tags
data
language
normalized

:competitor_map
yaml https://yaml.org
json json.org

Purex

Extension to pure. File extension: .purex

• ! - Include path
• ~a8 - Inline binary with a8(hex) bytes
• $lang - Code in language lang

Related

• cue - Data validation and configuration language
• Dhall - Typed configuration language
• HCL - HashiCorp configuration language
• Jsonnet - JSON-based templating language
• KDL - Human-readable document language
• Nickel - Typed configuration with functions
• RON - Rust-oriented data format
• Starlark - Deterministic Python-like configuration language
• TOML - Minimal human-readable config format
• YAML - Human-readable data serialization

TODO

• True / false key character instead of value
• Is purex a good idea?
• Pseudocode implementation
• + instead of %?
• Extra datatype for URLs?