SyConf

SyConf is a simple configuration language that keeps your configuration lean.

Documentation

Why?

Why another configuration language?

Software configuration is getting larger and more complex. Modern software require more complex and large configurations. Configuration files of Kubernetes, Prometheus, AlertManager, Concourse, Vector.dev, etc. tend to get large, repetitive, clumsy, and at the end less maintainable.

Existing configuration languages either do not support templating and functions, like JSON, YAML, TOML, or have complex grammars and multiple features that often go beyond the scope of a configuration language, e.g. JsonNet, Dhall.

SyConf is a simple, pure functional, not Turing complete configuration language that supports user defined functions, powerful string interpolation, and helps to keep complex configurations concise and maintainable.

Features

• Very simple grammar

  let house_number = 53
  let my_hash = {
      name: "john",
      age: 33,
  }
  in
  {
      person: my_hash,
      address: "Awesome Street ${house_number}",
  }
  

• Function support

  let add2 = (x) => x + 2
  in
  [3, add2(4)]
  

• String interpolation

  let name = "alexei"
  in
  "hello ${name}"
  

• It is possible to split configuration in multiple files

  let instance_x = import "./instance_x.sy"
  let instance_y = import "./instance_y.sy"
  in
  {
      instances: [instance_x, instance_y]
  }
  

• SyConf objects have a small set of meaningful methods

  let hash = {
      aa: 3,
      bb: 4,
  }
  in
  
  hash.map( (key, value) => [key, value * 10])
  

• There are helpful functions to interact with the environment

  let region = getenv("AWS_REGION", "eu-central-1")
  let prom_config = read_file("prometheus.yaml").parse_yaml()
  

• Conditionals

  let x = if getenv("MY_VAR", "default_value") == "abc" then "this" else "that"
  

• Block expressions

  let func = (x) => {
      let a = x + 2
      in
      a * 4
  }
  

• There is no magic :-)

Name

The letters S and Y in the name SyConf are the first and the last letter in the word simplicity.

Previous Work

• JSON - simple, great for machines, not very human friendly.
• YAML - human friendly, supports entity references, error prone because of sensitivity to indentation.
• TOML - human friendly, eliminates some weaknesses of YAML, big files gets less readable.
• HCL - structured language, supports functions, does not support user defined functions.
• jsonnet - supports functions and templating, has complex syntax, does not support functions as first level citizens.
• dhall - has support for types, has complex syntax with a lot of features.
• JavaScript - heavy (compared to SyConf), full featured, Turing-complete language with side effects.

Licence

Apache 2.0