JSON is tedious to type for humans. YAML is error-prone and hard to parse. TOML is verbose for nested data structures.
Enter Zaml.
Zaml is the final form of configuration files. It's a zero-dependency, type-checking machine that points out your errors as graceful as a dove. It takes the pain out of your gain.
Never again deal with the boilerplate of validating config data structures. Make config easier for you and your users.
Zaml isn't even an ancronym. That's how good it is.
npm install zaml
or
yarn add zaml
or
Zaml's syntax is clean, effective, and not obsessed with your shift key:
import {parse} from 'zaml'
var schema = '{ fileRoots: { dev:list, prod:list } }'
var zamlContent = `
# This is Zaml!
# You'd probably load this from a file instead.
fileRoots {
dev {
$HOME/dev/services
$HOME/dev/states
}
prod {
$HOME/prod/services-2
$HOME/prod/states
}
}
`
var result = parse(zamlContent, schema, { vars: process.env })
console.log("Got result:", result)
(You can run this example in your browser)
Parsing the above will result in this data structure:
{
"fileRoots": {
"dev": [
"/home/alice/dev/services",
"/home/alice/dev/states"
],
"prod": [
"/home/alice/prod/services-2",
"/home/alice/prod/states"
]
}
}
No quotes, no commas, no colons. Only what you need and nothing else.
Your users also get nice, accurate error messages if they make a mistake writing their config. It's a win-win!
See the examples/ folder for more complete syntax & schema examples like the above.
Here are Zaml's features, each with an example use, from simplest to most complex.
A comment is any line that begins with #
# I am a comment
# title This is a comment
title This is # not a comment
A bool
accepts true
or false
# if your schema is {autoCleanup:bool}
autoCleanup true
#=> { "autoCleanup": true }
View this example in the online editor
A num
accepts a single numerical value.
# if your schema is {port:num}
port 3000
#=> { "port": 3000 }
View this example in the online editor
A str
is the default type of any unspecified schema key. It accepts any character until it reaches a newline.
If you need multiline strings, you can use triple quotes ("""
). Zaml will also un-indent your string automatically, just like Ruby squiggly heredoc.
# if your schema is {title,description} OR {title:str,description} OR etc.
title You, Yourself, and U
description """
A story
...
about your mirror.
"""
#=> { "title": "You, Yourself, and U", "description": "A story\n ...\nabout your mirror." }
View this example in the online editor
An enum
is a str with restricted options. It's useful for a "choose one" situation.
# if your schema is { paymentMode: enum(test,live) }
paymentMode test
#=> { "paymentMode": "test" }
Naturally, if the user provides a value outside the schema, Zaml will reject it and report a readable error message.
View this example in the online editor
A kv
is a set of key-value pairs. It requires a block.
# if your schema is {redirects:kv}
redirects {
/contact /contact-us
/profile/:user /u/:user
}
#=> { "redirects": { "/contact": "/contact-us", "/profile/:user": "/u/:user" } }
Please note Zaml is not indentation sensitive.
View this example in the online editor
A {}
block is a specified inner schema. It translates to a hash that only allows your specified keys.
# if your schema is { project: { title, private:bool } }
project {
title My Sweet App
private true
}
#=> { "project": { "title": "My Sweet App", "private": true } }
View this example in the online editor
You can also enhance basic types with blocks. This lets you specify cleaner config:
# if your schema is { page|multi: str {hide:bool} }
page index.html
page wip.html {
hide true
}
#=> { "page": [
# ["index.html", {}],
# ["index.html", { "hide": true }],
# ] }
View this example in the online editor
A list
is a sequence of values. A user can write lists either inline or with a block (but not both).
# if your schema is {tags:list} OR {tags:list(str)}
# Inline example
tags library, npm, with spaces, js
# Block example
tags {
library
npm
with spaces
js
}
#=> { "tags": ["library", "npm", "with spaces", "js"] }
View this example in the online editor
You can specify the type of items in your list by specifying it in parethesis. When no item type is specified, it is defaulted to str
, resulting in list(str)
.
# if your schema is { fav_nums:list(num) }
fav_nums 10, 20
#=> { "fav-nums": [10, 20] }
View this example in the online editor
The list item type may also be enhanced with a block. Note that you must specify an item type when you do!
# if your schema is { users:list(str { admin:bool }) }
users {
andy
beth {
admin true
}
carl
}
#=> { "users": [["andy", {}], ["beth", {admin: true}], ["carl", {}]] }
View this example in the online editor
Note how a block changes the shape of the above parsed result. This allows you to use destructuring for each list item:
var result = parse(source, schema)
for (let [user, options] of result.users) {
//
// `options` will be {admin: true} for beth,
// and undefined for andy & carl
//
}
NOTE: THIS FEATURE IS NOT IMPLEMENTED YET
An inline list a list that can only be written inline. This is useful when you want to extend your list with a block:
# if your schema is { env|multi: inline-list{require|multi} }
env development, test {
require lib-1
require lib-2
}
env production {
require lib-3
}
# => { "env": [
# [["development", "test"], { "require": ["lib-1","lib-2"] }],
# [["production"], { "require": ["lib-3"] }]
# ]}
Appending the |req
attribute to a key will make that key requried.
# if your schema is { access_token|req:str }
access_token abc123
Open that in your browser and see what happens when you remove the access_token
line.
If you specify a |req
within a basic-type hash block, it will make that block required.
# if your schema is { table: str{ id|req: enum(INT,VARCHAR) } }
# This is invalid! It requires a block
# table users
# This works
table users {
id INT
}
View this example in the online editor
Appending the |multi
attribute to a key allows your users to specify it more than once.
# if your schema is {project|multi:{title,type}}
project {
title A
}
project {
title B
type personal
}
#=> { "project": [{ "title": "A" }, { "title": "B", "type": "personal" }] }
View this example in the online editor
|multi
will also ensure your key is always present, even if the user does not provide any.
# if your schema is {project|multi:{title,type}}
# (intentionally left blank)
#=> { "project": [] }
View this example in the online editor
A tuple captures two or more values for a given key, separated by commas. You can specify one in the schema using parenthesis:
# if your schema is {redirect:(num,str,str)}
redirect 302, /old, /new
#=> { "redirect": [302, "old", "new"] }
View this example in the online editor
Please note that tuples may only contain basic types (str
, num
, bool
, and enum). However, you're free to mix tuples with other features:
# if your schema is {redirect|multi:(num,str,str){enableAt}}
redirect 301, /x, /y
redirect 302, /old, /new {
enableAt 2020-10-10
}
#=> { "redirect": [[301, "x", "y"], [302, "/old", "/new", { "enableAt": "2020-10-10" }]] }
View this example in the online editor
A []
block is an array of items from a specified schema. It translates to an array of key-value tuples.
# if your schema is {sidebar:[header,link:(str,str)]}
sidebar {
header Site
link Home /
link About Us, /about
header Account
link Settings, /account/settings
}
#=> { "sidebar": [
# ["header", "Site"], ["link", ["Home", "/"]], ["link", ["About Us", "/about"]],
# ["header", "Account"], ["link", ["Settings", "/account/settings"]]
# ] }
Note how a block changes the shape of the above parsed result. This allows you to use destructuring for each item:
View this example in the online editor
var result = parse(source, schema)
for (let [type, value] of result.sidebar) {
//
// `type` will be "header" or "link",
// whereas `value` will be a string (for header) or an array of size 2 (for link).
//
}
The primary way you interact with Zaml is through its parse
function. This is how you convert Zaml source to a JavaScript data structure.
parse
takes two, maybe three arguments:
- The Zaml source code (as a string)
- A schema to parse with (as a string)
- Options to enable extra features (optional parameter)
Here's a full example that includes reading from the file system. Assuming you have the following my-config.zaml
:
host www.example.com
port 443
disallow /admin
disallow /dashboard
You can parse the above with this node.js code:
var fs = require('fs')
var zaml = require('zaml')
var source = fs.readFileSync(__dirname + '/my-config.zaml')
var result = zaml.parse(source, 'host,port:num,disallow|multi')
result
//=> { "host": "www.example.com", "port": 443, "disallow": ["/admin", "/dashboard"] }
parseOptions, the third parameter to parse()
, has the following options:
Type: Record<string,string>
Each key-value in the provided vars
object will be made available for users to interpolate using the $
operator. Example:
let source = `
title Welcome to $APP_NAME
`
zaml.parse(source, 'title:str', {
vars: { APP_NAME: 'My App' }
})
//=> { "title": "Welcome to My App" }
Note that you can easily use this feature to provide the user access to their own environment variables in a node.js environment:
zaml.parse(source, 'title:str', {
vars: process.env
})
If vars
is set, then setting failOnUndefinedVars
to true
will ensure users cannot use variables that are not defined. This is useful if you want to e.g. ensure a key is always defined:
let source = `
title Welcome to $iDontExist
`
//
// This will throw an error!
//
zaml.parse(source, 'title:str', {
vars: { anyOtherKey: '' },
failOnUndefinedVars: true,
})
Setting this to true
will allow users to write their config keys in a case-insensitive manner.
Setting this to true
will allow users to write enum values in any case.
Zaml has not yet reached 1.0, so there is no spec as of yet. However, here's a rough ABNF grammar for the lexer:
line = statement | comment
statement = key rest ["{\n" *statement "}"] "\n"
key = string-with-no-whitespace
rest = string-with-no-newlines
comment = "#" rest ("\n" | EOF)
After lexing, the parser uses the schema to determine how to parse the rest
and *statement
for each statement.
- New regular expression type
- Allow inline
kv
? - Allow blocks on
kv
- New
json
type for arbitrarily-nested json-like data? - Multiline strings?
text
type? - Split
num
intoint
andfloat
? - Pluggable validation?
- Default values for tuple types?
- Command line interface?
Regarding the online editor:
Get code
button- Fancy explanations of schema on hover
Interested in contributing? There are several ways you can help:
- Open or discuss an issue for an item on the roadmap
- Implement Zaml in another programming language
- Report any bugs you come across
- Report a behavior that you think should be bug
- Help start a spec!
npm install
npm start # In a separate terminal
npm test