If you find this useful, please consider supporting my work with a donation.
Momoa is a general purpose JSON utility toolkit, containing:
- A tokenizer that allows you to separate a JSON string into its component parts.
- A ECMA-404 compliant parser that produces an abstract syntax tree (AST) representing everything inside of a JSON string.
- A traverser that visits an AST produced by the parser in order.
- A printer that can convert an AST produced by the parser back into a valid JSON string.
JavaScript defines the JSON
object with methods for both parsing strings into objects and converting objects into JSON-formatted strings. In most cases, this is exactly what you need and should use without question. However, these methods aren't useful for more fine-grained analysis of JSON structures. For instance, you'll never know if a JSON object contains two properties with the same names because JSON.parse()
will ignore the first one and return the value of the second. A tool like Momoa comes in handy when you want to know not just the result of JSON parsing, but exactly what is contained in the original JSON string.
You can install Momoa using npm or Yarn:
npm install @humanwhocodes/momoa --save
# or
yarn add @humanwhocodes/momoa
To parse a JSON string into an AST, use the parse()
function:
const { parse } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
The parse()
function accepts a second argument, which is an options object that may contain one or more of the following properties:
comments
- set totrue
if you want to parse C-style line and block comments inside of JSON.ranges
- set totrue
if you want each node to also have arange
property, which is an array containing the start and stop index for the syntax. Iftokens
is alsotrue
, then the tokens will also haverange
properties.tokens
- set totrue
to return atokens
property on the root node containing all of the tokens used to parse the code. Ifcomments
is alsotrue
, then the tokens include comment tokens.
Here's an example of passing options:
const { parse } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string, { tokens: true });
// root now has a tokens array
console.dir(ast.tokens);
To produce JSON tokens from a string, use the tokenize()
function:
const { tokenize } = require("@humanwhocodes/momoa");
for (const token of tokenize(some_json_string)) {
console.log(token.type);
console.log(token.value);
}
The tokenize()
function accepts a second parameter, which is an options object that may contain one or more of the following properties:
comments
- set totrue
if you want to tokenize C-style line and block comments inside of JSON.ranges
- set totrue
if you want each token to also have arange
property, which is an array containing the start and stop index for the syntax.
There are two ways to traverse an AST: iteration and traditional traversal.
Iteration uses a generator function to create an iterator over the AST:
const { parse, iterator } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
for (const { node, parent, phase } of iterator(ast)) {
console.log(node.type);
console.log(phase); // "enter" or "exit"
}
Each step of the iterator returns an object with three properties:
node
- the node that the traversal is currently visitingparent
- the parent node ofnode
phase
- a string indicating the phase of traversal ("enter"
when first visiting the node,"exit"
when leaving the node)
You can also filter the iterator by passing in a filter function. For instance, if you only want steps to be returned in the "enter"
phase, you can do this:
const { parse, iterator } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
for (const { node } of iterator(ast, ({ phase }) => phase === "enter")) {
console.log(node.type);
}
Traversing uses a function that accepts an object with enter
and exit
properties:
const { parse, traverse } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
traverse(ast, {
enter(node, parent) {
console.log("Entering", node.type);
},
exit(node, parent) {
console.log("Exiting", node.type);
}
});
To convert an AST into the JavaScript value it represents, use the evaluate()
function:
const { parse, evaluate } = require("@humanwhocodes/momoa");
// same as JSON.parse(some_json_string)
const ast = parse(some_json_string);
const value = evaluate(ast);
In this example, value
is the same result you would get from calling JSON.parse(some_json_string)
(ast
is the intermediate format representing the syntax).
To convert an AST back into a JSON string, use the print()
function:
const { parse, print } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
const text = print(ast);
Note: The printed AST will not produce the same result as the original JSON text as the AST does not preserve whitespace.
You can modify the output of the print()
function by passing in an object with an indent
option specifying the number of spaces to use for indentation. When the indent
option is passed, the text produced will automatically have newlines insert after each {
, }
, [
, ]
, and ,
characters.
const { parse, print } = require("@humanwhocodes/momoa");
const ast = parse(some_json_string);
const text = print(ast, { indent: 4 });
To work on Momoa, you'll need:
The first step is to clone the repository:
git clone https://github.com/humanwhocodes/momoa.git
Then, enter the directory and install the dependencies:
cd momoa
npm install
After that, you can run the tests via:
npm test
Note: Momoa builds itself into a single file for deployment. The npm test
command automatically rebuilds Momoa into that single file whenever it is run. If you are testing in a different way, then you may need to manually rebuild using the npm run build
command.
This project takes inspiration (but not code) from a number of other projects:
Esprima
inspired the package interface and AST format.json-to-ast
inspired the AST format.parseJson.js
inspired me by showing writing a parser isn't all that hard.
Apache 2.0
Momoa is the last name of American actor Jason Momoa. Because "JSON" is pronounced "Jason", I wanted a name that played off of this fact. The most obvious choice would have been something related to Jason and the Argonauts, as this movie is referenced in the JSON specification directly. However, both "Argo" and "Argonaut" were already used for open source projects. When I did a search for "Jason" online, Jason Momoa was the first result that came up. He always plays badass characters so it seemed to fit.
There are a number of programs that allow C-style comments in JSON files, most notably, configuration files for Visual Studio Code. As there seems to be a need for this functionality, I decided to add it out-of-the-box.
Unfortunately, Node.js still doesn't natively support ECMAScript Modules (ESM) and everyone generally expects npm packages to export things via CommonJS. As such, the source files are built (using Rollup) into a CommonJS package before publishing. To ensure that the published API is working correctly, it makes sense to write the tests in CommonJS and to pull in what would be the published package API.