acorn

ECMAScript parser

Acorn


Author funding status:

A tiny, fast JavaScript parser, written completely in JavaScript.

The easiest way to install acorn is with npm.

npm install acorn

Alternately, download the source.

git clone https://github.com/marijnh/acorn.git

When run in a CommonJS (node.js) or AMD environment, exported values appear in the interfaces exposed by the individual files, as usual. When loaded in the browser (Acorn works in any JS-enabled browser more recent than IE5) without any kind of module management, a single global object acorn will be defined, and all the exported properties will be added to that.

This is implemented in dist/acorn.js, and is what you get when you require("acorn") in node.js.

parse(input, options) is used to parse a JavaScript program. The input parameter is a string, options can be undefined or an object setting some of the options listed below. The return value will be an abstract syntax tree object as specified by the ESTree spec.

When encountering a syntax error, the parser will raise a SyntaxError object with a meaningful message. The error object will have a pos property that indicates the character offset at which the error occurred, and a loc object that contains a {line, column} object referring to that same position.

  • ecmaVersion: Indicates the ECMAScript version to parse. Must be either 3, 5, or 6. This influences support for strict mode, the set of reserved words, and support for new syntax features. Default is 5.

  • sourceType: Indicate the mode the code should be parsed in. Can be either "script" or "module".

  • onInsertedSemicolon: If given a callback, that callback will be called whenever a missing semicolon is inserted by the parser. The callback will be given the character offset of the point where the semicolon is inserted as argument, and if locations is on, also a {line, column} object representing this position.

  • onTrailingComma: Like onInsertedSemicolon, but for trailing commas.

  • allowReserved: If false, using a reserved word will generate an error. Defaults to true. When given the value "never", reserved words and keywords can also not be used as property names (as in Internet Explorer's old parser).

  • allowReturnOutsideFunction: By default, a return statement at the top level raises an error. Set this to true to accept such code.

  • allowImportExportEverywhere: By default, import and export declarations can only appear at a program's top level. Setting this option to true allows them anywhere where a statement is allowed.

  • allowHashBang: When this is enabled (off by default), if the code starts with the characters #! (as in a shellscript), the first line will be treated as a comment.

  • locations: When true, each node has a loc object attached with start and end subobjects, each of which contains the one-based line and zero-based column numbers in {line, column} form. Default is false.

  • onToken: If a function is passed for this option, each found token will be passed in same format as tokenize() returns.

    If array is passed, each found token is pushed to it.

    Note that you are not allowed to call the parser from the callback—that will corrupt its internal state.

  • onComment: If a function is passed for this option, whenever a comment is encountered the function will be called with the following parameters:

    • block: true if the comment is a block comment, false if it is a line comment.
    • text: The content of the comment.
    • start: Character offset of the start of the comment.
    • end: Character offset of the end of the comment.

    When the locations options is on, the {line, column} locations of the comment’s start and end are passed as two additional parameters.

    If array is passed for this option, each found comment is pushed to it as object in Esprima format:

    {
      "type": "Line" | "Block",
      "value": "comment text",
      "range": ...,
      "loc": ...
    }

    Note that you are not allowed to call the parser from the callback—that will corrupt its internal state.

  • ranges: Nodes have their start and end characters offsets recorded in start and end properties (directly on the node, rather than the loc object, which holds line/column data. To also add a semi-standardized range property holding a [start, end] array with the same numbers, set the ranges option to true.

  • program: It is possible to parse multiple files into a single AST by passing the tree produced by parsing the first file as the program option in subsequent parses. This will add the toplevel forms of the parsed file to the "Program" (top) node of an existing parse tree.

  • sourceFile: When the locations option is true, you can pass this option to add a source attribute in every node’s loc object. Note that the contents of this option are not examined or processed in any way; you are free to use whatever format you choose.

  • directSourceFile: Like sourceFile, but a sourceFile property will be added directly to the nodes, rather than the loc object.

  • preserveParens: If this option is true, parenthesized expressions are represented by (non-standard) ParenthesizedExpression nodes that have a single expression property containing the expression inside parentheses.

parseExpressionAt(input, offset, options) will parse a single expression in a string, and return its AST. It will not complain if there is more of the string left after the expression.

getLineInfo(input, offset) can be used to get a {line, column} object for a given program string and character offset.

tokenizer(input, options) returns an object with a getToken method that can be called repeatedly to get the next token, a {start, end, type, value} object (with added loc property when the locations option is enabled and range property when the ranges option is enabled). When the token's type is tokTypes.eof, you should stop calling the method, since it will keep returning that same token forever.

In ES6 environment, returned result can be used as any other protocol-compliant iterable:

for (let token of acorn.tokenize(str)) {
  // iterate over the tokens 
}
 
// transform code to array of tokens: 
var tokens = [...acorn.tokenize(str)];

tokTypes holds an object mapping names to the token type objects that end up in the type properties of tokens.

Note on using with Escodegen

Escodegen supports generating comments from AST, attached in Esprima-specific format. In order to simulate same format in Acorn, consider following example:

var comments = [], tokens = [];
 
var ast = acorn.parse('var x = 42; // answer', {
    // collect ranges for each node 
    ranges: true,
    // collect comments in Esprima's format 
    onComment: comments,
    // collect token ranges 
    onToken: tokens
});
 
// attach comments using collected information 
escodegen.attachComments(ast, comments, tokens);
 
// generate code 
console.log(escodegen.generate(ast, {comment: true}));
// > 'var x = 42;    // answer' 

Some contexts, such as Chrome Web Apps, disallow run-time code evaluation. Acorn uses new Function to generate fast functions that test whether a word is in a given set, and will trigger a security error when used in a context with such a Content Security Policy (see #90 and #123).

The dist/acorn_csp.js file in the distribution (which is built by the bin/without_eval script) has the generated code inlined, and can thus run without evaluating anything.

This file implements an error-tolerant parser. It exposes a single function. The loose parser is accessible in node.js via require("acorn/dist/acorn_loose").

parse_dammit(input, options) takes the same arguments and returns the same syntax tree as the parse function in acorn.js, but never raises an error, and will do its best to parse syntactically invalid code in as meaningful a way as it can. It'll insert identifier nodes with name "✖" as placeholders in places where it can't make sense of the input. Depends on acorn.js, because it uses the same tokenizer.

Implements an abstract syntax tree walker. Will store its interface in acorn.walk when loaded without a module system.

simple(node, visitors, base, state) does a 'simple' walk over a tree. node should be the AST node to walk, and visitors an object with properties whose names correspond to node types in the ESTree spec. The properties should contain functions that will be called with the node object and, if applicable the state at that point. The last two arguments are optional. base is a walker algorithm, and state is a start state. The default walker will simply visit all statements and expressions and not produce a meaningful state. (An example of a use of state it to track scope at each point in the tree.)

ancestor(node, visitors, base, state) does a 'simple' walk over a tree, building up an array of ancestor nodes (including the current node) and passing the array to callbacks in the state parameter.

recursive(node, state, functions, base) does a 'recursive' walk, where the walker functions are responsible for continuing the walk on the child nodes of their target node. state is the start state, and functions should contain an object that maps node types to walker functions. Such functions are called with (node, state, c) arguments, and can cause the walk to continue on a sub-node by calling the c argument on it with (node, state) arguments. The optional base argument provides the fallback walker functions for node types that aren't handled in the functions object. If not given, the default walkers will be used.

make(functions, base) builds a new walker object by using the walker functions in functions and filling in the missing ones by taking defaults from base.

findNodeAt(node, start, end, test, base, state) tries to locate a node in a tree at the given start and/or end offsets, which satisfies the predicate test. start end end can be either null (as wildcard) or a number. test may be a string (indicating a node type) or a function that takes (nodeType, node) arguments and returns a boolean indicating whether this node is interesting. base and state are optional, and can be used to specify a custom walker. Nodes are tested from inner to outer, so if two nodes match the boundaries, the inner one will be preferred.

findNodeAround(node, pos, test, base, state) is a lot like findNodeAt, but will match any node that exists 'around' (spanning) the given position.

findNodeAfter(node, pos, test, base, state) is similar to findNodeAround, but will match all nodes after the given position (testing outer nodes before inner nodes).

The bin/acorn utility can be used to parse a file from the command line. It accepts as arguments its input file and the following options:

  • --ecma3|--ecma5|--ecma6: Sets the ECMAScript version to parse. Default is version 5.

  • --locations: Attaches a "loc" object to each node with "start" and "end" subobjects, each of which contains the one-based line and zero-based column numbers in {line, column} form.

  • --allow-hash-bang: If the code starts with the characters #! (as in a shellscript), the first line will be treated as a comment.

  • --compact: No whitespace is used in the AST output.

  • --silent: Do not output the AST, just return the exit status.

  • --help: Print the usage information and quit.

The utility spits out the syntax tree as JSON data.

Acorn is written in ECMAScript 6, as a set of small modules, in the project's src directory, and compiled down to bigger ECMAScript 3 files in dist using Browserify and Babel. If you are already using Babel, you can consider including the modules directly.

The command-line test runner (npm test) uses the ES6 modules. The browser-based test page (test/index.html) uses the compiled modules. The bin/build-acorn.js script builds the latter from the former.

If you are working on Acorn, you'll probably want to try the code out directly, without an intermediate build step. In your scripts, you can register the Babel require shim like this:

require("babel-core/register")

That will allow you to directly require the ES6 modules.

Acorn is designed support allow plugins which, within reasonable bounds, redefine the way the parser works. Plugins can add new token types and new tokenizer contexts (if necessary), and extend methods in the parser object. This is not a clean, elegant API—using it requires an understanding of Acorn's internals, and plugins are likely to break whenever those internals are significantly changed. But still, it is possible, in this way, to create parsers for JavaScript dialects without forking all of Acorn. And in principle it is even possible to combine such plugins, so that if you have, for example, a plugin for parsing types and a plugin for parsing JSX-style XML literals, you could load them both and parse code with both JSX tags and types.

A plugin should register itself by adding a property to acorn.plugins, which holds a function. Calling acorn.parse, a plugins option can be passed, holding an object mapping plugin names to configuration values (or just true for plugins that don't take options). After the parser object has been created, the initialization functions for the chosen plugins are called with (parser, configValue) arguments. They are expected to use the parser.extend method to extend parser methods. For example, the readToken method could be extended like this:

parser.extend("readToken", function(nextMethod) {
  return function(code) {
    console.log("Reading a token!")
    return nextMethod.call(this, code)
  }
})

The nextMethod argument passed to extend's second argument is the previous value of this method, and should usually be called through to whenever the extended method does not handle the call itself.

There is a proof-of-concept JSX plugin in the jsx branch branch of the Github repository.