🏰 ghast.js v0.6.7 'FLAY'

Version Series License Documentation

ghast.js is an abstract syntax tree designed for use with Peggy/PEG.js.



npm install ghast.js

The ast function

ghast.js provides the AST class and ast helper function:

const { AST, ast } = require('ghast.js')

You probably won't need to interact with the AST class itself. The helper is a wrapper around new AST(). It takes an ID and zero-or-more syntax elements. A syntax element may be a string, another AST node or an array of these. Example:

  ast('Ident', 'foo'),
  "(", ast('String', '"', 'bar', '"'), ")"

This will return a small tree representing a call to function foo with one string as its argument, "bar".

The classify function

The ast.classify function takes a number of tags and returns a new helper function that automatically applies these tags to created nodes. For example:

const foo = ast.classify('foo')
const n1 = foo('Int', '55') // n1 will have the tag 'foo'

const bar = foo.classify('bar')
const n2 = bar('Str', 'foo') // n2 will be tagged 'foo bar'

The locate function

ast.locate takes a location function and returns a new helper function that automatically adds location data to any created nodes. In a grammar you would use it like this:

  const ast = options.ast.locate(location)

Any created nodes will capture location data from the rule where they're created, available as node.location. Note that all nodes created in an action will share the same location information; to get information on a portion of the match, create another rule for just that portion.

Using ghast.js in a Grammar File

To use ghast in a grammar file, create a parser and place the ast helper function in the parser's options. For example:

const peggy = require('peggy')
const { ast } = require('ghast.js')

const parser = peggy.generate(GRAMMAR)

const tree = parser.parse(INPUT, {ast})

The ast function will be available in your grammar:

  const ast = options.ast.locate(location)
  const node = ast.classify('Node')
  const val = ast.classify('Value')

Example = ex:Atom* { return ast('Example', ex) }

Atom = A / B / N / S / [ \n]
Sub = "(" Atom* ")"

A = x:("a" Sub / "a") { return node('A', x) }
B = x:("b" Sub / "b") { return node('B', x) }
N = n:$[0-9]+ { return val('Number', n) }
S = "'" C "'"
C = c:$[^']+ { return val('String', c) }

The parser will now return a ghast AST which can be used to manipulate the parsed syntax. This will remove all B elements directly below an A element:

const tree = parser.parse(INPUT, {ast})
tree.select("A", {id: "B", depth: 0}, b=> b.remove())


Complete API documentation is available. Below is an overview of common methods:

The each method is used to query the tree:

node.each()                       // return all descendants of `node`
node.each({self: true})           // return `node` and all of its descendants
node.each('Section')              // return all descendants with id `Section`
node.each({id: 'Section'})        // same as above
node.each({id: 'X', tag: 'y'})    // return all descendants with both id `X` and tag `y`
node.each({tag: 'val key'})       // return all descendants tagged `val` and `key`
node.each({id: 'A', first: true}) // return the first descendant with id `A`
node.each({leaf: true})           // return all descendant leaf nodes
node.each({stem: true})           // return all non-leaf descendant nodes
node.each({depth: 0})             // return all direct children of `node`
node.each({depth: 1})             // return all direct children and grandchildren
node.each({up: true})             // return all ancestors of `node`
node.ancestor()                   // same as above
node.each({up: true, tag: 'x'})   // return all ancestors of `node` tagged `x`
node.ancestor({tag: 'x'})         // same as above
node.climb(3)                     // return nth ancestor of `node`

The select method creates complex selections from multiple traverses, similar to CSS selectors. The following is similar to A .foo > B:

node.select('A', {tag: 'foo'}, {id: 'B', depth: 0})

The when method is used to visit nodes. Each visitor is an array of queries followed by a callback which will be called for each node matching any of its associated queries:

  [{id: 'A', tag: 'foo'}, n=> n.foo()],
  ['T', 'V', n=> n.bar()],
  [{tag: 'bar'}, {leaf: true}, n=> n.baz()],

The following methods exist to modify the tree:

// replace a child node with another:
node.replace(node.first(), ast('Test', 'test'))
// self-replace a node with another:
node.replace(ast('Test', 'test'))

// transform nodes in-place:
node.mutate({id: 'Foo', attrs: {x: 1}})
node.mutate({tags: 'x y z', syntax: ['foo']})

// remove a child node:
// self-remove a node:

Nodes can be tagged:

node.tag('foo')         // tag a node
node.tag('foo bar baz') // apply multiple tags at once
node.hasTags            // true if the node has any tags
node.hasTag('foo')      // true if the node has the tag `foo`
node.hasTag('bar baz')  // true if the node has all of the given tags

Nodes have attributes:

node.attr('a', 100)         // set a single attribute
node.attr({foo: 1, bar: 2}) // set one or more attributes
node.attrs.foo              // accessing attributes
node.attrs['foo']           // accessing attributes

The read method deep-reads attributes; it merges the attributes of this node with all of its descendants and returns the value of the given property:

const node = ast('Function',
  ast('Ident', 'foo').attr('foo', 1),
  "(", ast('String', '"', 'bar', '"').attr('bar', 2), ")"
node.read('foo') // returns 1
node.read('bar') // returns 2

Location data for a node can be set with loc:

node.loc({start: 1, end: 2})
node.location // read set location data


Two examples are provided:

  • ab - the nonsense example used in this README
  • ini - a simplistic ini parser


Available under the terms of the MIT license.

Copyright 2023 0E9B061F