Cherow
A very fast and lightweight, standards-compliant, self-hosted javascript parser with high focus on both performance and stability.
Features
- Conforms to the standard ECMAScript® 2020 (ECMA-262 10th Edition) language specification (draft)
- Support stage 3 proposals via option.
- Support for parsing with and without web compability (AnnexB)
- Support to only parse expressions
- Experimental feature support via option.
- Optionally track syntactic node locations
- Emits an ESTree-compatible abstract syntax tree.
- Very well tested (~25 000 unit tests with full code coverage))
- Supports all module loaders
- Lightweight - ~72 KB minified
ESNext features
Stage 3
features support. These need to be enabled with the next
option.
API
Cherow generates AST according to ESTree AST format, and can be used to perform syntactic analysis (parsing) of a JavaScript program, and with ES2015 and later a JavaScript program can be either a script or a module.
The parse
method exposed by Cherow takes an optional options
object which allows you to specify whether to parse in script
mode (the default) or in module
mode.
Here is a quick example to parse a script:
cherow.parseScript('foo = bar');
cherow.parse('x = async() => { for await (x of xs); }');
This will return when serialized in json:
{
body: [{
expression: {
left: {
name: 'x',
type: 'Identifier'
},
operator: '=',
right: {
async: true,
body: {
body: [{
await: true,
body: {
type: 'EmptyStatement',
},
left: {
name: 'x',
type: 'Identifier',
},
right: {
name: 'xs',
type: 'Identifier',
},
type: 'ForOfStatement',
}],
type: 'BlockStatement'
},
expression: false,
generator: false,
id: null,
params: [],
type: 'ArrowFunctionExpression'
},
type: 'AssignmentExpression'
},
type: 'ExpressionStatement'
}],
sourceType: 'script',
type: 'Program'
}
The API also allows you to use Cherow as a stand-alone expression parser and let you parse your code in either script or module goal code.
cherow.parseExpression('foo', { ranges: true });
This will return when serialized in json:
{
end: 3,
name: 'foo',
start: 0,
type: 'Identifier'
}
Options
The second argument allows you to specify various options:
Option | Description |
---|
module | Enable module syntax |
ranges | Append start and end offsets to each node |
globalReturn | Allow return in the global scope |
globalAwait | Allow await in the global scope |
webcompat | Enable web compability (AnnexB) |
impliedStrict | Enable strict mode initial enforcement |
next | Enable stage 3 support (ESNext) |
experimental | Enable experimental features |
parenthesizedExpr | Enable non-standard parenthesized expression node |
raw | Attach raw property to each literal node and identifier node |
directives | Enable directive prologue to each literal node |
onComment | Accept either callback or array to collect comment |
onToken | Accept either callback or array and returns each found token |
Experimental features
Cherow aim to support the same experimental features as found in V8 when enabling the 'experimental' flag. However. An exception has been made for the
numeric separators proposal. It's currently pending on Stage 2 after being "degraded" from Stage 3. The future of this proposal seems uncertain at the moment and will not be implemented.
Also the proposal-decorators are constantly changing so at current stage this will not be implemented.
Nightly builds
It's available on NPM.
Contributing
If you feel something could've been done better, please do feel free to file a pull request with the changes.
Read our guidelines here
Bug reporting
If you caught a bug, don't hesitate to report it in the issue tracker. From the moment I respond to you, it will take maximum 60 minutes before the bug is fixed.
Note that I will try to respond to you within one hour. Sometimes it can take a bit longer. I'm not allways online. And if I find out it
will take more then 60 minutes to solve your issue, you will be notified.
I know how irritating it can be if you are writing code and encounter bugs in your dependencies. And even more frustrating if you need to wait weeks or days.
Rationale
Existing parsers have many issues with them:
-
Acorn
is the most commonly used tool out there because of its support for recent ES standards, but it's slow and it often is too permissive in what it accepts. It's also not optimized for handheld devices.
-
Esprima
is a little faster than Acorn, but it's almost never updated, and their test suite has too many invalid tests. It also doesn't support recent ES standards.
-
Babylon
is tightly coupled to Babel, and is comparatively very slow.
Most of these parsers would not fare any chance against the official Test262 suite, and fail a substantial number of them.
We figured we could try do better. We are used in plural form because Cherow is developed by a main developer and two
others "behind the scenes" that contributes with their knowledge whenever it's necessary.
Cross-browser testing provided by: