Skip to main content

Build Status Downloads

Pure Python JavaScript Translator/Interpreter

Everything is done in 100% pure Python so it's extremely easy to install and use. Supports Python 2 & 3. Full support for ECMAScript 5.1, ECMA 6 support is still experimental.


Simple Example:

    >>> import js2py
    >>> js2py.eval_js('console.log( "Hello World!" )')
    'Hello World!'
    >>> add = js2py.eval_js('function add(a, b) {return a + b}')
    >>> add(1, 2) + 3
    6
    >>> add.constructor
    function Function() { [python code] }
    >>> js2py.require('underscore')
    'function _(obj) { [python code] }'

You can also import a big number of node modules as if they were written in Python! For example, here we import a pure JS library crypto-js:

    >>> CryptoJS = js2py.require('crypto-js')
    >>> data = [{'id': 1}, {'id': 2}]
    >>> JSON = js2py.eval_js('JSON')
    >>> ciphertext = CryptoJS.AES.encrypt(JSON.stringify(data), 'secret key 123')
    >>> bytes  = CryptoJS.AES.decrypt(ciphertext.toString(), 'secret key 123')
    >>> decryptedData = JSON.parse(bytes.toString(CryptoJS.enc.Utf8)).to_list()
    >>> decryptedData
    [{u'id': 1}, {u'id': 2}]

Now also supports JavaScript 6 (still experimental):

    >>> js2py.eval_js6('let a = () => 11; a()')
    11

JavaScript 6 support was achieved by using Js2Py to translate javascript library called Babel. Babel translates JS 6 to JS 5 and afterwards Js2Py translates JS 5 to Python. The only downside is that translated babel.js has about 4 MB and importing such a long Python file takes about 15 seconds!


Translating a JavaScript file:

    # this will translate example.js to example.py
    >>> js2py.translate_file('example.js', 'example.py')
    # example.py can be now imported and used!
    >>> from example import example
    >>> example.someFunction()
    ...

Every feature of ECMA 5.1 is implemented (except of 'with' statement):

>>> js2py.eval_js("Object.prototype.toString.call(Function('s', 'return s+arguments[1]')(new Date(), 7).__proto__)")
[object String]

Unfortunately even though Js2Py can be generally used to translate huge Js files (over 50k lines long), in rare cases you may encounter some unexpected problems (like javascript calling a function with 300 arguments - python allows only 255). These problems are very hard to fix with current translation approach. I will try to implement an interpreter in near future which will hopefully fix all the edge cases.

Installation

pip install js2py

More advanced usage example

It is possible to access all the variables from JS scope using EvalJs. Moreover, you can use Python objects from your JavaScript code if you add them to the scope. In this example we will be using Python built-in sum function to sum elements in JS array. It will stand under python_sum.

# Adding Python built-in sum to the JS context:
>>> context = js2py.EvalJs({'python_sum': sum})  
>>> js_code = '''
var a = 10
function f(x) {return x*x}
'''
>>> context.execute(js_code)
# Get value of variable a:
>>> context.a
10
# context.f behaves just like js function so you can supply more than 1 argument. '9'*'9' in javascript is 81.
>>> context.f('9', 0)  
81    
# context.f has all attributes of normal JavaScript object
>>> context.f.toString()
u'function f(x) { [python code] }'
>>> context.f.bind
function bind(thisArg) { [python code] }
# You can also add variables to the context:
>>> context.foo = [1,2,3]  # context.foo is now Js Array object and behaves just like javascript array!
>>> context.foo.push(4)  
4
>>> context.foo.to_list() # convert to python list
[1, 2, 3, 4]
# You can use Python objects that you put inside the context!
>>> context.eval('python_sum(new Array(1, 2, 3))')
6

You can also enable require support in JavaScript like this:

>>> context = js2py.EvalJs(enable_require=True)
>>> context.eval("require('esprima').parse('var a = 1')")

JavaScript 'VirtualMachine' in Python

As a fun experimental project I have also implemented a VM-based JavaScript (yes - there are 2 separate JS implementations in this repo). It is feature complete and faster than the translation based version. Below you can see a demo with a nice debug view (bytecode + execution sequence):

>>> from js2py.internals import seval
>>> seval.eval_js_vm("try {throw 3+3} catch (e) {console.log(e)}", debug=True)
[LOAD_UNDEFINED(),
 JUMP(4,),
 LABEL(1,),
 LOAD_UNDEFINED(),
 POP(),
 LOAD_NUMBER(3.0,),
 LOAD_NUMBER(3.0,),
 BINARY_OP('+',),
 THROW(),
 NOP(),
 LABEL(2,),
 LOAD_UNDEFINED(),
 POP(),
 LOAD('console',),
 LOAD('e',),
 LOAD_N_TUPLE(1,),
 CALL_METHOD_DOT('log',),
 NOP(),
 LABEL(3,),
 LOAD_UNDEFINED(),
 NOP(),
 LABEL(4,),
 TRY_CATCH_FINALLY(1, 2, 'e', 3, False, 4)]

0 LOAD_UNDEFINED()
1 JUMP(4,)
18 TRY_CATCH_FINALLY(1, 2, 'e', 3, False, 4)
  ctx entry (from:2, to:9)
  2 LOAD_UNDEFINED()
  3 POP()
  4 LOAD_NUMBER(3.0,)
  5 LOAD_NUMBER(3.0,)
  6 BINARY_OP('+',)
  7 THROW()
  ctx exit (js errors)
  ctx entry (from:9, to:16)
  9 LOAD_UNDEFINED()
  10 POP()
  11 LOAD('console',)
  12 LOAD('e',)
  13 LOAD_N_TUPLE(1,)
  14 CALL_METHOD_DOT('log',)
6
  15 NOP()
  ctx exit (normal)

This is just a curiosity and I do not recommend using VM in practice (requires more polishing).


Limitations

There are 3 main limitations:

  • "strict mode" is ignored
  • with statement is not supported
  • Indirect call to eval is treated as direct call to eval (hence always evals in local scope)

They are generally not a big issue in practice. In practice more problematic are minor edge cases that unfortunately sometimes do happen. Please report a bug if you find one.

Js2Py was able to successfully translate and run huge JS libraries like Babel (100k+ loc), esprima, crypto-js and more. You can try it yourself by importing any supported npm package via js2py.require('your_package').


Other Examples

In Js2Py all JavaScript objects are a subclass of PyJs object. For example JS Number is represented by PyJsNumber class. js2py.eval_js and js2py.EvalJs automatically tries to convert PyJs type to builtin python type. So for example if you execute:

>>> js2py.eval_js('var a = "hello"; a')

eval_js will return unicode type (u"hello"). However for complex types such conversion is impossible and JsObjectWrapper is returned. See the conversion table JsType -> PyType:

Boolean -> bool
String -> unicode (str in Python 3)
Number -> float (or int/long if whole number)
undefined -> None
null -> None
OTHER -> JsObjectWrapper

JsObjectWrapper supports: getitem, getattr, setitem, setattr, repr and call. Moreover it has to_list and to_dict methods if you want to convert it to builtin python type.

>>> js = js2py.eval_js('d = {a:1, b:2}')
>>> js
{a: 1, b: 2}  
>>> type(js)
<class 'js2py.base.JsObjectWrapper'>
>>> js.a
1
>>> js['a']
1
>>> js.b = 20
>>> js
{a: 1, b: 20}  
>>> js['c'] = 30
>>> js.to_dict()
{u'a': 1, 'c': 30, u'b': 20}

Also, of course you can use Js2Py to parse (tree is the same as in esprima.js) and translate JavaScript

Parsing:

>>> js2py.parse_js('var $ = 5')   
{
    "body": [
        {
            "declarations": [
                {
                    "id": {
                        "name": "$",
                        "type": "Identifier"
                    },
                    "init": {
                        "raw": "5",
                        "type": "Literal",
                        "value": 5
                    },
                    "type": "VariableDeclarator"
                }
            ],
            "kind": "var",
            "type": "VariableDeclaration"
        }
    ],
    "type": "Program"
}

Translating:

>>> print(js2py.translate_js('var $ = 5'))
from js2py.pyjs import *
# setting scope
var = Scope( JS_BUILTINS )
set_global_object(var)

# Code follows:
var.registers(['$'])
var.put('$', Js(5.0))

pyimport statement

Finally, Js2Py also supports importing any Python code from JavaScript using 'pyimport' statement:

>>> x = """pyimport urllib;
           var result = urllib.urlopen('https://www.google.com/').read();
           console.log(result.length)
        """
>>> js2py.eval_js(x)
18211
undefined

Metadata

Release files for Js2Py-3.14 0.74.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for Js2Py-3.14 0.74.2
File Size Uploaded
js2py_3_14-0.74.2.tar.gz 1.0 MB Details

Built distribution (wheel)

Table of built distributions (wheels) for Js2Py-3.14 0.74.2
File Interpreter ABI Platform
js2py_3_14-0.74.2-py3-none-any.whl Python 3 none any Details

Total release size: 2.1 MB

Release files / js2py_3_14-0.74.2.tar.gz

Download URL js2py_3_14-0.74.2.tar.gz
Size 1.0 MB
Tags Source
SHA-256 checksum
How to use checksums
e6f3a3b01bbf002b0683f5a6086eeab08204cbbe006a474f9019bffb19c72383
BLAKE2b-256 checksum
How to use checksums
f58644dc21f8529aafac5c50d720a5fb57ae4ea94266d024a0325b5f40f2cbe3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.0

Release files / js2py_3_14-0.74.2-py3-none-any.whl

Download URL js2py_3_14-0.74.2-py3-none-any.whl
Size 1.1 MB
Tags Python 3
SHA-256 checksum
How to use checksums
c2979caca55f72f0131bcecd627d841159415af699dc2241b1594a8b69992860
BLAKE2b-256 checksum
How to use checksums
c4e9c3fc21f9e03a5383d6b0c8a4f228ce5289306ee30669097a081c138f9e3b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.14.0

Release history Release notifications | RSS feed

This release

0.74.2 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page