Releases: Distributive-Network/PythonMonkey
Releases · Distributive-Network/PythonMonkey
v1.0.0
PythonMonkey v1.0.0
PythonMonkey is now stable for production, including Distributive's use-case of DCP! 🥳
Bug Fixes:
- calling a python method that has 0 parameters in JavaScript would cause a segfault rather than a
TypeError: Class.method() takes 0 positional arguments but 1 was given
exception - Using a python iterable as the yield of a JavaScript
function*
generator would cause a segfault rather than aTypeError: 'iterable' object is not an iterator
exception
v0.9.0
PythonMonkey v0.9.0
This release fixes the long annoying string corruption bug.
PythonMonkey is now super powerful and a lot more stable! 🦾 🦾 🦾
Generated release notes:
What's Changed
- Upgrade SpiderMonkey to mozilla-central commit
a283127a5d0aa005c54d339e8ca27414b55f079b
by @github-actions in #424 - Fix the string corruption bug by @Xmader in #428
- Philippe/final edits for first release by @philippedistributive in #427
- Make the wheel packages we build also support lower versions of macOS by @Xmader in #415
- Make easier debugging HTTP errors (such as
DCPError: no transports defined
) by @Xmader in #371 - Docs: add uninstallation instructions in the README by @Xmader in #431
- Add CI for publishing to ⊇istributive's archive server by @Xmader in #432
Full Changelog: v0.8.1...v0.9.0
v0.8.1
-
bug fixes:
- segfault could occur if
exit()
orquit()
were called asynchronously atob
did not include padding- pythonmonkey could not be installed on Windows if cargo was not installed beforehand
- segfault could occur if
-
performance improvements:
- improved memory-use, particularly regarding python strings passed to JavaScript
- increased the maximum JavaScript garbage-collected heap size to the maximum allowed (4 GB)
-
other:
- pythonmonkey now ships pre-built arm64 wheels
- removed Doxygen and Graphviz as mandatory prerequisites to install PythonMonkey (you still need them to build the docs, but you'll have to install them yourself if you wish to do so)
v0.8.0
- python
bytes
objects are now passed to JavaScript as immutableUint8Array
- python
dict
and user-defined class objects now use the prototype chain for following methods in JavaScript:Object.prototype.toString
Object.prototype.toLocaleString
Object.prototype.valueOf
- python iterator objects passed to JS now have the
[Symbol.toPrimitive]
and.valueOf
properties in JavaScript to support use withconsole.log
v0.7.1
v0.7.0
- added a new build type:
- Sanitize: same as Debug, except with AddressSanitizer enabled
- added a new tool inspired by wtfnode called WTFPythonMonkey to track down any hanging
setTimeout
/setInterval
timers that are still ref'd when you hit ctrl-C.- When using
pmjs
, to enable it you can simply pass the--wtf
flag, like so:
pmjs --wtf <filename>.js
- When using pythonmonkey from python, it can be enabled by using a context manager (
with
-statement) around the entry point of the event loop, like so:
import asyncio import pythonmonkey as pm from pythonmonkey.lib.wtfpm import WTF async def pythonmonkey_main(): pm.eval("setInterval(() => console.log(new Date), 500)") await pm.wait() with WTF(): asyncio.run(pythonmonkey_main())
- When using
- implemented JS-like function calling for python functions in JS. Similar to JS functions, you can now call python functions with too few or too many arguments without throwing an error.
- When too many arguments are supplied, those beyond the function's parameter count are ignored, e.g.:
def f(a, b): return [a, b] assert [1, 2] == pm.eval("(f) => f(1, 2, 3)")(f)
- When too few arguments are supplied, those beyond the number of supplied arguments are passed as
None
to match JS's behaviour of passingundefined
, e.g.:
def f(a, b): return [a, b] assert [1, None] == pm.eval("(f) => f(1)")(f)
- This also works for functions with default arguments, or varargs, e.g.:
def f(a, b, c=42, d=43, *args): return [a, b, c, d, *args] assert [1, 2, 3, 4, 5] == pm.eval("(f) => f(1, 2, 3, 4, 5)")(f) assert [1, 2, 3, 4 ] == pm.eval("(f) => f(1, 2, 3, 4)" )(f) assert [1, 2, 3, 43 ] == pm.eval("(f) => f(1, 2, 3)" )(f) assert [1, 2, 42, 43 ] == pm.eval("(f) => f(1, 2)" )(f) assert [1, None, 42, 43 ] == pm.eval("(f) => f(1)" )(f) assert [None, None, 42, 43 ] == pm.eval("(f) => f()" )(f)
- implemented the copy protocol (both
copy.copy
andcopy.deepcopy
) for JSStringProxies - using the aforementioned WTFPythonMonkey, we've fixed several bugs related to timers, including:
- the
Future object is not initialized
error and following segfault - heap-use-after-free in
timerJobWrapper
- hitting ctrl-C in
pmjs
printing out the entire PythonKeyboardInterrupt
traceback - intervals from
setInterval
were not being unref'd correctly
- the
- fixed a bug where uncaught JS Promise rejections would result in a
Future exception was never retrieved
Python error, rather than the actual JS error - added support for
HTTP-Keep-Alive
in our implementation ofXMLHttpRequest
- fixed a memory leak related to cross-language strings
- fixed a bug where attempting to install Pythonmonkey from source failed on Ubuntu 24.04
- PythonMonkey now uses the bleeding edge version of SpiderMonkey on this and all future releases
- we now build and distribute binaries for python 3.8 on amd64 Mac OS
v0.6.0
PythonMonkey v0.6.0
- significant performance improvements, particularly in memory usage
- improved build system and build time, including making building the docs optional, and the following build types:
- Release: stripped symbols, maximum optimizations (default, what gets published on pip)
- DRelease: same as Release, except symbols are not stripped
- Debug: minimal optimizations
- Profile: same as Debug, except profiling is enabled
- fixed a bug where users with particularly old versions of PythonMonkey were unable to update to the latest release using pip unless they completely uninstalled PythonMonkey first
v0.5.0
PythonMonkey v0.5.0
- fixed a bug where
pmjs -e
/pmjs -p
was not able to call functions that require the event loop - implemented setInterval and clearInterval
- implemented further cross-language support for iterators (more widespread use of iterators, such as the for-of loop for arrays, was already working in previous versions)
using a JS iterator in python:
import pythonmonkey as pm
myit = pm.eval('(function* () { yield 1; yield 2; })')()
print(next(myit)) # 1.0
print(next(myit)) # 2.0
print(next(myit)) # StopIteration exception
using a python iterator in JS:
import pythonmonkey as pm
myit = iter((1,2))
pm.eval("""
(myit) => {
console.log([...myit]); // [1, 2]
}
""")(myit)
v0.4.0
PythonMonkey v0.4.0
- fixed a bug where methods called on proxied JS objects would use
globalThis
for the value ofthis
- implemented proxying of arbitrary python objects in JavaScript, like so:
import pythonmonkey as pm
class Counter:
def __init__(self):
self.count = 0
def increment(self):
self.count = self.count + 1
counter = Counter()
pm.eval("""
(pyObject) => {
console.log(pyObject.count); // 0
pyObject.increment();
console.log(pyObject.count); // 1
}
""")(counter)
- implemented a new type called JSMethodProxy, which can be used to implement methods on python objects in JavaScript, like so:
import pythonmonkey as pm
jsFunc = pm.eval("(function() { this.count++; })")
class Counter:
def __init__(self):
self.count = 0
self.increment = pm.JSMethodProxy(jsFunc, self)
counter = Counter()
print(counter.count) # 0
counter.increment()
print(counter.count) # 1
- various garbage collection optimizations
- various memory leak fixes
- implemented complete cross-language stack traces
pm.eval
can now accept a file object as its first argument (such as an object returned by theopen()
python built-in), which is expected to be a javascript file- when calling
pm.require
(or other requires created bypm.createRequire
), .py CommonJS modules now have precedence over .js modules when there is a namespace collision setTimeout
now returns a Node.js-styleTimeout
class for the timeout id, with.ref()
and.unref()
methods- implemented
XMLHttpRequest.withCredentials
- implemented
"json"
support forXMLHttpRequest.responseType
- implemented remaining standard
console
methods
v0.3.0
Our JS Proxies now implement all Array and Object methods on Python Lists and Dicts, and our Python List and Dict subtypes implements operators and methods on JS Arrays and Objects.
One can now do
items = [1, 2, 3, 4, 5]
result = [None]
pm.eval("(result, arr) => {result[0] = arr.some((element) => element % 2 === 0)}")(result, items)
-> result[0] == True
and going the other way
a = pm.eval("([1,2,3,4,5,6])")
b = a[1:5]
-> b == [2,3,4,5]
We also have all the iterators returned by such methods as values() now working without making copies of the object being iterated upon