esoteric-python

[rapid sparrow]

opening

is it near yield_expr

maybe whatever comes after that? or maybe the problem is that nothing is supposed to follow it?

that looks kind of relevant as well

ahh the number annotation won't work either

because annotated expr starts with NAME

nooo my other change stopped working

[rapid sparrow]

Programs/_freeze_module importlib._bootstrap /data/media/0/src/python3/src/Lib/importlib/_bootstrap.py /data/media/0/src/python3/src/Python/frozen_modules/importlib__bootstrap.h Makefile:765: *** [Python/frozen_modules/importlib__bootstrap.h] Aborted (core dumped)

sometimes gotta be slapped to be sure changes are still happening

oh damn reslly?

what happens to them

how do you edit it?

do you have a python-aware/powered editor?

dang, why would my change not show up anymore

pyc cache?

mine is similarly unresuscitationable

i'm probably doing domeyhing stupid

like running the wrong progran

omg yes

i forgot to set RPATH on it

its using my system-installed libpython

that fixed it OMG

patchelf --set-rpath '$ORIGIN/:$ORIGIN/../lib/' build/python.exe

Syntax error

>>> [f for f in (d:="hello")] File "<stdin>", line 1 SyntaxError: assignment expression cannot be used in a comprehension iterable expression

ya

syntab.c and the file you told me to look at

have no idea what do do with it so i just started doing random changes, natural selection-style

lost? 😕

whycoz.. you have the source files still?

its just some dumb build issue

or run issue

im sure

[golden finch]

surely that can't parse, right?

[cerulean rivet]

brainfuck = [(sys:=__import__(SystemError.__name__[:-~-~-~0].lower())).setrecursionlimit(pow(_:=-~-~-~-~-~-~-~-~0,_)),(type(str().__add__(bin.__name__[0].__add__(range.__name__[:-~-~0]).__add__(int.__name__[:~0]).__add__(type(lambda:0).__name__[:-~-~0]).__add__(chr.__name__[0].__add__(KeyError.__name__[0]))),(),{(0).__init__.__name__:lambda s,c:(None,_:=lambda f,v:setattr(s,f.__name__[0],v),_(id,0),_(pow,0),_(chr,c),_(all,[0]),_(repr,[]),)[0],exec.__name__:lambda s:[lambda:0,lambda:[s._(s.c[s.i]),setattr(s,id.__name__[0],s.i.__add__(-~0),),s.exec()][~0]][s.i.__lt__(len(s.c))](),__import__.__name__[0]:lambda s,c:__import__(str().__add__(compile.__name__[:-~-~0].__add__(len.__name__[0].__add__(len.__name__[:~0].__add__(chr.__name__[0].__add__(TimeoutError.__name__[:-~-~0].lower().__add__(oct.__name__[0].__add__(next.__name__[0].__add__(str.__name__[0]))))))))).defaultdict(lambda:lambda:0,{chr(_:=ord(str(0)).__sub__(-~-~-~0)):lambda:[lambda:0,lambda:s.a.__setitem__(s.p,s.a[s.p].__sub__(-~0),),][s.a[s.p].__ne__(0)](),chr(_.__sub__(-~-~-0)):lambda:s.a.__setitem__(s.p,s.a[s.p].__add__(-~0)),chr(-~_):lambda:print(end=chr(s.a[s.p])),chr(_:=-~-~-~ord(str(-~-~-~-~-~-~-~-~-~0))):lambda:setattr(s,pow.__name__[0],s.p.__sub__(-~0)),chr(-~-~_):lambda:[lambda:[s.a.append(0),setattr(s,pow.__name__[0],s.p.__add__(-~0))],lambda:0,][s.p.__eq__(len(s.a))](),chr(_:=-~ord(zip.__name__[0].upper())):lambda:s.r.append((s.i,s.p)),chr(-~-~_):lambda:[_:=s.r.pop(),[lambda:0,lambda:[s.r.append(_),setattr(s,id.__name__[0],_[0])]][s.a[_[~0]].__gt__(0)]()]})[c]()}))][~0]

https://github.com/NiumXp/pyfuck

[golden finch]

I'm gonna be honest what does that do?

I see a lot of incrementing via unary

[snow beacon]

I'm adding a credits section to https://github.com/IFcoltransG/esoteric-python-guide and I've used your aphorism on the readme. How would you suggest I credit the quote?

[golden finch]

or decrementing

neat implementation

I think you can actually get rid of the need for the - via int.__neg__

wait, no

that won't work because it contains .

use getattr then, perhaps

though your implementation does in fact contain [].-,

[simple crystal]

ya you can just use strs of code and exec them

[snow beacon]

Or you can store the functions in the dict.

[rugged sparrow]

Do you want any write ups from me for that guide?

[snow beacon]

It would be great to have a writeup about Cpython specific things, and possibly also bytecode. The goal is that someone can read a public resource, then understand most of what's going on in this channel.

I don't currently understand a lot of Cpython or bytecode — I haven't bothered to learn — so I can't really make those guides.

I should really bother to write some more guides as well.

Do you mean a preexisting one?

I don't care about duplicating things that already exist. People have a lot of knowledge in their noggins that isn't already out in the world, which is most important to share.

[rugged sparrow]

>>> interned['abc'] = 'magic'
>>> 'abc'
'magic'
>>> 'abc' is 'magic'
<stdin>:1: SyntaxWarning: "is" with a literal. Did you mean "=="?
True
>>> ```

they do in 3.10

if they werent in 3.10 then the above code wouldnt have worked

PyAST_Optimize does it i think?

>>> interned['abc'] = 1
>>> 'abc'
1
>>> ``` 👀

!e ```py
from ctypes import util
from ctypes import *
import atexit

base_size = sizeof(c_void_p)
libc = cdll.LoadLibrary(util.find_library('c'))

PAGE_SIZE = libc.getpagesize()
MEM_READ = 1
MEM_WRITE = 2
MEM_EXEC = 4
ENDIAN = 'little' if memoryview(b'\1\0').cast('h')[0]==1 else 'big'

libc.mprotect.argtypes = (c_void_p, c_size_t, c_int)
libc.mprotect.restype = c_int

def mprotect(addr, size, flags):
addr_align = addr & ~(PAGE_SIZE - 1)
mem_end = (addr + size) & ~(PAGE_SIZE - 1)
if (addr + size) > mem_end:
mem_end += PAGE_SIZE
memlen = mem_end - addr_align
libc.mprotect(addr_align, memlen, flags)

def addr(cfunc):
ptr = c_void_p.from_address(addressof(cfunc))
return ptr.value

def hook(cfunc, restype=c_int, argtypes=()):
cfunctype = PYFUNCTYPE(restype, argtypes)
cfunc.restype, cfunc.argtypes = restype, argtypes
o_ptr = addr(cfunc)
mprotect(o_ptr, 5, MEM_READ | MEM_WRITE | MEM_EXEC)
mem = (c_ubyte5).from_address(o_ptr)
default = mem[:]
def wrapper(func):
@cfunctype
def injected(*args, **kwargs):
try:
mem[:] = default
return func(*args, **kwargs)
finally:
mem[:] = jmp
n_ptr = addr(injected)
offset = n_ptr - o_ptr - 5
jmp = b'\xe9' + (offset & ((1 << 32) - 1)).to_bytes(4, ENDIAN)
mem[:] = jmp
@atexit.register
def unhook():
mem[:] = default
injected.unhook = unhook
return injected
return wrapper

@hook(pythonapi.PyDict_SetDefault, restype=py_object, argtypes=[py_object]*3)
def setdefault(self, key, value):
if key == 'MAGICVAL':
return self
return pythonapi.PyDict_SetDefault(self, key, value)

pythonapi.PyUnicode_InternFromString.restype = py_object
interned = pythonapi.PyUnicode_InternFromString(b'MAGICVAL')
setdefault.unhook()
interned['abc'] = 1
print(eval("'abc'"))```

[night quarry]

@rugged sparrow :white_check_mark: Your eval job has completed with return code 0.

1

[rugged sparrow]

no i don't

only works on unix based OS, intel x86 cpu (unless the assembly happens to be the same on other platforms)

[rugged sparrow]

it actually should work if you rewrite mprotect to use the windows api

https://github.com/fdintino/python-doublescript/blob/master/doublescript/asm_hooks.py

[rugged sparrow]

weirdly this doesn't seem to work unless the string was in the interned dict previously

[golden finch]

strange

[thorny tapir]

!e print(chr(ord(chr(ord(chr(ord("a")))))))

[night quarry]

@thorny tapir :white_check_mark: Your eval job has completed with return code 0.

a

[thorny tapir]

not even esoteric but wtv

[knotty delta]

!e

from ctypes import c_void_p
class int2(int):
    def __add__(self, other):
        return self - other
c_void_p.from_address(id(int)+8).value = id(int2)

print(1 + 1)
print(int("1") - int("1"))

[night quarry]

@knotty delta :x: Your eval job has completed with return code 1.

001 | 2
002 | Traceback (most recent call last):
003 |   File "<string>", line 8, in <module>
004 | TypeError: 'int2' object is not callable

[knotty delta]

hm?

!e

class test:
    ...
print(callable(test))

[night quarry]

@knotty delta :white_check_mark: Your eval job has completed with return code 0.

True

[knotty delta]

!e

from ctypes import c_void_p
class int2(int):
    def __add__(self, other):
        return self - other
print(callable(int2))
c_void_p.from_address(id(int)+8).value = id(int2)

print(1 + 1)
print(int("1") - int("1"))

[night quarry]

@knotty delta :x: Your eval job has completed with return code 1.

001 | True
002 | 2
003 | Traceback (most recent call last):
004 |   File "<string>", line 9, in <module>
005 | TypeError: 'int2' object is not callable

[knotty delta]

oh nvm i probably don't want to use c_void_p

!e

print(callable(object))

[night quarry]

@knotty delta :white_check_mark: Your eval job has completed with return code 0.

True

[knotty delta]

hmm

[velvet lichen]

Is there a way to get the annotations of a function, where it is also includes untyped? I'm sure there's a way that combines __annotations__ with __code__, but all I can find in __code__ is the co_argcount, just the number of arguments or co_varnames, the names of all variables in the function, not just arguments?

e.g calling the function to get the annotations on this method

def test(x: int, y) -> float:

would return {'x': int, 'y': "", 'return': float} if the thing to indicate untyped was empty string.
I may just go with the inpect module, but I was sort of hoping to do it module-less

[tacit cobalt]

!e ```py
def test(x: int, y) -> float:
...

print({**{arg: None for arg in test.code.co_varnames[:test.code.co_argcount]}, **test.annotations})

[night quarry]

@tacit cobalt :white_check_mark: Your eval job has completed with return code 0.

{'x': <class 'int'>, 'y': None, 'return': <class 'float'>}

[velvet lichen]

Oh that's nice, thanks. So it always has the arguments first?

[tacit cobalt]

Yeah

[velvet lichen]

Great, thanks

[astral rover]

def test(x: int, y) -> float:
    ...

print(dict.fromkeys(test.__code__.co_varnames[:test.__code__.co_argcount]) | test.__annotations__)```might be more succinct if youre on 3.9+

[velvet lichen]

I haven't upgraded yet but that is very nice and I might as well upgrade now, so thanks, I'll use that.

[earnest wing]

.value expects a normal object, not an address

[knotty delta]

my bad

[velvet lichen]

The one thing that the typing doesn't cover is the return dictionary element, but that will be fine to add

[knotty delta]

!e

from ctypes import c_void_p
class int2(int):
    def add(self, other):
        return self - other
c_void_p.from_address(id(int)+8).value = int2
print(1 + 1)
print(int("1") - int("1"))

[night quarry]

@knotty delta :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 5, in <module>
003 | TypeError: cannot be converted to pointer

[knotty delta]

hm

[velvet lichen]

I also think None as in unassigned is interchangeable with the None type that can be indicated, which may be an issue

[tacit cobalt]

Docs say that

Note that None as a type hint is a special case and is replaced by type(None).
But

!e ```py
def test() -> None:
...

print(test.annotations["return"] == type(None))

[night quarry]

@tacit cobalt :white_check_mark: Your eval job has completed with return code 0.

False

[tacit cobalt]

https://docs.python.org/3/library/typing.html

[astral rover]

thats when you use

!d typing.get_type_hints

[night quarry]

typing.get\_type\_hints

typing.get_type_hints(obj, globalns=None, localns=None, include_extras=False)```
Return a dictionary containing type hints for a function, method, module or class object.

This is often the same as `obj.__annotations__`. In addition, forward references encoded as string literals are handled by evaluating them in `globals` and `locals` namespaces. If necessary, `Optional[t]` is added for function and method annotations if a default value equal to `None` is set. For a class `C`, return a dictionary constructed by merging all the `__annotations__` along `C.__mro__` in reverse order.

The function recursively replaces all `Annotated[T, ...]` with `T`, unless `include_extras` is set to `True` (see [`Annotated`](https://docs.python.org/3/library/typing.html#typing.Annotated "typing.Annotated") for more information). For example:

[tacit cobalt]

I guess Aster should just replace all None type hints with type(None)

[velvet lichen]

Thanks

Also, is there a way to get the types from a types.Union? I did a dir of it and found nothing obvious, and apparently you can't create an instance of it

[astral rover]

__args__

[velvet lichen]

Nice, thanks again

[velvet lichen]

How might I get a class' dunder method to be a method of the metaclass it is derived from? (preferably without the dynamic code gen)

[vivid prairie]

https://mail.python.org/pipermail/python-announce-list/2004-April/002982.html

it's a joke

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

True

[knotty delta]

!e print(exec(compile('1<>2','a','exec',4194304)))

[night quarry]

@knotty delta :white_check_mark: Your eval job has completed with return code 0.

None

[knotty delta]

!d str.replace

[night quarry]

str.replace

str.replace(old, new[, count])```
Return a copy of the string with all occurrences of substring *old* replaced by *new*. If the optional argument *count* is given, only the first *count* occurrences are replaced.

[knotty delta]

🤔

[night quarry]

Command Help

!eval [code]
Can also use: e

*Run Python code and get the results.

This command supports multiple lines of code, including code wrapped inside a formatted code
block. Code can be re-evaluated by editing the original message within 10 seconds and
clicking the reaction that subsequently appears.

We've done our best to make this sandboxed, but do let us know if you manage to find an
issue with it!*

[vestal solstice]

your appointment to FEMA will be finalized within the week

[earnest wing]

!e print(exec(compile('1<>2','a','eval',4194304)))

[night quarry]

@earnest wing :white_check_mark: Your eval job has completed with return code 0.

None

[shadow carbon]

Hey guys. Does anyone know how to program an event of detecting a windows toast notification?
I want python to detect a type of toast notification with specific keywords, etc to perform an action.

[grave rover]

cursed code screenshot of the day

[rugged sparrow]

are you using kotlins C integration to make C python modules?

[grave rover]

yes

and it's painful as heck

[rugged sparrow]

i can imagine

[grave rover]

I mean it definitely could be worse

it's mainly the fact that staticCFunction needs to be top-level

also strings are wacky

[sullen glen]

Hello

[grave rover]

o/

[sullen glen]

can anyone help me to get started with python

[grave rover]

wrong channel buddy

[sullen glen]

sorry

[grave rover]

crazy how people ask this channel of all places

[rugged sparrow]

its cause its active people ask in any channel that has activity

[knotty delta]

perhaps they want to learn esoteric python who knows

[grave rover]

I don't think you'd word it as "get started with python" in that case

[pure dew]

yo long time no see

[grave rover]

yooo

[grave rover]

what you been up to

[pure dew]

working a lot lately

but i've been doing stupid stuff with python before that

doing some more work on my language vm now

[grave rover]

oh god don't remind me

[pure dew]

??

oh you'll like thjis

https://totally-not.a-sketchy.site/AdbqcVj.png

@grave rover and this too

class A:
    ...


@derive(Debug, Colorize)
@dataclass
class CFrame(A):
    x: int
    y: int
    z: int

    rx: float
    ry: float
    rz: float

    @classmethod
    def origin(cls) -> "CFrame":
        ...

    def add(self, other: "CFrame") -> "CFrame":
        ...

    def translate(self, x: int, y: int, z: int):
        ...

    @staticmethod
    def zero() -> int:
        ...


cframe = CFrame(23, 4, 12, 98.2, 110.2, 0.002)

print("Debug:")

print(cframe.fmt())

print("\nColorize:")

print(cframe.bright_green().on_black())

https://totally-not.a-sketchy.site/yQPacJK.png

trait system for python

[earnest wing]

oh god

[pure dew]

i discovered many unsafe ways of monkeypatching object

[earnest wing]

Blanket impls

[grave rover]

what the actual fuck

[pure dew]

```py
def derive_all():
def format_wrapper(self):
return repr(self)

# WARNING: very sensitive - arbitrary things cause this to break
for _type in object.__subclasses__():
    if _type not in IMPLEMENTED_TYPES:
        if _type.__module__.startswith("traitor"):
            continue

        # print(f"{_type.__name__}: {_type.__module__}")
        impl(Debug >> _type)(type(f"Debug{_type.__name__.title()}", (object,), {"fmt": format_wrapper}))

this breaks on nearly every non-pure-python module

[grave rover]

I mean have you seen my port of spongepowered/mixin to python

[pure dew]

i think so

[earnest wing]

unsurprisingly, c-implemented types mess up

[pure dew]

but i can patch c-implemented types

but the blanket impl breaks it

so I just implemented it for object and let inheritance cover my ass

[grave rover]

sounds like #esoteric-python alright

[pure dew]

and then something else breaks...
AttributeError: 'method' object has no attribute '__code__'

[grave rover]

I wonder if I could use kotlin-reflect to make py types from kt classes

[pure dew]

this here is fun too

>>> range(10).map(lambda x: Nothing if x % 2 == 0 else Just(x)).collect()
[Nothing, Just(1), Nothing, Just(3), Nothing, Just(5), Nothing, Just(7), Nothing, Just(9)]

>>> _.map(lambda m: m.fmap(lambda n: n + 2)).collect()
[Nothing, Just(3), Nothing, Just(5), Nothing, Just(7), Nothing, Just(9), Nothing, Just(11)]

[grave rover]

oh no

[rugged sparrow]

What method are you using to patch c types?

[pure dew]

fishhook, under the hood

but there's a whole bunch of nasty above it

if "__sentinel" not in self.target.__dict__.keys():
            sentinel = MethodSentinel(self.target)

            @wraps(self.target.__getattribute__)
            def wrapper(s, name):
                return sentinel(s, name)

            hook(self.target, "__sentinel")(sentinel)

            hook(self.target, "__getattribute__")(wrapper)

        getattr(self.target, "__sentinel").add_impl(self.trait, self.impl)

[rugged sparrow]

whats the nasty for? (what does MethodSentinel do)

[pure dew]

it replaces __getattribute__ and manages multiple traits and possible overlaps

if you have two traits with the same method implemented for the same type, it'll catch it and make you disambiguate

like Rust

[rugged sparrow]

ahh clever

would you want my smaller hook implementation instead of fishhook?

[pure dew]

and also handles rebinding of methods to the new instance

if it works, sure

if you ever wanted a trial by fire this is it lol

[rugged sparrow]

https://paste.pythondiscord.com/uvasitecey.py

it should work

it toggles some type flags and lets python handle the inheritance (instead of how fishhook does it now)

[pure dew]

oh is this the posix-only thing?

[rugged sparrow]

no

*it shouldnt be

https://paste.pythondiscord.com/uvutacehow.py use this one

(also tbh i never tested fishhook on windows fyi)

[pure dew]

i dont know if i ever did or not

but i run linux so it doesnt really matter

this shouldnt be near a production machine to begin with

[rugged sparrow]

hahaha

https://github.com/chilaxan/pysnippets/blob/main/small_hook.py#L12-L29 small_hook does use an exploit to get memory rw instead of ctypes

[pure dew]

is this a drop-in for fishhook?

[rugged sparrow]

it doesnt have hook_cls, hook_cls_from_cls, or unhook cause i havent added those yet

[pure dew]

but hook(item, attr) should work the same?

[rugged sparrow]

ye

[pure dew]

ah, well, it segfaults

[rugged sparrow]

shoot

[pure dew]

python 3.9.5

[rugged sparrow]

hmmm

lemme test

[pure dew]

owait

this may be an issue with my code

[rugged sparrow]

haha

what hook segfaults it?

(also hook(cls, name, value) works with small_hook)

[pure dew]

its not so much the hook, but the recursive hooking

is that github file the same as what you sent earlier?

[rugged sparrow]

oh you dont need to hook recursively anymore cause small_hook handles inheritance right

yea

wait lemme check something

[pure dew]

oh

hooking into object breaks it

[rugged sparrow]

yup found the bug

gimme 5

[pure dew]

hooking into type does BadThings™️

<unknown>RecursionError: maximum recursion depth exceeded while calling a Python object

[rugged sparrow]

https://github.com/chilaxan/pysnippets/blob/main/small_hook.py

@pure dew fixed the hook object bug (i needed to recursively allocate new structures)

[pure dew]

hrm

[rugged sparrow]

originally i only allocated structs for the passed in class, so python was writing to invalid memory when it attempted to handle subclasses

fishhook only works with this cause i did the subclasses myself (so it passed my struct check and allocate code)

[pure dew]

doesnt seem to work for me

[rugged sparrow]

still segfaulting?

[pure dew]

yes

but i've narrowed down the location and it makes no sense

[rugged sparrow]

what location

[pure dew]

from traitor.traits.debug import Debug
print("done with imports")

class A:
    ...

print("set up A")

it seems to break on the A

oh yeah its whenever i declare a class

how...

[rugged sparrow]

and your only hook is __getattribute__?

[pure dew]

only hook that overrides any possible existing entries, yes afaik

if I remove my object hook it works fine

[rugged sparrow]

but fishhook works?

thats bizzare

guess i need to work on small_hook more

[pure dew]

ah yeah fishhook does work

but i probably shouldnt have a blank implementation anyway lol

[rugged sparrow]

if i had to guess i would assume that small_hook only sets the generic tp_getattr while fishhook sets all of them

[pure dew]

wacky business here https://totally-not.a-sketchy.site/8BRwSV7.png

as soon as i declare a class it segfaults

and thats with object hooked

i wonder if yours is trying to make the inheritance work too hard and screwing with A as its being setup

[rugged sparrow]

Internally fishhook and small hook should be doing the same things tho. Cause I wrote fishhook and small_hook from the same research

[shut trail]

You created fishhook??

Woah

How did you achieve changing builtins?

[rugged sparrow]

fishhook uses ctypes to change function pointers at runtime

[pure dew]

oh you wrote fishhook?

nice lol

[rugged sparrow]

yea lol last year

[shut trail]

c_void_p.address?

I tried doing that but it didn’t change the actual things itself

For example, I tried to overwrite int with a new class for c_void_p.address(id(int) + 8) but if I used integers like 2 and 3 they would still be an int

[pure dew]

overwriting literal types? now that is esoteric python

[rugged sparrow]

That would overwrite the type of int, not 2 or 3

[shut trail]

ah, how would you do the latter?

[rugged sparrow]

c_void_p.from_address(id(2)+8).value = new_cls

[pure dew]

is it possible to overwrite the type of a literal tho?

not to just overwrite the type of 3, but change the type of any literal integer

without hardcoding infinite numbers

[shut trail]

Ah, what if you wanted to change all integer types though?

[rugged sparrow]

That's trickier

[rugged sparrow]

Depends on what strategy you want to use

[shut trail]

What would be an example of a strategy?

[rugged sparrow]

you can abuse inheritance

also what would your goal be?

[shut trail]

to override the literal type to allow say a new dunder

like if I want to add support for a certain dunder

[rugged sparrow]

then you should just hook int

wdym a new dunder tho

like adding __getitem__ to numbers?

[shut trail]

it's a generic example, but say I wanted to add __iter__ and __next__ to int

and how would you implement it by itself

where you take new literals to implement those methods

[pure dew]

oh well thats easy enough

just patch builtin int

but i mean doing where type(5) != int

[shut trail]

how though

Without using a module like forbiddenfruit or fishhook

[pure dew]

you can't

but all they're doing is messing with C pointers

and it is pure python

[shut trail]

Yea so how would I try and do that

[pure dew]

i don't know, your best bet is to look at what they do

[rapid sparrow]

lol

classic

you have to use C I expect

to fill the various Type "slots"

maybe it could be done in python if you are good at computing offsets..

[rugged sparrow]

^ thats what fishhook and small_hook do, they compute offsets at runtime

[rapid sparrow]

oh that's clever, so you translate __repr__ to wherever the tp_repr is?

on this monster

and somehow give it the right type

too bad there's two different meanings for slots

does it work the same whether it's a static or heap allocated type

[rugged sparrow]

fishhook computes the actual offset for tp_repr, small_hook lets python do it

[rapid sparrow]

wondering how python could be off any help

[rugged sparrow]

well i toggle the flag that python checks to forbid modifying static classes (so python thinks its a heap type class)

then i just setattr

and then i flip the flag back

[rapid sparrow]

oh so python already knows how to do it basically

[rugged sparrow]

it has to inorder for user defined classes to work

[rapid sparrow]

huh, never thought of that

so when's the third hooking package coming

maybe large_hook?

so when you do __dict__ on a type, it had to be all generated from whatever's on these slots?

[rugged sparrow]

sorta?

slot_wrappers have an offset attribute that can be used to find the C function they wrap

[gilded spoke]

what is escoteric-python?

[snow beacon]

Intentionally weird code.

Things like trying to write your code with as few characters as possible, or making 2 + 2 = 5.

I bet that's valid Haskell code, including the >>>.

Actually those quotes might not work.

[pure hearth]

.

[still lily]

Quick question, has anyone tried replacing and and or in bytecode with function calls? I need that for a project of mine and I wouldn't want to reinvent the wheel

So that expr_a and expr_b becomes f(expr_a, expr_b)

I guess I could rotate the stack slightly to avoid the pain of inserting LOAD_CONST before the left expr

yeah

I could load f right after JUMP_IF_FALSE_OR_POP, then ROT_TWO and insert CALL_FUNCTION after expr_b is on the stack I guess?

Right, yeah

I'm thinking I could use JUMP's destination to find the point where expr_b is on the stack

I wonder how much code relies on the short-circuit aspect of it tbh

Thanks for the help!

[rugged sparrow]

@still lily depending on when your code is executed to mutate the bytecode, expr_b may not be on the stack yet

But you can use the jump to determine the segment of bytecode that corresponds to expr_b and move it as a block

[still lily]

yeah, that's what I'm doing right now

I'm assuming the jump target stands right after expr_b is fully loaded

[rugged sparrow]

It should

[still lily]

So I'm inserting CALL_FUNCTION immediately before the jump target

Though that's probably going to mess up further jumps, now that I think about it

[rugged sparrow]

you need to use a decompiler and recompiler to make sure you dont break any jumps

[still lily]

Could I perhaps fix further jumps by keeping track of the replacements count? Each replacement should insert two instructions, so I'm thinking about incrementing every non-relative jump's target accordingly

That seems to have mostly fixed the issue

[rugged sparrow]

that will work in the meantime but you could still run into issues

[still lily]

yeah, it's a wonky endeavor

[severe bison]

hey

how do i hide blocked messages

in discord

using python

[earnest wing]

sir this is a wendy's #esoteric-python

[severe bison]

where do i ask then

@earnest wing

[mortal thorn]

This sounds really interesting and weird. What kind of project makes you wanna do this?

[still lily]

I'm toying around with a proxy-ish kind of thing for LINQ-like syntax in python, though there's a LOT of hoops to jump through for more complicated cases

The goal is to allow for things like this:

with collection.item as i:
    squares = collection.select(i ** 2).where(i < 10)

i is an instance of said proxy class that first stores every operation performed on it, and then performs the same set of actions on any object you pass to its protected __eval method

Common operations (attribute access, method calls, calling the proxy itself, operators, so on and so forth) are trivial, but things become way harder when it comes to and, or, passing it as a function argument, and some other cases

I've made it around and and or with bytecode manipulation (If you want to support correct evaluation of and and or, you apply a proxify decorator to the function this takes place in), and I have some ideas on how to allow mixing proxies together (i.attr(i.other_attr)), but I'll probably have to put strict limitations on passing proxies as function arguments

Funnily enough, this whole project started from an entirely unrelated AttributeIndexer mapping that would allow you to query a sequence of items via any of its attributes

[fleet loom]

I noticed that when I use 2to3 on the simple script print "foo", it converts it to python3 with a different print behavior. Example below:

$: cat myscript.py
print "foo",

$: python2 myscript.py
foo

$: 2to3 -w -n --add-suffix=3 myscript.py 
RefactoringTool: Skipping optional fixer: buffer
RefactoringTool: Skipping optional fixer: idioms
RefactoringTool: Skipping optional fixer: set_literal
RefactoringTool: Skipping optional fixer: ws_comma
RefactoringTool: Refactored myscript.py
--- myscript.py    (original)
+++ myscript.py    (refactored)
@@ -1 +1 @@
-print "foo",
+print("foo", end=' ')
RefactoringTool: Writing converted myscript.py to myscript.py3.
RefactoringTool: Files that were modified:
RefactoringTool: myscript.py

$: cat myscript.py3 
print("foo", end=' ')

$: python3 myscript.py3
foo $:

That is, without a trailing newline. The print behavior is clearly different, but this exact case is covered by tests in the source code: https://github.com/python/cpython/blob/main/Lib/lib2to3/tests/test_fixers.py#L500

Why would it be this way? What am I missing?

[night quarry]

Lib/lib2to3/tests/test_fixers.py line 500

def test_trailing_comma_3(self):```

[snow beacon]

You might need to adjust some of the relative jumps, if they go over the replaced section.

[still lily]

true, gotta check if the jump target is inside the block and if it's not, adjust it accordingly

[sick hound]

the hell is dis?

mf seg faulted python how?

[bright hearth]

I've done that before

[sick hound]

that scares me

[bright hearth]

[a for a in ().__class__.__class__.__base__.__subclasses__() if a.__name__ == 'function'][0]([a for a in ().__class__.__class__.__base__.__subclasses__() if a.__name__ == 'code'][0](0, 0, 0, 0, 0, 0, b'please explode now :)', (), (), (), '', '', 0, b''), {})()```

[sick hound]

when python starts looking like in line assembly you know your from hell.

[bright hearth]

This is actually way simpler than it seems.

As you probably know, everything is an object - this extends to functions.

[sick hound]

I know basic 32 assembly but shit tf is yall doing in here lmao

[bright hearth]

So if functions are a kind of object, then maaaybe there's a function constructor, right?

There is.

[sick hound]

your making me cry.

[bright hearth]

A function constructor requires a name, and a piece of raw bytecode to execute when the function is called.

Raw bytecode is also a kind of object, of the code class It has its own constructor.

To construct a code object, you need byte code as bytes - and some smol metadata about how to do the setup for executing it.

All the zeros and empty tuples and shit you see in this snippet are just metadata I've ignored or provided junk values to. Importantly, this tries to set up the code with a stack size of zero.

So this constructs code which will try to run with a zero size stack - and builds a function which will execute that code

then it calls the function.

Since there's no stack, it crashes instantly.

[sick hound]

then it is outside the stack

ok gotcha

[bright hearth]

b'please explode now :)' is actually a red herring. this is the bytecode which will try to execute, and the instructions are junk.

It doesn't matter what instructions you put here because the code will always crash on entry without even running the instructions.

[sick hound]

im trying to ask a question but keep reading and getting more confused

[bright hearth]

Oh right there's a bit of other weird shit here huh

() is an object - it's an empty tuple literal.

So it has a class

and that class has parent classes.

().__class__.__class__.__base__ is the object base class.

[sick hound]

I only understand the part where it makes stack zero then crash

[bright hearth]

So ().__class__.__class__.__base__.__subclasses__() is every subclass of the object base class

which includes both the function class and the code class.

I just do a list comprehension to find them.

So [a for a in ().__class__.__class__.__base__.__subclasses__() if a.__name__ == 'function'][0] will get you the function class.

Because it filters out classes whose name is equal to 'function'

Only one class has that function, so the list comprehension builds a list with only one element - the function class.

[0] just pulls it out of the list ofc, that's just regular list use

().__class__.__class__.__base__.__subclasses__() if a.__name__ == 'code'][0] does the same thing, but for the code class instead.

Finding them lets me call their constructors - which I need to do since I need to make a code object, and a function object.

Try it in repl

[sick hound]

holy hell man this is hard to understand

[bright hearth]

This might be easier to look at

[sick hound]

yeah it breaks in mine too

also that helps a bit

[bright hearth]

This one is really cool but a lot harder to explain and only works in some python versions

import random
class E(BaseException):
    def __new__(cls, *args, **kwargs):
        return 1
def a(): yield
a().throw(E)```

it only works in some versions because unlike my crappy little joke code, this is actually due to a legitimate bug in the Python interpreter.

I think it works up until 3.9

In later versions, it raises a TypeError saying that an instance of Exception was expected, not int.

In older versions, it segfaults due to a null pointer dereference.

def a(): yield```
This little function just constructs and returns a generator

This generator has no items, so as soon as you try to call next() on it you'll raise StopIteration.

Generators have a little-known feature called throw, which allows you to set up a generator such that the next time someone tries to get a value from it, an exception of your choice is raised.

Actually it might just throw from the iterator right away, it's a weird feature...

>>> a().throw(ValueError("Boo!"))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<stdin>", line 1, in a
ValueError: Boo!
>>>```

[tacit cobalt]

!eval ```py
type(lambda:0)((lambda:0).code.replace(co_consts=()),{})()

[night quarry]

@tacit cobalt :warning: Your eval job has completed with return code 139 (SIGSEGV).

[No output]

[bright hearth]

Anyways, the important thing is that an exception is raised from the generator builtin type's code.

And that code is written in C.

But in Python you can control __new__, which lets you write the code for making new objects of a particular class.

__new__ has to return an object. This object is then used as self for the __init__ method.

Anyways, you'll notice that this E class - which inherits BaseException, returns an int from __new__.

So even though the E class inherits baseException, actually calling the constructor returns an int.

So here's how the crash works:
generator.throw (which is written in C) calls the constructor to build a new Python Object. It then uses an attribute of this object.

generator.throw fails to actually check the type of the object it built (in fixed versions it checks and raises TypeError)

So it gets a pointer to python int, not a pointer to python BaseException.

It then tries to use a member of BaseException. And since the pointed object is a python int which doesn't have that member, this is invalid. Junk memory is accessed and it fatally dies.

[tacit cobalt]

!eval ```py
def a():yield
a().throw(type("E",(BaseException,),{"new": lambda*a:1}))

[night quarry]

@tacit cobalt :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 2, in <module>
003 |   File "<string>", line 1, in a
004 | TypeError: calling <class '__main__.E'> should have returned an instance of BaseException, not int

[bright hearth]

Yup, that's how it should work.

Can you make the bot run the code in a specific version?

Try 3.5, I know it works in that.

[tacit cobalt]

!eval ```py
import sys
print(sys.version_info)

[night quarry]

@tacit cobalt :white_check_mark: Your eval job has completed with return code 0.

sys.version_info(major=3, minor=10, micro=0, releaselevel='final', serial=0)

[tacit cobalt]

Oh they updated

Works in 3.9.4 anyways

[tired umbra]

I feel like a beginner when I see this channel

anyways, esteric side of python looks cool

[night quarry]

@sick hound :warning: Your eval job has completed with return code 0.

[No output]

[tacit cobalt]

Still works with 1114113 constants

[knotty delta]

what are you guys trying haha

on which version?

ooo

lets try python 1.9

ok nvm cant find it, lets use 1.6.1

possibly won't work on windows 😢

😦

guess the only way is to wsl

but i don't know how

that is python 1.6.1 lol

ooo

lets see is it a lot of constants were on 0.9.1 already 😈

[tacit cobalt]

Still works with and

[knotty delta]

>>> sys.ps5
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'sys' has no attribute 'ps5'
``` 😦

there is no ps5

not yet

i'm still on 3.9

i dont have any c compilers now

are there some compiled binaries i can install

[tacit cobalt]

It actually just optimizes everything away and returns 0 so you end up with

  2           0 LOAD_CONST               1 (0)
              2 RETURN_VALUE

I tried with addition but it just does nothing

Or exceeds the maximum recursion depth

[sleek sphinx]

how do i get started with esoteric python this is my favourite channel in the whole server

[earnest wing]

you can also use any object that implements str for dynamic prompts

[rugged sparrow]

@sick hound with EXTENDED_ARG python can handle up to max 32 bit signed integer constants

Because internally the variable for oparg is 4 bytes

[earnest wing]

I have line numbers and local time, what wacky stuff should I fill my prompt with?

I was thinking terminal colors

[knotty delta]

and syntax highlight

[earnest wing]

Do that yourself

The prompt gets no direct info about the code executed.

[pure dew]

hook into sys and override the keyboard input code

read the line as its typed and highlight it

like how zsh highlighting works

[rugged sparrow]

You could hook sys.stdin

[near gust]

Is there a way to do slice assignment without using slice notation?

[proper vault]

!e

a = [1, 3, 6]
a[slice(2)] = [4, 6]
print(a)

[night quarry]

@proper vault :white_check_mark: Your eval job has completed with return code 0.

[4, 6, 6]

[rugged sparrow]

@grave rover https://bytecode.readthedocs.io/en/latest/usage.html this may be useful for your decompiler, it has CFG

[grave rover]

oh hey that;s basically pyasm

and uh

I'm actually almost done with my CFG impl

[sleek sphinx]

what is the opposite of golfing like making the most creative and long code to print hello world

[grave rover]

I think that's called code bowling

[near gust]

Nevermind I figured it out

I'm working on a compiler that compiles Python (from bytecode) to another language that can call Python functions. However, the language has the exact reverse of Python bytecode call function semantics, so I would have to reverse a slice of the stack (or I could create another stack and do a bunch of transferring back and forth)

[grave rover]

is it lua

[near gust]

No, it's an esoteric language that I made that will scar you for life

I made both a compiler (to python) and interpreter (in pure Python)

Quite literally the compiled version can import any installed python module and push any Python function to the stack and call it with positional arguments

I'm doing this as a joke to prove that the language is turing complete because there are too many languages that use simple languages to prove turing completeness, so I'm deciding to make the entirety (with some limitations) of Python.

Wait a minute... It is already turing complete because it is a pda with at least 2 stacks with the ability to push, pop, rot2, rot3 and transfer between stacks

[bright hearth]

I'm pretty sure MS Word Autocomplete is turing-complete

You just have to have a human to "power" the turing machine by clicking the prompt which accepts the suggested change.

[rugged sparrow]

Yea

You can also overflow oparg with EXTENDED_ARG

[astral rover]

Why is the number of constants signed?

That just seems like a waste

[bright hearth]

What, you're just going to assume I want at least zero constants?

I want -1 constants. When my code is defined, it undefines a constant located somewhere else.

[rugged sparrow]

ironically you can construct code that loads the -1 const

this ends up loading the size of the consts as a pointer to a py_object

(if you load -2 you get tuple)

!e py import dis f = lambda:None f.__code__ = f.__code__.replace(co_code=bytes([ dis.opmap['EXTENDED_ARG'], 255, dis.opmap['EXTENDED_ARG'], 255, dis.opmap['EXTENDED_ARG'], 255, dis.opmap['LOAD_CONST'], 254, dis.opmap['RETURN_VALUE'], 0 ])) print(f())

[night quarry]

@rugged sparrow :white_check_mark: Your eval job has completed with return code 0.

<class 'tuple'>

[rugged sparrow]

that is because internally, a tuple looks like this [ob_refcount, ob_type, ob_size, item0, item1, item...]

[rugged sparrow]

weird

i dont think it should crash (but i have no idea lol)

[pure dew]

ooh, neat lib

[rugged sparrow]

tbh wish i had found it sooner

[snow beacon]

I feel like I'm self-advertising a lot these days, but I made a guide at https://github.com/IFcoltransG/esoteric-python-guide

[sleek sphinx]

nice

[shadow dome]

are there no metaclasses in there?

I think there's not even dynamic class creation... And I don't think there's really a way to contribute to a gh wiki...

[bright hearth]

have a slightly esoteric fizz buzz

_ = any(map(lambda n: (False, print((lambda n: (n, "fizz", "buzz", "fizzbuzz")[int(0 == n % 3) + 2 * int(0 == n % 5)])(n)))[0], range(100)))```

that is better

[sick hound]

any good guide on bytecode?

Im asking cause I have only come across this one

https://towardsdatascience.com/understanding-python-bytecode-e7edaae8734d

[grave rover]

Guess I'll try writing a guide later

[sick hound]

thanks

[maiden river]

while studying lambda calculus, got an interesting idea, how about implementing it just using functions in python
like defining 'true' 'false' just using functions

and surprisingly i was successful in making gates and half adder, then a full adder and finally a 4 bit adder : )
i used 'l' for 1 and 'o' for 0

the aim was to implement binary logic from ground up, so can't use 1,0 because they are already reserved
(for fun and learning of course)

[proper vault]

yeah, that is pretty fun

there is however one difference in that python is strict, but lambda calculus is lazy

so you need a Z combinator instead of a Y combinator

[maiden river]

Y & Z combinator 👀

I did it like this

`def true(a,b): return a

def Not(a): return a(true,false)

def And(a,b): return a(b,a)
`

[ruby briar]

hello, i have an hw to do but i am big noob, can someone help me ?

[proper vault]

yup, that's the smart way

[ruby briar]

Write a program that asks to enter three integers defining a supposedly valid time (under
the form h, m, s) and that displays the time that it will be a second later.
Example: Don’t you have the time? 15
What about the minutes? 59 And the seconds? 59
In a second, it will be 4:00 p.m. 0s.

[proper vault]

@ruby briarhello, we generally don't help with graded homework, though you are free to ask about specific parts of the solution.

[strange basin]

[print(''.join(_(O)for _ in map(lambda H,I:lambda O:H*(0==O/I%1),('Fizz','Buzz'),(3,5)))or O)for O in range(1,101)]

[near gust]

Python but you can only use list comprehensions

And all of the builtin types and functions except __import__, exec, and eval

• All function calls must happen inside a list comp
• lambdas (or iterables containing a lambda) cannot be assigned to a variable

• Walrus is specifically allowed

Examples
Fib (in a single list comp)

[(a := [print(0), 0, 1]) if not i else (a := [print(a[2]), a[-1], a[-2] + a[-1]]) for i in range(10)]

[grand tree]

weights = [ 
    -73.0,  88.0, -11.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  1.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0, 
      0.0,   0.0,   0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  0.0,  2.0, 
     77.2,   0.0,   0.0, 70.0,  0.0,  0.0, 69.0, 80.0,  0.0, 83.0, 
     13.0,  83.1,   0.0, 76.9,  0.0,  0.0,  0.0,  0.0, 79.9,  0.0, 
      1.0, -88.0,  80.0, 83.0, 77.0, 69.0,  2.0 
] 
  
pos, transition = 0, [ -1, 1 ] 
epsilon, delta = (.001, 0.2) 
feedback = [] 
  
while weights[pos]: 
    weight = abs(weights[pos]) 
    transition[0] = int(weight) + ( 
            2 ** 5 - 1 if weights[pos] >= 0 else -1) + ( 
            weight - int(weight) + delta > 1.0) 
    transition[1] = transition[0] if abs( 
    weight - int(weight) - delta) < epsilon else 0 
    feedback.extend(transition) 
    pos = int(weight) 
  
print(''.join(chr(val) for val in feedback), end='') 

!e

[night quarry]

Command Help

!eval [code]
Can also use: e

*Run Python code and get the results.

This command supports multiple lines of code, including code wrapped inside a formatted code
block. Code can be re-evaluated by editing the original message within 10 seconds and
clicking the reaction that subsequently appears.

We've done our best to make this sandboxed, but do let us know if you manage to find an
issue with it!*

[grand tree]

!e ```py
weights = [
-73.0, 88.0, -11.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 2.0,
77.2, 0.0, 0.0, 70.0, 0.0, 0.0, 69.0, 80.0, 0.0, 83.0,
13.0, 83.1, 0.0, 76.9, 0.0, 0.0, 0.0, 0.0, 79.9, 0.0,
1.0, -88.0, 80.0, 83.0, 77.0, 69.0, 2.0
]

pos, transition = 0, [ -1, 1 ]
epsilon, delta = (.001, 0.2)
feedback = []

while weights[pos]:
weight = abs(weights[pos])
transition[0] = int(weight) + (
2 ** 5 - 1 if weights[pos] >= 0 else -1) + (
weight - int(weight) + delta > 1.0)
transition[1] = transition[0] if abs(
weight - int(weight) - delta) < epsilon else 0
feedback.extend(transition)
pos = int(weight)

print(''.join(chr(val) for val in feedback), end='')

[night quarry]

@grand tree :white_check_mark: Your eval job has completed with return code 0.

001 | H�e�llo�,� �w�o�r�l�d�!�
002 | �

[near gust]

!e

[bool(b := [1]) + bool(print([i for i in b if
    i < 1000 and (not b.append(i + 1)) and not
    [j for j in range(2, i) if not i % j]])) for _ in [1]]

[night quarry]

@near gust :white_check_mark: Your eval job has completed with return code 0.

[1, 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131, 137, 139, 149, 151, 157, 163, 167, 173, 179, 181, 191, 193, 197, 199, 211, 223, 227, 229, 233, 239, 241, 251, 257, 263, 269, 271, 277, 281, 283, 293, 307, 311, 313, 317, 331, 337, 347, 349, 353, 359, 367, 373, 379, 383, 389, 397, 401, 409, 419, 421, 431, 433, 439, 443, 449, 457, 461, 463, 467, 479, 487, 491, 499, 503, 509, 521, 523, 541, 547, 557, 563, 569, 571, 577, 587, 593, 599, 601, 607, 613, 617, 619, 631, 641, 643, 647, 653, 659, 661, 673, 677, 683, 691, 701, 709, 719, 727, 733, 739, 743, 751, 757, 761, 769, 773, 787, 797, 809, 811, 821, 823, 827, 829, 839, 853, 857, 859, 863, 877, 881, 883, 887, 907, 911, 919, 929, 937, 941, 947, 953, 967, 971, 977, 983, 991, 997]

[near gust]

Pog

[open grail]

!e

[night quarry]

@maiden river :white_check_mark: Your eval job has completed with return code 0.

001 | 0 XNOR 0 = 1
002 | 
003 | 
004 | Half Adder : 1+1 = 10

:incoming_envelope: :ok_hand: applied mute to @maiden river until <t:1634212383:f> (9 minutes and 59 seconds) (reason: newlines rule: sent 106 newlines in 10s).

[last locust]

!unmute @maiden river

[night quarry]

:incoming_envelope: :ok_hand: pardoned infraction mute for @maiden river.

[last locust]

!paste

[night quarry]

Pasting large amounts of code

If your code is too long to fit in a codeblock in discord, you can paste your code here:
https://paste.pythondiscord.com/

After pasting your code, save it by clicking the floppy disk icon in the top right, or by typing ctrl + S. After doing that, the URL should change. Copy the URL and post it here so others can see it.

[last locust]

Please use our paste service for sending large code ^^

[maiden river]

oh ok 😅

[frank dome]

print(print(print(print(print(print(print("hello world")))))))

[royal sun]

!e
print(print(print(print(print(print(print("hello world")))))))

[night quarry]

@royal sun :white_check_mark: Your eval job has completed with return code 0.

001 | hello world
002 | None
003 | None
004 | None
005 | None
006 | None
007 | None

[night quarry]

@maiden river :white_check_mark: Your eval job has completed with return code 0.

<function true at 0x7fe390028c10>

[wintry island]

Is there someone who would help me with this? (
For the given natural numbers a, b and c. Decide whether a2 + b2 = c2) Iam new to python, dont be mean thank you👉👈

[fleet bridge]

!e

def f(a, b, c):
    return a**2 + b**2 == c**2
# or
f = lambda a, b, c: a**2 + b**2 == c**2

print(f(0,0,0))
print(f(3,4,5))
print(f(1,1,1))

[night quarry]

@fleet bridge :white_check_mark: Your eval job has completed with return code 0.

001 | True
002 | True
003 | False

[velvet lichen]

I was just messing with metaclasses and tried to make this. The issue is, despite it overriding the constructor and not throwing any errors, it does not print No "__init__"s!. I then tried exec-ing the string for the code in, but it did exactly the same thing. Does anyone know what is going wrong here?

def __init__(self):
    print("No \"__init__\"s!")

BLACKLIST = {"__init__": __init__}

class BlackListDict(dict):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        for i in BLACKLIST.keys():
            self.__setitem__(i, BLACKLIST[i])

    def __setitem__(self, k, v):
        if k in BLACKLIST:
            return
        super().__setitem__(k, v)

class Uninitializable(type):
    @classmethod
    def __prepare__(metacls, __name, __bases, **kwds):
        return BlackListDict(super().__prepare__(__name, __bases, **kwds))


class Test(metaclass=Uninitializable):
    def __init__(self, x) -> None:
        self.x = x

a = Test()

[astral rover]

its cause prepare shouldnt be a classmethod

its not ever called

[velvet lichen]

Oh, I thought it had to be?

What should it be instead?

[astral rover]

just a normal method

[velvet lichen]

Okay, I am now getting TypeError: Uninitializable.__prepare__() missing 1 required positional argument: 'bases' and I don't really understand what this error is trying to say. Do you think you could help?

I do think it is a classmethod, as everything worked when I did that except that it didn't print anything. It did overwrite the existing init like it was supposed to though, and I also looked it up and the first example was a classmethod

[astral rover]

i think the stubs for this might be wrong

[earnest wing]

self.__setitem__(i, BLACKLIST[i]) will not update the type object

but rather the namespace

[astral rover]

the namespace should be the 3rd arg to type.new right?

Test () {'__module__': '__main__', '__qualname__': 'Test'}

[velvet lichen]

This is what I found on it

[earnest wing]

actually, hold on

could it be the super() call in prepare

wouldn't just a return BlackListDict() work

[astral rover]

doesnt look like it

[velvet lichen]

I have just tried that now but it is still the same

[earnest wing]

oh, duh

your manual __setitem__ is triggering the filter

so __init__ is not getting set

[velvet lichen]

You are so right, thanks

It is good now

Presumably exec into a namespace also calls __setitem__? I had tried that, expecting that to catch any of the issues but it still had the same when I tried it

[earnest wing]

The namespace is populated as the class body is executed

[velvet lichen]

OK, thanks

[near gust]

I wonder how different it would be if new classes and new methods for classes couldn't be created it deleted at runtime

I bet there would be a lot more room for Python to be optimized

[earnest wing]

Well at that point you could do all mro work at compile time

[grave rover]

My disasm project is going perfectly fine

then boom
UNPACK_SEQUENCE

[rugged sparrow]

UNPACK_SEQUENCE trips it up?

[grave rover]

well uh

it pushes variables on the stack

and there will be multiple STORE_FASTs after it

which I need to somehow join together as x, y, z = TOS_before_UNPACK_SEQUENCE

[golden finch]

hey, I'm back temporarily

how can I get the size of the call stack in one expression?

without importing inspect

[rugged sparrow]

like how many frames there are?

[golden finch]

yeah

equivalent to ```py
len(import('inspect').stack())

[grave rover]

I mean op, arg = unpack("BB", op[:2]) is UNPACK_SEQUENCE(2) STORE_FAST(arg) STORE_FAST(op)

but right now I'm stuck on figuring out how to convert my CFG to a tree view without breaking everything

no I mean like

right after

BUILD_TUPLE UNPACK_SEQUENCE isn't the issue since my decompiler handles it just fine

it's the multiple STORE_FASTs I had trouble with

[golden finch]

@rugged sparrow you there?

[rugged sparrow]

you can just traverse the stack frames

and count

[golden finch]

I'm not sure how to do that

it has to be an expression

[rugged sparrow]

do you have sys?

[grave rover]

more or less
but it's annoying to work with when you only know a variable "x" is on the stack and UNPACK_SEQUENCE has oparg 3

[golden finch]

I have sys

[rugged sparrow]

one sec

(f:=sys._getframe())and len([(f:=f.f_back)for _ in iter(lambda:f, None)])

[golden finch]

How does that work?

[rugged sparrow]

first it stores the current frame into f

then it uses iter(func, flag) to call lambda:f until it returns None

and on each iteration, it sets f to f.f_back

[golden finch]

I see

and then the and len gets you it as an integer

[grave rover]

@rugged sparrow wanna help me out real quick? I've got the blocks from the CFG, but now I need to turn it into a tree view (or something) to know what the stack is for each block, without having to potentially process blocks twice somehow

but child nodes also need to know both branches that point to them I think

[rugged sparrow]

@grave rover I can help when I get home

[grave rover]

sure thing

it's 2 AM for me rn so I'll be heading off to bed then :)

[rugged sparrow]

Ping me tomorrow then we can work on it then

[grave rover]

👍

[sick hound]

I got bored and made this: https://gist.github.com/dzshn/6ec24800669595b7ce5ce680cd9f55b8

should probably fix the aspect ratio but also too lazy to lol

[hybrid dagger]

Cool, is that Mandelbrot?

[sick hound]

yep it is

[keen sierra]

The following code is run in IDLE:

while True:
  action = input(" > ")
  if not action.isalnum():
    print("Unsafe input.")
  elif action not in functions:
    print("Not found.")
  else:
    exec(action + "()")
    break

Is it possible to run malicious code on this program through the exec?

[sick hound]

I don't see a way it would be, but you're probably better off with:

functions = {
  "somefunction": somefunction
  # ...
}
while True:
  action = input(' > ')
  if action not in functions:
    print('Not found.')
  else:
    functions[action]()
    break

[keen sierra]

Yeah, I wouldn't use eval in any serious code, this was just to check whether eval can be made safe.
Hence posted in esoteric python

[sick hound]

if you only allow alphanumeric characters then I guess it does work as so

if you didn't check it against functions, I think the worst that could happen is a SyntaxError

[keen sierra]

Yeah, isalnum should make it fully safe from anyone trying to run malicious code

[thin trout]

well, exit is a first problem

[sick hound]

but it's being checked from a predefined list

[keen sierra]

^ And exit won't be there

[sick hound]

"fully safe" is maybe an overstatement
I can't think of anything though that doesn't depend on breaking the runtime itself, and those would probably apply to normal programs

[keen sierra]

Yeah I'm currently looking at "fully safe", this program itself is useless but you need to find a way to crash it or run malicious code, if it's opened on IDLE and you can only input stuff on that IDLE screen

[sick hound]

just dump 1 gb of text on it lol

[keen sierra]

You can only input stuff on that idle screen

You'd need to go to a different screen to copy the stuff lol

[sick hound]

hmmm

[keen sierra]

Also, ctrl+c isn't allowed either

[sick hound]

but is ctrl+d

that would raise an EOFError probably

[keen sierra]

Uh, any stuff involving ctrl+something instead of inputting stuff and pressing enter is not allowed

[thin trout]

https://git.pydis.com/snekbox

you should use services like that to use eval in a safe way

[sick hound]

we're trying to break it

[keen sierra]

Haven't needed to use eval so far

[keen sierra]

^

[thin trout]

to break snekbox?

[keen sierra]

Nah, this simple code

[thin trout]

right

[keen sierra]

Is snekbox breakable?

[thin trout]

maybe

two people broke it in the past

[keen sierra]

snekbox is actually useful though

[sick hound]

it's hard to prove otherwise, so

[keen sierra]

this code example ^^ is pretty much useless, but can't be broken even though it uses exec

[sick hound]

only way it could be safe is to make your own python implementation without access to system stuff

[keen sierra]

This'll reject anything other than letters and numbers. I can't think of any code that works without symbols to exploit eval

You need _ and .

brackets

[sick hound]

mm I think just because of the checking against the list it's already probably impossible to exploit it with these conditions

[earnest wing]

If it used deny- rather than allow-filters, there would at least be a chance :>

[near gust]

But if you try to do anything, it usually segfaults

Source: I've tried messing with that stuff before

The object at the id of an object + 8 is the __class__ of the object

You can turn an empty tuple into a list this way

[wooden badge]

can someone teach me the basics of esoteric python?

ie making python bend to your will

[wooden badge]

how would you maybe make that in java?

[grave rover]

@rugged sparrow lmk when you're available :)

code snippet at https://hastebin.com/emubafogec.py, it's got a lot of stuff commented out since I wanted to completely redo it and it's kinda messy

[rugged sparrow]

@grave rover what was the reason you wanted to build the tree again? Was it to check stack state?

[grave rover]

it was to pass the current stack to both branches

if you have e.g. LOAD_CONST(1) LOAD_FAST(some_bool) POP_JUMP_IF_FALSE() then both branches need to know the 1 is there

and if you just parse it sequentially it won't know it exists

[cloud fossil]

Do you actually code in Python byte code?

[grave rover]

occasionally

but right now I'm trying to make a decompiler

[cloud fossil]

How do you run Python byte code on Windows?

[rugged sparrow]

Cant you just copy the stack when you add a branch? (And eventually when they re-merge you can assert that the stacks are equal)

[bright hearth]

You can't not run python bytecode when running a python program

Python source just gets compiled into bytecode

So you just use the interpreter. You can bind bytecode to a function and call the function, exec it in place with exec(), whatever

exec() can take a code object, not just python source as a str

To construct a code object, you just call the code class's constructor. It takes the raw bytes of instructions as a bytes object, and some metadata.

[near gust]

When the code for a module gets compiled, the .pyc file contains a header and then a marshalled code object which contains the code for the module

[bright hearth]

compiling expensive, caching good

[near gust]

In that code object are the bytecode, constants (that can contain more code objects), flags, stacksize, etc.

And you need all 16 parameters of the constructor for the code object to be able to effectively code in bytecode

After working with some esoteric languages, programming in bytecode feels like a breeze

[bright hearth]

Yeah ( is pretty tricky to get working

[near gust]

I wonder how many ways there are to segfault the interpreter

There are a lot

[bright hearth]

Since bytecode can be executed directly, I imagine as many ways as there are to segfault at all

So in a way there's exactly one way to segfault the interpreter

Violate segmentation.

[grave rover]

that's the problem, figuring out where branches are
also the hecking loops and if-else expressions and that stuff

[rugged sparrow]

Branches should just be anywhere where the instruction pointer can be 2 different values after the opcode

[cloud fossil]

Hey @bright hearth can you write a Hello World program in byte code and send it here?

[grave rover]

well yeah but how should I go about "remembering" the stack across 7 different blocks between both branches

[earnest wing]

!e ```py
def f():
print("Hello, world!")

import dis
dis.dis(f)

[night quarry]

@earnest wing :white_check_mark: Your eval job has completed with return code 0.

001 |   2           0 LOAD_GLOBAL              0 (print)
002 |               2 LOAD_CONST               1 ('Hello, world!')
003 |               4 CALL_FUNCTION            1
004 |               6 POP_TOP
005 |               8 LOAD_CONST               0 (None)
006 |              10 RETURN_VALUE

[cloud fossil]

Alright but whither can I write this byte code and run it

[grave rover]

You can with libraries like https://github.com/martmists-gh/pyasm (ignore the shameless self-plug :)

[cloud fossil]

No problem. Friend helps friend.

[grave rover]

(note: the pip install does not work yet)

[rugged sparrow]

You can maybe store a bytes object on every block with each byte representing the result of a opcode (like a LOAD_CONST adds the LOAD_CONST opcode onto the bytes obj, and a CALL_FUNCTION would remove a certain number of items from the bytecode string then add on the CALL_FUNCTION opcode to represent that the result of a function is there

[grave rover]

wha?

[rugged sparrow]

like each block has a calculated bytes object thats something to represent the resulting stack (up until a branch)

[grave rover]

hmm...

[rugged sparrow]

or you could represent the stack as its own tree

(which wouldnt be efficient cause it should have the same structure as the block tree)

[grave rover]

n = 0

if ((2 * (offset + n)) < len(self.code)):

op = self.code[(2 * (offset + n)):]
op, arg = unpack(BB, op[:2])
n = (n + 1)



    current = (current + ( if (op >= 90) else stack_effect(op)))
    self.max = max(self.max, current)
    if (op in hasbranch):

    new_jumped = (jumped + [arg, (current + n)])
    if (arg not in jumped):

    self.check_offset(arg, current, new_jumped)

        if ((current + n) not in jumped):

        self.check_offset((current + n), current, new_jumped)

            return

        if (op in hasjabs):

        if (arg not in jumped):

        self.check_offset(arg, current, (jumped + [arg]))
        return

            if (op in hasjrel):

            if (((offset + n) + arg) not in jumped):

            self.check_offset(((offset + n) + arg), current, (jumped + [((offset + n) + arg)]))
            return

                if (op == opmap[RETURN_VALUE]):

                continue_or_break
                continue_or_break

    return

not a terrible start so far

original function: ```py
def check_offset(self, offset: int, current: int, jumped: list):
n = 0
while 2 * (offset + n) < len(self.code):
op = self.code[2 * (offset + n):]
op, arg = unpack("BB", op[:2])
n += 1
current += stack_effect(op, arg) if op >= 90 else stack_effect(op)
self.max = max(self.max, current)

        # TODO: Something still goes wrong here
        # if current < 0:
        #     raise ValueError("Negative stack index!")

        if op in hasbranch:
            new_jumped = jumped + [arg, current + n]

            # do both branches
            if arg not in jumped:
                self.check_offset(arg, current, new_jumped)
            if current + n not in jumped:
                self.check_offset(current + n, current, new_jumped)
            return

        elif op in hasjabs and arg not in jumped:
            # do jump
            self.check_offset(arg, current, jumped + [arg])
            return
        elif op in hasjrel and offset + n + arg not in jumped:
            # do jump
            self.check_offset(offset + n + arg, current, jumped + [offset + n + arg])
            return

        if op == opmap["RETURN_VALUE"]:
            break

I wonder what's making the indents so wacky

[rugged sparrow]

current = (current + ( if (op >= 90) else stack_effect(op))) what happens here?

[grave rover]

wait why is that first part missing

the heck

oh wait I see

right

uhhh

crap

the problem is that that if-else expression is modified in two different branches

attempt 2: py n = 0 if ((2 * (offset + n)) < len(self.code)): op = self.code[(2 * (offset + n)):] op, arg = unpack(BB, op[:2]) n = (n + 1) current = (current + (stack_effect(op, arg) if (op >= 90) else stack_effect(op))) self.max = max(self.max, current) if (op in hasbranch): new_jumped = (jumped + [arg, (current + n)]) if (arg not in jumped): self.check_offset(arg, current, new_jumped) if ((current + n) not in jumped): self.check_offset((current + n), current, new_jumped) return if (op in hasjabs): if (arg not in jumped): self.check_offset(arg, current, (jumped + [arg])) return if (op in hasjrel): if (((offset + n) + arg) not in jumped): self.check_offset(((offset + n) + arg), current, (jumped + [((offset + n) + arg)])) return if (op == opmap[RETURN_VALUE]): continue_or_break continue_or_break return

I think this is actually correct

only the continue_or_break is left to figure out

aand done

[sick hound]

hello

[grave rover]

hello

[grave rover]

oh no I have to build classes now

[grave rover]

yeah I made it a lot simpler now

what is that disassembler func

[grave rover]

listcomps are pain

[astral rover]

what about type is variably sized so that it doesn't allow slots?

[rugged sparrow]

@sick hound link to your decompiler?

[rugged sparrow]

Wdym doesn't allow slots

[astral rover]

like you cant subclass type with slots that arent empty, i get tuple, bytes and int not allowing it

[rugged sparrow]

I assume that python doesn't have the logic to calculate the space it needs to add the slots in front of the data for the internal type

[tacit cobalt]

!e ```py
type(lambda:0)((lambda:0).code.replace(co_code=b"t\x00d\x01\x83\x01\x01\x00d\x00S", co_consts=(None, "Hello, World!"), co_names=("print",)), {})()

[night quarry]

@tacit cobalt :white_check_mark: Your eval job has completed with return code 0.

Hello, World!

[tacit cobalt]

!e ```py
type(lambda:0)((lambda:0).code.replace(co_code=b"t\x00d\x01\x83\x01S", co_consts=(None, "Hello, World!"), co_names=("print",)), {})()

[night quarry]

@tacit cobalt :white_check_mark: Your eval job has completed with return code 0.

Hello, World!

[tribal moon]

I was looking around on github and I found this https://github.com/ssize-t/inlinec think you might find it interesting

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

Hello, World!

[bleak ermine]

If you have a class attribute assigned to a function like so:

class someObject:
  def __init__(self):
    self.func = self.someFunction

  def someFunction(self):
    pass

And you get the value of self.func and call it like so:

self.func()

Does it run the someFunction() or does it try to look for a method named func()?
Is there any reasoning behind this, like how does Python decide what it should do?

Oh, possible to have it call someFunction() by using the variable?

oh wait

can you do

(self.func)()

[sick hound]

You know wha I find esoteric ? Lighting up a lamp through watering plants by IoT using PYTHON !
https://youtu.be/XW17p62AQa4

[honest locust]

is this the right place to talk about python code?

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

True

[acoustic gust]

!e print(bool(1))

[night quarry]

@acoustic gust :white_check_mark: Your eval job has completed with return code 0.

True

[grave rover]

I'm in pain that you can say that so easily

What the heck

[rapid sparrow]

iot has gone mainstream.. it might have sold its esoteric card by now

"Alexa, write some esoteric Python"

[pure dew]

iot isnt really esoteric at all

[spring depot]

!e
type(lambda:0)((lambda:0).code.replace(co_code=b"d\x01F\x00d\x00S", co_consts=(None, type('',(),{'repr':lambda x:"Hello, World!"})())),{})()

[night quarry]

@spring depot :white_check_mark: Your eval job has completed with return code 0.

Hello, World!

[normal vortex]

hi

pff

[gusty hill]

is it possible to replace None with 1, for example when a function doesn't have a return type it returns 1 instead of None

[cloud fossil]

You can do that with a decorator

Or put return 1 as the last line in the function

[gusty hill]

I mean with an esoteric way, using ctypes or something

[rugged sparrow]

!e ```py
def replace_const(oval, nval):
def wrapper(func):
consts = [*func.code.co_consts]
consts[consts.index(oval)] = nval
func.code = func.code.replace(
co_consts=tuple(consts)
)
return func
return wrapper

@replace_const(None, 1)
def test():
print('yeet')

print(test())

[night quarry]

@rugged sparrow :white_check_mark: Your eval job has completed with return code 0.

001 | yeet
002 | 1

[gusty hill]

Oh wow thank you

[grave rover]

note: that replaces None everywhere in the function and may lead to strange results

[steady lily]

a script that decodes base64 without importing base64?

[tribal moon]

!e ```py
import binascii

print(binascii.a2b_base64(b"aGk="))```

[night quarry]

@tribal moon :white_check_mark: Your eval job has completed with return code 0.

b'hi'

[steady lily]

!e

import binascii
print(getattr(binascii, dir(binascii)[7])(b"SGVsbG8gV29ybGQh="))

[night quarry]

@steady lily :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 2, in <module>
003 | TypeError: 'ModuleSpec' object is not callable

[steady lily]

this worked on my ide

[astral rover]

!e

import binascii
print(getattr(binascii, dir(binascii)[8])(b"SGVsbG8gV29ybGQh="))

[night quarry]

@astral rover :white_check_mark: Your eval job has completed with return code 0.

b'Hello World!'

[steady lily]

im running 3.9.0

[astral rover]

im on 3.10 and that worked for me

and so is the python bot now apparently

[grave rover]

!e py b64decode = lambda t:(lambda l:''.join(chr(int(l[i:i+8],2))for i in range(0,len(l),8)))(''.join(format('ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/'.index(c),'06b')for c in t.strip("="))) print(b64decode("aGVsbG8gd29ybGQh"))

[night quarry]

@grave rover :white_check_mark: Your eval job has completed with return code 0.

hello world!

[grave rover]

@steady lily

[steady lily]

woah

wth this channel is on another level

[grave rover]

I could make it a lot more readable but you did ask in #esoteric-python :)

[steady lily]

i think esoteric python is the coolest thing and shows a mastery in the language

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

001 | yeet
002 | 1

[sick hound]

Someone should write a c compiler in pure python

What?

i feel like doing unspeakable things in python, where can i get started?

Start by learning dirty tricks

Like defining multiple variables in one line and using lambda

Or idk some weird math to make code go brrr

[sick hound]

!e ```py
a=b=c=0

[night quarry]

@sick hound :warning: Your eval job has completed with return code 0.

[No output]

[sick hound]

I don't do esoteric programming for just making code look weird, I just like less lines and chars

Unlike these guys who are crazy segfaulting python

This is new

Perfect

[shut trail]

I would start on messing with the call stack muhaha

acutally that's a bad place to start

i would start on learning bytecode / messing with inspect

[iron hedge]

ImportError: cannot import name 'handle_database_command' from partially initialized module 'src.utils.handle_database_command' (most likely due to a circular import) why am i getting this error (this sounds really generic, but i've no idea of what code should i send to help you to answer me)

[iron hedge]

oh okay

im sorry

[floral meteor]

>>> {*[2+(int)(Math.random()*6)+(int)(Math.random()*6)for fart in range(100000)]}
{2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}
>>> 

rolling two dice in Java, just gotta make sure in python tho

me: writing in Java
Java: sucks
me: rewrites Java in python in order to test Java code
Java: successfully debugged

[rapid sparrow]

wait till you discover the promise of ctypes

[tepid otter]

x = [1, 2, 3] # (this being an example list-- any works)
y = [x[-1], *([0]*(len(x)-1))]
exec('i = 1;' + 'y[i] = x[i-1]; i += 1;'*(len(x)-1))
print(y)

not terribly esoteric compared to what ive done before, but i was just asked to help someone rotate a given list to the right without any loops, list methods or what have you (obviously * is technically iteration, but its notfor or while, and indexing is technically __getitem__ but its not explicit) and i just thought the resulting code was ugly in a funny way

[smoky palm]

!ban ConnnorS.

[pliant current]

hi

i been messing around:
rukidding = 1 and True is not False in [bool, str, True]

yes = True if None is not True else False

silly i know

butwait = 1 in list(range(2)) if 'more' in 'lessismore'.split() else None or True

[earnest wing]

True in [True] in [True]

(False)

[pliant current]

i've seen that ^, very confusing - in any rate it shouldn't be code

great = 1 or 4 if len('idontneedthis'.split()) == 10 else lambda y: y + 1 if y + 2 < 4 else 5

[pliant current]

in is not is since is checks the id so since first fails it should/doesn't? check the 2nd

[earnest wing]

It desugars to True in [True] and [True] in [True]

[pliant current]

I see

[earnest wing]

Because in is a comparison op like >=

[pliant current]

are = lambda x:{take:= x in range(x * 2) and 42 }.union(range(2))

[snow beacon]

y = [x[-1], *x[:-1]]
```This is a little shorter, but less ugly.

[grave rover]

https://pypi.org/project/pyasm/ it's here

look at the C code mostly

If you want to learn bytecode, you'll eventually memorize ceval.c (https://github.com/python/cpython/blob/main/Python/ceval.c)

where next_instr is a _Py_CODEUNIT*

After everything is turned to bytecode it calculates the offsets and puts it in the bytecode

fun fact, if your bytecode jumps more than 8.6GB worth of bytecode the compiler won't be able to handle it because EXTENDED_ARG can only be used 3 times, making the max jump 2^32 instructions or 2^33 bytes

[near gust]

!e

print (([]==[])<<-~-~([]==[]))

[night quarry]

@near gust :white_check_mark: Your eval job has completed with return code 0.

8

[wooden badge]

this is awesome! is there more?

!e print([ox_for x in bananas])

[night quarry]

@wooden badge :x: Your eval job has completed with return code 1.

001 |   File "<string>", line 1
002 |     print([ox_for x in bananas])
003 |            ^^^^^^^^^^^^^^^^^^^
004 | SyntaxError: invalid syntax. Perhaps you forgot a comma?

[wooden badge]

hmm

!e import sys print(sys.version)

[night quarry]

@wooden badge :white_check_mark: Your eval job has completed with return code 0.

3.10.0 (default, Oct  6 2021, 00:08:37) [GCC 8.3.0]

[wooden badge]

that makes sense

[worldly musk]

Is it possible to make this work? ```py

from ctypes import py_object
py_object.from_address(id(True)+8).value = False
True
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
PS C:\Users\me>```

[worldly musk]

It just exits the shell

[rugged sparrow]

!e print([0x_for x in bananas])

[night quarry]

@rugged sparrow :white_check_mark: Your eval job has completed with return code 0.

[15]

[wooden badge]

ohh right, i used ox instead of 0x

!e print([0x_for x in (()==()+()==())*print(e**3.12)*()==[]==()])

[night quarry]

@wooden badge :white_check_mark: Your eval job has completed with return code 0.

[15]

[wooden badge]

!e print(()==()+()==()*()==()+(()==()+()==())+(()==()+()==()))

!e print(()==()+()==()*()==()+(()==()+()==())+(()==()+()==()))

[night quarry]

@wooden badge :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 1, in <module>
003 | TypeError: can't multiply sequence by non-int of type 'tuple'

[wooden badge]

!e print((()==())+(()==())+(()==()))

[night quarry]

@wooden badge :white_check_mark: Your eval job has completed with return code 0.

3

[forest plaza]

not sure if this count but I find it funny:

!e

class A():
  def __init__(self,a):
    self.a=a
    
    class B():
      c=a
      def __init__(self,a):
        self.a = a
  
    self.b=B(self.a)
    self.c=B(self.a)
    
x=A('a')
y=A('b')
x.b.__class__.c = 'aa'
print(x.c.c)
print(x.c.__class__)
print(y.c.c)
print(y.c.__class__)

[night quarry]

@forest plaza :white_check_mark: Your eval job has completed with return code 0.

001 | aa
002 | <class '__main__.A.__init__.<locals>.B'>
003 | b
004 | <class '__main__.A.__init__.<locals>.B'>

[forest plaza]

its different classes but if I print __class__ I get the same output

I guess the <locals> hides some of the details?

[snow beacon]

Not yet, I've been a tad preoccupied.

[wooden badge]

I appreciate what youve done already tho! As far as i can see, no other guide is that good

[snow beacon]

I hope I inspire people to make better guides.

[near gust]

!e

list(range(100000000000))

[night quarry]

@near gust :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 1, in <module>
003 | MemoryError

[near gust]

pog

[grave rover]

I guess I'll give it a shot

[potent comet]

It doesn't no, In Python 2 it was. To help users in switching there's a special case message to suggest the fix.

[rapid sparrow]

if print can be either a function or not in python 2, does that mean it's a keyword?

[potent comet]

It's a keyword normally, but the function always exists (you can see it in the builtins module), using globals() etc.

The __future__ flag changes behaviour inside the parser and compiler so it is or is not a keyword.

[sick hound]

You guys have a bot that runs code right?

I wonder if it's possible to escape the this with some weird tricks

Get it to @ someone or trigger a command as admin lmao

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

Code block escape attempt detected; will not output result

Full output: https://paste.pythondiscord.com/irelexewer.txt?noredirect

[sick hound]

Damn that sounded fun tho

Does it read for specific cases?

Like if I putbit in hex or a weird thing

Does it just have that check at the end to see if it contains that character?

[rugged sparrow]

it just checks at the end

[sick hound]

Ahh ok

[vast cargo]

any reference how to calculate EXTENDED_ARG?

also is this the right channel to ask this question?

oparg is byte?

or an 32 bit integer?

[last locust]

class MyInt(int):
    """Custom `int` representiation with a `__iter__` method."""
    ...
```is there a way I can make python default to using `MyInt` instead of `int` for integers? I currently have to do like `for x in MyInt(3)` etc. but I'd like to be able to do just `for x in 3:`

[rapid sparrow]

cool to know, thanks

[rapid sparrow]

are you willing to do horrible acts

[last locust]

Anything to make it work

I assume there's some ctypes hackery

[rapid sparrow]

yes

actually it could be trickier than I thought. are you wanting all ints to be MyInt or just new ones created after your program starts

[last locust]

all ints

[rapid sparrow]

that's beyond by wheelhouse, then 😅

[thin trout]

use forbiddenfruit to patch int.__iter__ instead

don't ask me how though

[rapid sparrow]

for x in 3? are ints even iterable in unmodified python?

[thin trout]

!e iter(42)

[night quarry]

@thin trout :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 1, in <module>
003 | TypeError: 'int' object is not iterable

[twin dock]

Can't do that, int is immutable

[thin trout]

nothing is immutable for forbiddenfruit

[twin dock]

!e setattr(int, "iter", None)

[night quarry]

@twin dock :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 1, in <module>
003 | TypeError: cannot set '__iter__' attribute of immutable type 'int'

[rapid sparrow]

lol 🤣

[thin trout]

!e

from forbiddenfruit import curse


def hello(self):
    return "blah"


curse(str, "hello", classmethod(hello))

assert str.hello() == "blah"

[night quarry]

@thin trout :warning: Your eval job has completed with return code 0.

[No output]

[thin trout]

see?

[rapid sparrow]

how does it defy forbiddenfruit..

[twin dock]

oh nice

[thin trout]

!e

from forbiddenfruit import curse

def hello(self):
    for i in range(4): yield i


curse(int, "__iter__", classmethod(hello))

for i in 36:
    print(i)

[night quarry]

@thin trout :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 7, in <module>
003 |   File "/snekbox/user_base/lib/python3.10/site-packages/forbiddenfruit/__init__.py", line 425, in curse
004 |     _curse_special(klass, attr, value)
005 |   File "/snekbox/user_base/lib/python3.10/site-packages/forbiddenfruit/__init__.py", line 319, in _curse_special
006 |     assert callable(func)
007 | AssertionError

[thin trout]

oh, that's interesting

[twin dock]

seems pretty immutable to me

[thin trout]

oh, I am stupid that's it

!e ```py
from forbiddenfruit import curse

def iter(self):
return iter(range(self))

curse(int, "iter", iter)

for i in 5: print(i)```

[night quarry]

@thin trout :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 6, in <module>
003 |   File "/snekbox/user_base/lib/python3.10/site-packages/forbiddenfruit/__init__.py", line 425, in curse
004 |     _curse_special(klass, attr, value)
005 |   File "/snekbox/user_base/lib/python3.10/site-packages/forbiddenfruit/__init__.py", line 332, in _curse_special
006 |     tp_as_name, impl_method = override_dict[attr]
007 | KeyError: '__iter__'

[thin trout]

interesting

[twin dock]

Not even close

[thin trout]

https://github.com/clarete/forbiddenfruit/issues/51

[rapid sparrow]

o.O

[twin dock]

Meh

[thin trout]

https://github.com/clarete/forbiddenfruit/pull/63 will add support for it

[last locust]

So basically I can't do it currently?

[thin trout]

yup

[near gust]

The fact that forbidden fruit can even exist shows why Python needs to look up attributes every time they are accessed

[thin trout]

yup

[near gust]

If python did not need to do so much lookup, it would be a lot faster, but that's the price you have to pay to be able to create new classes and attributes at runtime

I believe I've only used the creation of new classes at runtime for non-esoteric reasons only a handful of times, and all of them were named tuples

And for adding methods at runtime, I only used once in a discord bot because I had many (like 75) slash commands that took the same form. So I stored the command info in named tuples and added a method to the cog for each instance of the special named tuple class. And it somehow worked

[thin trout]

I used it in a Django project to generate generic routes

[rugged sparrow]

@last locust fishhook can hook all dunders

[rugged sparrow]

Fishhook can hook __iter__

[thin trout]

oh nice

didn't know about that one

[rugged sparrow]

I got frustrated with forbiddenfruits limitations so I wrote my own lib, fishhook is the result

[last locust]

So how would that work?

[thin trout]

lmfao

[rugged sparrow]

!pypi fishhook

[night quarry]

fishhook v0.1.1

Allows for runtime hooking of static class functions

[last locust]

So ig```py
@hook(int)
def iter(self):
...

for x in 5:
print(x)```?

[rugged sparrow]

Yes

[last locust]

from fishhook import hook

@hook(int)
def __iter__(self):
    yield from range(self+1)


for x in 5:
    print(x)
```perfect, thanks @rugged sparrow 👍

[rugged sparrow]

No prob

[restive void]

Finally, loose typing in Python! :D

@hook(str)
def __add__(self, other):
    if isinstance(other, str):
        return orig(other)
    return self + str(other)

And the inverse of str.__mul__ is possible, too, amazing <3

[rugged sparrow]

type(obj).__sizeof__(obj) works on all objects. obj.__sizeof__() fails if obj is a class

wdym?

what other dunders would there be?

which tp_* slots can fishhook not hook?

__mro__ isnt a function tho?

>>> from fishhook import *
>>> @hook(int)
... def __dir__(self):
...     print('hooked')
...     return orig(self)
... 
>>> v = dir(1)
hooked
>>> ```

fishhook can still hook it

generate_slotmap() is only needed for dunders that exist as tp_* function pointers. __dir__ (and others) are just a dictionary lookup

yea thats what .__subclasses__() uses to get the currently active subclasses

https://docs.python.org/3/library/enum.html#enum.IntFlag

[rugged sparrow]

it doesnt need to, they are implemented by the same tp_* slot

yea

https://github.com/python/cpython/blob/8e5de40f90476249e9a2e5ef135143b5c6a0b512/Objects/abstract.c#L1068-L1085

yea

[steady lily]

I found this online and it gives the square root of 17. How do all those symbols give you a number?

__=((()==[])+(()==[]));___=(__**__);____=((___<<___));_____=((____<<(__**__)));______=((_____<<(__**__)));
_________=((___<<_____));__________=((((___<<_____))<<(__**__)));_=((__**__)+(______<<(__**__)));_______=(__**__)
for ________ in range((_____+(_________<<(__**__))+(__________<<(__**__)))):
    _______=_______-((_______**((___<<___))-_)/(((___<<___))*_______))
print(_______)

Is there an online guide of some sort or a video?

[last locust]

Well at the begging there'spy __=((()==[])+(()==[]))this evaluates topy __ = (False + False)which ispy __ = 0 + 0or justpy __ = 0

Then there's a load of stuff which is basically just that repeated but with different symbols (e.g. ** which is exponent)

That gives a bit of an insight @steady lily

[scarlet osprey]

hey everyone. a friend of mine has been sent a token stealer and i've been looking into it and it's a pyarmor encrypted pyinstaller packed exe. i've retrieved the code from the pyc and the pyarmor decrypter dll. running this, it complains about missing deps so i created dummy files for it to import and i've put my own code into it so i now believe i have my own code running after the decryption.
what i need now is to dump this running code so i can take a look at the decrypted original source however i'm not to great with the dis, inspect modules and all that. anyone got any ideas?

[nocturne patrol]

I made this tool https://github.com/kootenpv/cliche
I think it's funny that it heavily uses dynamic features of python and it's really "ugly" but the idea is that at least to the user it is maximally convenient 🙂

[snow beacon]

It might help to look into sys.settrace and try logging the inputs to a file.

[grave rover]

working on a simple website to host my esopython guides n stuff on

wondering if I should add running code snippets in a snekbox instance

[rugged sparrow]

send me what you end up writing im interested

[pure dew]

if this is an amazing and dead-simple as it looks, I'll be your biggest fan

[nocturne patrol]

Let me know 👌

[rugged sparrow]

You could collapse the attr_cache and method_cache into one dict

Also those additions kinda defeat the point of fishhook (its dynamic so it works automatically on most versions without hard coded offsets)

I did have an idea for how to do those hooks dynamically tho,

I'm working on an implementation of that rn

[grave rover]

time to write esopython docs

[rugged sparrow]

your edits are quite clever tho

[grave rover]

what part of esopython should I cover first

[rugged sparrow]

c_void_p.from_address(id(a) + 176).value

[rugged sparrow]

Nice

[steady lily]

Here's my first attempt at obfuscation

import binascii, __main__

O0OO0OOOo0000oO = b'WXMyazlwZG9NbUJHT0cw'

o0ooo0OOoOOoO0O = b'd2hpbGUgVHJ1ZToKICAg'

o00o00oOoOOoO0O0 = b'IHByaW50KCJ5b3VyZSBh'

o00O0oo0OoOo00Oo = b'IGR1bWJhc3MiKQoKI2RK'

getattr(getattr(__main__, [x for x in dir(__main__) if x.startswith('__b')][0]), (lambda: "ArithmeticError" and "AssertionError" and "AttributeError" and "BaseException" and "BlockingIOError" and "BrokenPipeError" and "BufferError" and "BytesWarning" and "exec")())(''.join([getattr(binascii, [x for x in dir(binascii) if x.startswith(chr(97)+str((()==())+([]==[]))+chr(98))][0])(globals().get(var_name)).decode() for var_name in ['o0ooo0OOoOOoO0O', 'o00o00oOoOOoO0O0', 'o00O0oo0OoOo00Oo']]))