esoteric-python

[versed eagle]

!e 3.11 ```py
from future import barry_as_FLUFL
print(1 <> 2) # same as 1 != 2

[night quarry]

@versed eagle :x: Your 3.12 eval job has completed with return code 1.

001 |   File "/home/main.py", line 2
002 |     print(1 <> 2) # same as 1 != 2
003 |             ^^
004 | SyntaxError: invalid syntax

[rugged sparrow]

it works with my local 3.12

[versed eagle]

oh well

check out https://peps.python.org/pep-0401/

[elfin ether]

Any other esoteric python things that modifies allowed syntax?

[versed eagle]

!e most of the other future statements dont do anything in modern python, because they're already implemented, but there is this

from __future__ import braces

[night quarry]

@versed eagle :x: Your 3.12 eval job has completed with return code 1.

001 |   File "/home/main.py", line 1
002 |     from __future__ import braces
003 |     ^
004 | SyntaxError: not a chance

[versed eagle]

its quite silly

[elfin ether]

ok, now how do i revert <>...

[versed eagle]

dont import barry as flufl

[elfin ether]

too late for that

[elfin ether]

any suggestions for how to achieve something like this:

class C:
    registry = {}
    ...
asdf = C()
assert C.registry['asdf'] is asdf

[elfin ether]

not very efficient, but this is working alright.

class C:
    _registry = {}
    @classmethod
    @property
    def registry(cls):
        for k, v in globals().items():
            if isinstance(v, C):
                C._registry[k] = v
        return C._registry
    def __repr__(self):
        for k, v in C.registry.items():
            if v is self:
                return f'C(name: {k})'

asdf = C()
print(asdf) # prints "C(name: asdf)"

[rugged sparrow]

you could add a flag so registry only goes searching if there are any new instances of C

[elfin ether]

good idea

[rugged sparrow]

@elfin ether take a look at this for finding containers of objects after they are created: https://paste.pythondiscord.com/2NEA

[fleet bridge]

random thought: sum(n) == n*(n-1)//2 if n is a nonnegative integer

[quartz wave]

i was confused for the first 10 seconds until i remembered what sum() takes

[fleet bridge]

hmm, what is the sum of increasing sequence of integers from 0 to -10?
it feels like the same formula should work, if we count each number -1 times

[rugged sparrow]

you got me beat then, sum(1) is invalid right?

[versed eagle]

do you mean sum of integers in [0, n)?

if yes, it can be generalised for negative numbers

[versed eagle]

!e ```py
def rangesum(n):
s = n and n // abs(n) # sign of n
return s * (n*(n-s) // 2)

for i in range(-100, 101):
# to avoid division by zero. rangesum returns 0 when given 0 as input though, so it still works
if i:
i_sign = i // abs(i) # make sure the range increments in the right direction
assert sum(range(0, i, i_sign)) == rangesum(i)

[night quarry]

@versed eagle :warning: Your 3.12 eval job has completed with return code 0.

[No output]

[versed eagle]

the trick is to account for the sign of n

[versed eagle]

i assume sum(n) means the sum of range(n), because of the following equation

[rugged sparrow]

ah yea fair

[proper vault]

You can use math.sign instead of the abs trick

[magic wraith]

are you sure? i've read a while ago somewhere on stack overflow that python developers could not agree on all the edge cases of such sign function and so that it was never implemented

[proper vault]

Oh yeah, you have just copy sign, my bad

[sick hound]

I am a software engineer and I use chat gpt4 to assist me as a tool

Nothing wrong with that

[versed eagle]

sure, but there is something wrong with copy and pasting its output here-- the wrong thing being that it violates the rules of this server

if you don't like the rules, talk to the moderators i guess ¯_(ツ)_/¯

[versed eagle]

til about math.copysign, thanks

(though imo it should instead be an attribute or method on numeric classes, i.e., int.sign, float.sign, and complex.sign)

[sick reef]

!e

print(((B:=lambda x,y:x+y),(C:=lambda x,y:x-y),(A:=B(x=50,y=50)),(__:=C(x=A-15,y=A-25)))[3])

[night quarry]

@sick reef :white_check_mark: Your 3.12 eval job has completed with return code 0.

10

[grave grail]

((50+50)-15)-((50+50)-25)

[unique prairie]

hey, i have a little script where i overload the __or__ method on builtin functions to be able to do function chaining like in bash (i'm using forbiddenfruit which is a library that allows to overload operators from builtin functions using ctypes)
Here's the code :

from forbiddenfruit import curse
from collections.abc import Iterable
from inspect import signature
from types import *

_func_types = [FunctionType, LambdaType, MethodType, BuiltinFunctionType, BuiltinMethodType, MethodDescriptorType, MethodWrapperType]

def _pipe(self, func):
    def wrapper(*args, **kwargs):
        """Wrapper for the function combination"""
        res = self(*args, **kwargs) if callable(self) else self
        try:
            if func == print or len(signature(func).parameters) > 1:
                if isinstance(res, dict):
                    return func(**res)
                if isinstance(res, Iterable):
                    return func(*res)
        except:
            pass
        return func(res)

    if func == print:
        wrapper()
    else:
        return wrapper

for t in _func_types:
    curse(t, "__or__", _pipe)

and here's how it works :

def double(x):
    return 2 * x

2 | double | double | double | print # 16

map(str.lower | str.capitalize, ["mEsSy", "tExT"]) | print # Messy Text

what are your thoughts about this and how would you make it better ?

[lunar juniper]

!e ```py
print(eval("print(print(print(print(print(print(print(print(print(print(print(print(print("Hello World"))))))))))))"));

[night quarry]

@lunar juniper :x: Your 3.12 eval job has completed with return code 1.

001 |   File "/home/main.py", line 1
002 |     print(eval("print(print(print(print(print(print(print(print(print(print(print(print(print("Hello World"))))))))))))"));
003 |                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
004 | SyntaxError: invalid syntax. Perhaps you forgot a comma?

[flint hollow]

hey eso types: i have a challenge for you: golf this nedbat code

bomb = """
(lambda fc=(
    lambda n: [
        c for c in
            ().__class__.__bases__[0].__subclasses__()
            if c.__name__ == n
        ][0]
    ):
    fc("function")(
        fc("code")(
            0,0,0,0,0,0,b"BOOM",(),(),(),"","","",0,b"",b"",(),()
        ),{}
    )()
)()
"""
eval(bomb, {"__builtins__": {}})

[rugged bane]

that code looks... disgusting 😶

[wraith island]

i am not good at golfing, but ().__class__.mro()[1].__subclasses__() has less characters than using __bases__ :)

[astral rover]

cant you just __base__.__subclasses__()?

[rugged sparrow]

is the challenge just what it does?

[flint hollow]

no, to make it smaller

its feature is nonetheless special:

[rugged sparrow]

so smallest segfault

[flint hollow]

!e

bomb = """
(lambda fc=(
    lambda n: [
        c for c in
            ().__class__.__bases__[0].__subclasses__()
            if c.__name__ == n
        ][0]
    ):
    fc("function")(
        fc("code")(
            0,0,0,0,0,0,b"BOOM",(),(),(),"","","",0,b"",b"",(),()
        ),{}
    )()
)()
"""
eval(bomb, {"__builtins__": {}})

[night quarry]

@flint hollow :warning: Your 3.12 eval job has completed with return code 139 (SIGSEGV).

[No output]

[rugged sparrow]

with no imports?

[flint hollow]

smallest bomb that can survive eval(bomb, {"__builtins__": {}})

[rugged sparrow]

wdym survive?

like segfault?

[flint hollow]

yes

[rugged sparrow]

!e py bomb = "(m:=lambda:m).__class__(m.__code__.replace(co_code=b''),{})()" eval(bomb, {"__builtins__": {}})

[night quarry]

@rugged sparrow :warning: Your 3.12 eval job has completed with return code 139 (SIGSEGV).

[No output]

[rugged sparrow]

@flint hollow ^ thats pretty small

[quartz wave]

why does it not crash the REPL

unless put in an exec(r"""...""")

[rugged sparrow]

It should crash even without the eval, the guy challenged to do it without buildins, so the eval was used to remove them

[quartz wave]

weird
it behaves differently when i tried to reproduce the noncrash

now it does crash

oh

wait a minute

[rugged sparrow]

It should either segfault or hit a breakpoint/trap

Depending on the flags used to build python

[quartz wave]

it's so weird

>py
Python 3.13.0a3 (tags/v3.13.0a3:f009305, Jan 17 2024, 13:25:20) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> bomb = "(m:=lambda:m).__class__(m.__code__.replace(co_code=b''),{})()"
>>> eval(bomb, {"__builtins__": {}})
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
    eval(bomb, {"__builtins__": {}})
    ~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<string>", line 1, in <module>
  File "<string>", line 1, in <lambda>
SystemError: no locals found when storing 'm'
>>> (m:=lambda:m).__class__(m.__code__.replace(co_code=b''),{})()

<crash>

[rugged sparrow]

Huh, looks like 3.13 eval doesn't init locals properly?

[quartz wave]

i'd say it's more proper now since it doesn't crash then

but i don't have enough of an idea of python internals to know if it is indeed proper

[quartz wave]

i think i might've done it okay

>>> sum(1)
0
>>> sum(2)
1
>>> sum(5)
10

first n integers from start (defaults to 0)

>>> sum(5, start=2)
20

[rugged sparrow]

no that's a system error, I would bet there is some optimization where it doesn't add locals if it doesn't think it needs to and that code somehow bugs it out. (Or it somehow got lucky and constructed the opcodes needed for that from the empty byte string). Can you test (m:=lambda:m) by itself?

[quartz wave]

works fine

[rugged sparrow]

Hmm try the original code, without the call

[quartz wave]

works fine

[rugged sparrow]

Weirder

[quartz wave]

produces a lambda

[rugged sparrow]

And with the call it gets the locals not found?

[quartz wave]

yes

but only if the call is inside the eval()

otherwise it crashes

[rugged sparrow]

Ooo even weirder lol

Try disassembling the resulting lambda? It might raise a weird exception

[quartz wave]

nope

[quartz wave]

no specific eval() call btw ```pycon

bomb = "(m:=lambda:m).class(m.code.replace(co_code=b''),{})"
t = eval(bomb, {"builtins": {}})
eval("t()")
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
eval("t()")

~~~~^^^^^^^

File "<string>", line 1, in <module>
File "<string>", line 1, in <lambda>
SystemError: no locals found when storing 'm'

it works on any as long as t is defined as that lambda

[rugged sparrow]

Huh

Interesting

But if you call t outside eval, segfault?

[quartz wave]

yes

[rugged sparrow]

Interesting

[quartz wave]

wait hold on this is weird

[rugged sparrow]

I'll dig into this more later today, will build 3.13 locally

[quartz wave]

actually it does produce the same error when t() is called outside

maybe i just did the code on 3.12

[rugged sparrow]

That makes it a bit less weird

[quartz wave]

i had a lot of python shells open

[rugged sparrow]

Still interesting

[quartz wave]

and it gets confusing sometimes

[rugged sparrow]

Yea fair

I'll dig in at the C source level later today and figure out a root cause

[lament pond]

Thing I made, most overly complex way to turn one string into another string using 1 line:
for x,l in enumerate("Franz"): (l := ord(l)) and print(f"{(chr(l-3))*(x==0)}{(chr(l+3))*(x==1)}{(chr(l+19))*(x==2)}{(chr(l-9))*(x==3)}{(chr(abs(l-89)))*(x==4)}",end="")

[grave grail]

Technically anything can be a one linear

If you allow exec/eval

[fleet bridge]

anything can be a one-liner even without exec/eval

[grave grail]

Really? Ok

[quartz wave]

theoretically

[versed eagle]

anything that doesn't do silly things like reading its source file

[quartz wave]

why not?

[versed eagle]

because the source would be different

so it could behave differently from the equivalent multi-liner code

[quartz wave]

true

but the source inside can always be adapted

[fleet bridge]

it can write to source file, and then change it back at the end of the process

[versed eagle]

assuming you have write permissions on the file, yeah

[fleet bridge]

well, you cant change anything without a way to observe it
so we should make a distinction that makes sense

looking into source code is not a normal thing for a program to do, so breaking it is fine, imo

even if you change def into lambda it could be observed (you might patch the code object though)

normal functions and equivalent lambdas are written in pretty different ways, so bytecode almost always will be different (and if you patch it - then it pretty much the same thing as doing exec/eval)

also, one-expression code usually allocates more objects than normal code, and you can observe it (it changes state of the allocator, so objects might be allocated in different places)

doing try-except or with in expression-only code is impossible
either code generation is required, or some external module (iirc, contextlib provides a way to solve this problem)

and this might change traceback

[versed eagle]

neither code generation nor external modules are required

[fleet bridge]

match-case is possible, but it requires a lot of verbose code

[fleet bridge]

example?

try:
    1/0
except:
    pass
``` please convert this code into expression

[versed eagle]

by chilaxan,

[fleet bridge]

read his code, bruh

[versed eagle]

it uses neither code generation nor external modules

[fleet bridge]

it uses contextlib which has non-expression code in it

[versed eagle]

contextlib is not an external module

its part of the stdlib

[fleet bridge]

this is basically the same as writing ```py
def try_(...):
try:
...
except:
...

[fleet bridge]

it is, relative to our file
im not talking about 3rd party modules

[versed eagle]

so does the builtins module, and we still use builtin functions

[fleet bridge]

that is boring

just write normal code then

[versed eagle]

the builtins module is boring?

i dont understand what you're saying

[fleet bridge]

writing normal code when you are supposed to write esoteric (one-line in this case) code is boring
dont forget that we are in #esoteric-python

[versed eagle]

what does this have to do with writing normal code?

[fleet bridge]

.

[versed eagle]

how so

[fleet bridge]

🤦‍♂️

bye

[versed eagle]

if that's equivalent, how is defining any function locally different than just importing it from another file?

[versed eagle]

oh

bye, have a nice day

guess ill talk to you later then

[rugged sparrow]

I think I have some code that abuses generators and the fact that they can sort of catch exceptions ish, to simulate exceptions, would need to dig it up

But it couldn't be fully generalized to catch all exceptions 🤷‍♂️

[lament pond]

hmm, would disabling my copy and paste keys while programming help me to actually make functions for repeated tasks instead of copying and pasting?

[grave grail]

?

[bright valve]

huh?

[rugged sparrow]

import pickle
import datetime
import decimal
import uuid
import collections
import re
import io
import sys

SAFE_TO_IMPORT = {
    'builtins.range',
    'builtins.complex',
    'builtins.set',
    'builtins.frozenset',
    'builtins.slice',
    'builtins.str',
    'builtins.bytes',
    'builtins.list',
    'builtins.tuple',
    'builtins.int',
    'builtins.float',
    'builtins.dict',
    'builtins.bool',
    'builtins.bin',
    'builtins.None',
    'datetime.datetime',
    'datetime.time',
    'datetime.timedelta',
    'decimal.Decimal',
    'uuid.UUID',
    'collections.namedtuple',
    'collections.OrderedDict',
    're.Pattern',
}

class RestrictedUnpickler(pickle.Unpickler):
    def find_class(self, module, name):
        module_dot_class = '{}.{}'.format(module, name)
        if module_dot_class in SAFE_TO_IMPORT:
            try:
                module_obj = sys.modules[module]
            except KeyError:
                raise Exception("Module Not Found") from None
            return getattr(module_obj, name)
        raise Exception("Forbidden Module")

    def persistent_load(self, persistent_id):
        if persistent_id == "<<NoneType>>":
            return type(None)

def challenge(input: str):
        return RestrictedUnpickler(io.BytesIO(input.encode('utf-8'))).load()
``` new challenge for y'all, get full code exec from `challenge(<input>)`

i am not 100% sure its possible, but if anyone can solve it its yall

@lunar marsh you might want to give it a shot ^

[lunar marsh]

those are some very specific imports 🤔

[rugged sparrow]

(its from a real world package)

[rugged sparrow]

if you can find a solve, we split the bounty lol

[lunar marsh]

haha alright bet

[rugged sparrow]

i have all of the path to hit that code path

[rugged sparrow]

and ik how to get code exec if we can get a reference to <type 'code'>

[left hinge]

imagine not stacking all of your imports on one line

[torn cypress]

!e

f=lambda c:c==1and'""'or f'f"{f(c-1)}"'

eval(f(152))

[night quarry]

@torn cypress :x: Your 3.12 eval job has completed with return code 1.

001 | /home/main.py:1: SyntaxWarning: invalid decimal literal
002 |   f=lambda c:c==1and'""'or f'f"{f(c-1)}"'
003 | Traceback (most recent call last):
004 |   File "/home/main.py", line 3, in <module>
005 |     eval(f(152))
006 |   File "<string>", line 1
007 |     f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f"f""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
008 |                                                                                                                                                                                                                                                                                 
... (truncated - too long)

Full output: https://paste.pythondiscord.com/KSXSAMKSBIJVRSNVROL2JBOXCY

[left hinge]

perish

[quartz wave]

#1219231910790823977 message
multiple starred tuples that don't error ```pycon

*(a, b), *(c, d) = 1, 2, 3, 4
a
1
b
2
c
3
d
4

[wooden minnow]

hello, there is someone who already used opencv library? I have some questions about my code

[primal crater]

this was not a productive use of my time.

import numpy as N;import matplotlib.pyplot as H;𑀇=int('001100',2);Δ=lambda 𑀇:len(𑀇);𝛬=lambda 𝛌:int(𝛌);p=lambda i:i[::1];q=lambda i:i[::-1];Ω=lambda x:N.sin(x);Θ=N.pi;Π=N.sqrt(𑀇);F=H.figure(figsize=(𑀇-Δ([𑀇,𑀇,𑀇])/𑀇/𑀇,𑀇/𑀇+𑀇/𑀇+𑀇/𑀇),facecolor=chr(𑀇*𑀇-(𑀇+𑀇+𑀇+(𑀇//𑀇))),dpi=𑀇*𑀇*Δ([𑀇,𑀇,𑀇])*Δ([𑀇,𑀇,𑀇])+𑀇/𑀇,edgecolor=(𝛬(𑀇/𑀇),𝛬(𑀇/𑀇),𝛬(𑀇/𑀇)),tight_layout=True);Z=N.linspace(-𑀇,𑀇*𑀇,𑀇**Δ([𑀇,𑀇])*Δ([𑀇,𑀇,𑀇]));𑀇Z=q(Z);Z𑀇=p(Z);S = N.linspace(𑀇*𑀇,-𑀇,𑀇**Δ([𑀇,𑀇])*Δ([𑀇,𑀇,𑀇]));M=(N.empty(Z.shape));𑀇M=q(M);M𑀇=p(M);𑀇S=q(S);S𑀇=p(S);W=(N.empty(S.shape));𑀇W=q(W);W𑀇=p(W);M[::𝛬((𑀇-𑀇-𑀇/𑀇))]=Ω(𑀇Z)/1-Ω(Z𑀇)/1;Z[::𝛬((𑀇//𑀇))]=Ω(𑀇M)/Π-Ω(𑀇M)/Π/2;M[::𝛬((𑀇-𑀇-𑀇/𑀇))]=Ω(𑀇M)/1-Ω(M𑀇)*+(𑀇/𑀇);M[::𝛬(𑀇//𑀇)]=Ω(𑀇M)*Π/2-Ω(M𑀇)*Π/2;W[::𝛬((𑀇-𑀇-𑀇/𑀇))]=Ω(𑀇S)/1-Ω(S𑀇)/1;S[::𝛬((𑀇/𑀇))]=Ω(𑀇W)/Θ-Ω(𑀇W)/Θ/2;W[::𝛬(𑀇-𑀇-𑀇/𑀇)]=Ω(𑀇W)/1-Ω(W𑀇)*+(𑀇/𑀇);P,=H.plot(W𑀇-𑀇M-N.tan(Z)*-Ω(S),Ω((M)[::𝛬((𑀇-𑀇-𑀇/𑀇))]/(Θ))+Ω((W)[::𝛬((𑀇/𑀇))]/(Π/2)));P.set_color(chr((𑀇*𑀇)-(𑀇+𑀇+(𑀇//((𑀇+𑀇)//𑀇)))));H.axis('off');H.show()

[unreal echo]

Craxylzy

[fleet bridge]

replace exec with print and you probably will be fine
and print will print the code that your friend tried to execute on your machine

[stark granite]

its not even syntatically valid

in the exec call the opening parenthes is doubled

[mint summit]

solved, forgot to delete it - thy for trying to help tho 🫶

[torn cypress]

is this a programming language?

yk... APL and all

[unique heath]

very productive use of time

[sinful scarab]

So I wrote this line last night (Tkinter code), but was surprised after that it worked/confused about how:

self.button = tk.Button(self.frame, text='Button', font=custom_font, bg='#334455', command=lambda: self.button_click(self.button))

The button appears to work/be stable. How can I create an anonymous lambda function passing/using self.button before it exists?

[versed eagle]

the lambda isnt called until after you've set self.button

[sinful scarab]

Yes, but I've read that lambdas take the value from the context at the moment of creation (so it would be getting the value of self.button prior to it being set to the new Button).

Lambdas are closures in Python.

A closure is a function that captures the environment in which it was defined, including any variables that were in scope at the time of its definition. This allows the function to access those variables even after they have gone out of scope or changed their values.

[fleet bridge]

it captures self, not self.button

[rocky leaf]

import this

!!!

im bored

[fleet bridge]

import that

[rocky leaf]

!e

print = p;s+m+h=0;s+1;if s==60:s=0;m+1;if m==60:h+1;if h==24:h=0;while h<24:
p(s,m,h)

[fleet bridge]

doesnt work

at least as python code

[night quarry]

Missing required argument

code

[rocky leaf]

!e print = p;s+m+h=0;s+1;if s==60:s=0;m+1;if m==60:h+1;if h==24:h=0;while h<24:
p(s,m,h)

[night quarry]

@rocky leaf :x: Your 3.12 eval job has completed with return code 1.

001 |   File "/home/main.py", line 1
002 |     print = p;s+m+h=0;s+1;if s==60:s=0;m+1;if m==60:h+1;if h==24:h=0;while h<24:
003 |               ^^^^^
004 | SyntaxError: cannot assign to expression here. Maybe you meant '==' instead of '='?

[rocky leaf]

boo

how do i make my python more esoteric

[torn cypress]

well, I can tell you that it doesn't happen simply by obfuscation, you kinda have to use some obscure, uncommon stuff

[night quarry]

Missing required argument

code

[torn cypress]

!e for example:

[real := False, forever := False, [type("", (), {(i := "__init__"): lambda t: [k := "t", setattr(t, "s", type("s", (), {i: lambda s: [setattr(s, k, 0)] and None, (c := "__call__"): lambda s, l, f=0: not (isinstance(s, t.s) and (print("%s:%s:%s" % (str(t.hi.t).zfill(2), str(t.mi.t).zfill(2), str(t.si.t).zfill(2))) or __import__("time").sleep(real))) and getattr(s, k) + 1 < 60 and (setattr(s, k, getattr(s, k) + 1) is None and s(t.mi)) or not setattr(s, k, 0) and (f and l())})), setattr(t, "m", type("m", (t.s,), {c: lambda s: not setattr(s, k, getattr(s, k) + 1) and (s.t < 60 or setattr(s, k, 0) or t.hi()) and t.si(s, 1)})), setattr(t, "h", type("h", (t.m,), {c: lambda s: 0})), setattr(t, "si", t.s()), setattr(t, "mi", t.m()), setattr(t, "hi", t.h()), *map(lambda x: [t.si(t.mi, 1), setattr(t.hi, k, x)], range(25))] and None})() for _ in forever and __import__("itertools").cycle([0]) or [1]]]

[night quarry]

@torn cypress :x: Your 3.12 eval job timed out or ran out of memory.

001 | 00:00:00
002 | 00:00:01
003 | 00:00:02
004 | 00:00:03
005 | 00:00:04
006 | 00:00:05
007 | 00:00:06
008 | 00:00:07
009 | 00:00:08
010 | 00:00:09
011 | 00:00:10
... (truncated - too many lines)

Full output: too long to upload

[torn cypress]

still too long? smh /s

[flint hollow]

clearly winning the code golfing competition i see

[torn cypress]

ah yes, when you need over 900 characters to simply print out some time in a 24 hour format in a loop...

for reference, this is like a playable game and it's less than 500 characters

||though I have a feeling you were trying to be sarcastic, I wasn't entirely sure, so I was sort of trying to be sarcastic back as well, but.... yeah, welll... now it's awkward||

[flint hollow]

[twilit igloo]

too esoteric

[unique heath]

you made it more awkward

[simple sphinx]

not 100% sure if its on topic or off topic but I made a thing I think some people here might like 🙂
https://www.codewars.com/kata/63d4b700bce90f0024a9ca19

its a question about how you can represent python code to use a low amount of unique chars

[restive void]

Very much on topic. In Python 2, the lowest amount of characters needed was 8, so with the same approach it would be 10 in Python 3: https://codegolf.stackexchange.com/a/11697

[simple sphinx]

ooh

ill go ahread and read it but nice too see people had the same idea 🙂

my sample answer is actualy 9 chars in python3

but its a very neat idea to use that trick

[restive void]

Interesting. Do you use ()?

[simple sphinx]

yea

you can solve it and see the exact solution if you want the one you linked is under 12 chars so it will pass the tests

just make sure to self compile before submitting the answer

(gets behind a lot of cheeky answers)

[lunar marsh]

You can get 8 in Python3 as well :) https://theblupper.github.io/pyfuck.html

[simple sphinx]

thats a nice idea

I think you can cut a letter out

but im not sure

[blissful oar]

how do i learn this

[versed eagle]

learn what

[atomic violet]

Is there a quick method for saving an array to disk and reading each element from the file without loading the entire array into RAM?

[versed eagle]

do you know the size of each element

[atomic violet]

yep

it's is n-D-Array of uint8

but it's dimensions are fix.

[versed eagle]

oh

[versed eagle]

then yes

just write it all into a file and then read it out piecewise

seek if you want random access instead of sequential

[atomic violet]

@versed eagle Yes, I Need Random access instead of sequential. & Python Loops are Blazigly fast.

Is there is an already implementation.

The array is an video.

[versed eagle]

.

seek if you want random access

[atomic violet]

What dose it means?

[versed eagle]

if you want random access, seek through the file to where-ever you want to read the data from instead of reading piecewise sequentially

[atomic violet]

How?

[versed eagle]

file.seek

[atomic violet]

Thanks

[digital dirge]

Can I just toss a codegolf challenge in here? This is my best result, 245 chars. I wonder how much shorter you folks could get it
||```py
n=int(input())
S=sum
l=[[[x*(i<b)for i in(0,1,2)]for x in(1,2,3,0)]for b in[n//3[3],n%3]]
e=[[0]*3]*2
[print("".join(" - ( O|' O|' )' ( /' /) "[i::4]for i in p))for p in[S(zip(*z),())for z in zip(S(l[::2],[])+e,S(l[1::2],e)+e)]if S(p)]

||The string and the list comprehensions are mad expensive but I had to settle for this eventually 😄||

Spoilered my solution in case someone wants to do it with a fresh mind, here's the challenge:

BEES? BEES!

Print this little fellow and his crew to the console.

  OO  //
-(||)(')
  ''' 

Your input per stdin or argv is a single integer >= 1 that determines how many bees should be printed.
Bees always fly in formation though:
Maximum number of bees in a row is 3. Each bee has a bee-width distance to the neighbouring bees in the same row. The first line of each new row of bees is the same line as the last line of the previous row, and every second row is offset by the width of one bee to make that possible (see examples below). The bees fill the rows from top to bottom and left to right.
Don't print any extraneous empty trailing or leading lines.
Some example inputs + outputs:

2
  OO  //          OO  //                        
-(||)(')        -(||)(')                        
  '''             '''                           

5
  OO  //          OO  //          OO  //        
-(||)(')        -(||)(')        -(||)(')        
  '''     OO  //  '''     OO  //  '''           
        -(||)(')        -(||)(')                
          '''             '''                   

13
  OO  //          OO  //          OO  //        
-(||)(')        -(||)(')        -(||)(')        
  '''     OO  //  '''     OO  //  '''     OO  //
        -(||)(')        -(||)(')        -(||)(')
  OO  //  '''     OO  //  '''     OO  //  '''   
-(||)(')        -(||)(')        -(||)(')        
  '''     OO  //  '''     OO  //  '''     OO  //
        -(||)(')        -(||)(')        -(||)(')
  OO  //  '''             '''             '''   
-(||)(')                                        
  '''                                           

Have fun!

[fleet bridge]

cute bees!

[digital dirge]

Thanks 😄

[torn cypress]

165 chars
||```py
j=(" "*8).join
c,r=divmod(int(input()),3)
for n in range(c+1):p=" "8(n%2);k=[3,r][n==c];print(p+j([" OO //"]*k),p+j(["-(|<remove this...>|)(')"]*k),p+j([" ''' "]*k),sep="\n")

I mean, that's like perfect, thanks discord...

just need to get rid of <remove this...>

because discord parses || in the string as end of the spoiler, smh

[versed eagle]

use a \

to escape it

nvm discord doesnt parse escapes inside codeblocks

because of course

[torn cypress]

exactly

[versed eagle]

ugh discords markdown parser is so cursed lol

[quartz wave]

153 chars ||```py
c,r=divmod(int(input()),3)
for n in range(c+1):f=lambdas:(" "8).join(n%2("",)+s[3,r][n==c]);print(f(" OO //"),f("-(|<remove this...>|)(')"),f(" ''' "),sep="\n")

143 chars 166 chars ||```py
c,r=divmod(int(input()),3)
for n in range(c+1):print(map(lambdas:(" "8).join(n%2("",)+s*[3,r][n==c]),[" OO //","-(|<remove this...>|)(')"," ''' "]),sep="\n")

137 chars 144 chars ||```py
c,r=divmod(int(input()),3)
for n in range(c+1):print([(" "8).join(n%2[""]+s[3,r][n==c])for*s,in[" OO //","-(|<remove this...>|)(')"," ''' "]],sep="\n")

126 chars 133 chars ||```py
c,r=divmod(int(input()),3)
for n in range(c+1):
fors,in" OO //","-(|<remove this...>|)(')"," ''' ":print((" "8).join(n%2[""]+s[3,r][n==c]))

129 chars ||```py
N=int(input())
for n in range(N//3+1):
fors,in" OO //","-(|<remove this...>|)(')"," ''' ":print((" "8).join(n%2[""]+s[3,N%3][n==N//3]))

~~128 chars ||```py
N=int(input())
for n in range(N//3+1):[print((" "8).join(n%2[""]+s*[3,N%3][n==N//3]))for*s,in" OO //","-(|<remove this...>|)(')"," ''' "]

128 chars ||```py
N=int(input());J=" "8
for n in range(N//3+1):
fors,in" OO //","-(|<remove this...>|)(')"," ''' ":print(n%2J+J.join(s[3,N%3][n==N//3]))

[atomic violet]

What the Hell is going on.

[clear venture]

golf 🏌️‍♂️

[dire galleon]

your code doesn't work though

3
                  O        O                          /        /                          O        O                          /        /                          O        O                          /        /
-        (        |        |        )        (        '        )        -        (        |        |        )        (        '        )        -        (        |        |        )        (        '        )
                  '        '        '                                                     '        '        '                                                     '        '        '                           

This doesn't look like the desired output

[clear venture]

looks more like a alaskan bull worm than a bee

[dire galleon]

it stopped working from this point onwards

[quartz wave]

hmm

[quartz wave]

130 chars fixed ||```py
N=int(input());J=" "8
for n in range(N//3+1):
for s in" OO //","-(|<remove this...>|)(')"," ''' ":print(n%2J+J.join([s]*[3,N%3][n==N//3]))

[gleaming linden]

123 ||```py
N=int(input());J=" "8;a=0
while N>0:
for s in" OO //","-(|<remove>|)(')"," ''' ":print(aJ+J.join([s]*min(N,3)))
a=1-a;N-=3

[atomic violet]

Why Not Sequence. typing.Sequence

[quartz wave]

122 chars alt 123 ||```py
N=int(input());J=" "8;a=0
while N>0:
for s in" OO //","-(|<remove>|)(')"," ''' ":print(-aJ+J.join([s]*min(N,3)))
a=~a;N-=3

[wheat river]

122

N=int(input());J=" "*8;a=0
while N>0:
 for s in"  OO  //","-(|<remove>|)(')","  '''   ":print(a*J+J.join([s]*min(N,3)))
 a^=1;N-=3

[wheat river]

another 122

N=int(input());J=" "*8;a=0
while N>0:
 for s in"  OO  //","-(||)(')","  '''   ":print(*[J*-a+s+J*-~a]*min(N,3))
 N-=3;a=~a

[wheat river]

121

N=int(input());a=1
while N>0:
 for s in"  oo  //","-(||)(')","  '''   ":print(*(J:=' '*8,*[s+J]*min(N,3))[a:])
 a^=1;N-=3

120

N=int(input());a=0;J=' '*8
while N>0:
 for s in"  oo  //","-(||)(')","  '''   ":print(*(J*a,*[s+J]*min(N,3)))
 a^=1;N-=3

[fleet bridge]

print(*(J*a,*[s+J]*min(N,3)))
print(J*a,*[s+J]*min(N,3))
``` ?

[wheat river]

oh yea

[fleet bridge]

  ..          .. 
(  —)       (  —)
 '''`   ..   '''`
      (  —)      
  ..   '''`   .. 
(  —)       (  —)
 '''`   ..   '''`
      (  —)      
       '''`      

[digital dirge]

Damn, you folks are insane 😄

Wait a minute, those are solutions to an easier challenge

Check the formation again, the difficulty is that the bees are overlapping 😄

Very cool stuff tho, I'll make that a part of the challenge on my server

Do you mean *args?
Because those are tuples.
If all elements have the same type:

def something(*values: int):
    pass
something(3, 5)

If the elements have different types you can typehint them like this since 3.12:

from typing import Unpack
def something(*values: Unpack[tuple[int, str]]):
    pass
something(3, "spam")

If you weren't talking about *args, the question doesn't make too much sense because it depends on whether you expect values to be a tuple or a list. The typehint doesn't change whether the passed object is mutable or not. If you don't care, you can use Sequence, like Soham said

[torn cypress]

what do you mean?

[fleet bridge]

1. incorrect
2. correct

[sick hound]

!e

((6,9,(0,1,0,0,()),(0),0,0,(0,((0),(0,(0))))),0+(0)),(c:=0.3*divmod(30,10)[0]),(d
:=c*36/len([0,0,0]),(k:=__import__("builtins"),(0,0,(0,(0,o:=eval("k.print"))),0)
),(c:=80.00000000000001*c),(c:=int(c)),(0,),(j:=""),(0,(0,0,0,(j:=j+chr(c)))),0),
(c:=1.4027777777777777*c),(0,0,(),0,(j:=j+chr(int(c)))),(j:=j+chr(ord(j[-1])+7)),
(j:=j+j[-1]),(0,0,()),(u:=j,(0,10,3,(0x3f))),*(u:=list(map(ord,u))),(u:=(u[-1])),
(0,1,2,([range(2-u)])),(c:=u+3),(0,0,(0,0,0)),(-1),(((),0),(0)),(j:=j+chr(c),24),
((),0,0,-1,divmod(30, 0xffc),(j:=j+" ")),(j:=j+chr(sum([c,u])//2-22)+"AAoAA"[2]),
(0,30,(n:=((c*0)+abs(60.3337/pow(2,2)-36))+94)),(0,3,5,7,c:=k.int),(n:=c(n)),(0),
(0,30,50,70),(([]),j:=j+chr(n)),(k:=k.chr),(j:=j+j[3]+k(87//2+57)+k(66//2)),o(j),

[night quarry]

@sick hound :white_check_mark: Your 3.12 eval job has completed with return code 0.

Hello World!

[rose tapir]

💀

[sick hound]

I made a idle game

https://github.com/dislux-hapfyl/pynksh check outh all 5 games

[fleet bridge]

Python 3.9.13 (tags/v3.9.13:6de2ca5, May 17 2022, 16:36:42) [MSC v.1929 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> __peg_parser__
  File "<stdin>", line 1
    __peg_parser__
    ^
SyntaxError: You found it!

[unreal echo]

Huh

!e
peg_parser

[night quarry]

@unreal echo :x: Your 3.12 eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "/home/main.py", line 1, in <module>
003 |     __peg_parser__
004 | NameError: name '__peg_parser__' is not defined

[restive void]

This was only a "feature" in one or two versions I believe; when the PEG parser was new.

[digital dirge]

I think "We'll pass a list and call it a tuple so people don't modify it unless they want to in which case they can ignore the typehint (with care)" is terrible design and any typechecker would agree with me

Of course, how bad it actually is depends on the type of code you're writing and who is expected to work with that, but I can't imagine a case where that would be the best or even a good solution

This could come across as a bit hostile, but be assured that I'm not trying to attack you in any way, I just feel strongly about using typehints correctly 😄

[snow panther]

wtf is this

[gleaming linden]

The type carries some implicit meanings as well (list = variable amount of homogeneous items, tuple = fixed count and fixed types), which could be considered more important. The problem of inadvertently mutating lists stays, though. I think a separate indicator (perhaps Mutated[list]) would be a better solution.

Nedbat has a related blog post: https://nedbatchelder.com/blog/201608/lists_vs_tuples.html

[dapper walrus]

How can I print the variable FLAG, via the variable pattern?
I have this snippet:

FLAG = "TEST"
pattern = '{{FLAG}}'
print(f"{pattern}")

I am looking to get TEST as output. Pattern variable must be a string, (it gets its value via input() function but for simplicity I hardcoded it). Also the rest of the code (print statement or FLAG variable) cannot change.

[dense mural]

# > [TEST] You must send me (PID 14657) the following signals, in exactly this order: ['SIGINT', 'SIGIT2']
# pid="14657"
pid=$(echo $line | python -c 'print(__import__("re").search(r"PID (\d+)\)", input()).group(1))')
# signals="-SIGINT -SIGIT2"
signals=$(echo $line | python -c 'print("-" + " -".join(eval(__import__("re").search(r"order: (.*)", input()).group(1))))')

Any Ideas To Make This Less Cursed? D: (In Bash)

[fleet bridge]

this is a wrong channel to ask for "less cursed" anything 😄

[dense mural]

True D:
But, I Thought Maybe If You Can Cook Cursed Code, You Can Uncurse It

[rugged sparrow]

if pattern is just the return value of input() (and nothing else happens to it before it is interposed via f"{pattern}") then there is no way to leak the FLAG value

if there was, then fstrings would be fundamentally insecure

[dapper walrus]

yep, I discusses the problem in #cybersecurity and the print statement had also a .format() function which was the actual exploit

[rugged sparrow]

ah yea that will make it vulnerable

[fluid pumice]

!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 Paste! button in the bottom left, or by pressing CTRL + S. After doing that, you will be navigated to the new paste's page. Copy the URL and post it here so others can see it.

[fluid pumice]

!e https://paste.pythondiscord.com/6YRQ

[wraith island]

wow so many bools

[fluid pumice]

yep

[unique heath]

woh let this man cook

[versed eagle]

(have you tried using kill)

[dense mural]

Its Specifically For That.
I Wanted To Uncurse The Parsing Such As: signals=$(echo $line | python -c 'print("-" + " -".join(eval(__import__("re").search(r"order: (.*)", input()).group(1))))')

Oh Well, As Long As It Works D:

[versed eagle]

Its Specifically For That.
are you implementing kill

[dense mural]

No,

WELCOME! This challenge makes the following asks of you:
- the challenge checks for a specific parent process : shellscript
- the challenge will require the parent to send number of signals : 5

[earnest snow]

hello guys, just made a 1-line padding maker in python

s = lambda L,i=0: ([lambda L: s(L.split("\n"), i + 1),lambda S: s((S[:], max(S, key=len)),i+1),lambda S: s((list([" " * (len(S[1]) - len(line)//2) + line + " " * ((len(S[1]) - len(line)//2) + len(line) % 2) for line in S[0]]), len(S[1])),i+1),lambda S: s("\n".join(["+" + "-" * S[1] + "+", *map(lambda L:"|"+L+"|", S[0]), "+" + "-" * S[1] + "+"]), i + 1)][i](L)) if i < 4 else L

the only problem is that the outer marks aren't getting the desired length ¿any ideas on how to debug this beast?

[rugged sparrow]

What is it supposed to do?

[earnest snow]

Mark text

+---------+
| Examole |
+---------+

[rugged sparrow]

ok

one sec

[digital dirge]

That's a fun curse

I suppose the bug is the operator precedence in the second lambda S:

[rugged sparrow]

for multiline input, should the lines be centered or left aligned?

[digital dirge]

I'd assume centered

[rugged sparrow]

bet, lets me shave off 2 chars

!e ```py
S = lambda L:'\n'.join([h:=f'+{"-"(N:=max(map(len,I:=L.splitlines()))+2)}+',[f'|{l:^{N}}|'for l in I],h])

print(S('abc'))

print(S('a\nabc\na'))```

[digital dirge]

!e```py
s = lambda L,i=0: ([lambda L: s(L.split("\n"), i + 1),lambda S: s((S[:], max(S, key=len)),i+1),lambda S: s((list([" " * ((len(S[1]) - len(line))//2) + line + " " * (((len(S[1]) - len(line))//2) + len(line) % 2) for line in S[0]]), len(S[1])),i+1),lambda S: s("\n".join(["+" + "-" * S[1] + "+", *map(lambda L:"|"+L+"|", S[0]), "+" + "-" * S[1] + "+"]), i + 1)]i) if i < 4 else L
print(s("""hello world
it is
i"""))

[night quarry]

@digital dirge :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | +-----------+
002 | |hello world |
003 | |   it is    |
004 | |     i      |
005 | +-----------+

@rugged sparrow :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | +-----+
002 | | abc |
003 | +-----+
004 | +-----+
005 | |  a  |
006 | | abc |
007 | |  a  |
008 | +-----+

[digital dirge]

ah darn

ah because odd number of characters in the longest line

[rugged sparrow]

use fstring specifiers

[digital dirge]

Oh, we're golfing? I thought the mission was to use lambdas in a really weird way

[rugged sparrow]

why not both

[digital dirge]

split("\n") is 2 chars shorter than splitlines()

[rugged sparrow]

you can't inline \n inside a fstring

at least on <3.12

I am still using 3.11

[digital dirge]

get with the times, old man

[grave grail]

!e 3.11

print(f'Hello\nSee ya')
print(f'Hi{"\n"}Bye')

[rugged sparrow]

➜  ~ python3.11
Python 3.11.8 (main, Feb  6 2024, 21:21:21) [Clang 15.0.0 (clang-1500.1.0.2.5)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> f'{"\n"}'
  File "<stdin>", line 1
    f'{"\n"}'
             ^
SyntaxError: f-string expression part cannot include a backslash
>>>```

[grave grail]

I want to describe first work but second isn't

can you show them that?

[earnest snow]

what did you change

oh got it

messed parenthesis

on padding

[earnest snow]

just upgraded my indexing-lambda tecniques and made a terminal game of guessing a number in 1 line of code 😄

it's scary big though

s = lambda L, i = 0: [lambda L: (()for()in()).throw(ValueError("Expected None as an argument")) if L != None else s(__import__("sys"), i + 1),lambda sys: (()for()in()).throw(ValueError("Expected Python version 3")) if sys.version_info[0]!=3 else s((sys, __import__("random")), i + 1),lambda modules: (lambda sys, rnd:(print("Welcome to the Python random number guesser!") or s(rnd.randint(0,40),i+1)))(*modules),lambda guesser: (lambda guessed_number:(print("Greater than that!") or s(guesser,i)) if guessed_number < guesser else ((print("Smaller than that!") or s(guesser, i) if guessed_number > guesser else (print("You got it!") or s(guesser,i + 1)))))((lambda inp:(print("guess a number!") or s(guesser,i)) if(not(inp().isdigit())) else int(inp()))((lambda v:lambda:v)(input("Guess a number")))),lambda guesser: (print("You want to play again? [y/yes/sure/true]") or (s(None) if input(">>> ").lower() in {"y", "yes", "sure", "true"} else s(True, 5)))][i](L) if i < 5 else L

to play it, just paste that shit on the terminal and call s(None)

could've made L a keyword argument now that I think of it

just noticed that thinks mess up when you ask it to play again

good luck debuggin that sh!t because I won't

[atomic violet]

Most easy to read code.

continue, Keep it Up.

https://tenor.com/view/noice-nice-click-gif-8843762

@earnest snow You Are Chad.

[earnest snow]

thanks

you more or less get how it works?

[atomic violet]

I have a Suggestion. use (_ for _ in ()).throw(...) instead of (()for()in()).throw(...)

[earnest snow]

or put a list in it that even though it does nothing but makes thing less readable

but my point is that it works in one line

[atomic violet]

Yep.

[earnest snow]

if I wanted to obfuscate it more, I would've made a string-obfuscator and I would've used horrible variable names

the funny thing is that it even does the imports in that line and makes sure that is running in python 3

for anyone that actually is trying to read that, this is but separated in the inner list:

s = lambda L, i = 0: [
    lambda L: (()for()in()).throw(ValueError("Expected None as an argument")) if L != None else s(__import__("sys"), i + 1),
    lambda sys: (()for()in()).throw(ValueError("Expected Python version 3")) if sys.version_info[0]!=3 else s((sys, __import__("random")), i + 1),
    lambda modules: (lambda sys, rnd:(print("Welcome to the Python random number guesser!") or s(rnd.randint(0,400),i+1)))(*modules),
    lambda guesser: (lambda guessed_number:(print("Greater than that!") or s(guesser,i)) if guessed_number < guesser else ((print("Smaller than that!") or s(guesser, i) if guessed_number > guesser else (print("You got it!") or s(guesser,i + 1)))))((lambda inp:(print("guess a number! >>> ") or s(guesser,i)) if(not(inp().isdigit())) else int(inp()))((lambda v:lambda:v)(input("Guess a number")))),
    lambda guesser: (print("You want to play again? [y/yes/sure/true]") or (s(None) if input(">>> ").lower() in {"y", "yes", "sure", "true"} else s(True, 5)))
][i](L) if i < 5 else L

[atomic violet]

the below thing is good for 1 liner.

def s(L, i=0):    [lambda L: (_ for _ in()).throw(ValueError("Expected None as an argument")) if L is None else s(__import__("sys"), i + 1),lambda sys: (_ for _ in()).throw(ValueError("Expected Python version 3")) if sys.version_info[0]!=3 else s((sys, __import__("random")), i + 1),lambda modules: (lambda sys, rnd:(print("Welcome to the Python random number guesser!") or s(rnd.randint(0,40),i+1)))(*modules),lambda guesser: (lambda guessed_number:(print("Greater than that!") or s(guesser,i)) if guessed_number < guesser else ((print("Smaller than that!") or s(guesser, i) if guessed_number > guesser else (print("You got it!") or s(guesser,i + 1)))))((lambda inp:(print("guess a number!") or s(guesser,i)) if(not(inp().isdigit())) else int(inp()))((lambda v:lambda:v)(input("Guess a number")))),lambda guesser: (print("You want to play again? [y/yes/sure/true]") or (s(None) if input(">>> ").lower() in {"y", "yes", "sure", "true"} else s(True, 5)))][i](L) if i < 5 else L

[earnest snow]

I could change the lower and upper limits to recursive lambdas to make sure you input an integer so that you can choose them ¿no?

s = lambda L, i = 0: [lambda L: (()for()in()).throw(ValueError("Expected None as an argument")) if L != None else s(__import__("sys"), i + 1),lambda sys: (()for()in()).throw(ValueError("Expected Python version 3")) if sys.version_info[0]!=3 else s((sys, __import__("random")), i + 1),lambda modules: (lambda sys, rnd:(print("Welcome to the Python random number guesser!") or s(rnd.randint(s("choose the lower limit >>> ", 5),s("choose the upper limit >>> ", 5)),i+1)))(*modules),lambda guesser: (lambda guessed_number:(print("Greater than that!") or s(guesser,i)) if guessed_number < guesser else ((print("Smaller than that!") or s(guesser, i) if guessed_number > guesser else (print("You got it!") or s(guesser,i + 1)))))(s("Guess a number >>> ", 5)),lambda guesser: (print("You want to play again? [y/yes/sure/true]") or (s(None) if input(">>> ").lower() in {"y", "yes", "sure", "true"} else s(True, 8))),lambda prompt:(lambda inp:(print("guess a number! not that!  ") or s(prompt,5)) if(not(inp().isdigit())) else int(inp()))((lambda v:lambda:v)(input(prompt)))][i](L) if i < 6 else L

[earnest snow]

I moved the input as s(prompt, 5) and reused it to choose about the limits on the randint call

[elfin ether]

I'm trying to use the forbiddenfruit library https://github.com/clarete/forbiddenfruit/blob/master/forbiddenfruit/__init__.py

to override tuple.__new__ without success. Anyone care to take a look?

[vocal quartz]

I've been slaving over how to shorten this code past 68 bytes:py import numpy;print(*map(chr,numpy.random.randint(32,127,39)),sep='') There's a second one that is also 68 bytes:```py
from numpy import*;print(*map(chr,random.randint(32,127,39)),sep='')

The code generates a random string with length 39 with each character being a random ascii value between 32 and 126, inclusive. Here's a readable code example:```py
import random

string = ''
for i in range(39):
    string += chr(random.randint(32, 126))
print(string)

If anyone has any ideas, I would love to know.

It's been weeks and I can't find a way to shorten it further.

[gleaming linden]

from numpy import*;print("%c"*39%(*random.randint(32,127,39),))
``` 63 using string formatting to convert into chars

[vocal quartz]

I had no idea that existed, thank you so much!

Do you have any resources where I can further learn about that?

[gleaming linden]

It's basically C's printf-style formatting

print("template" % (var1, var2)) is roughly printf("template", var1, var2) in C

it's mostly superseded by f-strings and new format strings though

https://realpython.com/lessons/formatting/ here's an article demonstrating it

[vocal quartz]

That is very helpful, thanks a lot!

[restive void]

I still find old-style formatting very useful when strings naturally contain lots of braces (e.g. LaTeX templates), or when doing multi-step formatting.

[elfin ether]

Does executing this code: 1,2 call tuple.__new__ or tuple.__call__ (or both, and in what order?)

[earnest wing]

Neither

The compiler embeds a tuple literal into the code object

[elfin ether]

damn...

is there any way to hook into "literal embedding" from within python?

!e

import dis
def print_bytecode(st):
    bytecode = dis.Bytecode(st)
    for instr in bytecode:
        print(instr)
print_bytecode('(x,2)')
print()
print_bytecode('tuple(x,2)')

[night quarry]

@elfin ether :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | Instruction(opname='RESUME', opcode=151, arg=0, argval=0, argrepr='', offset=0, starts_line=0, is_jump_target=False, positions=Positions(lineno=0, end_lineno=1, col_offset=0, end_col_offset=0))
002 | Instruction(opname='LOAD_NAME', opcode=101, arg=0, argval='x', argrepr='x', offset=2, starts_line=1, is_jump_target=False, positions=Positions(lineno=1, end_lineno=1, col_offset=1, end_col_offset=2))
003 | Instruction(opname='LOAD_CONST', opcode=100, arg=0, argval=2, argrepr='2', offset=4, starts_line=None, is_jump_target=False, positions=Positions(lineno=1, end_lineno=1, col_offset=3, end_col_offset=4))
004 | Instruction(opname='BUILD_TUPLE', opcode=102, arg=2, argval=2, argrepr='', offset=6, starts_line=None, is_jump_target=False, positions=Positions(lineno=1, end_lineno=1, col_offset=0, end_col_offset=5))
005 | Instruction(opname='RETURN_VALUE', opcode=83, arg=None, argval=None, argrepr='', offset=8, starts_line=None, is_jump_target=False, positions=Positions(lineno=1, end_lineno=
... (truncated - too long, too many lines)

Full output: https://paste.pythondiscord.com/GYFYQOAUNNC5GC4DHWY3D7RQRM

[elfin ether]

Is there any way to intercept BUILD_TUPLE?

[rugged sparrow]

if you can get the address of _PyTuple_FromArraySteal (used here: https://github.com/python/cpython/blob/main/Python/bytecodes.c#L1589-L1592), then you can use fishhook.asm (if you are on x86 intel, i am still working on cross platform asm hooks)

[night quarry]

Python/bytecodes.c lines 1589 to 1592

inst(BUILD_TUPLE, (values[oparg] -- tup)) {
    tup = _PyTuple_FromArraySteal(values, oparg);
    ERROR_IF(tup == NULL, error);
}```

[rugged sparrow]

you could also use a bytecode tracing function, depending on what you want to be able to do

[elfin ether]

I want to look at the arguments of a tuple when it is created and take some special action based on the args.

[rugged sparrow]

yea an opcode tracing function would work for that, you would just need to grab a reference to the stack to look at the args

acc im gonna write some code to use the new monitoring api to do this

gimme like 15, ill ping you when its done

[elfin ether]

Here's what I'm trying to do:

_ = object()
z = (lambda x, y: x + y), 3, 4, _
assert z == 7

basically i want to check if the last arg to a tuple is some sentinel and then attempt to evaluate the expression described by the other args

[rugged sparrow]

ah, that is much trickier

will need to inspect the stack and modify it

[elfin ether]

if i can replace the BUILD_TUPLE opcode with CALL_FUNCTION I have a way of hooking into tuple.__call__ and can take the desired action. Is swapping out opcodes a possibility

?

[rugged sparrow]

swapping opcodes can be doable, but is also tricky due to cpython optimizers and opcodes needing different amounts of cache

[elfin ether]

hmm

[rugged sparrow]

gimme a bit, ill do the stack strategy

[elfin ether]

thanks m8

[rugged sparrow]

>>> x = 1
>>> x,x,x
Stack[0]: (1, 1, 1)
Stack[1]: 1
Stack[2]: 1
(1, 1, 1)
>>> 
``` @elfin ether  I have this working, but for some reason python unsets the stacktop inside the frame object before I can read it

so its difficult to know what the current offset into the stack should be

https://paste.pythondiscord.com/UFIQ here is what I have so far, ill keep thinking on how to efficiently retrieve the stack pointer

[elfin ether]

thanks. this isn't super important or anything, so don't spend too much time unless you're enjoying yourself

[rugged sparrow]

eh i always have fun writing cursed python

[elfin ether]

my python 3.12 doesn't have sys.monitoring ...

[rugged sparrow]

mine is on 3.12.2

[elfin ether]

installing 3.13 now

[grave grail]

!d sys.monitoring

[night quarry]

sys.monitoring

New in version 3.12.

Note

sys.monitoring is a namespace within the sys module, not an independent module, so there is no need to import sys.monitoring, simply import sys and then use sys.monitoring.

This namespace provides access to the functions and constants necessary to activate and control event monitoring.

As programs execute, events occur that might be of interest to tools that monitor execution. The sys.monitoring namespace provides means to receive callbacks when events of interest occur.

The monitoring API consists of three components...

[earnest snow]

I'll let y'all burn in anger with this

C:\>python --version
Python 3.10.5

[versed eagle]

for me python --version is python 2

[magic wraith]

can we see your code though

what have you tried that doesn't work?

[earnest snow]

welp, you got me in that

[fleet bridge]

C:\Users\denba>py -3.7 -V
Python 3.7.9

C:\Users\denba>py -3.8 -V
Python 3.8.10

C:\Users\denba>py -3.9 -V
Python 3.9.13

C:\Users\denba>py -3.10 -V
Python 3.10.11

C:\Users\denba>py -3.11 -V
Python 3.11.5

C:\Users\denba>py -3.12 -V
Python 3.12.0

C:\Users\denba>py -3.13 -V
Python 3.13.0a1

[rugged sparrow]

Forbidden fruit had issues with hooking __new__, and built-in type constructors aren't typically called when constructed using literals

[versed eagle]

lol nice

[arctic skiff]

E:\co>py -3.12 -V
Python 3.12.0

E:\co>py -3.10 -V
Python 3.10.11

E:\co>py -3.11 -V
Python 3.11.0

E:\co>py -3.7 -V
Python 3.7.0

E:\co>py -3.6 -V
Python 3.6.0

E:\co>py -3.5 -V
Python 3.5.0

[magic wraith]

you also can have 32 bit and 64 bit python
py -3.12-64 -V

[rugged sparrow]

one of my boxes has both arm and intel python installed

[arctic skiff]

yes but I don't need them

E:\co>py -0
 -V:3.12 *        Python 3.12 (64-bit)
 -V:3.11          Python 3.11 (64-bit)
 -V:3.10          Python 3.10 (64-bit)
 -V:3.7           Python 3.7 (64-bit)
 -V:3.6           Python 3.6 (64-bit)
 -V:3.5           Python 3.5

E:\co>```this is all I got

[rugged sparrow]

@elfin ether https://paste.pythondiscord.com/4JLA this should work for basic stuff, there are still some edge cases in how I compute the stack top, and sometimes its just flat wrong lol

sadly the stack top value gets yanked out of the stack frame before monitor/trace functions are called, so i need to recreate it (i walk the entire code object and psudo sim the bytecode for that) or grab it off of the C level frame stack

[versed wagon]

wow what is this

[rugged sparrow]

➜  ~/Desktop python3 -i opcode_hook.py
>>> v = (SENTINEL,)
Evaluating: (<object object at 0x104198830>,)
>>>```

[arctic skiff]

the versions of python I have

[restive void]

Why are you defining empty Structures and filling _fields_ afterwards?

[karmic pumice]

consistency for when need types that reference themselves i guess

oh yeah they do reference themselves

[restive void]

Ah, right..

[rugged sparrow]

@quartz wave yo i am trying to fix up my cross platform fishhook.asm, did you ever test fishhook.asm on windows and did it work?

[quartz wave]

i tried adapting it for windows but it doesn't work

there's the problem with libc (and therefore libc.mprotect()) not being available and there's the problem with windll.kernel32.VirtualProtect() not working (it causing a DEP violation)

there's also a negative offset but i don't know if that's a problem

[rugged sparrow]

ah negative offset will prob be fine in my new system, need to fix the VirtualProtect issue tho

my new system compiles the assembly on demand using KeyStone, so itll generate the correct relative jump instructions

just working on getting the virtual protect stuff correct (i think windows was complaining because memory cant be RWX at the same time)

also testing on x86 is a pain for me rn, my main device is an Arm Macbook

but if i get the VirtualProtect stuff to work on windows arm, it should be correct for x86, and the weird cache stuff that I am working on now is only relevant for arm, x86 invalidates automatically as a side effect of the architecture

[versed eagle]

are you able to use a vm?

[rugged sparrow]

the eventual goal is to be able to specify asm injection points using a psudo language, and use the Capstone decompiler to find indexes

[rugged sparrow]

a lot of JIT code like this relies on physical architecture behavior, which are not things that VM authors tend to implement fully

[versed eagle]

ah

[rugged sparrow]

i am able to test on ARM windows, because it is ran by apples hypervisor on the actual arm chip, which means it has the correct quirks

[versed eagle]

i have an x86 machine if you want me to test stuff for you

[rugged sparrow]

but x86 code is ran via rosetta or another emulation/translation engine

[rugged sparrow]

its all good, i have an x86 laptop, i just need to fix it up lol

[versed eagle]

ah good

[rugged sparrow]

end product from this project will be cool tho, goal is cross platform asm injection via python. (and stretch goal is for it to work on other executables via LD_PRELOAD or a harness program)

so really low level C level hooks, but you can write them in python

[versed eagle]

what architectures will be supported?

[rugged sparrow]

the goal is ARM64 and x86-64, and Windows, MacOS/UnixLike operating systems

[quartz wave]

also i'm an amateur at this so i just sort of merged cpython's source code with the current asm code

[versed eagle]

fun

[rugged sparrow]

making progress, but will need to spend some more hours on it

also rn uses GCC's cache invaliator, so ill prob cross compile a bunch of wheels once it works

[rugged sparrow]

it makes sense that libc.mprotect was not available, its a unix specific feature

I ended up having to move to a C module because when you are changing protections of memory ranges willy nilly, you really want to make sure you don't accidentally call any functions that are currently marked RW (read/write), instead of RX (read/execute)

[drowsy cobalt]

https://pointers.zintensity.dev/

[rugged sparrow]

@quartz wave making progress, got my 64bit arm trampoline to work, now to write my x86 one

arm64_trampoline = '''
    sub sp, sp, #0x10;          # make stack space for 2 64bit registers on stack (16 bytes)
    stp x28, lr, [sp];          # save x28, lr onto stack
    ldr x28, =0x{address:x};    # load 64bit address into x28
    blr x28;                    # branch and link to [x28]
    ldp x28, lr, [sp];          # restore x28, lr from stack
    add sp, sp, #0x10;          # clear stack space (16 bytes)
    ret;
'''

[rugged sparrow]

x86_trampoline = '''
    push r15;                   # save r15 register on stack
    mov r15, 0x{address:x};     # move 64bit address into r15
    call r15;                   # call r15
    pop r15;                    # restore r15
    ret;                        # return to caller
'''
``` woot x86 works

now to test windows builds, and then v1 of the new fishhook.asm will be done

[rugged sparrow]

also gonna shrink the arm asm, have some ideas to make it more efficient

Once the underlying tooling for cross platform injection works, I'm gonna start working on dynamic injection points and other fun tricks

[rugged sparrow]

ok so hooks on x86 windows are sorta functional, it gets inside the hook fine, just crashes after leaving it

weird

[fleet bridge]

is there any REPL for asm?

i know it is crazy, but still...

nvm, i just had to google it...

[rugged sparrow]

i don't have a ton of time to debug windows today, will probably push this version since it should work on the bot later tonight

[rugged sparrow]

@quartz wave https://pypi.org/project/fishhook/ pushed 0.3, has the new asm code, sadly no windows support yet

@fleet bridge ^ you might want to take a look too

[fleet bridge]

im on windows 😄

Python 3.12.0 (tags/v3.12.0:0fb18b0, Oct  2 2023, 13:03:39) [MSC v.1935 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>>
>>> import fishhook
>>> import fishhook.asm
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "D:\Programs\Python\3.12\Lib\site-packages\fishhook\asm.py", line 5, in <module>
    raise RuntimeError('fishhook.asm does not currently work on Windows')
RuntimeError: fishhook.asm does not currently work on Windows
``` i can confirm your RuntimeError does work :)

>>> fishhook._asm.writeExecutableMemory(b'123',b'456')
``` i managed to crash the process

[rugged sparrow]

yea thatll do it lmao

[fleet bridge]

why is it a native extension? couldn't that be done using ctypes?

[rugged sparrow]

modern OS don't like rwx memory pages

so i toggle to rw, write the memory, and toggle to rx

but that means i am marking an entire 4Kb page of function memory as rw

using ctypes means that I can't control if functions inside that page run

[fleet bridge]

now i get it, it makes sense

[rugged sparrow]

ironically, the writeExecutableMemory works on windows

its just some other part of my chain gets upset

[versed eagle]

modern operating systems are so silly

i know its a security thing but like

[rugged sparrow]

yea gosh darm them wanting security

[versed eagle]

its so boring to not be able to write and execute

[tidal canyon]

openbsd when you try to W&X memory

[rugged sparrow]

well it works to just toggle really fast lmao

once this pushes to the bot tho i can send some really weird sample code

[rugged sparrow]

the amount of hoops i need to jump thru to make sure NULL pyobjects are handled properly due to this single set of 13 lines of code is insane: https://github.com/python/cpython/blob/main/Modules/_ctypes/cfield.c#L1077-L1089

[night quarry]

Modules/_ctypes/cfield.c lines 1077 to 1089

static PyObject *
O_get(void *ptr, Py_ssize_t size)
{
    PyObject *ob = *(PyObject **)ptr;
    if (ob == NULL) {
        if (!PyErr_Occurred())
            /* Set an error if not yet set */
            PyErr_SetString(PyExc_ValueError,
                            "PyObject is NULL");
        return NULL;
    }
    return Py_NewRef(ob);
}```

[fleet bridge]

just patch ValueError, so that it is no longer an exception class 🤷

and then whatever happens happens

this function is probably inlined everywhere, right? so it is not possible to patch

[rugged sparrow]

It's not inlined anywhere, I could patch it out, but I don't want to break other code that depends on the value error

[rugged sparrow]

Current plan is a custom C subclass of _ctypes._SimpleCData

And I'll write custom get and set functions in C

[rugged sparrow]

welp ctypes.h is not included in the default python includes, and the way that get and set functions work changed in 3.13

so i think i will leave handling of NULL py_objects up to the user, since it probably won't be a super common use

and if i think of a clean way to implement later Ill do that

in the meantime, next feature is direct register hooks

[rugged sparrow]

@fleet bridge now you can pass registers={<reg name>: <ctype>, ...} into the asm.hook decorator and use ```py
def hookname(*args, * , registers):
print(registers.<reg name>)

and you can set registers.<regname> and it will be set once your function returns

(it will be vastly unstable lol)

@quartz wave if you want to test on windows, the c code should work i think. It just crashes in a weird spot and I did not have time to debug properly, if you can find anything regarding the crash I can probably fix it

you just need to patch out the os check in fishhook.asm

[sick hound]

hi

[rugged sparrow]

!e @quartz wave @fleet bridge can now try out fishhook.asm inside the snekbox, just need some weird metacode to bypass 3.12 deprecating distutils: ```py
(lambda S:import('sys').modules.update({'distutils':S(),'distutils.sysconfig':S(),'pkg_resources':S()}))(type('',(),dict(getattr=lambda s,a:s,call=lambda s,*a:'')))

from fishhook.asm import get_interned_strings_dict
print(get_interned_strings_dict())```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.12 eval job has completed with return code 0.

{'__subclasses__': '__subclasses__', '__basicsize__': '__basicsize__', '__itemsize__': '__itemsize__', '__flags__': '__flags__', '__weakrefoffset__': '__weakrefoffset__', '__base__': '__base__', '__text_signature__': '__text_signature__', 'ag_running': 'ag_running', 'ag_await': 'ag_await', 'ag_frame': 'ag_frame', 'ag_code': 'ag_code', 'ag_suspended': 'ag_suspended', '__alloc__': '__alloc__', 'capitalize': 'capitalize', 'center': 'center', 'expandtabs': 'expandtabs', 'fromhex': 'fromhex', 'hex': 'hex', 'isalpha': 'isalpha', 'isascii': 'isascii', 'isdigit': 'isdigit', 'islower': 'islower', 'isspace': 'isspace', 'istitle': 'istitle', 'isupper': 'isupper', 'ljust': 'ljust', 'removesuffix': 'removesuffix', 'rindex': 'rindex', 'rjust': 'rjust', 'swapcase': 'swapcase', 'title': 'title', 'zfill': 'zfill', '__self__': '__self__', 'cell_contents': 'cell_contents', '__objclass__': '__objclass__', '__func__': '__func__', '__wrapped__': '__wrapped__', 'co_lines': 'co_lines', 'co_positions': 'co_pos
... (truncated - too long)

Full output: too long to upload

[rugged sparrow]

capstone depends on distutils for right now, newer version should remove dependency when they distribute it

[north wraith]

# New Object
Woman = type('Woman', (object,), {'content':{}})

[unreal echo]

We getting certain political affiliations with this one 🗣️🗣️🗣️🔥💯

[unique heath]

lmao

[dense mural]

Why Political Tho? Can Be Done For Anything Else Too 😶‍🌫️

[arctic skiff]

why did you inherit it from object

it is already object

though why 'content' is a empty dict

[dapper walrus]

because man know nothing about woman?

[dense mural]

Plot Twist: Woman Wrote The Code

[fair quartz]

hello all! do you guys remember the evil trick which replaced the globals dict type with a custom object which would just return the dict's key??

# evil trick
@type.__call__
class Globloglabgalab:
  def __getitem__(self, name):
    # try get from globals, but if we cant, 
    # return, the key's name
    globals()[name] = name

# would not raise an error about the var not being found
print(hello)
# >>> hello

[north wraith]

I think you already know

[rugged sparrow]

yea there was more to it then that, i think its here if you scroll up

[restive void]

E.g. #esoteric-python message

or #esoteric-python message

[arctic skiff]

no I don't now tell me

else...

https://cdn.discordapp.com/emojis/854337001976758302.webp?size=48

[versed eagle]

((they wouldn't ask if they already knew))

[quartz wave]

DEP violation

that's what it says

and i don't really know how to use a debugger on an assembly file

[fair quartz]

THANKS!

pure eivl ahaah

[rugged sparrow]

Ooo ok, that actually tells me a lot. Means some memory is not being set as executable for some reason

[arctic kettle]

lol why do people prefer something weird like tuple.__itemsize__ over just 8

[versed eagle]

because 8 is not correct for every machine

tuple.__itemsize__ is sizeof(PyObject *)

pointer size varies between machines

[versed eagle]

(some older machines even have different types of pointers with different sizes!)

[earnest snow]

how did fuckit do it so that the module is callable? if module attributes can be changed from within the module itself, I've got quite an evil idea.

[earnest wing]

you can set sys.modules[yourmodulename] to any object

[languid hare]

https://github.com/ajalt/fuckitpy/blob/0b9c0df8e597d89cab4d8c12eec9b8a0e2f2b4fc/fuckit.py#L207

[night quarry]

fuckit.py line 207

sys.modules[__name__] = _fuckit('fuckit', __doc__)```

[languid hare]

what olivia said

[earnest snow]

Indeed

In a while I'll send an absolute catastrophy using that

[rugged sparrow]

@dry mirage I remember a while back we were talking about fishhook.asm and it didn't work on your ubuntu machine, pushed an update with a brand new asm hook strategy that now builds an absolute address for the jump, it works on x86 and ARM64 Unix like (tested on macos, kali, and ubuntu), working on fixing up windows now

[dry mirage]

Oh neat, would love to try. Did you have to do different strategies for arm64 addresses?

[rugged sparrow]

https://github.com/chilaxan/fishhook/blob/master/fishhook/jit.py#L49-L70 I build a assembly trampoline, based off of what a given hook needs

using fragments for arm64/x86

[rugged sparrow]

https://github.com/chilaxan/fishhook/blob/master/fishhook/jit.py#L90-L115 this is where the actual trampoline is built

basically for both, it backs up a temporary register, loads the 64bit constant into a temporary register, calls it, then restores the temp register

it also generates asm instructions to allow for reading and writing some registers via a hook

[rugged sparrow]

!e ```py
from fishhook.jit import *

address = 0xdeadbeefdeadbeef
payload = '\n'.join([
inst('push', TMP_REG),
inst('load_const', TMP_REG, address),
inst('call', TMP_REG),
inst('pop', TMP_REG),
inst('ret')
])
print(payload)```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | push r15;
002 | mov r15, 0xdeadbeefdeadbeef;
003 | call r15;
004 | pop r15;
005 | ret;

[quartz wave]

does your get_interned_strings_dict() mechanism require an exception be raised?

[rugged sparrow]

it shouldn't

[quartz wave]

because one thing i noticed in the debugger is that an exception was apparently raised and attempted to be displayed

but got suppressed by the fatal error from the DEP

[rugged sparrow]

python exception or windows exception

[quartz wave]

idk

[rugged sparrow]

hmm

i have theories on why its not working, been doing research to see if there is a memory protection cache

and its also possible that the memory areas are too close together

the weird thing I see is that the hook successfully fires (like if you add a print, it outputs)

[quartz wave]

ok i got the contents

it's an OSError about an access violation writing to some address

[rugged sparrow]

Ohhhh what?

That's weirder

[quartz wave]

it's sandwiched between 2 fatal exceptions btw

[rugged sparrow]

Of course it is lmao

[quartz wave]

ah

okay

[rugged sparrow]

no that was me being sarcastic lol

this testing is really helpful

can you change the third arg of https://github.com/chilaxan/fishhook/blob/master/fishhook/asm.py#L43 and https://github.com/chilaxan/fishhook/blob/master/fishhook/asm.py#L48 to PREAD | PWRITE | PEXEC (import them from fishhook._asm)

[night quarry]

fishhook/asm.py line 43

writeExecutableMemory(mem, default)```
`fishhook/asm.py` line 48
```py
writeExecutableMemory(mem, trampoline)```

[rugged sparrow]

that will set the memory after writing to the most permissive

its possible there is a section of memory that I am marking as Read/Execute that still needs to be Writeable as well

[quartz wave]

oki

[rugged sparrow]

i think its a keyword arg prot_after

so: py writeExecutableMemory(mem, default, prot_after = PREAD | PWRITE | PEXEC) and py writeExecutableMemory(mem, trampoline, prot_after = PREAD | PWRITE | PEXEC)

[quartz wave]

oh wait

i might've done a weird thing that triggered whatever i found

yeah

um

[rugged sparrow]

that still might be a good thing to try 🤷‍♂️

[quartz wave]

discard the 2 fatal errors with the OSError

i did a weird thing that caused it :p

sori

[rugged sparrow]

tbh the failures it has on windows are likely to cause DEP fatal errors

your all good

[quartz wave]

mkay same thing

as like before

[rugged sparrow]

i am gonna add some code to _asmmodule.c at somepoint, rn it doesnt call https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-flushinstructioncache because I assumed that x86 does it automatically

i might have been wrong

[quartz wave]

it goes through ctypes then errors at some place which i assume is the changed memory

[rugged sparrow]

this https://learn.microsoft.com/en-us/windows/win32/api/memoryapi/nf-memoryapi-virtualprotect also might need to be the base address of a memory page, right now its the address of the injected instructions

[rugged sparrow]

this is with the prot_after = PREAD | PWRITE | PEXEC?

ill make some updates to the C code tonight after I eat dinner, ill ping when there is an update on github (you will have to install from source tho cause I don't wan't to push new wheels till i fix this)

[quartz wave]

yea

same thing as like before

[rugged sparrow]

ok, i expected as much tbh

ill ping when I push the C updates

[quartz wave]

could just use this snippet of code i think
https://github.com/python/cpython/blob/main/Python/jit.c#L95-L103

[night quarry]

Python/jit.c lines 95 to 103

    // Do NOT ever leave the memory writable! Also, don't forget to flush the
    // i-cache (I cannot begin to tell you how horrible that is to debug):
#ifdef MS_WINDOWS
    if (!FlushInstructionCache(GetCurrentProcess(), memory, size)) {
        jit_error("unable to flush instruction cache");
        return -1;
    }
    int old;
    int failed = !VirtualProtect(memory, size, PAGE_EXECUTE_READ, &old);```

[rugged sparrow]

haha looks like someone has gone thru this before

excellent

[quartz wave]

thank you cpython jit devs

[rugged sparrow]

@quartz wave just pushed the addition, not at my windows machine to test compile, but if you pull the git and test that would be great

ill test on my vm, but it has been unreliable in the past with fishhook lol

[quartz wave]

my pc malfunctioned when i tried to see what changed

maybe i did something

it's not turning back on fully

maybe i should try to be more careful about this next time

[rugged sparrow]

its not turning on??

that def should not happen

it shouldn't have been able to do anything more then crash python

[quartz wave]

hmm

[quartz wave]

it's on a cycle of turning on for some time, off, repeat

monitor is unresponsive

so i can't test rn

[rugged sparrow]

do a full power cycle

[quartz wave]

huh?

[rugged sparrow]

like unplug from the wall, and turn off the powersupply

(if its a desktop)

[quartz wave]

yeah i did that

[rugged sparrow]

and its still failing to boot?

[quartz wave]

same thing happened so i just unplugged it for now

yea

[rugged sparrow]

thats concerning

I am gonna stick to testing on my own hardware now

i don't want to break anyone's stuff

[maiden blaze]

smh its like they dont want us to golf anymore

[gleaming linden]

fyi this has come up multiple times here before

[maiden blaze]

oh ok

[quartz wave]

!e also ```py
a = True
print(~a)

[night quarry]

@quartz wave :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | /home/main.py:2: DeprecationWarning: Bitwise inversion '~' on bool is deprecated. This returns the bitwise inversion of the underlying int object and is usually not what you expect from negating a bool. Use the 'not' operator for boolean negation or ~int(x) if you really want the bitwise inversion of the underlying int.
002 |   print(~a)
003 | -2

[quartz wave]

no worries

1 character isn't that much.. right?

[arctic skiff]

if there is a sign there is a story behind it

[arctic kettle]

is there a way to return something custom when I use something like MyClass and Myclass and MyClass or MyClass

rather not have it be very weird code just thought this is best place to ask

[restive void]

Not really. You can customize which side should be returned (by implementing __bool__), but not much more.

[versed eagle]

do you care if the object gets overwritten?

because you could have __bool__ check why it was called, and if it was because of an and or or operator, modify itself in memory to be a different object via ctypes
you'd have to be careful about the size of the object you return, but it should be possible

[earnest snow]

that's straight up trying to make python unsafe

[restive void]

You are aware what channel you are in? :)

[earnest snow]

I'm making a whole python library in one line, of course I am

[versed eagle]

wdym

its not going to bite you

you're perfectly safe

[earnest snow]

Memory safe by replacibg memory addresses using ctypes

[versed eagle]

its memory safe if you do it right

:3

[earnest snow]

Everything is memory safe if you do things right, the point is that you can mess everything up

[versed eagle]

Everything is memory safe if you do things right
exactly

people should stop making mistakes so often

[fleet bridge]

memory unsafe stuff is perfectly fine for this channel

[rugged sparrow]

@earnest snow would it make you feel better if we implement it by swapping out the return value on the frame stack itself?

i have demo code for that

[earnest snow]

The or/and thing? That's probably a better idea yeah

[rugged sparrow]

I'll write a proof of concept

[versed eagle]

actually thats probably a lot better than my idea

less destructive

[twilit igloo]

py

[quartz wave]

!e ```py
from fishhook import hook

@hook(str)
def eq(s,o):return 1

exec(r'''
if "School" == "Hell":
print("fuck")
else:
print('still bad but better I guess')
''')

[night quarry]

@quartz wave :white_check_mark: Your 3.12 eval job has completed with return code 0.

fuck

[twilit igloo]

LAMAO

[turbid dragon]

Why in an exec?

[long fulcrum]

any idea how to golf this further

import sys
D=-1,0,1
E=enumerate
for Y,l in E(A:=sys.argv[1].split('\n')):print(''.join('.#'[4+('.'>c)>sum(len(l)>X+x>-1<Y+y<len(A)!='.'>A[Y+y][X+x]for y in D for x in D)>2]for X,c in E(l)))```

it simulates the next generation of a conways game of life game

[fleet bridge]

because of optimizer

[quartz wave]

optimizer doesn't optimize comparisons

i just tested it out in #bot-commands after that

[fleet bridge]

interesting...

[quartz wave]

i just had to do it ```py
ordinal=lambda n:str(n)+'tsnrhtdd'[(n//10%10+3!=4>n%10)*n%10::4]

[versed eagle]

string munging is so fun

[unreal echo]

Just free the memory properly 🗣️🗣️🔥

[weak mist]

https://github.com/disentcorp/dateroll sweet library to make date math easier

[versed eagle]

exactly

[tight remnant]

hi

i am new

i am new to coding

i love python

[arctic skiff]

if you are new, you aren't supposed to be in this channel because this channel is filled with horrible nightmares of python, you should explore other channels like #python-discussion

!res for you

[night quarry]

Resources

The Resources page on our website contains a list of hand-selected learning resources that we regularly recommend to both beginners and experts.

[gentle plover]

Welcome to the best channel, where you get 100% taught real, correct and idiomatic Python

[versed eagle]

here we do things the Pythonic way

[unique heath]

daterewind

bitter library to make date english harder

[patent walrus]

I need to track the existence of an extension type that doesn't allow me to create weak references to it. Any ideas?

Background is, I'm working with pyodbc.Connection objects, and something somewhere is leaking them. I've been pulling my hair out trying to figure out what, so I'd like to track the existence of those objects directly to figure out which ones are sticking around longer than they are supposed to. The obvious way to do this would be with weak references, but when I try I get cannot create weak reference to 'pyodbc.Connection' object, so that's out. Is there some other way? I could save the id of each Connection object, which is just the memory address IIUC, but is there a way to safely inspect that memory address later on and determine whether this is still the same object?

I can't assign new properties to the Connection object the usual way, but maybe there's a hacky way around that? Then I could attach a marker object to the Connection and just watch for it to be cleaned up?

[fleet bridge]

you can override pyodbc.Connection.__del__ and implement some logic there

[patent walrus]

I can't seem to override properties on either the instance or the class itself

and I don't think I can subclass it because there's no way to tell pyodbc to give me a different class when I ask it for a connection

ok, just ran across this, which might constitute selling my soul to the devil but will also probably work? https://stackoverflow.com/questions/24497316/set-a-read-only-attribute-in-python/24498525#24498525

[fleet bridge]

have you tried fishhook or forbiddenfruit ?

[rugged sparrow]

Fishhook will work for extension classes, might even let you add a __weakref__ property to enable weakrefs

[patent walrus]

Good to know, thanks

I was not aware of those libs, glad I asked 😁

[fleet bridge]

in normal cases __weakref__ is a slot, and slots require some space in object struct
i dont think you can add legit slots with fishhook, but adding __weakref__ descriptor should be possible (but it would require some external storage, because there is no space in the object itself)

[rugged sparrow]

Yea I meant the descriptor, I haven't made fishhook so cursed that it allocs new slotvars quite yet

[unique heath]

do it

[digital mesa]

Thought of a fun one: programs without literals

[versed eagle]

there arent any cases i can think of where that would affect things overly much

[formal sandal]

Here's a fun question. Is it possible to create a generator object or a coroutine object with a single expression using these rules?

• The only available builtins are: str, int, float, randint (from random) and foo (a custom Python function)
• Attributes starting with func_ or __ are not allowed
• Attributes format and format_map are not allowed
• Lambdas are not allowed
• Generator expressions are not allowed

I know this is weirdly specific... it is

[versed eagle]

i think disallowing generator expressions is redundant since they're already disallowed by containing multiple subexpressions

[formal sandal]

No no, subexpressions are allowed

So str(str(str([x for x in [1, 2, 3]]))) is good

[versed eagle]

with a single expression

[formal sandal]

That's a single expression, a function call

I meant that you can't have import, def, or just multiple separate expression statements

sorry if that was unclear

[versed eagle]

oh
i thought you meant grammar expressions

lol

[formal sandal]

So what you can do is create literals, use operators, and use methods/attributes that meet the above rules

[unique heath]

til you can make a generator without a genexpr or import

[cloud fossil]

You mean, the good ol' yield inside a function?

[unique heath]

oh

do i have long-term memory loss

[fleet bridge]

do i have short-term memory loss

do i have short-term memory loss

[formal sandal]

Oh I don't know if it's actually possible, I want to find out 🙂

[unique heath]

lmao

[unique heath]

lmao

[quartz wave]

i don't think it's possible with those restrictions

[quartz wave]

with lambda yes

[formal sandal]

huh?

Oh yeah

[quartz wave]

with no lambda there is nothing we can do

[formal sandal]

!e

print(lambda: (yield))

[night quarry]

@formal sandal :white_check_mark: Your 3.12 eval job has completed with return code 0.

<function <lambda> at 0x7fd42ad22a20>

[quartz wave]

what does foo do

[formal sandal]

def foo(x):
    """docstring"""
    return int(random.random()*x)

[quartz wave]

dynamic literals ([], (), {}, "")

[formal sandal]

list(), tuple(), dict(), str()

[quartz wave]

yeah but constructing them with elements will take more lines of code than usual

[formal sandal]

yeah that's true

1 is (list()==list()), and you can get strings with chr

so string literals will be kind of brainfuckesque

[quartz wave]

t = 1, 2, 3
``` `1, 2, 3` here is technically a literal

i mean this sort of works ```py
T = lambda *args: args
t = T(1, 2, 3)

assuming 1, 2, and 3 are substituted with their non-literal equivs

here's how to create a list ```py
*t, = T(1, 2, 3)

[twilit igloo]

print((1, 2, 3))

ez

[arctic skiff]

that just writes it to stdout that doesn't return it

[versed eagle]

equivalent ones can be constructed without literals

[grave grail]

Are there any way to get eval(not in string, in function) without using compile, exec, eval and use as few amount of alphabet character as possible

[bright valve]

programming warcrimes in yt

[grave grail]

Unrelated

[tranquil yarrow]

that's how i wrote a print("Thanks!") program ```py
import collections
import contextlib
import io
import itertools

with contextlib.redirect_stdout(this := io.StringIO()):
import("this")

(*(0, 1, 36, 57, 286, 296, 855),) # ?????

magic = collections.Counter(loader.get_code(name).co_consts[~3])
print(*itertools.compress(this.getvalue(), map(magic.get, itertools.count())), sep="")

[tranquil yarrow]

are imports allowed?

[formal sandal]

no, since import x is not an expression

[fleet bridge]

i especially like the (*(0, 1, 36, 57, 286, 296, 855),) # ????? part ❤️

[tranquil yarrow]

but i meant __import__().

i guess it's not in the

[formal sandal]

that's not in the "builtins" so that's not available

[tranquil yarrow]

"only allowed" builtins

ok

xd

[fleet bridge]

btw if you dont have __import__, then import statement would not work either

[tranquil yarrow]

[tranquil yarrow]

is random defined in the scope of foo?

[fleet bridge]

no, only randint

[tranquil yarrow]

so that function basically doesn't work, interesting

[fleet bridge]

wait, are we allowed to write code in foo?

[tranquil yarrow]

excluding attributes starting with __ makes this so much harder :D

[tranquil yarrow]

i believe it's a builtin

we aren't

it is possible to create a generator object if only str is directly available, but dunder attrs ain't disallowed xd

i'll actually dump my thinking here

as i solve that challenge

[fleet bridge]

>>> print(*[(i,c) for i,c in enumerate(str.__base__.__subclasses__())],sep='\n')
(0, type)
(1, async_generator)
(2, bytearray_iterator)
(3, bytearray)
(4, bytes_iterator)
(5, bytes)
(6, builtin_function_or_method)
(7, callable_iterator)
(8, PyCapsule)
(9, cell)
(10, classmethod_descriptor)
(11, classmethod)
(12, code)
(13, complex)
(14, _contextvars.Token)
(15, _contextvars.ContextVar)
(16, _contextvars.Context)
(17, coroutine)
(18, dict_items)
(19, dict_itemiterator)
(20, dict_keyiterator)
(21, dict_valueiterator)
(22, dict_keys)
(23, mappingproxy)
(24, dict_reverseitemiterator)
(25, dict_reversekeyiterator)
(26, dict_reversevalueiterator)
(27, dict_values)
(28, dict)
(29, ellipsis)
(30, enumerate)
(31, filter)
(32, float)
(33, frame)
(34, frozenset)
(35, function)
(36, generator)
...
``` just take 12th `object` subclass, create a `code` object out of it, then make `function` object (35th `object` subclass), then call it to get a generator

i dont think this challenge is possible

[unique heath]

you cant use dunders

[fleet bridge]

obviously

[night quarry]

:incoming_envelope: :ok_hand: applied timeout to @tranquil yarrow until <t:1712968871:f> (10 minutes) (reason: newlines spam - sent 126 newlines).

The <@&831776746206265384> have been alerted for review.

[bronze escarp]

!unmute 712654007876976750

[night quarry]

:incoming_envelope: :ok_hand: pardoned infraction timeout for @tranquil yarrow.

[bronze escarp]

!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 Paste! button in the bottom left, or by pressing CTRL + S. After doing that, you will be navigated to the new paste's page. Copy the URL and post it here so others can see it.

[bronze escarp]

pls

[tranquil yarrow]

thanks, sorry.

[tranquil yarrow]

yes, it is possible, but on Python <3.8 (the question didn't state what exactly the version has to be and whether it must be before EOL).

Python 3.7 example generator object created with a single expression using the rules:

[(yield) for _ in ()]  # <generator object <listcomp> at ...>

the ability of using yield expressions inside implicit scopes (which includes list/dict/set comprehensions) was removed in 3.8.
moreover, afaik list/dict/set comprehensions no longer create implicit scopes since Python 3.12 due to inlined comprehensions.

on Python 3.8+, here's my modest progress as for now. cc @formal sandal

after a closer look, I observed a few things.

first, I assume some of the rules are just to distract.
there are no func_* attributes that could possibly help solving the problem.
I am also not sure how <I guess str>.format or <I guess str>.format_map would help, but I didn't check.
foo seems completely useless either and it was not clear whether it even worked (is random defined in foo.<locals>?)
I assumed we might somehow get something useful from the attributes matching the constraints, but I didn't find anything useful:

https://paste.pythondiscord.com/YBWA
as you can see, nothing interesting here.
if you have any more ideas/hints where I can explore further, let me know.

cc @fleet bridge

[fleet bridge]

i think you forgot about .mro()
this allows you to get reference to object
not sure what to do with it though

>>> int.mro()
[int, object]
>>> int.mro()[1]
object

[tranquil yarrow]

i don't see any useful attributes in object.

[fleet bridge]

yeah, there are only dunders and .mro() once again

[tranquil yarrow]

regardless, mro() was a good catch!

I also tried to check for some existing generators in gc.get_objects() and determine where they come from.
I found some spots in the stdlib where I can create generators, but they are not accessible in any legal (as in the question) way.
I was hoping I will get there through randint, but I can't.
nothing really useful there.

@formal sandal is random defined at the global scope?

that would probably simplify a lot

because then I could use random._os.walk()

and voilla, we have a generator.

[unique heath]

WAUT THAT WASNT A DUNDER>

[tranquil yarrow]

it's both a type method and dunder attribute

[fleet bridge]

there is a __mro__ descriptor and mro method on classes

[fleet bridge]

>>> [*bytes.maketrans(b'', b'')] == [*range(256)]
True

https://paste.pythondiscord.com/WXBQ

this code ^ also observes objects that are results of calling methods

i would also recommend adding this somewhere before the code ^, it adds nice reprs to some builtin objects that dont have nice reprs: https://paste.pythondiscord.com/HYIA

{self: str,
 'capitalize': {self: str.capitalize, str.capitalize(): ''},
 'casefold': {self: str.casefold, str.casefold(): ''},
 'center': {self: str.center},
 'count': {self: str.count, str.count(, ): 1},
 'encode': {self: str.encode, str.encode(): b''},
 'endswith': {self: str.endswith,
              str.endswith(, ()): False,
              str.endswith(, ): True},
 'expandtabs': {self: str.expandtabs, str.expandtabs(): ''},
 'find': {self: str.find, str.find(, ): 0},
 'index': {self: str.index, str.index(, ): 0},
 'isalnum': {self: str.isalnum, str.isalnum(): False},
 'isalpha': {self: str.isalpha, str.isalpha(): False},
 'isascii': {self: str.isascii, str.isascii(): True},

>>> object.__dict__
mappingproxy({   '__new__': object.__new__,
                 '__repr__': sc._@__repr__,
                 '__hash__': object.__hash__,
                 '__str__': object.__str__,
                 '__getattribute__': object.__getattribute__,
                 '__setattr__': object.__setattr__,
                 '__delattr__': object.__delattr__,
                 '__le__': object.__le__,
                 '__eq__': object.__eq__,
                 '__ne__': object.__ne__,
                 '__gt__': object.__gt__,
                 '__ge__': object.__ge__,
                 '__init__': object.__init__,
                 '__reduce_ex__': object.__reduce_ex__,
                 '__reduce__': object.__reduce__,
                 '__getstate__': object.__getstate__,
                 '__subclasshook__': object.__subclasshook__,
                 '__init_subclass__': object.__init_subclass__,
                 '__format__': object.__format__,
                 '__sizeof__': object.__sizeof__,
                 '__dir__': object.__dir__,
                 '__class__': <slot object.__class__>,
                 '__doc__': 'The base class of the class hierarchy.\n\nWhen called, it accepts no arguments and returns a new featureless\ninstance that has no instance attributes and cannot be given any.\n',

[rugged sparrow]

You could use str.format to access any attribute inside of foo's global scope, but repr/str will be called on it

[fleet bridge]

unless getting this attribute has side effect, it is useless
we can get arbitrary strings by just writing string literal

also, *.format is forbidden

[rugged sparrow]

Oh welp

[sick hound]

Can sm1 help me with an phyton project

[quartz wave]

yeah but much would change by then

[quartz wave]

not equivalent but whatever makes you happy i guess

[quartz wave]

do coroutines count?

oh wait i guess they aren't generators by themselves

but they do share the same object structure as generators..

[long fulcrum]

anyone have an idea on how to golf this further

import sys
for b in sys.argv:
    r=[a:=0]*99
    for i in b:b+='\n'+' '*a+'a+=1 a-=1 r[a]+=1 r[a]-=1 print(end=chr(r[a])) while-r[a]: 0'.split()[n:='><+-.['.find(i)];a+=n//5
    exec("#"+b)```

ive been trying to make the big string shorter recently but cant figure out anything

[formal sandal]

no

[formal sandal]

Coroutines do count yes

actually let me add that

[median cloak]

wtf am i seeing

[finite blaze]

first time here?

[left hinge]

every time i consider trying hard drugs, i look at this channel to scare myself out of it

[median cloak]

nah

its like acid

trippy af

[formal sandal]

Oh that's very interesting. Good finding

[unique heath]

lmao

[unique heath]

decrease indentation i guess

[long fulcrum]

indentation is tabs discord just automatically changes it to spaces

[dire galleon]

8

[long fulcrum]

?

[languid hare]

this is for that sandbox thing i assume?

[formal sandal]

yeah 🙂

[modest bough]

@fleet bridge > bytes.maketrans(b'', b'')
that was fun

[fleet bridge]

thanks 😊

i am tempted to write code that will do random stuff with random objects and keep track of results, but im scared it will rm -rf / my machine 🙂

since i want to get results, killing a system every time is not a good option
even VM wont help much

is it possible to fork the process into different physical machine? 🤔

[bright valve]

you could do a vm and clone it each time you brick it

or wsl

you can make copies easily

[fleet bridge]

well, if i brick it before i observe useful results, then it is useless

[bright valve]

hmm

idk what to suggest then

[rugged sparrow]

You could use something like qemu

[versed eagle]

what do you mean by [do random stuff with random objects]

[fleet bridge]

literally that
get some attributes, call objects, construct new objects, etc...
it might be useful in challenges, where some restrictions are present. And such tool can help us observe space of possibilities

[versed eagle]

why do you worry about removing your root directory then

as long as you're not calling a few specific functions you'd be fine

[fleet bridge]

the whole point of the thing is to not do the search manually

and i dont think it is possible to reliably tell if operation X will blow your system

even simple a.b can invoke a descriptlr, which will call os.system(...)

[versed eagle]

you dont need to reliably tell

you just need to unconditionally block things which could blow your system

e.g. os.system

[versed eagle]

probably os.* actually

(really posix.*!)

(or whatever the windows equivalent is i guess)

[fleet bridge]

ok
but should i block open(...)? pathlib.*? socket.*? ctypes? _ctypes? threading? multiprocessing?
ok, what about normal stuff? a:int; b: int; s: str; l: list: a*b? a**b? a<<b? s*a? l*a?

fyi, it is theoretically possible to crash the process with ast.safe_eval (or rather anything that performs code parsing, since parsing untrusted input can crash)

and in this case, everything is "untrusted", because it comes from unknown source and can be of any value (because it is constructed randomly from random objects)

[versed eagle]

block things that read from or otherwise interact with the file system or processes and such in a noncontrolled way

[fleet bridge]

second row of examples don't necessarily destroy the system, but they can take up entire memory of your machine

[versed eagle]

but they can take up entire memory of your machine
that can be prevented

so can all of the first row too, but it'd be more work probably since there are more of them

[fleet bridge]

how would you block that? with trace function? it can be messed with/deleted/...

[versed eagle]

you can tell the operating system to not give you very much memory
hook malloc to fail early
etc.

[fleet bridge]

>>> import _testcapi
>>> _testcapi.crash_no_current_thread()
Fatal Python error: PyThreadState_Get: the function must be called with the GIL held, after Python initialization and before Python finalization, but the GIL is released (the current Python thread state is NULL)
Python runtime state: initialized

Thread 0x000040cc (most recent call first):
  File "D:\GitHub/cp_exts\sc.py", line 175 in target
  File "D:\Programs\Python\3.12\Lib\threading.py", line 989 in run
  File "D:\Programs\Python\3.12\Lib\threading.py", line 1052 in _bootstrap_inner
  File "D:\Programs\Python\3.12\Lib\threading.py", line 1009 in _bootstrap

Current thread 0x00001254 (most recent call first):
  File "<stdin>", line 1 in <module>

Extension modules: _xxsubinterpreters, _xxinterpchannels, psutil._psutil_windows, _wmi, _testcapi (total: 5)

[process exited with code 3221226505 (0xc0000409)]

[versed eagle]

lots of ways to do things

[versed eagle]

i dont know what this is meant to show

that you should also block _testcapi i guess?

[fleet bridge]

that there are too many ways to do stuff, and it is pretty hard to block them all

[versed eagle]

pretty hard, yes
but you wanted to prevent it from deleting your files, not from crashing the process
the one has a lot less ways of doing things than the other

ultimately to read files, it's going to have to execute a syscall

[rugged sparrow]

is your goal to determine code constructs with odd side effects?

[fleet bridge]

my goal is to automatically solve challenges like this ^

[rugged sparrow]

ah

in my experience challenges like that depend on such specific side effects that you need to be very specific with them

idk if a random generator is gonna find successful techniques

[versed eagle]

i think randomly generating would be less useful than a breadth-first exhaustive search
obviously it'd go on forever but you might find some interesting stuff after 2 or 3 layers

[rugged sparrow]

fair fair

[rugged sparrow]

also do we know if that challenge is actually possible?

[rugged sparrow]

does this challenge actually have a solution (on newer python versions that disallow [(yield) for () in ()])?

[versed eagle]

i dont believe so

[formal sandal]

It's not a riddle, it's more of an exploratory question 🙂

There's a small but somewhat popular library that I've kind of... deconstructed, and it could have a serious vulnerability if it's possible to complete the challenge

Luckily the listcomp trick doesn't work on it

[rugged sparrow]

If you're willing to DM me the library, there may be other side effects that can be abused to bypass the syntax whitelist you gave

[covert rivet]

!e

match = 1
case = 1

match match:
    case case:
        print("match case")

[night quarry]

@covert rivet :white_check_mark: Your 3.12 eval job has completed with return code 0.

match case

[covert rivet]

should be illegal

[fleet bridge]

match case:
 case match:
  ...

[covert rivet]

and you made it from illegal to war crime I think

[unique heath]

lloool

[quartz wave]

!e ```py
match = 7
case = 5

match match-case:
case match:
print("ooh")

print(case-match)

[night quarry]

@quartz wave :white_check_mark: Your 3.12 eval job has completed with return code 0.

001 | ooh
002 | 3

[vivid rock]

the syntax highlighting hurts

[quartz wave]

ok i got my pc back

the changes don't work btw

it's the last thing i tried before my pc entered limbo

[rugged sparrow]

I figured

[crude raft]

if only this could be done for all keywords

[restive void]

We had a discussion a while back, about which keywords could be made soft. Turns out it's almost all of them.

[lament pond]

l=[9,5,7,4,0,2,6,3,1,8]
while any(l[x]<l[x-1]and(a:=l[x],l.insert(x-1,a),l.pop(x+1))for x in range(1,len(l))):print(l)```
Bubble sort successfully golfed, what next?

[unique heath]

quick

[versed eagle]

blocksort could be fun

[torn cypress]

blobsort
ha, can't golf it if it doesn't exist
proceeds to invent (or rather, find? ) a new sorting algorithm to then golf it

[versed eagle]

blobsort seems like it would be a variation of bucket sorting

[magic wraith]

blobsort 🤔

[night quarry]

@magic wraith :white_check_mark: Your 3.12 eval job has completed with return code 0.

[2, 3, 5, 5, 6, 6, 9, 10, 13, 16, 26, 33, 44]

[arctic skiff]

make golfed bogo-bogo-bogo sort

[magic wraith]

was actual bogo sort ever golfed here?

!e py from random import* from itertools import* a=[7,6,1,3,8,4] while 1-all(i<=j for i,j in pairwise(a)):shuffle(a) print(a)

[night quarry]

@magic wraith :white_check_mark: Your 3.12 eval job has completed with return code 0.

[1, 3, 4, 6, 7, 8]

[magic wraith]

if imports are not negligible, then py while 1-all(i<=j for i,j in zip(a,a[1:])):shuffle(a)

[versed eagle]

if imports are negligible, then

while 1-all(map(ge,pairwise(a))):shuffle(a)

*le

actually it should be starmap also

[fleet bridge]

this is a surprisingly good RNG: py [id(x)%9 for x in [[]for _ in range(1000)]] ```py

n = 100_000
x = [id(x)%9 for x in [[] for _ in range(n)]]
c = collections.Counter(x)
{i: round(c[i] / n, 3) for i in range(9)}
{0: 0.111, 1: 0.111, 2: 0.111, 3: 0.111, 4: 0.111, 5: 0.111, 6: 0.111, 7: 0.111, 8: 0.111}

[quartz wave]

there seems to be a pattern in there

[fleet bridge]

>>> d = lambda x,p: (c:=__import__('collections').Counter(x),n:=sum(c.values()),{i: round(c[i] / n, 3) for i in range(p)})[-1]
>>> d([id(x)%9 for x in [[] for _ in range(10_000)]], 9)
{0: 0.111, 1: 0.111, 2: 0.111, 3: 0.111, 4: 0.111, 5: 0.111, 6: 0.111, 7: 0.111, 8: 0.112}
>>> d([id([])%9 for _ in range(10_000)], 9)
{0: 0.0, 1: 0.0, 2: 0.0, 3: 1.0, 4: 0.0, 5: 0.0, 6: 0.0, 7: 0.0, 8: 0.0}
>>> d([id(_ * _)%9 for _ in range(10_000)], 9)
{0: 0.337, 1: 0.0, 2: 0.001, 3: 0.325, 4: 0.337, 5: 0.0, 6: 0.0, 7: 0.0, 8: 0.0}
>>> d([id(_ ** _)%9 for _ in range(10_000)], 9)
{0: 0.013, 1: 0.073, 2: 0.052, 3: 0.057, 4: 0.193, 5: 0.315, 6: 0.212, 7: 0.013, 8: 0.072}
>>> d([id(_ ** _)%9 for _ in range(10_000)], 9)
{0: 0.121, 1: 0.028, 2: 0.14, 3: 0.024, 4: 0.068, 5: 0.239, 6: 0.233, 7: 0.104, 8: 0.044}

>>> d = lambda x,p: (c:=__import__('collections').Counter(x),n:=sum(c.values()),{i: round(c[i] / n, 3) for i in range(p)})[-1]
i ran this a bunch of times:
>>> d([id([[],[],[],[],[],[],[]]*_)%9 for _ in range(10_000)], 9)
this:
{0: 0.111, 1: 0.111, 2: 0.445, 3: 0.222, 4: 0.0, 5: 0.0, 6: 0.0, 7: 0.0, 8: 0.111}
gets printed as this (and also with less spaces):
      1         1         4         2         0       0       0       0       1
results:
1 1 4 2 0 0 0 0 1
2 1 1 2 0 0 1 1 1
3 0 1 1 0 1 0 2 1
1 0 1 1 0 2 1 1 2
1 0 1 0 0 2 1 1 3
1 0 1 1 0 2 2 1 1
1 0 2 1 0 2 0 1 2
0 0 2 1 0 2 1 1 2
1 2 1 1 0 2 0 1 1
0 3 0 2 0 1 2 0 1
1 2 0 3 0 0 1 0 2
0 2 0 2 1 0 1 0 3

i am doing this in repl, and repl performs a bunch of allocations. This is why results are changing

d = lambda x,p: (c:=__import__('collections').Counter(x),n:=sum(c.values()),{i: round(c[i] / n, 3) for i in range(p)})[-1]

for _ in range(10):
    x = d([id([[],[],[],[],[],[],[]]*_)%9 for _ in range(10_000)], 9)
    x = {i: round(v * 9) for i, v in x.items()}
    print(' '.join(map(str, x.values())))

# results (same every time, and not uniform):
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1
0 1 1 0 1 2 2 1 1

with some allocations to break the pattern: ```py
d = lambda x,p: (c:=import('collections').Counter(x),n:=sum(c.values()),{i: round(c[i] / n, 3) for i in range(p)})[-1]

for _ in range(10):
import gc
gc.collect()
import random
_ = [[] for _ in range(random.randint(1, 1000))]

x = d([id([[],[],[],[],[],[],[]]*_)%9 for _ in range(10_000)], 9)
x = {i: round(v * 9) for i, v in x.items()}
print(' '.join(map(str, x.values())))

results (different every time, more uniform):

1 1 1 1 1 1 0 2 1
0 2 1 1 1 1 1 1 1
0 2 1 1 1 1 1 1 1
1 2 1 0 0 2 1 2 0
2 1 1 1 1 1 1 1 0
0 1 1 1 1 1 1 1 2
1 1 0 2 1 1 1 1 1
1 2 1 1 2 0 1 1 0
1 1 1 1 1 1 1 0 2
0 2 1 1 1 1 1 1 1

[slate wasp]

Wonder what could be the most esoteric DFS implementation

Has to be a recursive one for sure idk how to make it more esoteric

def dfs(g, s, v=None): 
    return (v := v or set()).union(*(dfs(g, n, v | {s}) for n in g[s] if n not in v))

[gleaming linden]

while any(map(gt,a,a[1:])):shuffle(a)

[versed eagle]

right i forgot map can do that

thats funny

[lament pond]

for _ in(answer:=__import__("random").randint(1,100),g:=-1):(x:=lambda g=g:(g:=int(input("Guess: ")),(print("Higher"),x(g))if g<answer else(print("Lower"),x(g))if g>answer else print("Correct")))and x(g) guess the number game compacted into 1 line of code

[quartz wave]

does 2 games but okay otherwise

[lament pond]

oh yeah because of for loop

[karmic pumice]

(f:=lambda a,b=__import__('random').randint(1,100):(a>b)-(a<b),sum(iter(lambda:print(["Correct","Lower","Higher"][x:=f(int(input()))])or x,0)))

[pallid coral]

Not code golf, but still pretty weird. Without running the code, what do you think this would output?

# 1
x = []
y = []

x.append(y)
y.append(x)

try:
    print("1")
    print(x == y)
except Exception as e:
    print(e.__class__.__name__)
print(f"x: {x}; y: {y}\n")

# 2
x = []
y = []

x.append(x)
y.append(y)

try:
    print("2")
    print(x == y)
except Exception as e:
    print(e.__class__.__name__)
print(f"x: {x}; y: {y}\n")

# 3
x = []
y = []

x.append(y)
y.append(y)

try:
    print("3")
    print(x == y)
except Exception as e:
    print(e.__class__.__name__)
print(f"x: {x}; y: {y}\n")

Answer:
||```
1
RecursionError
x: [[[...]]]; y: [[[...]]]

2
RecursionError
x: [[...]]; y: [[...]]

3
True
x: [[[...]]]; y: [[...]]

[fleet bridge]

my attempt: ||```py
1
RecursionError (or RubtimeError)
x: [[[...]]]; y: [[[...]]]
2
same error as before
x: [[...]]; y: [[...]]
3
True
x: [[[...]]]; y: [[...]]

yay, im correct

[pallid coral]

Lol, maybe it wasn’t that great of a gotcha

[versed eagle]

the hardest part is telling x and y apart

dyslexia moment

[pallid coral]

What’s your rationale as to why the last one ||returns True||?

[fleet bridge]

because i remembered that list.__eq__ has a fast way for cases where self is other

so, roughly, x == y first performs x is y check, it fails, then it compares elements: x[0] == y[0] which is equivalent to y == y, which can be evaluated immediately because y is y

[fleet bridge]

RubtimeError
just noticed 😄

[pallid coral]

Ah that’s interesting, so it still does the fast check for nested elements. It was confusing behavior for me, thank you

[fleet bridge]

!e ```py
nan = float('nan')
print([nan] == [nan])

[night quarry]

@fleet bridge :white_check_mark: Your 3.12 eval job has completed with return code 0.

True

[fleet bridge]

yeah, it first does is check and only then ==

[quartz wave]

||```
1
<can't figure this out>

2
<can't figure this out>

3
True
x: [[[...]]]; y: [[...]]

i don't get how python makes a ||RecursionError||

oh

i missed a crucial C api function call

[versed eagle]

|| isnt list eq basically just all(map(operator.eq, self, other)) ||

[quartz wave]

nope

it's merged with __lt__ etc. so

def list_richcmp(self, other, op):
    if not isinstance(self, list) or not isinstance(other, list):
        return NotImplemented

    if len(self) != len(other) and (op == _EQ or op == _NE):
        # Shortcut: if the lengths differ, the lists differ
        return op == _NE

    # Search for the first index where items are different
    for x, y in zip(self, other):
        if x is y:
            continue

        k = richcmp(x, y, _EQ)
        if not k:
            break
    else:
        # No more items to compare -- compare sizes
        # (Py_RETURN_RICHCOMPARE() expanded and optimized below)
        res = len(self) == len(other)
        if op == _EQ:
            return res
        if op == _NE:
            return not res
        res2 = len(self) > len(other)
        if op == _GT:
            return not res and res2
        if op == _GE:
            return res or res2
        if op == _LT:
            return not res and not res2
        if op == _LTE:
            return res or not res2

    # We have an item that differs -- shortcuts for EQ/NE
    if op == _EQ:
        return False
    if op == Py_NE:
        return True

    # Compare the final item again using the proper operator
    return richcmp(self, other, op)

[versed eagle]

oh right the whole rich comparison thing

[quartz wave]

simplified for equals ```py
def list_eq(self, other):
if not isinstance(self, list) or not isinstance(other, list):
return NotImplemented

if len(self) != len(other):
    # Shortcut: if the lengths differ, the lists differ
    return False

# Search for the first index where items are different
for x, y in zip(self, other):
    if x is y:
        continue

    if not x == y:
        break
else:
    # No more items to compare -- compare sizes
    return len(self) == len(other)

# We have an item that differs -- the lists aren't equal
return False

[versed eagle]

||its this, right? ```py
k = richcmp(x, y, _EQ)
if not k:

[versed eagle]

i never understood why they didn't just have them be separate