esoteric-python

[fleet bridge]

Use .__iadd__() 🤯

I heard that it is faster because it copies less

[mortal grotto]

This is actually pretty new, was working fine in 3.9 as far as I recall correctly

[distant salmon]

Really doubt that. I think its even in Python2

Python really wants to keep track of which variables are local or not

Ye I just checked, it is an error even as far back as Python 2.7.18 (and probably much further back than that)

[mortal grotto]

Then I messed something up

[burnt pasture]

this is trying to update a global in a function. you should avoid that in any case.

[burnt pasture]

definitely don't use .__iadd__(), though this is esopy, so all the rules are out the window

[distant salmon]

It is relatively common to see people do .append to a non-local list. I don't think there is anything wrong with that

[burnt pasture]

i don't know what you mean by "relatively common". Modifying globals from within functions is rare.

[distant salmon]

I'm talking about Python implementations of algorithms and data structures. I would say it is a relatively common to see people do .append to non-local lists there.

A basic example would be building graphs

For example this is a natural way to build directed graphs with capacities (used for max-flow)

V = []
C = []
F = []
def add_edge(u, v, cap, rcap = 0):
    V.append(v)
    V.append(u)
    C.append(cap)
    C.append(rcap)

[burnt pasture]

maybe we are talking about different things: small self-contained examples vs production code.

[distant salmon]

I'm definitely the small self-contained examples kind of guy

[burnt pasture]

ok, that explains it.

[earnest wing]

my production code is exclusively global variables using interior mutability, also dunders

[quartz wave]

well there's a reason
global A
exists

[burnt pasture]

makes a note to avoid Olivia's code :)

[quartz wave]

although if the functional exact equivalent has to add this in order to do whatever you're doing then whatever you're doing is probably not the best idea for good code

[vast wave]

shit like this is why i don't enjoy python as much as other languages

the scope definition of certain things is wonky at best

[mortal grotto]

python often tries to implement some things, which prohibit unwished behaviour. Same as sum(Iterable[str], begin="")

So it would work, but is a bad practice and so python says nuh uh

[restive void]

!e just as bad and not prevented :)

from functools import *

x = ["foo", "bar", "bat"]
print(reduce(str.__add__, x))

[night quarry]

@restive void :white_check_mark: Your 3.11 eval job has completed with return code 0.

foobarbat

[true basalt]

!e this is what I call the f'in hell

import time
_f = time.sleep
_f_ = int
f = _f_(f"5")
_f(_f_(_f_(_f_(f"{f'{_f_(f)}'}"))))

[night quarry]

@true basalt :warning: Your 3.11 eval job has completed with return code 0.

[No output]

[mint rain]

!e

A = [0,0,0,0]
x = A[x] = 3
print(A)```

[night quarry]

@mint rain :white_check_mark: Your 3.11 eval job has completed with return code 0.

[0, 0, 0, 3]

[sonic birch]

Wait..what? Why it does not raise an error?

[mint rain]

this is why i love python

only learned this trick bc i saw it somewhere up in this channel

x[i] = i idk when ill use it tho

[sonic birch]

Even chatgpt is confused) Looks like Mandela effect to me

Oh, I wouldn't be so sure:)

[turbid dragon]

!e

A = [0,0,0,0,0,0,0,0]
x = A[x] = 3
print(A)```

[night quarry]

@turbid dragon :white_check_mark: Your 3.11 eval job has completed with return code 0.

[0, 0, 0, 3, 0, 0, 0, 0]

[turbid dragon]

ahhh

[sonic birch]

Guys, does anybody know a way how to override getting attribute in super class (when calling super().attribute?
I tried

class Meta(type):

  def __getattr__(cls, key):
    print('meta getattribute', key)

    return None


class Parent(metaclass=Meta):
  def __getattr__(cls, key):    
    print('parent getattribute', key)

    return None

class Child(Parent):
  def foo(self):
    super().foo()

child = Child()

child.foo()

But I get AttributeError: 'super' object has no attribute 'foo'

[burnt pasture]

is this to solve a real problem, or to do esoteric things?

[sonic birch]

It is real. I want to have a cooperative parent class (aka root) which I always can include as a last Mixin, so if a sibling calls super().something() the code will not crash

[burnt pasture]

hmm, i'm not sure how to do that.

[lime wave]

how can I space words in python

import re

text = "hellobrotherhowareyou"

# Split the string into words based on camelCase or PascalCase conventions
split_text = re.findall(r'[A-Za-z][a-z]*', text)

# Join the words with a space between them
spaced_text = " ".join(split_text)

print(spaced_text)
``` I asked an ai but it  just printed "hellobrotherhowareyou"

[burnt pasture]

read the comment it included for you. Is that what you wanted?

[burnt pasture]

also, who is to say what are words? These are both reasonable answers: "hello brother how are you" and "hello bro the rho ware you"

[karmic pumice]

!e

(hgjntyr5bu6f56htgr79b8nyfj65htg79bn8jryf:=lambda:(u7oiyghtvbu89hybg7ijo:={},bhjnm_kg:=3,EeeeeeeeeEeeeeeeeeeEeeeeeeeeeeEeeeeeeeeeeeeeeeEeeeeeeeeeeeEeeeeeeeeeEeeeeeeeeEeeeeeeeee:=5,Htgjryu5f6i:="fizz",V5c67bgFsuxizjtd3475Vvc:="buzz",u7oiyghtvbu89hybg7ijo.__setitem__(bhjnm_kg, Htgjryu5f6i),u7oiyghtvbu89hybg7ijo.__setitem__(EeeeeeeeeEeeeeeeeeeEeeeeeeeeeeEeeeeeeeeeeeeeeeEeeeeeeeeeeeEeeeeeeeeeEeeeeeeeeEeeeeeeeee, V5c67bgFsuxizjtd3475Vvc),func_func_func_for_func_funcfuncfunc_vbcgdfxhe4:=lambda S7c4ALBz8WjVBNse5raIrfKzyrGwiDXAWjUmuwsG: ''.join([u7oiyghtvbu89hybg7ijo[DdzlL7BsfRlW0TqER0HV5JDd3j7uu750AefzOB2k] for DdzlL7BsfRlW0TqER0HV5JDd3j7uu750AefzOB2k in u7oiyghtvbu89hybg7ijo if not(S7c4ALBz8WjVBNse5raIrfKzyrGwiDXAWjUmuwsG%DdzlL7BsfRlW0TqER0HV5JDd3j7uu750AefzOB2k)]) or str(S7c4ALBz8WjVBNse5raIrfKzyrGwiDXAWjUmuwsG),list(map(print, map(func_func_func_for_func_funcfuncfunc_vbcgdfxhe4, range(1, 101)))),None)[-1])()

[night quarry]

@karmic pumice :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 1
002 | 2
003 | fizz
004 | 4
005 | buzz
006 | fizz
007 | 7
008 | 8
009 | fizz
010 | buzz
011 | 11
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/F5OP6MVIZRMK5CC2RG4WRCO3QU

[vast wave]

time.sleep(5)

[vast wave]

your regex looks for an uppercase letter, then any amount of lowercase letters

so HelloBrotherHowAreYou would be processed correctly

splitting a string just based on words is hard if not impossible to do without ai

we as humans have previous experience with words and, more importantly, misspelling of words, which is why you can read hellobroterhowru

how do you tell a computer to read that

just taking a wordlist and using that to split the string also wouldnt work properly, since certain words in that wordlist may appear in the string you want to split, where it contextually wouldnt make sense

eg hireggie might be recognized as "hire ggie"

that's the entire point of even using ai, letting the computer figure out how to do it for us

so ai might be your best bet here, but it'd still be hard to train and use a neural net in the first place

[restive void]

The task is also fundamentally impossible to do unambiguously, AI or not. There are many strings that can be split in several legitimate word lists.
E.g. "nappythonghost" could be nap-python-ghost or nappy-thong-host.

[odd moat]

also the regex from the ai's answer would not do that, the corect one would be: '[A-Z][a-z]*' (i think)

[restive void]

The one they use would work as well, and it could then also split someCamelCaseWords

[odd moat]

o really? man regex is confusing

[restive void]

It's because * is greedy by default. If it was *? you'd be right

(but then it wouldn't correctly get the words)

[sonic birch]

!e

def wrapper(*args, **kwargs):
    return None

def root():
    class MetaRoot(type):
        def new(cls, *args, **kwargs):
            if args:
                for c in args[1][:-1]:
                    methods = [m for m in c.dict if not m.startswith('')]
                    for m in methods:
                        setattr(args[1][-1], m, wrapper)
            return super().__new(cls, *args, **kwargs)
    return MetaRoot('Root', (), {})

class Parent:
    pass
class Child:
    def foo(self):
        return super().foo()
class Mixed(Child, Parent, root()):
    pass
m = Mixed()
print(m.foo())

The only way I have found till now is this

UDP: strange, in works in python 3.10 on my local

[night quarry]

@sonic birch :x: Your 3.11 eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "/home/main.py", line 23, in <module>
003 |     print(m.foo())
004 |           ^^^^^^^
005 |   File "/home/main.py", line 19, in foo
006 |     return super().foo()
007 |            ^^^^^^^^^^^
008 | AttributeError: 'super' object has no attribute 'foo'

[past oak]

what on earth

[brazen geyser]

if you're feeling adventurous, could try replacing super 🙃

something like this to start off with

!e

import builtins
import sys

def patched_super():
  caller_locals = sys._getframe().f_back.f_locals
  self = caller_locals['self']
  cls: typing.Type = caller_locals['__class__']
  parent = cls.mro()[1]
  parent_getattr = parent.__getattr__

  class Proxy:
    def __getattribute__(proxu_self, item):
      return parent_getattr(self, item)

  return Proxy()

builtins.super = patched_super

class Parent:
  def __getattr__(self, item):
    print('parent getattr', item)
    return None

class Child(Parent):
  def foo(self):
    super().foo()

child = Child()

child.foo()

[night quarry]

@brazen geyser :x: Your 3.11 eval job has completed with return code 1.

001 | parent getattr foo
002 | Traceback (most recent call last):
003 |   File "/home/main.py", line 30, in <module>
004 |     child.foo()
005 |   File "/home/main.py", line 26, in foo
006 |     super().foo()
007 | TypeError: 'NoneType' object is not callable

[restive void]

Make sure to really call the patched super super though or it will not work ;)

[sonic birch]

You suggest overridinng built-in's super? Looks not safe for other parts of project where this behaviour is not needed

[sonic birch]

!e

def wrapper(*args, **kwargs):
    return None

def root():
    class MetaRoot(type):
        def __new__(cls, *args, **kwargs):
            print(args)
            if args:
                for c in args[1][:-1]:
                    methods = [m for m in c.__dict__ if not m.startswith('_')]
                    for m in methods:
                        setattr(args[1][-1], m, wrapper)
            return super().__new__(cls, *args, **kwargs)
    return MetaRoot('Root', (), {})

class Parent:
    pass
class Child:
    def foo(self):
        return super().foo()
class Mixed(Child, Parent, root()):
    pass
m = Mixed()
print(m.foo())

Fixed!

[night quarry]

@sonic birch :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | ('Root', (), {})
002 | ('Mixed', (<class '__main__.Child'>, <class '__main__.Parent'>, <class '__main__.Root'>), {'__module__': '__main__', '__qualname__': 'Mixed'})
003 | None

[versed eagle]

!e

from ctypes import (
    py_object as PyObject_p,
    c_int as int32_t,
    c_ssize_t as Py_ssize_t,
    sizeof as sizeof,
)

true = True
false = False

true_v = int32_t.from_address(
    id(true) + sizeof(Py_ssize_t * 2) + sizeof(PyObject_p)
)
false_v = int32_t.from_address(
    id(false) + sizeof(Py_ssize_t * 2) + sizeof(PyObject_p)
)

print(f"""\
before change:
    {true_v = }
    {false_v = }
""")

true_v.value = 0
false_v.value = 1

print(f"""\
after change:
    {true_v = }
    {false_v = }
""")

print(f"""\
True: {+true}
False: {+false}
""")

[night quarry]

@versed eagle :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | before change:
002 |     true_v = c_int(1)
003 |     false_v = c_int(0)
004 | 
005 | after change:
006 |     true_v = c_int(0)
007 |     false_v = c_int(1)
008 | 
009 | True: 0
010 | False: 0

[versed eagle]

why doesnt the value of false change also?

[rugged sparrow]

!e booleans are a subclass of int, so they have a ob_size field as well. You would need to change that (ignoring the fact that internally python checks bools by identity and not value)

from ctypes import *
class Boolean(Structure):
    _fields_ = [
            ('ob_refcnt', c_ssize_t),
            ('ob_type', py_object),
            ('ob_size', c_ssize_t),
            ('bool_val', c_int)
    ]
    def __repr__(self):
      ob_refcnt = self.ob_refcnt
      ob_type = self.ob_type
      ob_size = self.ob_size
      bool_val = self.bool_val
      return f'Boolean({ob_refcnt=}, {ob_type=}, {ob_size=}, {bool_val=})'

T_struct = Boolean.from_address(id(True))
F_struct = Boolean.from_address(id(False))
print(f'{T_struct = }')
print(f'{F_struct = }')

T_struct.bool_val = 0
F_struct.ob_size = 1
F_struct.bool_val = 1

true, false = True, False

print(f"""\
True: {+true}
False: {+false}
""")

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | T_struct = Boolean(ob_refcnt=56, ob_type=<class 'bool'>, ob_size=1, bool_val=1)
002 | F_struct = Boolean(ob_refcnt=79, ob_type=<class 'bool'>, ob_size=0, bool_val=0)
003 | True: 0
004 | False: 1

[versed eagle]

i know they're checked by identity and not value
that's why i used + to convert it to an int for printing and whatnot

why does the size need to be changed though

im not increasing it past 2**32 -1

[rugged sparrow]

if ob_size is zero, there is assumed to be no numerical data to calculate the value.

[versed eagle]

why would the size be zero though

oh

nvm

[rugged sparrow]

there are several shortcuts for zero (and False) that use ob_size being zero as a flag

[sage viper]

Is there a clear academic distinction between the terms "parameters" and "arguments"?
We have function "parameters," but "keyword arguments", *args, and **kwargs
This doc uses both terms
https://docs.python.org/3/reference/compound_stmts.html#function-definitions

Is it a parameter in the function declaration, and an argument at the call-site?

For a parameter with a default value, the corresponding argument may be omitted from a call, in which case the parameter’s default value is substituted.

A parameter is a variable in a function definition. It is a placeholder and hence does not have a concrete value. An argument is a value passed during function invocation.

answered my own question

[rugged sparrow]

@dry mirage have you had any issues updating einspect's hooking capability to 3.12+?

i think my setattr trick is broken

[quartz wave]

i think all of my C extensions are gonna break in 3.12

that's not good

should've listened to godlygeek

[rugged sparrow]

i figured it out, setattr expects normal dict locations and now the internal types have their dicts stored in a different location

[quartz wave]

all of my bytecode things too

[rugged sparrow]

oh yea those are gonna break hard

[dry mirage]

you mean the TPFLAGS_IMMUTABLETYPE unlock?

thought that still worked last I tried

[rugged sparrow]

there is some funky stuff involving type dicts now

[dry mirage]

the managed instance dictionary?

[rugged sparrow]

yea

i think?

in my testing at one point it manifested a second instance dict for int

one that had my hooks and one that it had before hand

setattr when tp_dict is null seems to manifest a dict, even if the type has a managed dict stored in the interpreter state

[dry mirage]

I had to do some logic for it for the einspect << move to actually move the instance dict as well https://github.com/ionite34/einspect/blob/main/src/einspect/structs/py_object.py#L248

[night quarry]

src/einspect/structs/py_object.py line 248

def instance_dict(self) -> ptr[PyObject[dict, Any, Any]] | None:```

[rugged sparrow]

i just fixed mine to clear out the manifested dict so that the class state remained consistent

probably not the best solve but itll work

[dry mirage]

wonder if it affects c extension types as well, might be an upstream bug in setattr

[rugged sparrow]

it shouldn't since the only reason these types were broken is because they have cached dicts that are shared across interpreters

[brazen geyser]

oh yeah definitely super unsafe. and potentially a fun #esoteric-python adventure. :)

[rugged sparrow]

https://github.com/python/cpython/blob/main/Objects/typeobject.c#L224-L229 this is where the issue begins to manifest

[night quarry]

Objects/typeobject.c lines 224 to 229

if (self->tp_flags & _Py_TPFLAGS_STATIC_BUILTIN) {
    PyInterpreterState *interp = _PyInterpreterState_GET();
    static_builtin_state *state = _PyStaticType_GetState(interp, self);
    assert(state != NULL);
    return state->tp_dict;
}```

[rugged sparrow]

since its an internal type flag _Py_TPFLAGS_STATIC_BUILTIN i don't think its a bug in setattr

[stark granite]

i dont understand lol

>>> (lambda:(lambda t:t(None,f).run())(t:=__import__("threading").Thread),t(None,f).run())()
<stdin>:1: SyntaxWarning: 'tuple' object is not callable; perhaps you missed a comma?
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
NameError: name 't' is not defined
>>> 

in what place is it not defined

wait

i think i realised what was wrong

, has higher priority

so you need to put () around lambda's code

[pastel stump]

I'm trying to implement __setattr__ for a class that also uses getters and setters using the property class with the following implementation:

def __setattr__(self, name: str, value):
  if isinstance(value, property) or callable(value):
    super().__setattr__(name, value)
  else:
    getattr(self, "__dynamic_set_"+name)(self, self, value)

The __dynamic_set_ function is a predefined setter function that is used in conjunction with the property() class. The issue is that when i set the variable with setattr and i print said variable, the output will always be <property object at 0x*********>.

The getattr portion works perfectly fine in another portion of the script. What on earth am I doing wrong here?

[versed eagle]

a little bit unrelated but why do you pass self twice into the dynamic set?

[versed eagle]

can you provide more context

like, an example usecase

and what you want it to do vs. what it does

[pastel stump]

I'm trying to implement a class that will keep track of certain modified variables. Basically: You add a property using the following function def __add_tracked_property(self, property_type, property_name, property_value=None), and it automagically creates the variable, including the property() statement in order to add a setter that will keep track of modified variables.

The specific usecase is a generic API client that will self-populate it's variables. In order to save on bandwidth/database usage, I only want to send variables that have been modified without bothering the end-user/programmer with keeping track of everything

Because of the dynamic nature of it all, I still need access to setattr without overriding the property() instance.

[versed eagle]

call .__set__ on the property instance

directly

at least, thats what it sounds like you want

[pastel stump]

So basically replace the getattr line with getattr(self, name).__set__(self, value)?

[versed eagle]

if its a property

yeah

property objects have __set__ and __get__ for managing themselves

[pastel stump]

def __setattr__(self, name: str, value):
  if hasattr(self, name) and isinstance(getattr(self, name), property):
    getattr(self, name).__set__(self, value)
  else:
    super().__setattr__(name, value)

More like this then?

Because that does not work either, and i still get just the <property object at 0x****> when i print

[versed eagle]

you still havent explained to me what you actually want to do

[versed eagle]

.

[versed eagle]

.

[pastel stump]

As stated before, i want to be able to use the property class in conjunction with setattr.

If I assign a setter for a certain variable, and use setattr(obj, thatcertainvariable, somevalue), it will override the property that was just assigned, and replace it with the actual value

I want to be able to use setattr on all variables, property or normal variable

[versed eagle]

the idea behind the class isn't the most relevant; what i need to know is

• what the intended behaviour of your function is
• what it actually does
• an example usecase of this, so i can see how the function is being invoked

[pastel stump]

Hang on, debugging atm. Something is not quite right

[fleet bridge]

Not twice, but three times! 1st is bounded by method, two more are passed as arguments

[versed eagle]

we don't know that its a bound method!

could be a static or class method ;)

[pastel stump]

That one is a bit weird, i know. Give me a minute 😛 Aparently the setter is not being registered somehow

[fleet bridge]

Probably you have mangling issues

Dont use __names if not necessary

[pastel stump]

Somehow in a refactor, it broke. Not sure why -_-

So the setter was never called. I'll post a full code example in a minute, with the reasoning behind it 😛

Riiiight..... I'm just kinda running in circles 😛

class DynamicVariableMixin():
    '''
    Used to be able to store and keep track of dynamic variables
    '''
    _updated_vars=None
    _dynamic_variables = {}

    def __add_method__(self, property_type, property_name):
        def set_attribute(self, obj, property_value, **kwargs):
            print("Setting")
            try:
                if property_type == type(property_value):
                    self._dynamic_variables[property_name] = property_value
                else:
                    raise TypeError
            finally:
                pass
        def get_attribute(self, *args, **kwargs):
            
            return self._dynamic_variables[property_name]
        setattr(self.__class__, '__dynamic_set_'+ property_name, set_attribute)
        setattr(self.__class__, '__dynamic_get_'+ property_name, get_attribute)
        setattr(self.__class__, property_name, 
            property(
                getattr(self, '__dynamic_get_'+ property_name), 
                getattr(self, '__dynamic_set_'+ property_name)
            )
        )

    def __add_tracked_property__(self, property_type, property_name):
        self.__add_method__(property_type, property_name)

This works, only on a class level (which i dont want because of obvious reasons). When i remove the .__class__ from setattr in order to set the getter/setter, it completely breaks and no longer uses the getter/setter

Implementation example:

class Test(DynamicVariableMixin):
    a = "Hello"

t1 = Test()
t2 = Test()

t1.__add_tracked_property__(str, "myvariable")
t1.myvariable = "Hello!"
print(t2.myvariable)

t2.myvariable currently resolves without any issue to "Hello!", which shouldn't happen.

Why the convoluted mess? I'm declaring some classes dynamically using input from an API. Object properties are set dynamically using the API response, and kept track of using the class above (WIP of course).

This is so the end-user/programmer does not have to fiddle with keeping track of modified properties of said object/instance when saving (patching) back to the API

[fleet bridge]

Why are you even setring __dynamic_set/get attributes? You are not using them

!e ```py
n = 15

s = f'<{n}>'
while True:
# print(f'{s} [{len(s)}]')
if '|0' in s:
s = s.replace('|0', '0||||||||||', 1)
elif '.>' in s:
s = s.replace('.>', '>', 1)
elif '.|' in s:
s = s.replace('.|', '-.', 1)
elif '0-' in s:
s = s.replace('0-', '1', 1)
elif '1-' in s:
s = s.replace('1-', '2', 1)
elif '2-' in s:
s = s.replace('2-', '3', 1)
elif '3-' in s:
s = s.replace('3-', '4', 1)
elif '4-' in s:
s = s.replace('4-', '5', 1)
elif '5-' in s:
s = s.replace('5-', '6', 1)
elif '6-' in s:
s = s.replace('6-', '7', 1)
elif '7-' in s:
s = s.replace('7-', '8', 1)
elif '8-' in s:
s = s.replace('8-', '9', 1)
elif '9-' in s:
s = s.replace('9-', '-0', 1)
elif '<-' in s:
s = s.replace('<-', '<1', 1)
if '-' not in s:
break
elif '9' in s:
s = s.replace('9', '8|', 1)
elif '8' in s:
s = s.replace('8', '7|', 1)
elif '7' in s:
s = s.replace('7', '6|', 1)
elif '6' in s:
s = s.replace('6', '5|', 1)
elif '5' in s:
s = s.replace('5', '4|', 1)
elif '4' in s:
s = s.replace('4', '3|', 1)
elif '3' in s:
s = s.replace('3', '2|', 1)
elif '2' in s:
s = s.replace('2', '1|', 1)
elif '1' in s:
s = s.replace('1', '0|', 1)
elif '00|' in s:
s = s.replace('00|', '0|', 1)
elif '0|' in s:
s = s.replace('0|', '0.||', 1)
else:
assert 0

n = int(s[1:-1])
print(n) # 16

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

16

[fleet bridge]

Markov algo that does x+1

It isnot the best markov algo for this task, it can be a lot simpler (so it is kinda esoteric 😝)

(and it is bugged for some numbers)

(strictly speaking it is not a markov algo because of custom termination behaviour, but im too lazy to fix it)

[pastel stump]

They are used when setting property, using another setattr call

[fleet bridge]

You are setting these values and immediately using them, which is equivalent to not setting them at all

[pastel stump]

Right. Corrected that with

setattr(self.__class__, property_name, 
  property(
    get_attribute, 
    set_attribute
  )
)

Only issue now is that with the code above, object t2 still has the myvariable attribute due to the self.__class__ being used in setattr. If i remove the __class__, the property is somehow being overwritten when i assign a new value. The Setting in set_attribute never gets printed

[tough willow]

What does that do?

[rugged owl]

adds one to n

[fleet bridge]

Exactly

[tough willow]

Why Not n += 1

[fleet bridge]

Because it is wrong channel for n+=1

[tough willow]

True

[proper vault]

!e ```py
def foo():
for a in [1,2,3]:
print(a)
for a in "abc":
print(a)
def trace(fr, ev, arg):
if fr.f_lineno == 3: fr.f_lineno = 5
return trace
import sys
sys.settrace(trace)

foo()

[night quarry]

@proper vault :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 1
002 | 2
003 | 3

[restive void]

Works in 3.9. Before it would complain about being unable to jump inside of a loop?

[proper vault]

    if fr.f_lineno == 3: fr.f_lineno = 5
ValueError: can't jump into the middle of a block

ye

[restive void]

Funnily it still does this sometimes

!e first a variant to show a bit more what's going on:

def foo():
    for a in [1,2,3]:
        print(a, "is a number")
    for a in None:
        print(a, "is a None?")

def trace(fr, ev, arg):
    if fr.f_lineno == 3:
        fr.f_lineno = 5
        fr.f_trace = None
    return trace

import sys
sys.settrace(trace)

foo()

[night quarry]

@restive void :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 1 is a None?
002 | 2 is a None?
003 | 3 is a None?

[proper vault]

yeah, the entire block stack must match to allow the jump, or you must be jumping out of a block

[restive void]

Block stack?

[proper vault]

block stack as in nested blocks, where a block is a for loop, with handler, etc

[restive void]

!e if the first loop has a break it seems to, well, break:

def foo():
    for a in [1,2,3]:
        print(a, "is a number")
        break
    for a in None:
        print(a, "is a None?")

def trace(fr, ev, arg):
    if fr.f_lineno == 3:
        fr.f_lineno = 6
        fr.f_trace = None
    return trace

import sys
sys.settrace(trace)

foo()

[night quarry]

@restive void :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 1 is a None?
002 | 2 is a None?
003 | 3 is a None?

[restive void]

huh. At least on 3.9 it does, whereas the previous code works

[proper vault]

IG no one really checks debugger jumps for backwards compat

or even debuggers in general tbh

[restive void]

There are tests

[proper vault]

ye, they get updated every version to match the behaviour of the new version

the code for checking jumps is about here ish https://github.com/python/cpython/blob/3.10/Objects/frameobject.c#L124

[night quarry]

Objects/frameobject.c line 124

markblocks(PyCodeObject *code_obj, int len)```

[proper vault]

or well, the interesting part

[restive void]

(That reminds me, I should add a link to my talk on our internal company wiki now that the videos were released)

[proper vault]

you did a talk?

[restive void]

Yeah, back in May, at PyCon Italia; about debuggers (live-coding a tiny one): https://www.youtube.com/watch?v=Cl6U9VoPn_M

[proper vault]

noice

[calm cove]

!e

def weird(l):
    return max(list(filter(lambda x: x % 2 != 0, (sum(n) for n in zip(*l)))) or [0])


print(weird([[1, 2, 3], [7, 8, 9]]))
print(weird([[1, 2, 3], [4, 5, 6], [7, 8, 9]]))
print(weird([[1, 2, 3], [4, 5, 6], [7, 8, 9], [9, 10, 11]]))

[night quarry]

@calm cove :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 0
002 | 15
003 | 29

[calm cove]

This finds the largest odd sum of integer values of a column in a 2d list. The tricky part was figuring out how to do the check for the case of no uneven number and then use [0] instead. My initial approach of

intermediary = filter(lambda x: x % 2 != 0, (sum(n) for n in zip(*l)))
list(intermediary) if list(intermediary) != [] else [0]

or anything similar to that did not work as the call list() on the filter object, which is an iterator consumes it. And because you can't assign variables in a lambda function, I had to use

list(intermediary) or [0]

because bool(list(intermediary)) is False in the case that intermediary is empty. So then [0] is used

[distant salmon]

!e

def weird(A):
    return max((sum(col) for col in zip(*A) if sum(col) % 2), default = 0)

print(weird([[1, 2, 3], [7, 8, 9]]))
print(weird([[1, 2, 3], [4, 5, 6], [7, 8, 9]]))
print(weird([[1, 2, 3], [4, 5, 6], [7, 8, 9], [9, 10, 11]]))

I would have probably just done something like this ^.

[night quarry]

@distant salmon :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 0
002 | 15
003 | 29

[icy cypress]

is there a way to build this list using a list comprehension? maybe using the walrus operator?

new_lst = []
prev = 0
for i in lst:
    new_lst.append(i + prev)
    prev = i
    
>>> lst
[1, 1, 4, 1, 6, 1, 1, 3, 1, 3, 1, 4]
>>> new_lst
[1, 2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

[fleet bridge]

Bruh

!e ```py
l = [1,1,4,1,6,1,1,3,1,3,1,4]
l2 = [*map(int.add, l, l[1:]+[0])]
print(l2)

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

[2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5, 4]

[fleet bridge]

Almost

[icy cypress]

map + int.add, thats beautiful

thanks

wait why is the last opposite

[fleet bridge]

!e ```py
l = [1,1,4,1,6,1,1,3,1,3,1,4]
l2 = [*map(int.add, l, [0]+l[:-1])]
print(l2)

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

[1, 2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

[fleet bridge]

Fixed

[icy cypress]

wait maybe itertools has something

oh nice

do you think there is a more eficcient solution? that maybe avoids copying the list?

[fleet bridge]

Yeah, itertools.tee can dupe list iterator, then you can prepend 0 to one of them and iterate over using map

Idk how map behaves on inputs of unequal length

!d map

[night quarry]

map

map(function, iterable, *iterables)```
Return an iterator that applies *function* to every item of *iterable*, yielding the results. If additional *iterables* arguments are passed, *function* must take that many arguments and is applied to the items from all iterables in parallel. With multiple iterables, the iterator stops when the shortest iterable is exhausted. For cases where the function inputs are already arranged into argument tuples, see [`itertools.starmap()`](https://docs.python.org/3/library/itertools.html#itertools.starmap).

Similar names: 2to3fixer.map

[restive void]

Strange that it didn't also get a strict kwarg

[icy cypress]

there is itertools.pairwise but the default element for the first pair is the same element (and you cant override it apparently)

[fleet bridge]

!e ```py
import itertools as it
l = [1,1,4,1,6,1,1,3,1,3,1,4]
l2 = [*map(int.add, (i:=it.tee(l))[0], it.chain([0], i[1]))]
print(l2)

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

[1, 2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

[fleet bridge]

Nice

[restive void]

!e

from itertools import*
l = [1,1,4,1,6,1,1,3,1,3,1,4]
l2 = [*map(sum, pairwise([0]+l))]
print(l2)

[night quarry]

@restive void :white_check_mark: Your 3.11 eval job has completed with return code 0.

[1, 2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

@half escarp :white_check_mark: Your 3.11 eval job has completed with return code 0.

[2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

[fleet bridge]

Oh, i though about starmap'ping tuples into int.add, but completely forgot about sum 😄

[icy cypress]

Thanks all for the solutions

I will benchmark them and see whats faster

[distant salmon]

If you care about speed then probably just do something like this

!e

l = [1,1,4,1,6,1,1,3,1,3,1,4]
l2 = l[:]
for i in range(1, len(l)):
  l2[i] += l[i - 1]
print(l2)

[night quarry]

@distant salmon :white_check_mark: Your 3.11 eval job has completed with return code 0.

[1, 2, 5, 5, 7, 7, 2, 4, 4, 4, 4, 5]

[versed eagle]

if we're we micro-optimising, make the first one a tuple instead of a list because the tuple can be loaded as a const

then again, if we're micro-optimising, a generator or comprehension might be better than duplicating the list and then adding to it
ill go benchmark stuff once im on my computer

[fleet bridge]

If we are micro-optimizing, just print the result as string because it is known beforehand

[versed eagle]

shush

lmao

if we're micro-optimising then we should use smaller variable names and less whitespace to minimise file-read and parsing times /j

[fleet bridge]

Just dont do anything and imagine correct result being printed

[versed eagle]

if we precalculate and preprint the result, then wait a really long time before benchmarking it, we can have a negative runtime (endtime will be before starttime so when calculating runtime endtime-starttime will be below 0)

[versed eagle]

the longer you wait, the more performant your program is

[distant salmon]

The way I look at python is that anything in Python is slow apart from the built in stuff.

Copying a list is one of those built in operations that runs blazingly fast

However, anything involving generators is Python code and will run rather slow

I've also never seen a performance gain from using tuples over lists. I would be really surprised if that could actually make it faster.

[fleet bridge]

I guess iterating over tuple can be a bit (tiny tiny bit) faster than iteration over list
Because tuple stores data in the object itself, but list allocates new memory and uses pointer to reach it
So less memory jumps => less time

[fleet bridge]

!e ```py
from dis import dis
dis('x in [1,2,3]')
dis('x in {1,2,3}')

Bot died?

[languid hare]

!e ```py
from dis import dis
dis('x in [1,2,3]')
dis('x in {1,2,3}')

[night quarry]

@languid hare :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 |   0           0 RESUME                   0
002 | 
003 |   1           2 LOAD_NAME                0 (x)
004 |               4 LOAD_CONST               0 ((1, 2, 3))
005 |               6 CONTAINS_OP              0
006 |               8 RETURN_VALUE
007 |   0           0 RESUME                   0
008 | 
009 |   1           2 LOAD_NAME                0 (x)
010 |               4 LOAD_CONST               0 (frozenset({1, 2, 3}))
011 |               6 CONTAINS_OP              0
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/7M4V7PH5M5NCV5TLYIN6MOTZK4

[fleet bridge]

Peepholes optimized these checks to not construct new list/set at runtime, so there is no difference in using literal tuples or literal lists

But in other cases if your list is known at compile time, you can use tuple instead

[quartz wave]

that's adding to a list

a list might actually iterate faster than a tuple and i still have no idea why

[distant salmon]

Does a tuple really store "data in the object itself"?

[quartz wave]

if you consider the data allocated together as a whole with the object, yes

[restive void]

there's one less level of indirection

[quartz wave]

they're still like converted into direct memory offsets anyway aren't they?

unless the pointer has to be loaded or something i think

[restive void]

AFAIR (could be wrong though) the tuple still only consists of an array of pointers. But that array is stored at the memory location of the tuple, whereas a list stores a pointer to an array somewhere else that has pointers to the actual objects (because it can be resized)

[distant salmon]

Ah ok

[versed eagle]

tuples can be LOAD_CONSTed

[simple moat]

for i in input():d="cv"[i in"aeiouyAEIOUY"];print(end=(i,(d,d.upper())[i.isupper()])[i.isalpha()])

is it still possible to reduce length of this?

[quartz wave]

for i in input():print(end=(i,chr(b"CV"[i in"aeiouyAEIOUY"]|ord(i)&32))[i.isalpha()])

[simple moat]

lol, thats crazy, thx

[sick hound]

Is pre-run time normally part of benchmarking?

[fleet bridge]

I think no

[dense crypt]

...this might be the wrong place to ask but this is Esoteric so here goes. I want to smash a portion of my animation and physics math AIO in a submodule named Colliction, the Color, Collision, and Projection Collection... is this too left brained?

basically it handles math functions between two tuples each less than length 4 with a factory approach

[earnest wing]

colliction is great, consider also adding another word into the mix

[dense crypt]

most of the words I would want to end in ion I would assume, but I definitwly havent actually thought of that, lol

I'm not an engineer, but I've been thinking about it for a while, and I think it could be a good educational and supplemental module

[dense crypt]

do you know what the other word is?

[fleet bridge]

Consider making all the words start with "col"

[dense crypt]

infeasible, this is essentially piece-meal math functions from core and core adjacent libraries, literally the biggest problem is naming things because like, half of it might be a vector and I have numpy so what if it just supports rich comparison?

why isnt it just a tuple?

[dense crypt]

@fleet bridge I'm trying to keep semi-natural names for things so that I can use Structural Pattern Matching to make it semi-lingual

[dense forum]

I prebake normals and my code instantly doubles in speed.

[stark granite]

my new milestone in oneliners!

(lambda data: (Path('analysis.txt').write_text('\n'.join(f"{char} {frequency:.3f}" for frequency, chars in sorted(((frequency, sorted(map(lambda char: char[0], filter(lambda info: info[1] == frequency, data.items())), key=ord)) for frequency in set(data.values())), key=lambda group: group[0], reverse=True) for char in chars))))((lambda total_length: {letter: occurrences / total_length for letter, occurrences in __import__('collections').Counter(s).items()})(len(s := ''.join(filter(lambda c: c in __import__('string').ascii_lowercase, map(str.lower, (Path := __import__('pathlib').Path)('text.txt').read_text()))))))

and without a single ; (or even :) hehe

prob there is a way to make that shorter (aside from shortening variable names obv and removing unnecessary spaces)

[last locust]

(There's loads of :)

[stark granite]

...

:= dont count

and in strings too

ah wait

[distant salmon]

There are lambda: too

[stark granite]

lambdas'

[last locust]

lambda info: ...

[stark granite]

...

[stark granite]

yea

i forgot about existence of those

lets just uhh....

ignore('lambda') :)

[last locust]

Then it's info: :)

ignore('lambda.*:') :)

[distant salmon]

One small codegolf I can see is to change the sorted(..., reverse=True) to sorted(...)[::-1]

But that would add more :

[stark granite]

hm, indeed

but like

the original code was separated into 3-4 function calls

(functions were oneliners)

so like

lets say i only did "thin" LTO

[distant salmon]

Here is one more codegolf

c in __import__('string').ascii_lowercase

can be changed to

'`'<c<'{'

It is also possible to do ```py
'a'<=c<='z'

[versed eagle]

no

i was joking

[versed eagle]

(Path := __import__('pathlib').Path)('text.txt').read_text() -> open('text.txt').read()

or, if its guaranteed that there's only one line in the file, *open('text.txt')

actually

why do you map str.lower

[earnest wing]

Big news! Development of hoopy is back!!!

[versed eagle]

i don't know what that is but yay!

oo you're doing codec stuff

✨

you can put the bootstrap in a .pth file and then run python3 main_hoopy.py directly

[versed eagle]

''.join(filter(lambda c: c in __import__('string').ascii_lowercase,map(str.lower, (Path := __import__('pathlib').Path)('text.txt').read_text())))

->

"".join(filter(lambda c:'`'<c<'{',open('text.txt').read().lower()))
# if the file only has 1 line,    str.lower(*open('text.txt')) is shorter

[earnest wing]

I know, it is also cross incompatible and fragile

[versed eagle]

wdym by [cross incompatible]

[earnest wing]

there is no universal way to do it on install across platforms

[versed eagle]

ah

[earnest wing]

you can account for the most common use cases though

[versed eagle]

you should be able to get the correct directories via sysconfig

[earnest wing]

That's funny

sysconfig.get_path("platlib")

I guess the interpreter has to know about them

I'll probably make it an opt-in feature for the library then

[stark granite]

there is not

[stark granite]

hm, interesting

i actually kinda just wanted to make use of pathlib... for no reason really

[stark granite]

because i was doing that after .splitlines() ._.

[versed eagle]

ah

[quartz wave]

where

[humble blade]

!e

from fishhook import hook

@hook(object)
def __matmul__(self, f):
  return f(self)

x = 2
sq = lambda x: x**2
x @= sq
print(x)

[night quarry]

@humble blade :white_check_mark: Your 3.11 eval job has completed with return code 0.

4

[humble blade]

heh

[fleet bridge]

!e

from fishhook import hook
@hook(object)
def __matmul__(self, f):
  return f(self)

x = 2
x @= 3 .__rpow__
print(x) # 2**3 == 8

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

8

[fleet bridge]

I like that

[rugged sparrow]

!e ```py
from fishhook import *
@hook(object)
def matmul(self, func):
return func(self)

print(123 @ str @ (lambda l:map(int, l)) @ list)
print(3456 @ str @ (lambda l:map(int, l)) @ list)```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | [1, 2, 3]
002 | [3, 4, 5, 6]

[quartz wave]

i did it with modified cpython ```pycon

.\cpython-main\PCbuild\amd64\python_d.exe
Python 3.13.0a0 (main, Aug 15 2023, 12:02:56) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

x = 2
x @= 5 .radd
x
7

i just realized this is piping

[languid hare]

indeed it is

where's that one sebastiaan pycon talk

[quartz wave]

this?

ok so for some reason i couldn't have tp_as_number populated in object

and it was supposed to originally be called a "decorator operator" but i just realized it's basically a pipe operator only now

nevertheless "decorator operator" sounds cooler and rhymes better

[rugged sparrow]

did it cause a segfault when making subclasses?

if so you hit the same bug that is the reason for fishhook's patch_object function

[stark granite]

*somewhere

[quartz wave]

it segfaulted while building

so i couldn't even get python to run with it

[rugged sparrow]

most likely the same bug

python assumes a lot of things about object when calculating method resolution and slots inheritence, and if object has one of its tp_as_* structs populated it attempts to look at its base (which normally is NULL) which causes the crash

[quartz wave]

where does it do that?

[rugged sparrow]

https://github.com/python/cpython/blob/main/Objects/typeobject.c#L6848-L6851

[night quarry]

Objects/typeobject.c lines 6848 to 6851

if (type->tp_as_number != NULL && base->tp_as_number != NULL) {
    basebase = base->tp_base;
    if (basebase->tp_as_number == NULL)
        basebase = NULL;```

[rugged sparrow]

basebase is never checked against being NULL

[quartz wave]

simple fix then

yep

done

[rugged sparrow]

you also probably need to add the check to all of the tp_as_* sections

[quartz wave]

should it be PR'd to cpython or have you already done that?

[rugged sparrow]

it doesn't need to be fixed, its a correct assumption

[quartz wave]

ok then

[rugged sparrow]

it only breaks when you try and apply custom operations to object

[rugged sparrow]

yea its simple when you know where the issue is, when i was first adding support to fishhook for hooking the base object type i spent so many hours tracking down the root cause of the crash

and then more working out how to patch around the issue while maintaining stability

[languid hare]

entering real esoteric territory now

[fleet bridge]

i like that even more 👍

[arctic skiff]

you seriously named version 3.13.0a0

[restive void]

That's just the current version of CPython on main

[arctic skiff]

huh, when did they start 3.13 dev?

[restive void]

Ever since feature-freezing 3.12 I'd imagine

[quartz wave]

since 3.12 had beta

[arctic skiff]

well

[quartz wave]

they also did that with 3.11 beta and 3.12 dev

[arctic skiff]

didn't know about it

[quartz wave]

i could make it python 4.0.0 tbh

hold on

yay ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 4.0.0 (main, Aug 15 2023, 14:40:15) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

[languid hare]

[arctic skiff]

nah

[stark granite]

i suppose that version has JIT, no GIL, strong 💪 typing mode, better typing system than typescript and faster than java, and has an ability to act as a systems language, am i right?

[versed eagle]

why would we want strong typing

[vapid magnet]

Can anyone explain why it should work?

[astral rover]

that doesnt work does it?

[vapid magnet]

Yes, it does

[astral rover]

til wtf

[vapid magnet]

First print is '3', second is '<class type>'

[astral rover]

second isnt hard to explain, its just returning the class type

[vapid magnet]

I tried to find more info about classes, but I dont find anything helpfull. Maybe I should try find more info about scopes

[astral rover]

i cant behave the firsts behaviour im really not sure why it loads a global there

i get where the naming is coming from though

[distant salmon]

!e

_Animal__type = 3

class Animal():
  def get_type_first(self):
    return __type
  def get_type_second(self):
    return type

anim = Animal()
print(anim.get_type_first())
print(anim.get_type_second())

[night quarry]

@distant salmon :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 3
002 | <class 'type'>

[versed eagle]

its because of name mangling

[versed eagle]

https://docs.python.org/3.11/tutorial/classes.html?highlight=mangling#private-variables

[astral rover]

i didnt realise it was as simple as a find and replace sorta thing for mangled names

[vapid magnet]

Thanks a lot, I missed it when I read this first time

[quartz wave]

to be done

[stark granite]

fix like quite a lot of bugs before even running?

[fleet bridge]

https://youtu.be/0hrEaA3N3lk

Good video about mangling tricks

[quartz wave]

TIL __A__.__x gets mangled to __A__._A____x

!e ```py
@lambda x: globals().update(ﾠ=x)or x
class ﾠ:
def init(_, x):
_.__x = x

print(ﾠ(...)._ﾠ____x)

[night quarry]

@quartz wave :white_check_mark: Your 3.11 eval job has completed with return code 0.

Ellipsis

[fleet bridge]

Wut

!e ```py
@lambda x: print(ord(x.name[2]))
class ﾠ:
...

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

4448

[fleet bridge]

!charinfo 4448

[night quarry]

Character Info

\u0034 : DIGIT FOUR - 4
\u0034 : DIGIT FOUR - 4
\u0034 : DIGIT FOUR - 4
\u0038 : DIGIT EIGHT - 8

Full Raw Text

\u0034\u0034\u0034\u0038

[fleet bridge]

!charinfo ﾠ

[night quarry]

Character Info

\uffa0 : HALFWIDTH HANGUL FILLER - ﾠ

[quartz wave]

4448??

0xffa0 is 65440

[fleet bridge]

Idk

[stark granite]

what

[vast wave]

GOD i love python's class system

genuinely, how can you make something this convoluted

[humble blade]

this is more a name mangling thing than anything actually related to "core" OOP

[quartz wave]

>>> from dis import dis
>>> dis(r"""
... class __A__:
...     def __init__(_):
...             import __a.b
... """)

 ...

Disassembly of <code object __init__ at 0x0000016667C8F050, file "<dis>", line 3>:
  3           0 RESUME                   0

  4           2 LOAD_CONST               1 (0)
              4 LOAD_CONST               0 (None)
              6 IMPORT_NAME              0 (__a.b)
              8 STORE_FAST               1 (_A____a)

             10 RETURN_CONST             0 (None)
``` `__a.b` is imported in class `__A__` but the result gets stored into `_A____a`

1. _A____a does not contain the class name (__A__)
2. __a.b is left unmangled when importing, but it is stored into a mangled name (_A____a) anyway

[lime wave]

hello I have this anti bad word system something like this

badwords = ["badword", "bad"]
message = "bad"

if badwords in message.lower():
 print("true")

but people can bypass it by doing "badbad" how can I fix it

[restive void]

You can't truly fix this

[lime wave]

oh.

[distant salmon]

Ehm isnt that if statement the wrong way around?

!e

badwords = ["badword", "bad"]
message = "bad"

if badwords in message.lower():
 print("true")

[night quarry]

@distant salmon :x: Your 3.11 eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "/home/main.py", line 4, in <module>
003 |     if badwords in message.lower():
004 |        ^^^^^^^^^^^^^^^^^^^^^^^^^^^
005 | TypeError: 'in <string>' requires string as left operand, not list

[distant salmon]

Ye it is. You should have

!e

badwords = ["badword", "bad"]
message = "bad"

if message.lower() in badwords:
 print("true")

[night quarry]

@distant salmon :white_check_mark: Your 3.11 eval job has completed with return code 0.

true

[fleet bridge]

That's will catch only messages that are bad entirely
"very bad badword!" will not be captured

!e

badwords = ["badword", "bad"]
message = "bad".lower()

for bw in badwords:
 if bw in message:
  print("it is bad")

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

it is bad

[quartz wave]

for anyone who missed ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 4.0.0 (main, Aug 17 2023, 12:01:58) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

5 |> 2 ** $
32
$
File "<stdin>", line 1
$
^
SyntaxError: template not in righthand side of pipe operation

piping and substitution

[versed eagle]

fun

[fleet bridge]

🤯

nice

[hollow moth]

def counter():
    x = 0
    def add():
        nonlocal x
        x += 1
        return x
    return add

add = counter()

print(add()) #  1
print(add()) #  2

[languid hare]

very nice

[dusk breach]

intron=''''''
exon='''
intron = input() or intron
print(f"{intron=:'^{len(intron)+6}}")
print(f"{exon=:'^{len(exon)+6}}")
print("exec(exon)")
'''
exec(exon)

[stark granite]

is it an actual thing in 3.12/3.13?

[stark granite]

my username "Spaceginner" is banned on roblox :(

prob due to last 6 letters

[restive void]

or reference to alcohol ("gin")

[stark granite]

wait

bruh

[quartz wave]

no

i just made it

[lean lodge]

Hello

[quartz wave]

hi

[rugged owl]

Did you release the modified version anywhere?

[quartz wave]

not really

[rugged owl]

Pretty sure \x is used for string representations, while 0x is for hexadecimal nums

[clear venture]

^

hexadecimal literals can be defined with 0x

[dusk breach]

intron = input() or intron
print(f"{intron=:'^{len(intron)+6}}")
print(f"{exon=:'^{len(exon)+6}}")
print("exec(exon)")
``` a little more clarity of the exon

[quartz wave]

discord highlighter is having problems

[fleet bridge]

It is always having problems

[dense mango]

It's not really convoluted, just the result of hidden behaviour.

Like structurally it's really simple. Animal is just a normal class with no special features that make it different from any other class. Structurally the example is no different from:

!e

global_var = 3

class Animal():
  def foo(self):
    return global_var
  def bar(self):
    return dict

anim = Animal()
print(anim.foo())
print(anim.bar())

[night quarry]

@dense mango :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 3
002 | <class 'dict'>

[dense mango]

All I've done here is change some names, the structure is identical.

It's just that Python has some surprising rules about renaming stuff that starts with a __ in class scope.

It has nothing to do with the type system

When you type __example inside a class MyClass: ..., Python sees __MyClass_example

So there just happens to be a global (module-level) variable with that name

[fleet bridge]

Attribute lookups are inconsistent for types and non-types.
Suppose X is a class and x is an instance of X, and X.a is some descriptor object in X dict. Both X.a and x.a will execute descriptor protocol, but x.b will not (assuming it is a descriptor object in instance dict)

I expect X.a and x.b to behave in the same way

[dense mango]

There's no attribute lookup involved in that example. The only attribute lookup is the method calls.

[fleet bridge]

it is not related to your example, it is just my thoughts

[dense mango]

__Animal_type and type are not attributes of Animal

Oh my mistake, I thought you were replying to me

[fleet bridge]

I'm sorry that i interrupted you

Name mangling is also inconsistent

!e ```py
class X: __a=1
class X: __a=1
class _X: __a=1
print(vars(X))
print(vars(X))
print(vars(_X))

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | {'__module__': '__main__', '_X__a': 1, '__dict__': <attribute '__dict__' of 'X' objects>, '__weakref__': <attribute '__weakref__' of 'X' objects>, '__doc__': None}
002 | {'__module__': '__main__', '_X___a': 1, '__dict__': <attribute '__dict__' of '_X_' objects>, '__weakref__': <attribute '__weakref__' of '_X_' objects>, '__doc__': None}
003 | {'__module__': '__main__', '_X__a': 1, '__dict__': <attribute '__dict__' of '_X' objects>, '__weakref__': <attribute '__weakref__' of '_X' objects>, '__doc__': None}

[fleet bridge]

First and third mangled names are the same, if you look closely

[dense mango]

Yep

[fleet bridge]

Because it lstrip's underscores from class name

[dense mango]

I think it would be nice if descriptors worked the same when the descriptor is in the class dict vs the instance dict.

[fleet bridge]

!e ```py
x=type('',(),dict(init=lambda self:setattr(self,'dict',globals())))()
print(x.x is x)
a = 1
x.b = 2
print(x.a + b)

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | True
002 | 3

[fleet bridge]

Here x is an object that provides convenient way to use global vars through attributes on x.
Idk if it is useful, but it is definitely interesting

!e ```py
X=type('',(dict,),dict(init=lambda self,*a,**b:setattr(self,'dict',self)))
x = X(a=1, b=2)
print(x)
x[3] = x['a'] + x.b
print(x)

Sadly dict initialization is broken here

!e ```py
X=type('',(dict,),dict(init=lambda self,*a,**b:[setattr(self,'dict',self),dict.init(self,*a,**b)][0]))
x = X(a=1, b=2)
print(x)
x[3] = x['a'] + x.b
print(x)

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | {'a': 1, 'b': 2}
002 | {'a': 1, 'b': 2, 3: 3}

[fleet bridge]

Yay

Nice

[restive void]

!e copying from ot0: Lifetimes in Python!

import sys
import threading
from time import sleep
from typing import Annotated, get_origin, get_args
from collections import namedtuple

timeout = namedtuple("timeout", "s")
def deleter(seconds, name, scope):
    sleep(seconds)
    exec(f"del {name}", scope.f_globals, scope.f_locals)
class A:
    def __init__(self):
        self.d = {}
    def __setitem__(self, name, val):
        origin = get_origin(val)
        if origin is Annotated:
            *_, deadline = get_args(val)
            if isinstance(deadline, timeout):
                frame = sys._getframe(1)
                threading.Thread(
                    target=deleter,
                    args=(deadline.s, name, frame)
                ).start()
        self.d[name] = val

    def __getitem__(self, item):
        return self.d[item]

__annotations__ = A()

x: Annotated[int, timeout(1)] = 23
print(x)
sleep(2)
print(x)

[night quarry]

@restive void :x: Your 3.11 eval job has completed with return code 1.

001 | 23
002 | Traceback (most recent call last):
003 |   File "/home/main.py", line 34, in <module>
004 |     print(x)
005 |           ^
006 | NameError: name 'x' is not defined

[hollow vigil]

I just wrote this one-line image to ascii text converter:

(lambda u,r,n:(lambda i,k:(lambda j:[print("".join([" ░▒▓█"[int((1-j.getpixel((x,y))/255)*4)]for x in r(k*2)]))for y in r(k)])(i.open(u("sys").argv[1]).convert("L").resize((k*2,k),1)))(u("PIL.Image",n,n,("PIL")),u("os").get_terminal_size().lines))(__import__,range,None)

what do you guys think?

[wheat river]

!e

import ast

with open(__file__) as source_file:
    source = iter(ast.parse(source_file.read()).body)


def givex():
    for node in source:
        if isinstance(node, ast.Assign):
            if node.value.func.id != "givex":
                continue
            if not isinstance(node.targets[0], ast.Tuple):
                continue

            return range(len(node.targets[0].elts))

    return (0,)


a, b = givex()
c, d, e, f = givex()

print(a, b)
print(c, d, e, f)

[night quarry]

@wheat river :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 0 1
002 | 0 1 2 3

[languid hare]

wouldnt it be return 0 at the end

since x, = givex() would be covered by the tuple case

otherwise x = givex() would be x = (0,)

[wheat river]

Yea ur right

[fleet bridge]

codon (python implementation) has pipe operator:
https://docs.exaloop.io/codon/advanced/pipelines

[quartz wave]

cool

[sullen cargo]

Wait, "get pixel"? Is that from a module or something?

Ohhhhhhhhhhhhh. "PIL"

I need to look into that. I've always wanted to do that.

The closest I got to messing with graphics is using features in pygame.

Nice work.

[quartz wave]

ideas for cpython

[unique heath]

make faster

[quartz wave]

i don know how to do that safely

n easily

[unique heath]

make slower

[quartz wave]

no

[fleet bridge]

make reprs better, so they more frequently return same value when eval'ed:

# functions/lambdas:
>>> f = lambda x: lambda y: x + y
>>> f(1)
<lambda <lambda>@<lambda> x=1>
>>> f
<lambda>
>>>
>>> def g(x):
...   def h(y):
...     return x + y
...   return h
...
>>> g
g
>>> g(1)
<function g@h x=1>

# methods/bound methods:
>>> list.append
list.append
>>> [].append
[].append
>>> {1:2}.keys.__call__
{1: 2}.keys.__call__

# descriptors:
>>> object.__dict__['__class__']
<slot object.__class__>
>>> type.__dict__['__bases__']
<slot type.__bases__>

# modules:
>>> import math; math
math
>>> import os.path
>>> os.path
ntpath

# types/instances:
>>> int
int
>>> class Meta(type): ...
...
>>> class X(metaclass=Meta): ...
...
>>> X
<Meta 'X'>
>>> X()
<<Meta 'X'> at 0x001A1ECC61A50>
>>>
>>> class Y: ...
...
>>> Y
Y
>>> Y()
<Y at 0x001A1ECD88690>

# properties/classmethods/staticmethods:
>>> @property
... def x(): ...
...
>>> @x.setter
... def x(): ...
...
>>> @x.deleter
... def x(): ...
...
>>> x
property(fget=x, fset=x, fdel=x)
>>>
>>> classmethod(lambda: ...)
classmethod(<lambda>)

[unique heath]

make

[unique heath]

<lambda <lambda>@<lambda> x=1>
>>> f
<lambda>``` cursed

[fleet bridge]

>>> object.__dict__
mappingproxy({   '__new__': object.__new__,
                 '__repr__': sitecustomize._@__repr__,
                 '__hash__': object.__hash__,
                 '__str__': object.__str__,
                 '__getattribute__': object.__getattribute__,
                 '__setattr__': object.__setattr__,
                 '__delattr__': object.__delattr__,
                 '__lt__': object.__lt__,
                 '__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'
                            '\n'
                            'When called, it accepts no arguments and returns a new featureless\n'
                            'instance that has no instance attributes and cannot be given any.\n'})

[fleet bridge]

>>> lambda: ...
<function <lambda> at 0x0000028278EDF4C0>
``` isnt any better

[unique heath]

this isnt html tho dude

[quartz wave]

eh sure

[fleet bridge]

what do you suggest?

[fleet bridge]

i have impl for all of this using fishhook, do you want it?

[quartz wave]

yea

[fleet bridge]

!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.

[fleet bridge]

https://paste.pythondiscord.com/4OCA

also do this in your cpython fork: ```py
del builtins.quit
del builtins.exit
del builtins.copyright
del builtins.credits
del builtins.license

these things are very anoying

[quartz wave]

why

[fleet bridge]

because of this mess ```py

builtins.dict
{ 'name': 'builtins',
...,
'open': io.open,
'quit': Use quit() or Ctrl-Z plus Return to exit,
'exit': Use exit() or Ctrl-Z plus Return to exit,
'copyright': Copyright (c) 2001-2022 Python Software Foundation.
All Rights Reserved.

Copyright (c) 2000 BeOpen.com.
All Rights Reserved.

Copyright (c) 1995-2001 Corporation for National Research Initiatives.
All Rights Reserved.

Copyright (c) 1991-1995 Stichting Mathematisch Centrum, Amsterdam.
All Rights Reserved.,
'credits': Thanks to CWI, CNRI, BeOpen.com, Zope Corporation and a cast of thousands
for supporting Python development. See www.python.org for more information.,
'license': Type license() to see the full license text,
'help': Type help() for interactive help, or help(object) for help about object.}

[quartz wave]

eh maybe

[fleet bridge]

x? syntax for repl is pretty cool
it runs help(x)

>>> int?
Help on class int in module builtins:

class int(object)
 |  int([x]) -> integer
 |  int(x, base=10) -> integer
...

(in my repl it also works if i type >>> ?i???n????????????t?)

[fleet bridge]

@quartz wave can your inline module inline functions?

my old 3.10 impl couldnt do it

i have 3.11 code that can do that

[quartz wave]

i don't think i did that

[hasty temple]

where clauses for comprehensions

e.g.

items = [
  (partition[0], partition[2])
  for item in list_of_strings
  where partition = item.partition('.')
]

[proper vault]

for partition in [item.partition('.')]

[hasty temple]

well yeah obviously there's other ways to do it

[proper vault]

it would be neat to have though

everytime you write for a in [b], you are making your code more incomprehensible

[quartz wave]

i might just make where a = x equivalent to for a in [x]

[proper vault]

on that note, I read for a in [b] is optimised into a normal assignment, is that actually a thing?

[quartz wave]

part 1 done ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 3.13.0a0 (heads/main-dirty:ef67ea5, Aug 19 2023, 22:20:40) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

f = lambda x: lambda y: x + y
f(1)
<lambda <lambda>@<lambda> x=1>
f
<lambda>
def g(x):
... def h(y): return x + y
... return h
...
g
g
g(1)
<function g@h x=1>

[quartz wave]

yep

i'm gonna continue development on monday since i can't use the pc on sundays

[fleet bridge]

Nice!

Do you want my function inlining code? (It is awful, as usual)

Implement tagged breaks/continues in your cpython
(And optionally goto)

[versed eagle]

goto would be nice

[fleet bridge]

Implement goto from one function to another

Also consider creating some object for ecery target. And we should be able to get this object, pass it to goto statement and jump to place where goto object was defined

def f():
  label a
  print("a")
  return
  label b
  print("b")

print(f.a) # <goto to file "..." at line 123>

goto f.a # jumls to label a in f, prints a

goto [f.a, f.b][random.randint(0,1)] # haha

[fleet bridge]

Make implicit __result__ variable that holds function return value
return (and implicit return at the end) actually does return __result__
return x does __result__ = x; return __result__

__result__ initial value is None

Make return and raise not statements, but expressions syntactically. They doesn't return anything because it doesn't make sense, but you can use them in expressions.

And assert too*

[meager zinc]

we need manual memory management

[fleet bridge]

while True:
    malloc(1)

Cursed

[versed eagle]

while malloc(1):
    ...

[meager zinc]

what if python made the gc async

[fleet bridge]

!e ```py
def f():
try:f()
except:f()
f()

[night quarry]

@fleet bridge :warning: Your 3.11 eval job timed out or ran out of memory.

[No output]

[fleet bridge]

This should crash

[fleet bridge]

What if it made everything async?

[meager zinc]

😮

genius

[fleet bridge]

await print('hello world')

Wait no

await print(await 'hello world')

Fixed

Wait no

await (await print)(await 'hello world')

Fixed

[astral rover]

why arent you awaiting the await smh?

[fleet bridge]

Because it is keyword

Imagine async compilation

[dreamy pier]

await async await def await function(await a, await b):
    ...

[unreal echo]

Bruh had to change

https://paste.pythondiscord.com/6BSQ function overloading

Had to make it support stuff like tuple[int|str]|list[list]]

Also why is typing.Optional like 3 different types in different scenarios

[versed eagle]

what does it do when it gets multiple overloads that are possible?

ooo you have like a scoring system to check how closely it matches the desired type

[unreal echo]

Demerit points like driving

[unreal echo]

It just picks the latest one that fits

I thought about making it prefer more strict arguments, but that wouldnt make sense

Since the user would be responsible for making different functions, why make two functions that have the same argument types even for overloading?

[versed eagle]

anything that can be passed to a function which takes an int can in theory be passed to a function which takes typing.Any

but if the user passes an int, they probably want it to use the int one

[versed eagle]

wdym

[quartz wave]

that's really hard

[quartz wave]

2nd one partially implemented ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 3.13.0a0 (heads/main-dirty:ef67ea5, Aug 19 2023, 22:20:40) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

[].append
[].append
list.append
<method 'append' of 'list' objects>
min.call
min.call
[].getitem
[].getitem
{1: 2}.keys
{1: 2}.keys
{1: 2}.keys.call
{1: 2}.keys.call

[fleet bridge]

nice

[quartz wave]

2nd one fully implemented, 3rd one fully implemented, 4th one fully implemented, 5th one partially implemented ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 3.13.0a0 (heads/main-dirty:ef67ea5, Aug 19 2023, 22:20:40) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

list.append
list.append
object.dict['class']
<slot object.class>
type.dict['bases']
<slot type.bases>
import math; math
math
import os.path
os.path
ntpath
int
int
class Meta(type): ...
...
class X(metaclass=Meta): ...
...
X
<Meta 'X'>
X()
<<Meta 'X'> at 0x00000258CA58BE60>
class Y: ...
...
Y
Y
Y()
<Y at 0x00000258CA58BE60>
classmethod(lambda: ...)
classmethod(<lambda>)

[quartz wave]

5th one fully implemented ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 3.13.0a0 (heads/main-dirty:ef67ea5, Aug 19 2023, 22:20:40) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

@property
... def x(): ...
...
@x.setter
... def x(): ...
...
@x.deleter
... def x(): ...
...
x
property(fget=x, fset=x, fdel=x)
classmethod(lambda: ...)
classmethod(<lambda>)
class A(classmethod): ...
...
A(lambda: ...)
A(<lambda>)

• additional in the file ```pycon

.\cpython-main\PCbuild\amd64\python.exe
Python 3.13.0a0 (heads/main-dirty:ef67ea5, Aug 19 2023, 22:20:40) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

import struct
struct.Struct('d')
_struct.Struct(b'd')

[quartz wave]

i already have goto in scopes but goto with arbitrary labels across scopes is gonna be a bit hard

[rugged sparrow]

im trying to write that rn

[quartz wave]

in fact really difficult

especially doing that with the current goto implementation i have rn

[quartz wave]

anyway i pushed the changes into my cpython repo

i'm gonna do frozendict

[rugged sparrow]

My idea for implementation is using sys.settrace to basically do a delayed jump and watch until execution is inside of the target function, then call the target function yourself after setting whatever global signals what the trace function should look for

[quartz wave]

my problem is i already have a working goto implementation

and the names are separate from variable names so things like these are possible ```py
a = 5
label a
a *= 2
if a < 100:
goto a
print(a)

and they translate into jumps and jump offsets instead of objects/instructions

[rugged sparrow]

Ah mine is a native version, so labels and jumps look like label .a and goto .func.label

[versed eagle]

shouldn't int be <type 'int'> then?

since type is the metaclass

[quartz wave]

idk ask @fleet bridge

[quartz wave]

it's a little painful and tedious but i did it

[fleet bridge]

That would be redundant, since every class is an instance pf type.
My goal was to make reprs that more frequently evaluate to the same object

[quartz wave]

i think this works ```pycon

.\cpython-main\PCbuild\amd64\python_d.exe
Python 3.13.0a0 (heads/main-dirty:4d56f51, Aug 21 2023, 13:11:15) [MSC v.1937 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.

[x for x in [1, 2, 3] where x = x + 3]
[4, 5, 6]

[fleet bridge]

So it is just an intermediate assignment statement

Cool

[quartz wave]

what is this for

[fleet bridge]

Other langs have that feature, so why not?
#esoteric-python message (read below)

[lament pond]

just a calculator: for vars in [[int(input("Number 1: ")),input("+, -, * or /? "),int(input("Number 2: "))]]: print(vars[0]+vars[2] if vars[1] == "+" else vars[0]-vars[2] if vars[1] == "-" else vars[0]*vars[2] if vars[1] == "*" else vars[0]/vars[2] if vars[1] == "/" else print('Error'))

[quartz wave]

I=input
a=int(I((i:='Number %d: ')%1))
print(eval('%d%s%d'%(a,o,int(I(i%2))))if(o:=I('+, -, * or /? '))in{*'+-*/'}else'Error')

[lament pond]

Number 1: 6
+, -, * or /? -
Number 2: 7
6666666

[quartz wave]

fixed

[lament pond]

Number 1: 6
+, -, * or /? -
Number 2: 7
42

[quartz wave]

eh

i'll fix it a second time

done

[lament pond]

SyntaxError: '(' was never closed

[quartz wave]

fixed it a 3rd time

[lament pond]

Number 1: 1
+, -, * or /? /
Number 2: 0
Traceback (most recent call last):
ZeroDivisionError: division by zero

[quartz wave]

works fine

[lament pond]

Number 1: 5.5
Traceback (most recent call last):
ValueError: invalid literal for int() with base 10: '5.5'

[quartz wave]

works just like the original

[lament pond]

oh, if thats what you are comparing it to

gimme a sec

[stark granite]

so dat is basically this:

operand_a = int(input("Number 1: "))
operation = input("+, -, * or /? ")
operand_b = int(input("Number 2: "))

match operation:
    case "+":
        print(operand_a + operand_b)
    case "-":
        print(operand_a - operand_b)
    case "*":
        print(operand_a * operand_b)
    case "/":
        print(operand_a / operand_b)
    case _:
        print('Error')

[lament pond]

yep

[stark granite]

:D

[lament pond]

squeezed into 1 line

[stark granite]

i think i might be able to do better?

idk

[lament pond]

that smile makes me so happy!

[stark granite]

ah wait

actually i doubt..?

[lament pond]

[stark granite]

https://cdn.discordapp.com/emojis/1104736344275697795.webp?size=48&name=shirohug&quality=lossless

[lament pond]

hugs :)

[stark granite]

ye i like em

[lament pond]

same

[lament pond]

[stark granite]

i should try "reverse engineering" more one-liners tbh

quite an interesting task

[quartz wave]

really?

are you sure?

[stark granite]

yes.

[quartz wave]

i'll cook up a challenge then

[stark granite]

lol

[lament pond]

@polar plover SHOW HIM YOUR 1 LINER

[quartz wave]

i reverse-one-linered that i think

import pygame as pg
import pygame.locals
import sys # bonus! for exit purposes
from random import randint as ri

sc = pg.display.set_mode((600, 600))
cl = pg.time.Clock()
sp = pg.Vector2(10, 10)
sb = []
di = pg.Vector2(10, 0)

def nf():
    global fo
    fo = pg.Vector2(ri(0, 59) * 10, ri(0, 59) * 10)

while True:
    sb.append(pg.Vector2(sp))
    sb.pop(0)
    sp += di
    if sp in sb or not (0 <= sp.x <= 600 and 0 <= sp.y <= 600):
        sb.clear()
        sp.update(10, 10)
    if sp == fo:
        sb.insert(0, pg.Vector2(-10, -10))
        nf()
    sc.fill((0, 0, 0))
    pg.draw.rect(sc, (255, 0, 0), (fo.x, fo.y, 10, 10))
    pg.draw.rect(sc, (255, 255, 255), (sp.x, sp.y, 10, 10))
    for p in sb:
        pg.draw.rect(sc, (255, 255, 255), (p.x, p.y, 10, 10))
    for ev in pg.event.get():
        if ev.type == pg.locals.QUIT:
            sys.exit(0) # 1 / 0
        elif ev.type == pg.locals.KEYDOWN:
            if ev.key == pg.locals.K_w: di.update(0, -10)
            elif ev.key == pg.locals.K_s: di.update(0, 10)
            elif ev.key == pg.locals.K_a: di.update(-10, 0)
            elif ev.key == pg.locals.K_d: di.update(10, 0)
    pg.display.update()
    cl.tick(10)

[lament pond]

this one? ```py
(pg:=import("pygame"),import("pygame.locals"),ri:=import("random").randint,sc:=pg.display.set_mode((600, 600)),cl:=pg.time.Clock(),sp:=pg.Vector2(10,10),sb:=[],di:=pg.Vector2(10,0),nf:=lambda:globals().update(fo=pg.Vector2(ri(0,59)*10,ri(0,59)*10)),nf(),[(sb.append(pg.Vector2(sp)),sb.pop(0),sp.iadd(di),(sb.clear(),sp.update(10,10))if(sp in sb or not(0<=sp.x<=600and 0<=sp.y<=600))else None,(sb.insert(0,pg.Vector2(-10,-10)),nf())if sp==fo else None,sc.fill((0,0,0)),pg.draw.rect(sc,(255,0,0),(fo.x,fo.y,10,10)),pg.draw.rect(sc,(255,255,255),(sp.x,sp.y,10,10)),[pg.draw.rect(sc,(255,255,255),(p.x,p.y,10,10))for p in sb],[((1/0)if ev.type==pg.locals.QUIT else di.update({pg.locals.K_w:(0,-10),pg.locals.K_s:(0,10),pg.locals.K_a:(-10,0),pg.locals.K_d:(10,0)}.get(ev.key,di))if ev.type==pg.locals.KEYDOWN else None)for ev in pg.event.get()],pg.display.update(),cl.tick(10),)for _ in iter(int,1)])

[quartz wave]

yep

that one

[quartz wave]

__builtins__.__getattribute__((_:=__name__.__class__.__class__.__doc__.__getitem__((___:=(_:=__name__.__getitem__((__:=__name__.__class__().__len__()))).__add__(_).__len__())).__add__((_____:=(_____:=__name__.__ne__(__name__).__invert__()).__neg__().__truediv__(_____.__add__(_____).__neg__()).__rpow__(_____).__class__).__doc__.__getitem__((______:=__name__.__eq__(__name__).__pos__()))).__add__(__name__.__len__().__class__.__doc__.__getitem__(__)).__add__(__name__.__len__().__class__.__doc__.__getitem__(______)).__add__(__name__.__class__.__class__.__doc__.__getitem__(__))))(___________ if(___________:=(________:=__builtins__.__getattribute__((_______:=__name__.__len__().__class__.__doc__.__getitem__(__).__add__(__name__.__len__().__class__.__doc__.__getitem__(______)).__add__(__name__.__class__.__class__.__doc__.__getitem__(___)).__add__(__name__.__len__().__class__.__module__.__getitem__(______)).__add__(__name__.__class__.__class__.__doc__.__getitem__(__))))()).__mul__(__________:=__builtins__.__getattribute__((_________:=__name__.__len__().__class__.__doc__.__getitem__(__).__add__(__name__.__len__().__class__.__doc__.__getitem__(______)).__add__(__name__.__class__.__class__.__doc__.__getitem__(__))))(__builtins__.__getattribute__(_______)())))!=(_________________:=__builtins__.__getattribute__(__builtins__.__dir__().__getitem__((____________:=(_____________:=(____________:=__name__.__getitem__(__)).__add__(____________).__add__(____________).__len__()).__mul__((________________:=(______________:=___.__mul__(___)).__mul__(_____________).__add__((_______________:=___.__mul__(___).__invert__().__neg__()))))))))else __________.__rmul__(________))

try that

[lament pond]

obfuscator detected

[lament pond]

@stark granite try this too

[stark granite]

wow

[stark granite]

wtf does it do?

[quartz wave]

something

try testing it

[stark granite]

wow

[stark granite]

i tried running it but uhh

it did nothing?

[quartz wave]

try providing some inputs

it should be pretty quick once you figure it out

[lament pond]

this one is snake

[stark granite]

is it a calculator?

[quartz wave]

no it's simpler than one

[stark granite]

multiplicator?

[lament pond]

I like vivax, cool friend

[stark granite]

tho i doubt

uh

oh

ic

basically this: input() * int(input())?

[quartz wave]

missing a call

[stark granite]

ah

print(input() * int(input()))

[quartz wave]

yep

[stark granite]

why tf does it need __doc__ and __len__ and __rmul__ (just as well __add__)?

wait i gotta understand wtf does it do

[fallow leaf]

Does this count as esoteric ```py
def inject(clazz, otherclazz=None):
if otherclazz==None:
for name, obj in globals().copy().items():
if isinstance(obj, type) and not issubclass(obj, clazz):
try:
for method in getMethods(clazz):
setattr(obj, method, getattr(clazz,method))
for variable in getVariables(clazz):
setattr(obj, variable, getattr(clazz,variable))
globals()[name] = obj
print(obj.name)
except Exception as e:
print(e)
else:
obj = globals()[otherclass.name]
for method in getMethods(clazz):
setattr(obj, method, getattr(clazz,method))
for variable in getVariables(clazz):
setattr(obj, variable, getattr(clazz,variable))
globals()[otherclass.name] = obj

Because I doubt alot of people would understand it 😂

[fallow leaf]

The methods also work to get self and the class they are in even though the are instances of another classes method

[gentle plover]

Not quite IMO. It's unholy, don't get me wrong, but still not cursed enough to be truly esoteric.

[restive void]

globals()[name] = obj this is unnecessary; you're modifying the other class(es) in-place

[quartz wave]

327 ```py
def a(c,o):
for v in dir(c):
if v not in dir(type)and(callable(x:=getattr(c,v))or"weakref"!=v):setattr(o,v,x)
def inject(c,o):
if o==None:
for x in globals().values():
if isinstance(x,type)and not issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)
else:a(c,globals()[o.name])

[fallow leaf]

Mines is somewhat readable still, but that is on a whole other level 💀

[quartz wave]

321 ```py
def a(c,o):
for v in dir(c):
if v not in dir(type)and(callable(x:=getattr(c,v))or"weakref"!=v):setattr(o,v,x)
def inject(c,o):
if o==None:
for x in globals().values():
if isinstance(x,type)&~issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)
else:a(c,globals()[o.name])

309 ```py
def a(c,o):
for v in{*dir(c)}-{*dir(type)}:
if callable(x:=getattr(c,v))or"weakref"!=v:setattr(o,v,x)
def inject(c,o,g=globals()):
if o==None:
for x in g.values():
if isinstance(x,type)&~issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)
else:a(c,g[o.name])

[fallow leaf]

What are you using to do this?

[quartz wave]

mind

306 might continue tmr ```py
def a(c,o):
for v in{*dir(c)}-{*dir(type)}:
if callable(x:=getattr(c,v))or"weakref"!=v:setattr(o,v,x)
def inject(c,o,g=globals()):
if o==None:
for x in g:
if isinstance(x:=g[x],type)&~issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)
else:a(c,g[o.name])

[stark granite]

shouldnt you compare o to None using is instead of ==?

[restive void]

Takes more bytes

This is not exactly idiomatic code anyways

[stark granite]

yes i know

actually what is the reason behind using is instead of ==?

[languid hare]

!e

import numpy as np
x = np.array([1, 2, 3])

print(x is None)
print(None == x)
print(x == None)

[night quarry]

@languid hare :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | False
002 | [False False False]
003 | [False False False]

[languid hare]

there is no guarantee the == works how you expect it to
of course, for builtins they're much more well behaved, but you shouldn't rely on it

[stark granite]

huh

[stark granite]

ok

[rugged sparrow]

@fleet bridge native goto that can jump into a new function scope

it uses sys.settrace

[fleet bridge]

Yooo

Nice

[rugged sparrow]

basically just sets a trace function to perform the jump as soon as execution enters a frame with a specific code object, then calls the function

[fleet bridge]

That's smart

[rugged sparrow]

https://paste.pythondiscord.com/XNKA

rn the function jump doesnt share a its jump code with the normal goto code, im trying to generalize rn

[rugged sparrow]

https://paste.pythondiscord.com/MF5A now with a more shared code base and dynamic goto's for function labels

[versed eagle]

identity vs equality

[quartz wave]

301 ```py
def a(c,o):[setattr(o,v,x)for v in{*dir(c)}-{*dir(type)}if callable(x:=getattr(c,v))or"weakref"!=v]
def inject(c,o,g=globals()):
if o==None:
for x in g:
if isinstance(x:=g[x],type)&~issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)
else:a(c,g[o.name])

[versed eagle]

def a(c,o):[setattr(o,v,x)for v in{*dir(c)}-{*dir(type)}if callable(x:=getattr(c,v))or"__weakref__"!=v]
def inject(c,o,g=globals()):
 if o!=None:return a(c,g[o.__name__])
 for x in g:
  if isinstance(x:=g[x],type)&~issubclass(x,c):
   try:a(c,x),print(x.__name__)
   except Exception as e:print(e)```

297

def a(c,o):[setattr(o,v,x)for v in{*dir(c)}-{*dir(type)}if callable(x:=getattr(c,v))or"__weakref__"!=v]
def inject(c,o,g=vars()):
 if o!=None:return a(c,g[o.__name__])
 for x in g:
  if isinstance(x:=g[x],type)&~issubclass(x,c):
   try:a(c,x),print(x.__name__)
   except Exception as e:print(e)```

294

[versed eagle]

this only works if the function is defined at global scope though

292, if you assume .__weakref__ will never be callable (which is almost always a reasonable assumption)

def a(c,o):[setattr(o,v,x)for v in{*dir(c)}-{*dir(type)}-{"__weakref__"}if callable(x:=getattr(c,v))]
def inject(c,o,g=vars()):
 if o!=None:return a(c,g[o.__name__])
 for x in g:
  if isinstance(x:=g[x],type)&~issubclass(x,c):
   try:a(c,x),print(x.__name__)
   except Exception as e:print(e)```

thats all i can think of for now

might try to make it shorter tomorrow

wait actually

def a(c,o):[setattr(o,v,x)for v in{*dir(c)}-{*dir(type),"__weakref__"}if callable(x:=getattr(c,v))]
def inject(c,o,g=vars()):
 if o!=None:return a(c,g[o.__name__])
 for x in g:
  if isinstance(x:=g[x],type)&~issubclass(x,c):
   try:a(c,x),print(x.__name__)
   except Exception as e:print(e)```

290

ok now im done for today

no more ideas

[quartz wave]

that'll not include the things that aren't callable

it's either callable or not callable and is not .__weakref__

[versed eagle]

oh true

im too tired for this lmao

[quartz wave]

so 273 ```py
def a(c,o):[setattr(o,v,getattr(c,v))for v in{*dir(c)}-{*dir(type),"weakref"}]
def inject(c,o,g=vars()):
if o!=None:return a(c,g[o.name])
for x in g:
if isinstance(x:=g[x],type)&~issubclass(x,c):
try:a(c,x),print(x.name)
except Exception as e:print(e)

[versed eagle]

✨

i should probably go sleep now

[quartz wave]

ok gn

[versed eagle]

gn

[quartz wave]

can use exec() for a possible improvement

268 ```py
def inject(c,o,g=vars()):
def a(o):[setattr(o,v,getattr(c,v))for v in{*dir(c)}-{*dir(type),"weakref"}]
if o!=None:return a(g[o.name])
for x in g:
if isinstance(x:=g[x],type)&~issubclass(x,c):
try:a(x),print(x.name)
except Exception as e:print(e)

266 ```py
def inject(c,o,g=vars()):
def a(o):[setattr(o,v,getattr(c,v))for v in{*dir(c)}-{*dir(type),"weakref"}]
if o!=None:a(g[o.name]);g=[]
for x in g:
if isinstance(x:=g[x],type)&~issubclass(x,c):
try:a(x),print(x.name)
except Exception as e:print(e)

[compact raven]

I hardly ever use is, except for testing/debugging.

[stark granite]

you are supposed tho

[unique heath]

okay but now make it work with ctx managers

[rugged sparrow]

Like jump into a context manager block?

[rugged sparrow]

You could implement that by jumping to the beginning of the setup line, then installing a handler to jump to the specific label inside the ctx manager block as soon as execution enters the block

[earnest wing]

Hah, that's some gimped goto

[rugged sparrow]

had to still execute the ctx manager to maintain stack levels, but was able to jump around upon entry

[restive void]

Eh, that's way too well-behaved... Where's the fun if it actually sets up the resource :D

Trace functions don't let you jump into a context manager though, right?

[rugged sparrow]

they technically dont

but my implementation uses trace functions

[restive void]

Yeah, you jump in front of it, and then immediately jump within it, right?

[rugged sparrow]

yup

but i implemented it in such a way that also allows jumps into and out of for loops

[restive void]

(Reminds me of the weird hack I had to pull with dont)

Nice, I'm always impressed by your stunts :)

[earnest wing]

I like the oxymoron of a resource respecting goto

[rugged sparrow]

it seemed like a neat and cursed way of handling goto

[restive void]

Do you set up/tear down the trace function stuff before and after each jump?

[rugged sparrow]

yea

[proper vault]

Jumping out of a with block will skip it's exit, right?

[rugged sparrow]

not currently

actually yes it does

unintentionally

[restive void]

Alternative syntax idea:

goto(my_label)

my_label:  # (half-width hangul filler)

Ah, I guess goto(my_label) doesn't work if the tf isn't active yet (unless the label has been declared already). And it would make it more complicated inside functions.. too bad

[rugged sparrow]

yea, i played around with a few syntaxes, and keyword .label was the easiest one to scan bytecode for (LOAD_NAME | LOAD_GLOBAL + LOAD_ATTR)

[restive void]

The label I think is doable now that I think about it. It's just a SETUP_ANNOTATIONS at the beginning, and then somewhere a LOAD_NAME, LOAD_NAME ("__annotations__"), LOAD_CONST, STORE_SUBSCR.

And technically the goto(my_label) is also doable by hooking into globals/builtins __missing__ (or equivalent).

[rugged sparrow]

the issue is it gets stripped away inside functions

!e ```py
import dis
def foo():
label: here

dis.dis(foo)```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 |   2           0 RESUME                   0
002 |               2 LOAD_CONST               0 (None)
003 |               4 RETURN_VALUE

[serene stratus]

Even if you turn off all optimizations?

[rugged sparrow]

im pretty sure they are always stripped

[rugged sparrow]

any suggestions for syntax for function jumps with args (rn i have goto .function.with_args(...).label or goto *function.with_args(...).label)

[restive void]

goto &function.args(...).label?

Or just goto &function(...).label?

[rugged sparrow]

i like this one but its tricky because rn that would just call the function right there

it would work for goto .function(...).label but would not for goto *function(...).label (which is my syntax for dynamic jumps)

[fallow leaf]

How would I modify on a part of a built-in classes methods body

Or any classes methods body

Cant find anything for this anywhere

[dull tusk]

Can't you just create a new type that inherits from builin one?

[rugged sparrow]

!e like this? ```py
from fishhook import hook, orig

@hook(list)
def sub(self, other):
new = self.copy()
for item in other:
if item in new:
new.remove(item)
return new

print([1,2,3] - [1])```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

[2, 3]

[fallow leaf]

Exactly💀

But I wouldn't want to rewrite the entire body only part of it like changing the bytecode of the return value etc

[rugged sparrow]

can you show me a psudocode example?

[dull tusk]

i belive you can override any method you want like that, but I have not tested it:
MyClass.my_method = lambda self: completely_different_function()

[rugged sparrow]

you can do that for classes defined in python, not classes defined in C
fishhook lets you do it for C classes as well

[fallow leaf]

Original```
class myClass:
def say(self,d):
print(d)
if d == 10:
return "yes"
else:
return "no"

Modified 

class myClass:
def say(self,d):
print(d)
if d == 10:
return 10 <- changed value and not the whole body
else:
return "no"

[fallow leaf]

That's actually cool

[rugged sparrow]

^ in that specific example, you could swap out the constant value "yes" with 10

[fallow leaf]

Can you give me a example?

[rugged sparrow]

!e ```py
class myClass:
def say(self,d):
print(d)
if d == 10:
return "yes"
else:
return "no"

orig_consts = list(myClass.say.code.co_consts)
orig_consts[orig_consts.index("yes")] = 10
myClass.say.code = myClass.say.code.replace(co_consts=tuple(orig_consts))

m = myClass()
print(m.say(10))```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | 10
002 | 10

[fallow leaf]

.code is a thing? 💀 that's so much easier than what I was thinking

But does it work for built-in classes?

[fleet bridge]

!e ```py
class X:
def foo(self):
print('foo')

from fishhook import hook
@hook(X)
def foo(self):
print('bar')

X().foo()

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

bar

[fleet bridge]

cool

fishhook can even patch non-C classes

[rugged sparrow]

builtin classes are defined in C, you will need to use a tool like ctypes, fishhook or einspect to mess with them

[rugged sparrow]

in theory orig should even work in that case

tbh i don't remember if i tested orig with python classes

[fallow leaf]

Yeah that's what I thought that's why I asked in #c-extensions and then they told me to go here 😐

[fleet bridge]

!e ```py
class X:
def foo(self):
print('foo')

from fishhook import hook, orig
@hook(X)
def foo(self):
print('bar')
orig(self)

X().foo()

[night quarry]

@fleet bridge :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | bar
002 | foo

[fleet bridge]

cool

[fleet bridge]

it works

[rugged sparrow]

sweet

[rugged sparrow]

ah you are trying to write mixins

[fallow leaf]

Yes

Minecraft like mixins, @mixin(target, etc)

[rugged sparrow]

seems every few months or so someone is in here trying to write python mixins lol

i have an implementation for 3.10 somewhere

[fallow leaf]

Nah I am really gonna do it

[rugged sparrow]

but so much changes with the bytecode every version it is very difficult to handle every situation

and mixins for builtin classes, do you mean like assembly patching?

[fallow leaf]

Possibly

[rugged sparrow]

ah yea, that stuff gets very finiky very fast

i am working on fishhook.asm currently to make assembly patching way easier, (currently it only supports complete hooking of C functions on x86 intel)

[fleet bridge]

is there any fishhook documentation?
i would like to see examples of how to use fishhook.asm.hook

[rugged sparrow]

i was avoiding writing docs for it, since i plan on changing it a ton when i make it more generalized, tbh only added that first version so i could use it for hacky stuff here

[fleet bridge]

can you just drop some examples somewhere?

[rugged sparrow]

!e ```py
from ctypes import *
from fishhook.asm import hook

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

pythonapi.PyUnicode_InternFromString.restype = py_object
interned = pythonapi.PyUnicode_InternFromString(b'MAGICVAL')
setdefault.unhook()
print(interned)```

[night quarry]

@rugged sparrow :white_check_mark: Your 3.11 eval job has completed with return code 0.

{'type': 'type', 'AttributeError': 'AttributeError', '__qualname__': '__qualname__', 'obj': 'obj', 'update': 'update', '__dict__': '__dict__', 'getattr': 'getattr', 'setattr': 'setattr', 'hasattr': 'hasattr', '__doc__': '__doc__', '__name__': '__name__', '__module__': '__module__', 'replace': 'replace', 'old': 'old', 'new': 'new', 'sys': 'sys', 'name': 'name', '_DeadlockError': '_DeadlockError', 'waiters': 'waiters', 'count': 'count', 'owner': 'owner', 'wakeup': 'wakeup', 'lock': 'lock', 'allocate_lock': 'allocate_lock', '_thread': '_thread', 'self': 'self', 'add': 'add', 'get': 'get', '_blocking_on': '_blocking_on', 'set': 'set', 'get_ident': 'get_ident', 'seen': 'seen', 'tid': 'tid', 'me': 'me', 'release': 'release', 'acquire': 'acquire', 'has_deadlock': 'has_deadlock', 'RuntimeError': 'RuntimeError', 'id': 'id', 'format': 'format', '__repr__': '__repr__', '__init__': '__init__', '_ModuleLock': '_ModuleLock', '_DummyModuleLock': '_DummyModuleLock', '_lock': '_lock', '_name': '_name',
... (truncated - too long)

Full output: https://paste.pythondiscord.com/DVSANU65G53QEDR4OSQUBBHXZU

[rugged sparrow]

heres an example i made to grab the interned dict

it doesnt work with every function, some are optimized in ways that bypass the things it patches

[fleet bridge]

i also would like to see type annotations in fishhook
are they planned in the future?
do you need any help? i think im pretty good at this, i can help you

[rugged sparrow]

honestly, i would love help with that, I have never realy felt the need to add type hints to any of my projects

also i really need to write a full docs for fishhook, there are a few features that are not documented at all (like fishhook.asm and @hook(cls1 | cls2))

if you want to make pull requests, be my guest, i look thru and merge them

[fleet bridge]

• is it ok to run black on your code?
• import typing, import typing as t or from typing import ...?

also, which python versions are supported by fishhook?

python_requires='>=3.8', i see...

[rugged sparrow]

sure you can run black, and for the typing import i don't have a preference

[rugged sparrow]

yea i first wrote fishhook for 3.8 so i wanted to retain support

[fallow leaf]

Isn't _code_.co_code the bytecode of the function?

[fleet bridge]

it is

[fallow leaf]

Coul have just told me that 😐

Now I can just change the bytecode

https://rushter.com/blog/python-bytecode-patch/

[fleet bridge]

i discovered bug in fishhook
if you use same wrapper for two methods with different names, it will overwrite wrong method

from fishhook import hook, orig
class X:
    def foo(self):
        print('foo@1')
    def bar(self):
        print('bar@1')

h = hook(X)

@h
def foo(self):
    print('foo@2')
    orig(self)

@h
def bar(self):
    print('bar@2')
    orig(self)


x = X()
x.foo()
print()
x.bar()

[rugged sparrow]

yea that makes sense, the wrapper returned by hook has some internal state stuff

[rugged sparrow]

that tutorial is quite out of date, just a warning

[rugged sparrow]

https://github.com/chilaxan/fishhook/blob/master/fishhook/fishhook.py#L321 this line is the reason, the first use of the wrapper sets name so code.co_name is never regrabbed from the second function object

[night quarry]

fishhook/fishhook.py line 321

name = name or code.co_name```

[fleet bridge]

yes, i know

i discovered this bug by looking at this exact line

fishhook is actually very hard to type-annotate...

especially stuff like @hook.cls

[rugged sparrow]

Yea it has some weird code to facilitate the nicer syntax

Working on the better fishhook.asm using keystone and capstone for disassembly and reassembly

[fleet bridge]

https://pastebin.com/6S2DsaM0

first attempt to annotate fishhook.py

im not satisfied with this at all

[astral rover]

whens the mypy plugin happening?

[fallow leaf]

import dis
import types

def my_function(x):
    if x > 0:
        return x * 2
    else:
        return x

original_code = my_function.__code__

new_instructions = []
instructions = list(dis.get_instructions(original_code))
before_instr = None
newopcode = dis.opmap["BUILD_TUPLE"]
for instr in instructions:
    if before_instr is not None and before_instr.opcode == dis.opmap["LOAD_CONST"] and instr.opcode == dis.opmap['RETURN_VALUE']:
        new_instr = dis.Instruction(
            opcode=newopcode,
            opname=dis.opname[newopcode],
            arg=0,
            argval=None,
            argrepr="()",
            offset=instr.offset,
            starts_line=instr.starts_line,
            is_jump_target=instr.is_jump_target
        )
        new_instructions.append(new_instr)
    else:
        new_instructions.append(instr)
    before_instr = instr

modified_code = types.CodeType(
    original_code.co_argcount,
    original_code.co_kwonlyargcount,
    original_code.co_nlocals,
    original_code.co_stacksize,
    original_code.co_flags,
    b''.join(instr.opcode.to_bytes(1,"little") for instr in new_instructions),
    original_code.co_consts,
    original_code.co_names,
    original_code.co_varnames,
    original_code.co_filename,
    original_code.co_name,
    original_code.co_firstlineno,
    original_code.co_lnotab,
    original_code.co_freevars,
    original_code.co_cellvars
)

modified_function = types.FunctionType(modified_code, globals())

result = modified_function(3)
print(result) 
``` don't know what's wrong

#bot-commands message

^ Error here

Thats on 3.11 but on 3.6.6 its Traceback (most recent call last): File "/storage/emulated/0/qpython/projects3/mixins/.last_tmp.py", line 134, in <module> result = modified_function(3) File "/storage/emulated/0/qpython/projects3/mixins/.last_tmp.py", line 80, in my_function if x > 0: IndexError: tuple index out of range

This way I can actually append a whole methods body to the instructions

[rugged sparrow]

Ah all of your jump targets are going to need to be recomputed if you are changing the length of the byte code

[rugged sparrow]

Honestly, it looks pretty good until you get to hook, then you can tell very quickly that fishhook does not like typing

[fallow leaf]

ok ..

[rugged sparrow]

#esoteric-python message here is a 3.10 mixin implementation I wrote a while back

Oh wait the paste got wiped

[versed eagle]

old pastes do that :(

[rugged sparrow]

I'll go find the source

[fleet bridge]

Annotating build_orig (and everything inside) is painful, because there is a lot of magic happening, so we should slap Any almost everywhere

[rugged sparrow]

Yea that makes sense, build_orig does a ton of weird stuff to try and ensure it works in any situation

[fleet bridge]

I think annotating only public api (and some easy non-public functions) makes more sense, than annotating everything

So the most painful part is hook.* things

[fallow leaf]

No idea how I would do that 💀

At this point I'ma just use fishhook for mixins and use that for the modifying logic

[rugged sparrow]

i found the source but it doesnt even work on 3.10 anymore rip

[fallow leaf]

Alright so when I do something like if x==z: print() that's a jump correct? And the jump target is print correct?

Cuz idk

[rugged sparrow]

yea

and the jump target in bytecode is going to be an offset

[quartz wave]

um
this doesn't possibly skip a return does it?

[exotic stirrup]

Guys, I have a question, is it possible to get the media list on the last call of the function when the number of calls is unknown?

import random


media = []


def containing_function():
    media.append('PHOTO')

    print(media)


for _ in range(random.randint(1, 5)):
    containing_function()

[quartz wave]

it does skip a return

[quartz wave]

import dis
import types

def my_function(x):
    if x > 0:
        return x * 2
    else:
        return x

original_code = my_function.__code__

new_instructions = []
instructions = list(dis.get_instructions(original_code))
before_instr = None
newopcode = dis.opmap["BUILD_TUPLE"]
for instr in instructions:
    if before_instr is not None and before_instr.opcode == dis.opmap["LOAD_CONST"] and instr.opcode == dis.opmap['RETURN_VALUE']:
        new_instr = dis.Instruction(
            opcode=newopcode,
            opname=dis.opname[newopcode],
            arg=0,
            argval=None,
            argrepr="()",
            offset=instr.offset,
            starts_line=instr.starts_line,
            is_jump_target=instr.is_jump_target
        )
        new_instructions.append(new_instr)
    new_instructions.append(instr)
    before_instr = instr

modified_code = types.CodeType(
    original_code.co_argcount,
    original_code.co_kwonlyargcount,
    original_code.co_nlocals,
    original_code.co_stacksize,
    original_code.co_flags,
    b''.join(instr.opcode.to_bytes(1,"little")+(instr.arg.to_bytes(1,"little") if instr.arg is not None else b"\0") for instr in new_instructions),
    original_code.co_consts,
    original_code.co_names,
    original_code.co_varnames,
    original_code.co_filename,
    original_code.co_name,
    original_code.co_firstlineno,
    original_code.co_lnotab,
    original_code.co_freevars,
    original_code.co_cellvars
)
``` should work for 3.6 now

[fallow leaf]

It shouldn't

[quartz wave]

it does

tested out the working thing before i changed this part ```diff
new_instructions.append(new_instr)

• else:

•    new_instructions.append(instr)
  

• new_instructions.append(instr)

[quartz wave]

jump offsets aren't affected for this specific case

[fallow leaf]

Ok what if I added more than just a different return value? Does that change it? Like what if I change the return to another if-return

[fallow leaf]

6

For 3

So doesn't

[quartz wave]

here's the modified code from above ```pycon
5 0 LOAD_FAST 0 (x)
2 LOAD_CONST 1 (0)
4 COMPARE_OP 4 (>)
6 POP_JUMP_IF_FALSE 16

6 8 LOAD_FAST 0 (x)
10 LOAD_CONST 2 (2)
12 BINARY_MULTIPLY
14 RETURN_VALUE

8 >> 16 LOAD_FAST 0 (x)
18 RETURN_VALUE
20 LOAD_CONST 0 (None)
22 BUILD_TUPLE 0
24 RETURN_VALUE
``` if the highlighted LOAD_FAST 0 (x) moves from position 16 by instruction addition/removal, the argument to POP_JUMP_IF_FALSE must be changed

[quartz wave]

what should be the output?

[fallow leaf]

()

[quartz wave]

nothing jumps to the final BUILD_TUPLE and RETURN_VALUE

[fallow leaf]

💀

[quartz wave]

which is the only place that contains LOAD_CONST + RETURN_VALUE

[fallow leaf]

Hold on

Ohhh the value is before the RETURN_VALUE opcode?

[rugged sparrow]

@fleet bridge a lot of operating systems don't let me map READ/WRITE/EXEC memory anymore, i can only do pairs of READ/WRITE or READ/EXEC. normally that would be fine, but since i have to remap entire memory regions to write in new assembly, sometimes there are functions in the same region that need to get called while the region is not marked executable. (i think im going to have to write my entire mprotect(R/W), writeASM, mprotect(R/X) in assembly

[quartz wave]

i might have to use slightly different syntax to implement dynamic goto/label in cpython

[versed eagle]

are you doing computed goto?

[quartz wave]

maybe

jumping in and out of anything that needs init code is hard

[rugged sparrow]

#esoteric-python message I got it working up here

Do you want my source?

[quartz wave]

that's with tracing and stuff

[rugged sparrow]

True but it's something to start with right?

[quartz wave]

maybe

i might use that after i'm done with the math.factorize() thing

[maiden blaze]

golf: https://leetcode.com/problems/longest-common-prefix/description/
my best so far ||return "".join(x for x,in __import__("itertools").takewhile(lambda g:len(g)==1,map(set,zip(*strs))))||

[quartz wave]

base 136 ||```py
class Solution:longestCommonPrefix=lambda _,s:"".join(x for x,in import("itertools").takewhile(lambda g:len(g)==1,map(set,zip(*s))))

136 -> 116 has an issue ||```py
class Solution:longestCommonPrefix=lambda _,s:s[0][:next((i for i,x in enumerate([*zip(*s),(1,2)])if~-len({*x})),0)]

136 -> 112 turns out it was not an issue ||```py
class Solution:longestCommonPrefix=lambda _,s:s[0][:next(i for i,x in enumerate([*zip(*s),(1,2)])if~-len({*x}))]

[maiden blaze]

whats the (1, 2) for

[quartz wave]

112 -> 110 good point, replaced it with a better one ||```py
class Solution:longestCommonPrefix=lambda _,s:s[0][:next(i for i,x in enumerate([*zip(s),s0])if~-len({*x}))]

it's a stopper

[maiden blaze]

[] is shorter than s*0 but i dont get what its for

[quartz wave]

oh yeah

110 -> 108 s*0 -> () + next() -> listcomp[0] ||```py
class Solution:longestCommonPrefix=lambda _,s:s[0][:[i for i,x in enumerate([*zip(*s),()])if~-len({*x})][0]]

[quartz wave]

preventing out-of-range things

[maiden blaze]

what case does it fail for without it

[quartz wave]

list of strings are empty

[maiden blaze]

s[0] would fail for that anyway

[quartz wave]

hmm true

[maiden blaze]

oh wait if a single string is empty

[quartz wave]

oh

no shared prefix

[maiden blaze]

no ["x", "y"] works without it

["x", ""] errors without it

[quartz wave]

hmm yea

[maiden blaze]

["x", "xy"] too

dafuq

ok ill just accep its necessary

[quartz wave]

it's when the string of minimum length gets consumed whole

i.e. all the other strings start with it

[maiden blaze]

a right

[maiden blaze]

@quartz wave 97

class Solution:longestCommonPrefix=lambda _,s:s[0][:[1==len({*x})for x in[*zip(*s),()]].index(0)]

[fleet bridge]

:(

Cant you:

1. mark memory as W/E
2. change code
3. mark as R/E
4. execute code

[rugged sparrow]

Can't have a segment be Writeable and Executable at the same time

So I'm gonna write my writeASM stub in assembly

[fleet bridge]

1. mark memory as W
2. change code
3. mark as R/E
4. execute code

Fixed

[rugged sparrow]

The problem with doing that in python is that I have to mark an entire 16k page, so if another function is in that page and is called during the patch when the page isn't executable I get a bus error

[fleet bridge]

I see

Yeah, doing that in asm now makes sense to me

---

Is it possible to have object that is not an instance of object?

x = ...
assert not isinstance(x, object)

[rugged sparrow]

Via memory tricks you can make a new base type

Then I think the resulting objects would not be instances of object

[fleet bridge]

What is the least amount of hacks you have to do to achieve that result?

class X(object): ...
x = X()
X.__base__ = X # memory trick requred
assert not isinstance(x, object)

Im not sure that this will work

Probably you also should modify bases, mro and something else mro-related

This will always be funny to me: ```py

class M(type): ...
...
class X(M, metaclass=M): ...
...
X.class = X
X is type(X) is X.class
True
isinstance(X, X)
True
issubclass(X, X)
True

[vast wave]

doesnt python have the GIL

[rugged sparrow]

yea, thats why using direct asm would work

[zenith bluff]

i wonder if it would be possible for lambda to operate on a whole list at once

[rugged sparrow]

Like map a lambda across the list?

[quartz wave]

85 cw problem [pdsn base](#python-discussion message) ```py
minimum_number=m=lambda n:not(t:=sum(n))>1==all(t%i for i in range(2,t))and-~m(n+[1])

84 ```py
minimum_number=m=lambda n:not~-(t:=sum(n))*all(t%i for i in range(2,t))and-~m(n+[1])

still 84 but performance boost ```py
minimum_number=m=lambda n:not~-(t:=sum(n))*all(t%i for i in range(2,t))and-~m([t+1])

[sick hound]

how do I join this cult

[distant salmon]

81 but with a performance degradation

minimum_number=m=lambda n:(all(sum(n)%i for i in range(2,sum(n)))or 2+m(n+[1]))-1

[quartz wave]

79 ```py
minimum_number=m=lambda n:~-all(sum(n)%i for i in range(2,sum(n)))and-~m(n+[1])

[unique heath]

whatcha golfing

[quartz wave]

.

[unique heath]

wait where r you checking prime

[quartz wave]

the major part of the program ```py
all(sum(n)%i for i in range(2,sum(n)))

[unique heath]

oh thats the prime check

[sick hound]

/run