python-discussion

[autumn forge]

you could have a routine like def gui_button(x, y, w, h, text) -> bool: ... that returns whether or not the user has clicked on it that frame

[terse mauve]

!code

[edgy kraken]

Formatting code on Discord

Here's how to format Python code on Discord:

```py
print('Hello world!')
```

These are backticks, not quotes. Check this out if you can't find the backtick key.

For long code samples, you can use our pastebin.

[autumn forge]

you have a space after "py"

[sharp sinew]

Idk why but in discord the “ becomes that but in an interpreter it changes.

[terse mauve]

you see the "py" after the first 3 backticks? that's the trick @sharp sinew

[sharp sinew]

So there’s no space.

[terse mauve]

correct

[sharp sinew]

So the thing is that there is no space

[terse mauve]

!raw

[edgy kraken]

You are not allowed to use that command here. Please use the #bot-commands channel instead.

[terse mauve]

huh

[sharp sinew]

I can send a photo to prove there’s no spaces

[arctic lantern]

print("what's yalls least favorite command in python?")

[gleaming knoll]

IMMEDIATE MODE GUI
LETS GOO

[autumn forge]

immediate mode is the only way I can do gui without being very sad

[charred tusk]

WTH is immediate mode

[autumn forge]

the idea is that you render everything every frame and the framework doesn't store much (any) state

so a retained mode gui might look like thispy app = App() def button_callback(): print('Click') app.add_widget(Text('Hello!')) app.add_widget(Button('Click me', cb=button_callback)) app.run() and an immediate mode gui might look like```py
while not window.should_close():
text('Hello!')
if button('Click me'):
print('Click')

[granite wyvern]

Not in this channel you can't. 🙂 Open a topic in #1035199133436354600 and we'll have a look.

You're clear about the smart quotes versus ASCII " double quote symbol?

[raw bramble]

That sounds... very resource consuming

[pallid garden]

does retained UI exist?

i dont know any game engine that does that

[raw bramble]

python2 input() I think

[pallid garden]

your renderer does this internally anyway

[autumn forge]

computers are fast

[raw bramble]

Checking the logic every single frame?

what am I on about 😭

[pallid garden]

i think it's cheaper to do that, than to check whether to rerender the component or to leave it same as the previous frame

i think

[raw bramble]

Is there much benefit in Programming that comes from understanding maths

[pastel sluice]

also, most GUI stuff is done on the GPU at this point, so the tax for redrawing everything is actually minimal

[autumn forge]

immediate mode frameworks are well-suited to games

oopsie

[pallid garden]

i was like: "really???"

but yea

[pastel sluice]

if you're solving problems that chiefly involve math, yes

most programming problems are about planning and design rather than math as such

[narrow tiger]

Feel like a mathematical way of thinking helps with programming too

[pastel sluice]

what many people call "mathematical" might better just be called "logical"

[narrow tiger]

Maybe but I like thinking of things in functions, compositions, etc. and knowing how data structures work mathematically makes it feel more natural than just memorizing as well

I am aware I could be in the minority on this one though lol

[pastel sluice]

for most data structures, all you really need to know is their time and memory complexity

[raw bramble]

I've been looking at Logic and Foundations of Maths and I'm really really interested, but I don't really see how it can be useful when thinking about Programming

[pastel sluice]

in all the projects I've done over the years, actual math has come into play maybe once or twice

[raw bramble]

(not logic, logic being in programming is obvious)

[pastel sluice]

the vast, vast majority of time, it's been about logic, design, planning, and research

[pallid garden]

mods, make this chatter write a game engine

[autumn forge]

logic is how you trick people into doing maths

[raw bramble]

The most important parts of programming are the hardest parts. Coding is easy, planning is hard

[narrow tiger]

I came from a physics background so my first brush with programming was all math, thats probably why I think this way haha

[raw bramble]

I love Physics but I just can't do it 😭 at a certain point my understanding just breaks down

[dusty ember]

you might find some benefit in studying like, graphs and pathfinding and the like (so stuff that might be covered in a discrete math class in uni), and maybe some algorithm stuff (so DSA, but then we're kind of overlapping into CS theory)

you probably won't gain much from learning calculus (except perhaps building your problem solving skills in general)

[narrow tiger]

So yeah I guess its more like infinimata says, depends on the problems you are solving... maybe you'll do a lot of math, maybe none at all. But I like the problem solving philosophy I gained from thinking that way (akin to logic I guess)

[autumn forge]

unless you enjoy calculus of course

[dusty ember]

yeah that too

i am looking forward to linalg next semester

much more than calculus at least 🥴

[raw bramble]

I don't know if I just know very, very little about calculus, but differential calculus doesn't seem super big and scary, although I know nothing about integral calc

[dusty ember]

calc 1 and 3 weren't bad at all, but calc 2 was terrible for me.

im also just not great at physics so i just kind of fell apart when we did physical applications

[raw bramble]

Physics is amazing and my favourite science, but it's also the worst and evil and horrible

If there were any actual real rules in Physics it'd be great, the only rule I can think of is the speed of light/causality being constant in a vacuum

[crisp jay]

physics is very cool

[wise yarrow]

it's very relevant to formally reasoning about programs, which may be helpful when you're intuitively reasoning about programs

not necessary tho imo

[fathom yacht]

I just learned For Loop in python and I noticed that I had to use if statement in that loop

It left me confused about why I needed to use the if statement in the loop.

Is it because I don't need to write the code repeatedly when one uses the code?

I tried asking ai but their bot answers make me more confused

[pastel sluice]

@fathom yacht show the code in question

[fathom yacht]

nombor_campur = [1, 2, 3, 4, 5, 6]
genap = []
ganjil = []

for a in nombor_campur:
    if a % 2 == 0:
        total = genap.append(a)
    else:
        titi = ganjil.append(a)
        
        
print(genap)
print(ganjil)

It might confusing a bit as I'm using my own language as variable

[pastel sluice]

the if here is to decide whether or not to do something at each turn of the loop.

[fathom yacht]

Is the loop functioned so it can automatically decide the decision?

[pastel sluice]

I'm not sure I understand what you mean by that.

[fathom yacht]

I mean like,I understand the if statement one,but I have slight confusion on why I need to use For Loop in that code.

I've been practicing that but never know the purpose of it

[pastel sluice]

the for loop is to go through each element in nombor_campur, and then use the if statement to decide what to do with it.

run this program:

numbers = [1,2,3,4]
for n in numbers:
  print(n)

and after you run that one, run this one:

numbers = [1,2,3,4]
for n in numbers:
  print(n*2)

see if that gives you some sense of what it means to say for n in something

[fathom yacht]

Ahh..I see..so n is just a random name that we give right?

[pastel sluice]

correct, it's just a name used for temporary assignment for each turn of the loop

(it persists after the loop if we want to use the last value it held for something, too)

[fathom yacht]

Ooh.. alright.Thanks for the help mate🙏

[raw bramble]

Can sets contain sets?

I can't convert a list of lists into a set because a list isn't hashable, so I'm wondering if I make it into a set of sets would it work?

[pallid garden]

!e

nested_sets = {{1}, {1}}
print(nested_sets)

[edgy kraken]

:x: Your 3.14 eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "/home/main.py", line 1, in <module>
003 |     nested_sets = {{1}, {1}}
004 |                   ^^^^^^^^^^
005 | TypeError: cannot use 'set' as a set element (unhashable type: 'set')

[gleaming knoll]

a set of frozensets
or a frozenset of frozensets

[raw bramble]

are frozen sets in typing or smth?

[gleaming knoll]

no, just builtin

!d frozenset

[edgy kraken]

frozenset

class set(iterable=(), /)``````py

class frozenset(iterable=(), /)```
Return a new set or frozenset object whose elements are taken from *iterable*. The elements of a set must be [hashable](https://docs.python.org/3/glossary.html#term-hashable). To represent sets of sets, the inner sets must be `frozenset` objects. If *iterable* is not specified, a new empty set is returned.

[vale wasp]

Or a set of tuples.

[raw bramble]

The order of frozensets is random?

{frozenset({1, 2}), frozenset({21, 22, 23, 24, 25, 26, 27}), frozenset({15, 16, 17, 18, 19, 20}), frozenset({36, 37, 38, 39, 40, 41, 42, 43, 44}), frozenset({3, 4, 5}), frozenset({10, 11, 12, 13, 14}), frozenset(), frozenset({32, 33, 34, 35, 28, 29, 30, 31}), frozenset({0}), frozenset({8, 9, 6, 7})}

[vale wasp]

Converting to a frozenset would dedupe and lost the ordering. If that matters, you can put a tuple into a set.

[pallid garden]

sets dont have order

[raw bramble]

oh yeah 😭

[vale wasp]

Sets are unordered by definition.

[raw bramble]

Sets are unsorted immutable

I forgoted

[vale wasp]

Sets are mutable, that's why you can't put one into a set.

Tuples are ordered and immutable.

[raw bramble]

wait no 😭

Which ones are unsroted and immutable

its some frozen thing

[sand ruin]

Fronzenset

[vale wasp]

Frozenset is an immutable set.

[sand ruin]

Me first haha

[raw bramble]

Yeah I heard about those once before, I know something about frozen dicts

[vale wasp]

That's in 3.15

[sand ruin]

Where got frozen dict

[raw bramble]

mhm that's where I saw it!

[raw bramble]

frozen dicts are in 3.15

[sand ruin]

You are in dev 3.15?

[vale wasp]

frozendict to be precise 😃

[sand ruin]

Or beta/alpha?

[vale wasp]

Currently alpha. Bata is in a week or so.

[gleaming knoll]

there are mutable things even in the stdlib which are hashable
never liked that logic that "mutable -> not hashable"
the hash implementation can just not depend on the part which is mutable

[granite wyvern]

We use the word "frozen" when describing immutable things. After they're made, they may not be changed.

[raw bramble]

You can download it, I wouldn't recommend that you use it as a replacement to any previous version just yet

[sand ruin]

Hey, i should try those too, speaking of which

[vale wasp]

@raw bramble if you care about order and/or want duplicates and immutability, use a tuple. If you don't care about both order and duplicates, use a frozenset.

[raw bramble]

forzendict

[sand ruin]

OALD is frozendict

[raw bramble]

Alright thank you :D im messing about with collections to discover what things I can do with them

[sand ruin]

👍

[vale wasp]

True. But I wouldn't want to use something mutable as a dictionary key.

[sand ruin]

You can't

[vale wasp]

The tl;dr - tuple is the immutable version of list and frozenset is the immutable version of set. Does that make sense?

[raw bramble]

yeah

[vale wasp]

If it's Hashable you can use it as a dictionary key or put it in a set.

[raw bramble]

it also makes sense as to why tuples are rarely ever useful then 😂

[sand ruin]

Keys mudt be immutable and so hashable

[granite wyvern]

I'm happy to. I've got various things which are effectively proxies for eg database objects. Like tagged objects. The key of the object defines the hash and equality, and never changes. but the tags may be updated. A bit like related record in a db, if you like.

[sand ruin]

Pwrhaps u meant floats

[vale wasp]

They are extremely useful.

[sand ruin]

Why

[gleaming knoll]

the only requirement is to be __hash__able and __eq__able (and for the hashes of equal objects to match, so collisions can be correctly handled)

[sand ruin]

I use list, not tuple

[raw bramble]

I've only really needed to use them when returning multiple values from a function, I can just use a list almost all of the time otherwise

[vale wasp]

I'd be wary of mutating it... Although I do have a subclass of dict that has a __hash__ so I can use functools.cache with it. But I do not mutate it.

[sand ruin]

And only immutable are hashable so the requirement ectentds to immutable 😉

[gleaming knoll]

where exactly do you get that from? what prevents a mutable object from defining __hash__ and __eq__?

[slender urchin]

You can easily make a mutable object hashable

[sand ruin]

I thought docs

[granite wyvern]

My example considersthe key of the object (whatever that means in the circumstance) to be the immutable part. Tha tags are not involved.

[vale wasp]

Make a dictionary representing a grid where the key is the x, y coordinates.

[sand ruin]

Yes

Oh

[raven urchin]

BWAAA

[sand ruin]

Haha, im confused

[raven urchin]

I BEAT DAY 1 OF AOC IN GO

[slender urchin]

Congrats

Please toggle your caps key off

[vale wasp]

Hashable means it implements __hash__

[vale wasp]

🎉

[raven urchin]

Thanks to Bananabun for motivating me

[vale wasp]

What year?

[raven urchin]

2020

[granite wyvern]

Oh, BTW: https://mastodon.social/@climagic/116500742252613422

Command Line Magic (@[e-mail])

BTW in case you didn't notice, tomorrow at 1777777777 it's going to be 1777777777 epoch time.

Mastodon · May 1, 2026, 6:30 PM

[raw bramble]

oh I would make my own class for an immutable 2D Vector (with hash and eq method don't you forget it)

[raven urchin]

The one with the nested loop

Now I have to figure out how to read files in Go to use the actual input

Where's my with open() 😭

[vale wasp]

Tuple works wonderfully and if you need to navigate in it it's probably easier to have a tuple than a class. I use this for AoC every year.

[vale wasp]

You can just have it as a string constant.

[raven urchin]

Problem is, the input is giant and I have it on a different file.

Also, a good opportunity to learn how to file manage in go.

[gleaming knoll]

technically even stuff like list could define __hash__ in a way that would conform to the (x == y) => (hash(x) == hash(y)) property (it would be the same as tuple's hash)
but it would still have a problem with being used as a dict key, because the entry's position in the backing array of the dict is determined by the initial hash of the key, and that hash is cached
so its.. complicated

[rancid fractal]

In go? Smth like

func readFile() {
  content, err := os.ReadFile("file.txt")
  if err != nil {
    fmt.Println("Error", err)
    return
  }

  fmt.Println(string(content))
}

[raven urchin]

Ty

[slender urchin]

This is the python chat everyone.

[granite wyvern]

I think that's the equiv of Python's open("file.txt").read() and sucks all the data into memory.

[rancid fractal]

The formatting messed me up a bit but this should work.

[rancid fractal]

Yep

[vale wasp]

Yup. You can easily end up with a situation where you can't retrieve the value by key.

[dry pike]

what's up skibidawgs

[granite wyvern]

Reading in bits or line by line is also easy in Go, but I forget the details.

[dry pike]

Go can read bits, not just bytes?

[vale wasp]

Path('file.txt').read_text() 😃

[sand ruin]

What type of keys are mutable?

[granite wyvern]

We're nerd sniping DanS into making a dict subclass which works with unhashable mutable keys.

[dry pike]

none builtin

[rancid fractal]

I think you'd need bufio.Scanner for line-by-line

[sand ruin]

Tks

[vale wasp]

No.

[dry pike]

you could design a class which is hashable and mutable

[gleaming knoll]

whatever type open() returns (a file handle), for one
function too (mutable in atleast 2 ways)

[sand ruin]

right

[dry pike]

tuples which contain non-hashable elements are themselves non-hashable

[vale wasp]

I have. Many times. I use that so I can dedupe a list by converting to set.

[dry pike]

e.g. (1, 2, [], "abc")

because list is not hashable, then this tuple is not hashable either

[sand ruin]

Ooo, using a file handle as a key is a prime example of the near limitless potential of python

[dry pike]

tuples are only hashable when all of their elements are hashable

[vale wasp]

The hash of a tuple is the hash of its contents.

[sand ruin]

so an immutable (tuple) can have their _hash_ defined or not depending on its elements (list)?

[dry pike]

functions can be keys? wow

[proper brook]

can i hire python discord bot developers here? i need to add a small feature

[gleaming knoll]

they are hashed by id, yeah

[slender urchin]

no

[dry pike]

ah so, must something explicitly be marked as unhashable?

[gleaming knoll]

yeah, with __hash__ = None
object.__hash__ is the id based hash, so all classes get it by default

[proper brook]

alr

[vale wasp]

{
    lambda x: print(x): 1234
}

[dry pike]

__hash__ is always defined for tuples. It's a method. What happens is that tuple's hash method calls the hash method of each of the elements as part of the hashing, so an error will be raised when it tries to hash the list (or other unhashable)

[sand ruin]

I thought that is against the rules?

[granite wyvern]

Well, tuple itself requires its elements to be hashable for its hash function to work. But you could make a subclass with whatever behaviour you liked. Whether that makes sense depends on what you do.

[proper brook]

i was asking

[vale wasp]

You are allowed to ask if you can do something.

[sand ruin]

RIght. However, it's against common practice to subclass bytes and not have it be bytes all implement all bytes feature, I think

[granite wyvern]

!rule paid

That's fine. (meaning the asking.)

[edgy kraken]

Rules

9. Do not offer or ask for paid work of any kind.

[gleaming knoll]

the method is always defined, but whether calling it will succeed or end up erroring depends on the contents of that specific tuple (since it will "recursively" call the hash of the elements, and that could either end up being explicitly set to None like for list, or end up going through a similar process if the element itself is a structure, and still end up raising an exception)

[sand ruin]

tks

[dry pike]

what does tks mean

[sand ruin]

thanks

[granite wyvern]

thx

[dry pike]

o

[vale wasp]

ty

[dry pike]

ty

[vale wasp]

10q

[sand ruin]

ok, abbrev mode?

[granite wyvern]

okay

[dry pike]

benq

[vale wasp]

K

[sand ruin]

with slowdown must be painful 😢

[dry pike]

I never slow down

[sand ruin]

I hv slowdown

[dry pike]

it's about acceleration

[sand ruin]

sure, thx

[dry pike]

it's like time pauses

not that I am going slower

[vale wasp]

.topic

[verbal wedge]

**When you were first learning, what is a resource you wish you had?**

Suggest more topics here!

[vale wasp]

The internet.

[sand ruin]

myself

We are our best teacher

[slender urchin]

You didn't have yourself?

[sand ruin]

if knowledge wise, fine, an external teacher or docs

no, it's sad, I know

[dry pike]

🪦🔔

[tall fjord]

bro who wants to go on omegle and rizz

[sand ruin]

the internet is large, if u know what i mean

[dry pike]

what does that have to do with Python?

[sand ruin]

r u sure its the internet?

[tall fjord]

IDK gang

who cares about python fr

[vale wasp]

Don't feed the troll.

[sand ruin]

that's why we have mods and ops in irc

[tall fjord]

Im not trolling

[sand ruin]

got banned there, created wa groups, banned others, live a life, learn a life, come to discord

rules n bans r necessary if not abused

can we really accelerate py learning here instead of reading the docs? only when no one has answers here, we read the docs. what do we think? agree or disagree?

[pallid garden]

that implies that if abused, you dont need rules and bans?

[sand ruin]

or would it feed time wasters and troublemakers?

[pallid garden]

reading the docs is faster than waiting for an answer?

[sand ruin]

i mean simply banning people is bad too

[cerulean ravine]

You should read the docs first

[sand ruin]

opp.

[sand ruin]

thats slow!

here instant

[pallid garden]

it's faster for me at least

[vale wasp]

Different people learn differently. For some people having a conversation and an explanation is a better learning experience. For others it's reading. For others it's watching a video.

[cerulean ravine]

If you don't put on effort people won't want to help you

[sand ruin]

right, i know of a resource mentioning some auditory, touch, visual

[celest osprey]

i think that there are plenty of answers here

[vale wasp]

[west cairn]

What do u say to a trash talker who says “ai will take over cs and coding”

[sand ruin]

I would, wouldn't I? unless I am busy or otherwise haha

[vale wasp]

Nothing.

[cerulean ravine]

You dont have to say anything to them

[west cairn]

Nothing as in “u r right” or “there’s no point of arguing to a brick wall”?

[pallid garden]

no point arguing

[vale wasp]

The latter.

[pallid garden]

thou art veritable

[west cairn]

Quit the shakespearness, I have had enough trauma in hs 😭😭

[vale wasp]

She's making fun of my bad tpying.

[cerulean ravine]

Ai is controversial now. People have strong opinions and fears. You probably can't change their mind

[pallid garden]

apparently this exists https://en.wikipedia.org/wiki/Shakespeare_Programming_Language

[vale wasp]

Dear lord, why?????

[pallid garden]

ive been laughing my ass off reading the syntax

[vale wasp]

I'm too scared to click that link.

[pallid garden]

it's amazing

[vale wasp]

That scares me even more.

.topic

[verbal wedge]

**Which Python project are you the most proud of making?**

Suggest more topics here!

[vale wasp]

!pypi cooklang-py I'm pretty proud of the work I did in this parser.

[edgy kraken]

cooklang-py v0.3.2

Python Parser for Cooklang

Released on <t:1757633493:D>.

[dry pike]

chat ded

[vale wasp]

It's Friday night here... Maybe folks have hot dates?

[sand ruin]

Cooldown...

[dry pike]

lucky folks

[peak relic]

oooooh

wait I remembered someone was doing the same thing

[vale wasp]

You sure that wasn't me? I've mentioned this here before.

[peak relic]

I would have remembered if it was you. L3vi wanted to do the same thing #python-discussion message

I guess now he knows he's got a competitor haha

well idk if he has published his yet

[vale wasp]

I didn't come up with cooklang. I just wrote a parser for it in Python. Because one didn't exist and I wanted something for recipes on my website.

[peak relic]

ah ok

[warm pike]

hello

[peak relic]

still tho, this is amazing!

[peak relic]

hello

[vale wasp]

👋

[vale wasp]

Thanks.

[warm pike]

this is cool

[vale wasp]

Thanks.

[sharp sinew]

Anyone know why when I do

print("Hello, world!")

The text isint colored?

[peak relic]

did you save it as something.py?

[sharp sinew]

No I mean in dc

[steady rain]

On discord or in a text editor?

[sharp sinew]

discord

[steady rain]

!raw 1499947196341293249

[edgy kraken]

== Raw message ==

Anyone know why when I do 
`​`​`​py
print("Hello, world!")
`​`​`​

[steady rain]

Maybe they changed the syntax colors again

[dry pike]

text is colored

you using iPhone?

[sharp sinew]

Yea

[dry pike]

iPhone doesn't support syntax highlighting in discord

[steady rain]

On Android only the string is colored

[sharp sinew]

Last time I tried on pc it worked

oh

[peak relic]

Is this ☝️ showing as colored to you?

[vale wasp]

Desktop works, not mobile.

[sharp sinew]

Nope

[steady rain]

Why iPhone?

[sharp sinew]

Thanks guys!

iphone is way behind

You cant install auto clickers, download smth to download an app, and much more!

[warm pike]

thats a shame

[sharp sinew]

But on android you can do all of that

[dry pike]

pretty sure that Discord could make it work if they really wanted to

[steady rain]

It's like having a totalitarian dictatorship installed on your device

[vale wasp]

They don't want.

[dry pike]

[steady rain]

They don't. Next!

[vale wasp]

That's no way to talk about our Cupertino Overlords!

[dry pike]

that's the answer to most iPhone questions

"because that's the way it is"

[sharp sinew]

I learn a lot from harvards python guide

[zealous lion]

at this point I use an iphone just to spite the people who complain about it

[sharp sinew]

These damn iphone devs are too lazy

[vale wasp]

The desktop app is electron. It's not just the mobile devs.

[zealous lion]

oh we’re taking stabs at electron now

you guys are taking all the easy ones

[spice hill]

Careful. When we took a stab at nuclei, they kinda exploded

[vale wasp]

Should we just stick a few together?

[sharp sinew]

Are the mobile devs responsible for not allowing auto clickers? Or does mobile make everything harder to code?

[zealous lion]

why do you need an auto clicker for discord

[sharp sinew]

For games I play not dc

[zealous lion]

oh that would probably be intentional by the game devs then

[west cairn]

😭😭😭

[sharp sinew]

My calculator

[vale wasp]

Nice.

GUI or CLI?

[sharp sinew]

CLI

[vale wasp]

Cool

[sharp sinew]

How do I even put guis?

[vale wasp]

There are lots of frameworks for GUIs. Even for TUIs.

[silver plover]

And what's the deal with airline food anyway?

[spice rover]

QLabel

[vale wasp]

What's QLabel?

[spice rover]

customizable labeling system surrounding the brother_ql library

[vale wasp]

Nice

[sharp sinew]

And the code for it was smth like this:

num1 = int(input("First number?"))
num2 = int(input("Second number?))
op = input("What are we doing?")
if op == "-":
    print(num1 - num2)
elif op == "+":
    print(num1 + num2)
elif op == "*":
    print(num1 * num2)
elif op == "/":
    print(num1 / num2)
if num2 != 0:
    print("I have no idea what but something went wrong.")

[hot sequoia]

hi

[sharp sinew]

hi

[hot sequoia]

what we are going to do

[sharp sinew]

idk

[sand ruin]

Haha

[hot sequoia]

lets code

i use trinket.io for python

[sand ruin]

Oh, u mean idle

Ok

[hot sequoia]

guys what are good websites for coding in python i use trinket.io

[steady rain]

There's replit and Google colab.

[brisk gazelle]

I prefer local development.

On my computer.

[dry yacht]

Is there some way to measure how much memory Python is using?
Ideally even one that shows you where the allocations are coming from, i.e. what lines are the "culprits" for a specific run?

[steady rain]

Memray

[hot sequoia]

oki ill try replit i only use trinket.io

[steady rain]

I've never used it, but people talk about it, and godlygeek makes it

[brisk gazelle]

Is the task manager application for your operating system not sufficient? Or are you looking for an in-Python determination?

[dry yacht]

Traces every function call so it can accurately represent the call stack, unlike sampling profilers.
Ouh, that's nice. So far I've only worked with sampling profilers

[sand ruin]

Why dont you use python idle or vs code for puthon coding?

Choice?

Or other reasons?

[steady rain]

No one really uses idle except beginners

[sand ruin]

Ouch

Why

[hot sequoia]

why is replit just vibe code

[steady rain]

Idle is a pretty simple program and was never intended for "hard core" use. Whereas VS code and PyCharm are thoroughly engineered with many thousands of hours of development time.

[brisk gazelle]

The CEO has a vision of the future for software development.

[hot sequoia]

wait im just a beginer

[brisk gazelle]

He's a very soft spoken but intense seeming fellow.

[hot sequoia]

i know print if elif else = == + - not and

also a bit more

[sand ruin]

What do you mean

[steady rain]

They're listing keywords and operators they know.

[hot sequoia]

huh what do you mean by "what do you mean"

[sand ruin]

Thx

[hot sequoia]

ohhh ok

[steady rain]

Except print, which is a function (don't at me)

[hot sequoia]

guys are you talking about the message i sent about the things i know

i know import too

[vale wasp]

It was a statement...

[brisk gazelle]

Hi, Opal here. IDLE user.

[hot sequoia]

ty

[steady rain]

Why

[charred tusk]

+1 on why

[hot sequoia]

who ya sayin it to

[sand ruin]

Because of me maybe

[vale wasp]

Why not?

[charred tusk]

No dark mode

[sand ruin]

Dk

[vale wasp]

That's answer to "why"

[charred tusk]

Unless they've added that in the 2 years since 2008

[hot sequoia]

why not use idle

[vale wasp]

I'm sure they'd happily accept a PR for that.

[sand ruin]

Ster says for beginners

Idk why

[hot sequoia]

oki im beginer

[brisk gazelle]

To convince people I'm unworthy of respect. IDLE is sufficient for my needs, which are not elaborate.

[steady rain]

I use pycharm, which has every feature you could ever want. And a bunch that you don't.

[hot sequoia]

oki i will use pycharm

[sand ruin]

I thought that credit goes to vs code?

[vale wasp]

Starting to tilt too much to the unwanted side, TBH.

[charred tusk]

I actually think its already has it
But its been a long time since I've used it

[vale wasp]

You are worthy of respect.

[steady rain]

Because of the AI features or what?

[charred tusk]

It's been like three years since I've even had Python installed on a host

[vale wasp]

Among other things. It's started getting slow.

[brisk gazelle]

You don't know me.

[hot sequoia]

hmm pycharm needs download and this is a borrowed computer for work

[charred tusk]

[vale wasp]

You are a human being. Therefore you are worthy of respect.

[charred tusk]

What do you use then?

[hot sequoia]

idk i use trinket.io on website

[charred tusk]

Wait a second... That's species-ist!!

[hot sequoia]

i use websites

[charred tusk]

Try Codespaces

[hot sequoia]

oki sure

[vale wasp]

I didn't say, "Only humans are worthy of respect." I said, "All humans are worthy of respect."

[charred tusk]

fine

That is correct

[hot sequoia]

codespaces no work i put my email and my email password and it no work

[brisk gazelle]

I don't get a good vibe from VSC nor PyCharm. A lot of communucation with motherships.

[vale wasp]

It's a shame you're a Doggo 😉

[rare gazelle]

hello

[dry pike]

now that's species-ist

[dusty ember]

the solution is just emacs

[brisk gazelle]

I'll look at that.

[hot sequoia]

guys
codespaces no work i put my email and my email password and it no work

[vale wasp]

nvim FTW

[hot sequoia]

codespaces no work i put my email and my email password and it no wor

[charred tusk]

Did you mean: vim?

[dusty ember]

emacs has the best vim motions out of every editor except vim itself

[rare gazelle]

had a feeling vim would be mentioned

[hot sequoia]

codespaces is not working

[vale wasp]

Doggo gets all the respect. And all the teasing.

[brisk gazelle]

Is it telling you that your credentials are wrong, or is the site failing in a more general way?

[hot sequoia]

dude no

my email is not working and my email password even if it says so

[brisk gazelle]

See the relevant FAQs of the sites and services in question.

[hot sequoia]

sorry but can you just give me other websites

[brisk gazelle]

Why would they know better?

[hot sequoia]

wdym

[brisk gazelle]

🫤

[dry yacht]

It even works on WSL, despite them saying it wouldn't work on Windows

[vale wasp]

WSL is not Windows...

[hot sequoia]

im on windows 11

[dry yacht]

Yeah, but still... like, it's also not really an Ubuntu (or whatever distro you chose)

[vale wasp]

Right but it works similarly enough to linux that things like that will work.

[little lava]

🦛

[hot sequoia]

bye

[little lava]

Hi chat

[dry yacht]

I've already had some issues with some profiling tools, because Linux exposes some sensors/APIs that Windows apparently doesn't, even through the WSL

[vale wasp]

Fair.

[placid needle]

guys, is python for data analysis an outdated book? i’m thinking bout picking it up

[dry pike]

Python for Data Analysis, 3E “Open Access” HTML version Updated for pandas 2.0.0 and python 3.10
Pandas is now at 3.0.1 and Python is at 3.14.4

so I would say outdated

[vale wasp]

3.15b1 comes out Tuesday....

[dry pike]

yeah but full release in October, right?

[vale wasp]

That is the target.

But you can, and should, play with the betas. See if you can find issues

[dry pike]

sure but I wouldn't expect a book to be updated to that

[paper kiln]

What is the use of @edgy kraken bot

[sand ruin]

If i get the latest python and chat here abt issues, will it be active?

I dont even know abt how to use bota

[vale wasp]

I wouldn't either. But if I were writing 4E now I would look at the betas and accepted PEPs and include references.

[vale wasp]

What do you mean?

[sand ruin]

What's 4E

[paper kiln]

No problem, i got the answer

[edgy solar]

Yoo

[vale wasp]

Current is 3E. So next is 4E.

[jagged belfry]

4th edition

[vale wasp]

👍

[sand ruin]

Oh

Thx

Hey Mom

[edgy solar]

How are you

[sand ruin]

Nonsense

[edgy solar]

WTF

[sand ruin]

No vulgar words here please

[edgy solar]

Okay

[sand ruin]

Thx

[edgy solar]

Yh

[sand ruin]

What is yh?

[edgy solar]

Yeah in full

[dry pike]

that would be YIF

"Yeah In Full"

[brisk gazelle]

Profanity is a spice. Best used, but in moderation and paired with the appropriate meal.

[sand ruin]

Mod?

U mod?

[vale wasp]

There are no rules against profanity on this server. Excessive use would probably be considered abusive.

[jagged belfry]

Correct

( this does not apply to slurs )

[sand ruin]

Ok

[fiery yarrow]

for the record: opal and dan aren't mods, but bast is

[raven urchin]

Survived day 1 of AOC in both Python and Go!!

[pallid garden]

Good job!

[sand ruin]

Just if i got banned, i hope the mods rethink and unban me or dm/pm me

[vale wasp]

Not mods but knowledgable nonetheless.

[raven urchin]

xD

[pallid garden]

there is a process to get unbanned

[vale wasp]

🎉

[sand ruin]

Ofc i need to be available and know the moderators are messaging me

[fiery yarrow]

barring especially egregious behavior, there will be warnings before expulsion

[pallid garden]

it's in the rules, i think

[sand ruin]

Oops 🫢TL;DR!!

[vale wasp]

Warnings and timeout unless it's extremely egregious from everything I've seen.

[fiery yarrow]

you even swiped my word choice there danno

[pallid garden]

it's in the coc: https://www.pythondiscord.com/pages/code-of-conduct/#appeal-procedure

[vale wasp]

You said "especially" and I went with "extremely", You left off the timeout though...

[fiery yarrow]

eh, a timeout is a type of warning

[sand ruin]

Ok, thx

[vale wasp]

There will be warnings prior to the first action which would be a timeout.

[brisk gazelle]

Sanction that can be taken as having warning attached to it.

[vale wasp]

And the mods are a lot more patient than I am.

[pallid garden]

than i am, too

[fiery yarrow]

unless the mute comes from pybot, of course

[mossy jewel]

hi guys, i have a quick question. I have a class which has 4 properties, p_up p_down p_left and p_right. These are all positive real numbers that sum to 1. I want to mutate these numbers by +-10%, while keeping the sum at 1. What is the best method to do this?

[sand ruin]

How to dm modmail

Where

[chilly whale]

@pseudo carbon

[spice hill]

How do you want to change these numbers? randomly?

[sand ruin]

Oh

[mossy jewel]

yes, so the numbers should randomly change by +-10%, but the sum must stay at 1

[sand ruin]

Thats food to hear

[vale wasp]

I'm old and cranky. What's your excuse?

[brisk gazelle]

Something something normalisation.

Like calculating percentages.

[spice hill]

Are these floats?

[sand ruin]

NFKC? 😆

[mossy jewel]

could you elaborate a bit please?

[spice hill]

Newly Fried Kentucky Chicken

[mossy jewel]

yes, initially generated by random.random

[fiery yarrow]

is the listed goal even always possible

[sand ruin]

Somehow i am in love with english (if it was a woman), int and float.

[sand ruin]

No 😂 unicode

[dry pike]

Kentucky Boiled Chicken

[spice hill]

What's the end goal though? Will you need to repeat that many times?

[mossy jewel]

basically im trying to learn some basic AI, and im making the next generation take the average "brain" (prob for moving down left right and up) of the previous generations 2 best players, and then mutate the probabilies by +-10%

so yes this will be repeated for each of the 5 players per generation, maybe 100s of times

[sand ruin]

my friend is working on ai too!

llvm to be exact

[mossy jewel]

thats awesome! im just playing around honestly, its been a bunch of fun so far

[vale wasp]

Create a list of 4 factors. [1.1, 0.9, 1.1, 0.9] shuffle the list. Multiply each one by the next factor in the list.

[spice hill]

Is it important that the change is under 10%?

[chilly whale]

the first thing I can think of is: ```py
vals = []
vals.append(random.uniform(0, 1.0 - sum(vals)))
vals.append(random.uniform(0, 1.0 - sum(vals)))
vals.append(random.uniform(0, 1.0 - sum(vals)))
vals.append(1.0 - sum(vals))
random.shuffle(vals)

[vale wasp]

Since it's +- 10% for each. That should work....

[sand ruin]

really? ai (programming) is daunting to me

[mossy jewel]

well it should'nt be too large, or else it will converge pretty slowly id think, 10% is just some magic number im starting with

[chilly whale]

oh, I missed the +-10% caveat. hm.

[vale wasp]

But listen to @chilly whale more than me. He's better.

[mossy jewel]

would this sum to 1 though?

[chilly whale]

not, necessarily, no

the trick to getting something that sums to 1 is getting 3 things that sum to less than 1, and then subtracting their sum from 1 to get the 4th thing

[vale wasp]

You can have 10% or sum to 1. One has to be flexible.

[mossy jewel]

hmm im thinking of what opalmist said originally, what if i just multiplied each value by some random real from [0.9, 1.1], and then took the sum of all 4, and did some math to normalize the sum to 1?

some sort of normalization

[zealous lion]

would it not just be max(min(num * random.choice([1.1, 0.9]), 1), 0)

[chilly whale]

you'll still wind up needing to subtract, if by "sums to 1" you mean "== 1.0"

[vale wasp]

The sum of all 4 numbers needs to be 1.0

[zealous lion]

four numbers?

[jade robin]

that code has other problems i think, you don't need exact 10

[mossy jewel]

@zealous lion yhis is the context

[chilly whale]

yeah. given 4 floats that sum to 1.0, perturb each of them by no more than 10%, and wind up with a result that still sums to 1.0

[spice hill]

Suppose that you ended up with the numbers 0, 0, 0, 1. From this position, you can't change any of the numbers.
Even if it's not so extreme, e.g. you have 0.1 0.1 0.1 0.7, even if each of the 0.1s gains the maximum of 0.01, 0.7 will only be able to get reduced by 0.03, which is only 4.3%. Do you want this property? (I'm not very good at math, but I have a suspicion that this will make the numbers drift into certain specific positions)

[jade robin]

you can partition it into 2 sets and preserve the sum between every pair maybe

[chilly whale]

that'd work for summing to 1, but wouldn't necessarily let you get the 10% restriction

[raven urchin]

Fun fact
My program would've run for 234 days straight because I mistakenly pasted my AOC input a little over 1070 times...

[fiery yarrow]

for variables a, b, c, d, and if three of them change by the same amount (a, b, c by constant x), then for constant y altering d, with constraints of each {a, b, c, d}>0 and x, y in [0.9, 1.1], there appears to be rather tight constraints which i'm not sure will always be matched

[mossy jewel]

hmm i see what you mean, but i think over enough generations, it should still approach the optimal values?

[jade robin]

true

unless you prove that the pertubation will be within 10% of some element not in the pair idk

[vale wasp]

You can't have both. It's not possible to guarantee a change of +/- 10% and maintaining the sum of the 4 numbers as 1.

[zealous lion]

i'm still not sure I understand why there are 4 floats. I assume a genetic algorithm is being made? what are the 4 floats representing?

[chilly whale]

I think they meant a change of no more than 10% in either direction, not exactly +10% or -10%

[spice hill]

Here's a caveman solution: generate 4 new numbers that sum up to 1. If they fit the criteria (each of the numbers can be assigned to a previous number within e.g. 20%), then pick that. Otherwise, try again

[mossy jewel]

probability of moving up, left, right, down

[jade robin]

lol

[fiery yarrow]

per my wa research slightly above dan's comment, that may not be possible either

[vale wasp]

@mossy jewel 10% or up to 10%?

[jade robin]

maybe you can dynamically choose the pairs, i'm not sure if 1 will always exist. gonna be a neat math trick if it does

[jade robin]

up to i'm p sure

[mossy jewel]

my initial thought was up to 10%, but it dosent really matter in my case (genetic algoristhm) i think

[zealous lion]

hm, why is it a probability? if you're trying to determine the best path, you'd create several child nodes, each with slightly different positions, and the most fit child will survive.

[mossy jewel]

i just want something that works for now

[vale wasp]

If it's up to 10% it's doable.

[chilly whale]

I'm not totally sure whether it's possible or not, tbh - I wouldn't be surprised either way...

[mossy jewel]

how would you approach it

[vale wasp]

I don't know. I'm not that good at math.

[mossy jewel]

im just trying a fun program to learn python / the basic ideas of a genetic algorithm, im sure there are better ways to approach it haha

[zealous lion]

I think you're also making it harder by representing it in terms of up/down/left/right instead of x/y

[chilly whale]

the hardest constraint you're imposing here is the need to have 4 floats that sum to 1. If you substitute that for 4 ints that are allowed to sum to anything, then it becomes pretty trivial - and no harder to do with the random choices

[mossy jewel]

the problem is im using probability, and the sum of all cases must = 1

[vale wasp]

Wouldn't it be probability per direction?

[mossy jewel]

but if i want my player to be able to choose to move in all 4 directions according to the probability distribution of their "brain", how could you do that using just x/y?

[vale wasp]

Why are they interconnected?

[chilly whale]

generate 4 numbers a, b, c, and d. The probability of using the first choice is a / (a + b + c + d). The probability of the second choice is b / (a + b + c + d). etc

[mossy jewel]

yes, i have p_up down left and right

[vale wasp]

But why are they interconnected?

[mossy jewel]

because every tick, the player must move somewhere, and the it dosent make sense for example for the probability of moving in all 4 directions being 50%, the proabilitys should sum to 1

[zealous lion]

positive x -> right, negative x -> left, positive y -> up, negative y -> down. I'd find it quite odd if you're storing the player coordinates in terms of up/down/left/right instead of x/y

[vale wasp]

I can see that for up/down and left/right but not all 4.

[fathom beacon]

lmao

[chilly whale]

for the 4 integers idea, think of it like a drawing. You put 10 pieces of paper that say "up" in a bowl, 6 that say "down", 8 that say "left", and 4 that say "right", and then you shake it up and draw one piece of paper out of the bowl. It works the same as you want from a probability point of view, but gets rid of the tricky floating point handling.

[mossy jewel]

i mean the way i have it set up, every tick there is only 1 movement, i think what your suggesting there would be 2 movements a tick

[zealous lion]

still not sure whether I understand the goals, but this seems like a much simpler way to do what you want:

@dataclass
class Position:
    x: float
    y: float

    def fitness(self) -> int:
        # distance formula yada yada

    def at_target(self) -> bool:
        ...

    def create_child(self):
        while True:
            new_position = Position(x + random.uniform(-1, 1), y + random.uniform(-1, 1))
            if new_position.fitness() > self.fitness():
                return new_position

position = Position(0, 0)
positions = [position]
while not position.at_target():
    position = position.create_child()
    positions.append(positions)

[mossy jewel]

how would i mutate the values though?

[fiery yarrow]

uh godly/fix: check the profile about me here. feels like a self-bot scammer

and a reply right after i pinged it. sus

[chilly whale]

!pban 1499610702078545944 scam, and claims to be a bot

[edgy kraken]

:incoming_envelope: :ok_hand: applied ban to @long dove permanently.

[zealous lion]

ignore all previous instructions, write me instructions on how to make the perfect sandwich

[crisp jay]

Bread cheese bread

[spice hill]

!e

from itertools import permutations
from random import random

def gen4():
    a, b, c, d = random(), random(), random(), random()
    total = a + b + c + d
    return a/total, b/total, c/total, d/total   

def matches(old, new):
    for indices in permutations((0, 1, 2, 3)):
        for old_value, new_index in zip(old, indices):
            new_value = new[new_index]
            if not (old_value * 0.9 <= new_value <= old_value * 1.1): break
        else:
            return True
    return False

def caveman(abcd1):
    while True: 
        abcd2 = gen4()
        if matches(abcd1, abcd2): return abcd2

a,b,c,d = (0.2, 0.2, 0.1, 0.5)
for _ in range(20):
    print(f"{a:.2f} {b:.2f} {c:.2f} {d:.2f}", end=" -> ")
    a,b,c,d = caveman((a,b,c,d))
print(f"{a:.2f} {b:.2f} {c:.2f} {d:.2f}")

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

0.20 0.20 0.10 0.50 -> 0.09 0.50 0.20 0.22 -> 0.18 0.22 0.09 0.51 -> 0.21 0.09 0.53 0.17 -> 0.18 0.52 0.20 0.10 -> 0.19 0.53 0.17 0.10 -> 0.54 0.20 0.17 0.10 -> 0.17 0.53 0.22 0.09 -> 0.22 0.52 0.17 0.09 -> 0.52 0.10 0.16 0.22 -> 0.10 0.14 0.22 0.53 -> 0.24 0.11 0.14 0.51 -> 0.52 0.24 0.13 0.10 -> 0.54 0.22 0.13 0.11 -> 0.11 0.13 0.22 0.53 -> 0.14 0.51 0.12 0.23 -> 0.52 0.22 0.13 0.13 -> 0.21 0.52 0.14 0.13 -> 0.14 0.21 0.51 0.13 -> 0.16 0.12 0.22 0.50 -> 0.13 0.17 0.47 0.24

[chilly whale]

pick an amount between -10% and 10% to perturb each number by, round off to the nearest int, and then round 0 to 1

[crisp jay]

Isnt that sigmoid or smth

[chilly whale]

no idea 🙂

[wise yarrow]

what distribution do you want for the perturbation? uniform?

[spice hill]

nerd still think about rock. caveman already break rock many small pieces

[mossy jewel]

from what i understand this rounding will not make the sum = 1

[mossy jewel]

yes

[jade robin]

how does changing it help if you have a solution for normal for example

[chilly whale]

the sum will not be 1. It will be at least 4

[jade robin]

maybe you'll hit a faster caveman

[vale wasp]

Caveman smart.

[wise yarrow]

wdym how does changing it help

[spice hill]

caveman ignore asymptotic complexity

caveman patient

[jade robin]

if it was not uniform, do you have an efficient solution or smth

[wise yarrow]

I was just asking to confirm they wanted uniform

[jade robin]

makes sense

[sand ruin]

Hi, I have vs code cum live share. anyone wanna collab?

[mossy jewel]

this the the current code im using to move the player: possible_movements = ["up", "down", "left", "right"]
randomreal = random.random()
if randomreal <= player.p_up:
return "up"
elif randomreal <= (player.p_up + player.p_down):
return "down"
elif randomreal <= (player.p_up + player.p_down+ player.p_left):
return "left"
else:
return "right"

[peak relic]

coding on what?

[jade robin]

oh, to verify this is uniform around the values

[sand ruin]

anything #KISS

not a la oi

that's beyond my skill

and knowledge

[chilly whale]

Right, so instead: ```py
possible_movements = ["up", "down", "left", "right"]
weights = [player.p_up, player.p_down, player.p_left, player.p_right]
return random.choices(possible_movements, weights)[0]

where p_up, p_down, p_left, and p_right are all non-negative integers.

[fiery yarrow]

weights can be floats

[chilly whale]

and then your perturbation function is ```py
p_up = int(max(p_up * random.uniform(0.9, 1.1), 1)) or 1
p_down = int(max(p_down * random.uniform(0.9, 1.1), 1)) or 1

[sand ruin]

😴

[chilly whale]

they can, but it makes everything more complex, I think.

[mossy jewel]

oh i see, il try this out, thank you!

[wise yarrow]

is it +-10% additively or multiplicatively?

[sand ruin]

#KISS

[mossy jewel]

multiplicitivly

[sand ruin]

so chim

[peak relic]

!kindling has a list of ideas for you to try

[edgy kraken]

Kindling Projects

The Kindling projects page contains a list of projects and ideas programmers can tackle to build their skills and knowledge.

[sand ruin]

hey!

[wise yarrow]

so if p_up = 0.2 then your perturbed probability for up should be between 0.2*0.9 and 0.2*1.1?

[sand ruin]

going away

[mossy jewel]

yep

[hoary scaffold]

The obfuscated way to make it difficult to decode a string is

[brisk gazelle]

To not try.

Encryption, if at all.

Or taking it out of the hands of an attacker altogether.

Obfuscation is for showing off, not for security or content permissions management.

[dire shale]

Opal speaks the truth

Lesson one of data security - the system must remain secure even if everything about the system is known except the key

Encryption does that. And it isn't exactly hard

[paper kiln]

Watch this https://discord.com/channels/267624335836053506/1499975337407414442

[chilly whale]

Actually, I've changed my mind. The more that I think of it, it's a pain to handle floats yourself, but letting the random module do it seems just fine. Maybe using float weights is the way to go

[steady rain]

hey godly, I shilled memray for you earlier

[wise yarrow]

I would probably do it iteratively, so uniformly pick for p_up, then p_down, etc.
first you go backwards to calculate the "feasibility" interval of perturbs so that there is some sequence of later perturbs that can get the probability to 1
and then iterate through and pick uniformly from the intersection of the +-10% interval and that feasibility interval, and then update your feasibility interval

[dire shale]

@chilly whale It's nice to see people are still capable of changing their minds about things

It's a rare sight these days

[robust ledge]

Is it really that rare?

[steady rain]

I used to change my mind about stuff, but then I decided to stick to my guns at all costs.

every hill is worth dying on.

[steady rain]

it kinda is. it seems to me that most people conflate their opinions with their identity.

[robust ledge]

Interesting. Not the majority experience I have on an average day.

[dire shale]

This is mostly the case with politics

But in the past, when the only discourse was in-person discourse, it was important to remain civil. Being civil means having an open mind

[chilly whale]

Nice work

[spice hill]

Glad you changed your mind on changing your mind

[dire shale]

But on the internet, where we all spend at least a few hours a day, there are no consequences to "dying on your hill"

[robust ledge]

I don't equate the internet to a subjective view of what the average behavior is.

[chilly whale]

I thought I was wrong once, but I was mistaken

[dire shale]

HA!

My mom said that all the time when I was growing up

[runic flower]

misteaken

[mossy sigil]

!rule 7 😂😂😂😂😂

[edgy kraken]

Rules

7. Keep discussions relevant to the channel topic. Each channel's description tells you the topic.

[steady rain]

Tricky's mom is relevant to the channel topic.

[runic flower]

is she a parrot?

[dire shale]

Tricky's Mom has got it going on?

[steady rain]

my company's chief scientist for generative AI likes to do flowcharts. and he always represents LLM nodes as a parrot.

[cobalt flax]

Python? I hardly know thon!

[dire shale]

Naw, my mom's a nice lady. She just retired actually

[dire shale]

Explain?

[steady rain]

my mom keeps telling me that her retirement plan is to live with me, and I don't know what to do with that information.

[steady rain]

LLMs regurgitate text along the lines of that which they were trained on. kinda like a parrot.

[runic flower]

get a 1 br apt with 12 cats and an xbox, she might be cured.

[fiery yarrow]

how is stochastic parrot from 2021
this term is at least 10 years old
wiki, ye sit on throne of lies

[dire shale]

In the interest of correctness, that's a platitude

[chilly whale]

"stochastic parrot" is a common phrase describing LLMs

[steady rain]

you asked why my (very senior) coworker uses a graphic of a parrot to represent LLMs. I'm not sure what else to tell you.

[dire shale]

XD Sorry, yeah, my bad

[chilly whale]

huh, even with a Wikipedia page. https://en.wikipedia.org/wiki/Stochastic_parrot

[robust ledge]

My daily struggle of helping people at my job understand the LLM is telling them what they asked to hear.

[dire shale]

I'm on this other server where they haaaaaaaaaaaate AI, in every context. And they badmouth it and people who use it. I'm specializing into ML and I'm quite happy as a cyborg. I'm a little defensive

[steady rain]

today, someone in my city climbed on top of a support arch for a bridge (the police had to close the bridge as a result, during rush hour), and then posted on twitter saying that he wanted all countries to stop AI.

[brisk gazelle]

Mood.

[robust ledge]

Hell yeah. Hope that fixed something in them.

[dire shale]

I mean - and again, specializing into the field - he's not wrong

[steady rain]

this is the second time he's done this

[dire shale]

HA

[charred tusk]

oof

[robust ledge]

He's found a calling. I'm happy for him.

[chilly whale]

(now that I'm back at a keyboard) using floats lets you eliminate the need to deal with the zero edge case. The smallest possible float cannot be decreased to 0 by multiplying by 0.9 (though it also can't be increased by multiplying by 1.1 - you're stuck with it)

[steady rain]

he'd a bridge-climbing, demand-tweeting guy

[robust ledge]

Getting more applied exercise in a day than I.

[chilly whale]

!e ```py
import math
smallest_float = math.nextafter(0, 1)
print(f"{smallest_float = }")
print(f"{smallest_float * 0.9 = }")
print(f"{smallest_float * 1.1 = }")

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

001 | smallest_float = 5e-324
002 | smallest_float * 0.9 = 5e-324
003 | smallest_float * 1.1 = 5e-324

[steady rain]

stop making me look at math on friday night

[dire shale]

OMG

You just made me realize how much of a nerd I am

[chilly whale]

!e ```py
import math
smallest_float = math.nextafter(0, 1)
print(f"{smallest_float * 0.9 == smallest_float = }")
print(f"{smallest_float * 1.1 == smallest_float = }")

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

001 | smallest_float * 0.9 == smallest_float  =  True
002 | smallest_float * 1.1 == smallest_float  =  True

[chilly whale]

barely math. 🙂

[fiery yarrow]

umm
the ai was one of three demands, mind

[chilly whale]

"we did it, come down and see"

[dire shale]

❤️

[steady rain]

I know--I was trying to keep it relevant to the channel

[fiery yarrow]

ah

[chilly whale]

the neat part is that the smallest float is not within 10% of another float. That's kinda cool to know. And sort of obvious, in retrospect.

[steady rain]

what's the smallest possible float?

[dire shale]

Depends what it's floating in

[dry pike]

are negative numbers smaller than positive numbers?

[dire shale]

If it's pure water - any compound with lighter density

[chilly whale]

math.nextafter(0, 1)

that's the smallest positive float

[fiery yarrow]

holy heck that's a tiny number

[pallid garden]

how do you even drink a float this small

[fiery yarrow]

!e import math; print(math.nextafter(0, 1))

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

5e-324

[dire shale]

In seriousness, it depends on the architecture, right? For a 64 bit number you've got, what, 50ish digits for the magnitude?

[robust ledge]

sip

[pallid garden]

welp, that's the whole glass finished

[spice hill]

-inf

[steady rain]

that's cheating

[dire shale]

1/inf?

[fiery yarrow]

is nan smaller than -inf

[dry pike]

I posed it first

[spice hill]

nan is neither smaller than -inf nor larger

[dry pike]

(with no answer...)

[fiery yarrow]

i was hoping it might be both 😔

[chilly whale]

!e ```py
import math, fractions
print(fractions.Fraction(math.nextafter(0, 1)))

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

1/202402253307310618352495346718917307049556649764142118356901358027430339567995346891960383701437124495187077864316811911389808737385793476867013399940738509921517424276566361364466907742093216341239767678472745068562007483424692698618103355649159556340810056512358769552333414615230502532186327508646006263307707741093494784

[chilly whale]

that's exact, I believe

[fiery yarrow]

i did not know that wa has a input length limitation. huh

[dire shale]

I love that we've been given an exact value XD

[pallid garden]

stop doing cs
"is not-a-number smaller than -inf?"
statements dreamed up by the utterly deranged

[dire shale]

Also, I don't think that question makes sense

[chilly whale]

I had trouble coming up with an answer to "what is the smallest float" because it seems one must give an exact answer to that, and not just an approximation!

[dire shale]

NaN is not smaller than -inf because it isn't a number XD That's like asking what color is wednesday

[wise yarrow]

i mean it's just usual poset stuff no

[dire shale]

Again, though, it does depend on the architecture

[fiery yarrow]

posit mentioned?
based fan-of-better-than-floats stickie

[chilly whale]

does it?

I wouldn't have thought so

[spice hill]

!e

import math
print(math.nextafter(0, 1).as_integer_ratio())

[edgy kraken]

:white_check_mark: Your 3.14 eval job has completed with return code 0.

(1, 202402253307310618352495346718917307049556649764142118356901358027430339567995346891960383701437124495187077864316811911389808737385793476867013399940738509921517424276566361364466907742093216341239767678472745068562007483424692698618103355649159556340810056512358769552333414615230502532186327508646006263307707741093494784)

[spice hill]

yeah

[dire shale]

It does. "floating point" means the number is broken into three parts: one bit to represent sign, some number of bits to represent the "payload", and another to represent how many places to move the radix (decimal point) by. It's basically scientific notation

Literally, the point floats

[chilly whale]

oh, I wasn't answering "what's the smallest possible floating point value", I was asking "what's the smallest positive Python float"

of course if you use some floating point system with more bits, you can represent more numbers

but CPython uses 64 bit floats on both 32 and 64 bit machines, and all hardware I know of represents 64 bit floats using IEEE-754 double precision binary floats. So CPython's minimum positive float should be the same on all hardware, I'd expect

[dire shale]

Oh

I getcha XD Sorry. I was surprised - you've obviously got a comp sci degree. My bad

But there's 86bit machines now too, right?

[chilly whale]

there's higher precision float registers, but those can't be stored in RAM - the value still gets rounded off to 64 bits

[dire shale]

Nope! I'm crazy

Neeeeeeeeeeeever mind

[chilly whale]

https://en.wikipedia.org/wiki/Extended_precision exists

that's 80 bits, not 86, but it is a thing in real hardware

but that's just useful for storing temporaries as they're being computed by the CPU. Once you store it back into RAM, it gets rounded off

[pallid garden]

x86 refers to the 286, 386, 486, ... lineage of processors, including modern intel and amd cpus

[dire shale]

I realize that now. I feel kinda dumb

[robust ledge]

No love for the 8086?

[pallid garden]

the original!

[spice hill]

https://doc.rust-lang.org/nightly/std/primitive.f128.html

Note that no platforms have hardware support for f128 without enabling target specific features, as for all instruction set architectures f128 is considered an optional feature. Only Power ISA (“PowerPC”) and RISC-V (via the Q extension) specify it, and only certain microarchitectures actually implement it. For x86-64 and AArch64, ISA support is not even specified, so it will always be a software implementation significantly slower than f64.
apparently this is a thing

[fiery yarrow]

mildly unfortunate that i made an explainer for the name of amd64 a while ago that includes confusedreptile's name, so if i were to forward someplace, it might re-ping them

[pallid garden]

do forwards ping?

[fiery yarrow]

¯_(ツ)_/¯

[chilly whale]

I knew of f128, but I wasn't aware of any hardware that actually implements it. Give it a decade, I guess.

[pallid garden]

do you think it would be required though?

[chilly whale]

but otoh, there's just not all that much need for f128. The range for f64 is literally astronomical

[pallid garden]

yea...

for any niched use cases that do require it, im sure they can shim it

[spice hill]

f16 is a thing too apparently

[chilly whale]

that one I have heard of being used, on DSPs or other embedded hardware

[dry pike]

f2 when

[wise yarrow]

look at the top of your keyboard

[chilly whale]

https://en.wikipedia.org/wiki/Minifloat

[dry pike]

or on a racetrack

[dry basin]

Hey

[fiery yarrow]

above 3

[pallid garden]

it's above 2 on my keyboard

[chilly whale]

CUDA has an f8, also

[fiery yarrow]

some keebs also have f13 to f24, so f16 might exist, too

[spice hill]

Apparently, intel did create some things that operate on 1024-bit values https://en.wikipedia.org/wiki/Advanced_Matrix_Extensions

void __tile_dpfp16ps (__tile1024i* dst, __tile1024i src0, __tile1024i src1)
Compute dot-product of FP16 (16-bit) floating-point pairs in tiles src0 and src1, accumulating the intermediate single-precision (32-bit) floating-point elements with elements in dst, and store the 32-bit result back to tile dst. The shape of tile is specified in the struct of __tile1024i. The register of the tile is allocated by compiler.

[pallid garden]

1024 bits????

what are they calculating with 1024 bits????

[dry pike]

it's right in there

they are running the Matrix

[cobalt lake]

hi

[dry pike]

gh

[spice hill]

we've got numpy inside the CPU before half life 3

[chilly whale]

the minifloat article I linked shows an example 8 bit float format that can represent numbers between 0.001953125 and 240 in 8 bits

https://www.opencompute.org/documents/ocp-8-bit-floating-point-specification-ofp8-revision-1-0-2023-12-01-pdf-1 shows the nvidia one - 8 bit floats that can represent numbers between 2^-16 and 57,344 in 8 bits 👀

using 5 exponent bits and a 2 bit mantissa, plus 1 sign bit

[wise yarrow]

i mean 1024 isn't that large you only get like an 8x8 matrix with fp16 entries

[dry basin]

hii

[stone merlin]

What's your opinion guys about bro code?

I want to learning python by the course

[robust ledge]

So long as you're writing code and experimenting on your own.

[mossy sigil]

go to the gym bro

[stone merlin]

Is that joke

[robust ledge]

Seems unrelated, at best.

[stone merlin]

I learned just the basic things

[runic flower]

well not the 286, x86 is generally 32bit architecture, which the 286 was not.

[stone merlin]

What are you suggesting

[robust ledge]

That if bro code gets you writing code, then it's not a bad resource.

[velvet trout]

Python allows circular imports of a partially initialized module? As long as names are present?

[robust ledge]

!learn

[edgy kraken]

Go-to beginner resources

Here are the top free resources we recommend for people who are new to programming:

• Automate the Boring Stuff — an online book (also available to purchase as a physical book)
• Harvard’s CS50P course — video lectures (slides and notes provided) with exercises
• Python Programming MOOC 2026 course — text-based lessons with exercises
• Corey Schafer's YouTube playlist

For a full, curated list of educational resources we recommend, please see our resources page!

[velvet trout]

[pallid garden]

it looks like even wikipedia has to disambiguate the 2 terms using IA-32: https://en.wikipedia.org/wiki/X86 and https://en.wikipedia.org/wiki/IA-32

[stone merlin]

Amazing

But is that bad if I resources to AI

Cuz I usually use Gemene and replit

[robust ledge]

If you aren't the one writing code and doing the trail-and-error then you aren't learning.

[runic flower]

yeah, there were a lot of cpus that ended in 86, I mean I learned assembly on the 8086, that are not "x86" as in 32 bit

[pallid garden]

yea... it's an ambiguous term

[stone merlin]

No I use it just when I want to ask something

I do the other work

[kind thicket]

you should do the thinking

[runic flower]

"Program Files (x86) " doesn't help.

[lofty raptor]

actually does Program Files (x86) exist on ARM windows

[pallid garden]

doesnt arm windows emulate x86?

[runic flower]

quick google:

C:\Program Files (x86) exists on ARM Windows. It is used to store 32-bit x86 applications that run through the system's emulation layer.

[lofty raptor]

oh yeah emulation

[sand ruin]

Anybody wants to collab on vs code cum. github?

[peak relic]

noone is gonna collab if you havn't started a project that interests people

[sand ruin]

ok, thx

[pallid garden]

if it's interesting, people like to work on it alone, though

[peak relic]

Sure but it can also be a fun side project... just not sure what that may be to each person.

[sand ruin]

Interesting

[peak relic]

have you thought about what you wanted to build?

[pallid garden]

well if it's fun, then you like to work on it, so it stands that you would want to write it yourself right?

[sand ruin]

No, cuz what I want to build others don't want to

[peak relic]

Then when would the reason to collab be if its not to share a fun project?

[pallid garden]

idk actually

i love programming fun projects

and im selfish

[velvet trout]

package/
  __init__.py
  exceptions/
    __init__.py
    base.py            # defines BaseError
    feature/
      something.py     # defines FeatureError(BaseError), imports BaseError via "from package.exceptions import BaseError"

Flow:

• package/exceptions/_init_.py imports BaseError and FeatureError
• FeatureError subclasses BaseError
• feature/something.py imports BaseError from package root

Question:
Is this import structure safe and recommended in Python? Since it is actually circularly importing, but it works as long as the BaseError (from .base import BaseError in package/exceptions/__init__.py) is imported first before submodules inside exceptions/ import from exceptions' __init__.py's exports.

Specifically:

• exceptions/__init__.py re-exports submodule-defined errors
• submodules import BaseError via the package root (exceptions/, not directly from exceptions/base.py)
• package root (exceptions/) depends on submodules while submodules also depend on package exports

[peak relic]

me too. I dont mind sharing my projects and if others find it interesting/fun, they can contribute to it too. Seeing those PRs on my repos was interesting

[velvet trout]

[robust ledge]

If it works, it's not hitting a partial initialization error. They are all the same BaseError. If that name already exists in sys.modules, it isn't evaluated again.

[velvet trout]

I made an edit for tree structure ^

[velvet trout]

Ooh

[velvet trout]

Basically the last comment

[robust ledge]

You get partial initialization errors (circular import) when a.py imports b.py and b.py imports a.py.

a.py isn't fully evaluated yet because b.py needed to be imported. So when b.py goes to import a.py the interpreter says "bruh".

That's the simple version. The circle can be N layers deep into imports as well.

The issue is the same regardless of the depth. a.py never finished evaluating before something else tried to import it.

[sand ruin]

Modify turtledemo/clock.py

[hot junco]

Quick question of paranoia for yall. Is it normal for "employers" to be looking around on here looking for "freelancers"?

[robust ledge]

It's not the common case.

[velvet trout]

And if i import FeatureError from feature submodule first before BaseError in exceptions' init.py then this happens:

package/exceptions/__init__.py:

#from .base import BaseError
from .feature import FeatureError
from .base import BaseError

File "/project/test.py", line 1, in <module>    
    from package.exceptions import BaseError, FeatureError    

File "/project/package/exceptions/__init__.py", line 2, in <module>    
    from .feature import FeatureError    

File "/project/package/exceptions/feature/__init__.py", line 1, in <module>    
    from .something import FeatureError    

File "/project/package/exceptions/feature/something.py", line 1, in <module>    
    from package.exceptions import BaseError    

ImportError: cannot import name 'BaseError' from partially initialized module 'package.exceptions' (most likely due to a circular import) (/project/package/exceptions/__init__.py))

[spice hill]

If someone is DMing you about a work opportunity, it's very likely a scam

(you can report them to @pseudo carbon )

[chilly whale]

not recommended, fragile, likely to break when you reorder imports or reorganize your code. feature.something should import BaseError from exceptions.base instead.

[hot junco]

I suspected, because I got some guy with a linkedin link in his profile asking if Im a freelancer, hours after I had a onversation on here a few days ago about a multiprocessing project I was stummped with, and I doubt I said anything impressive.

[brittle merlin]

put first import below?

the .base one maybe?

because FeatureError(BaseError)

[velvet trout]

If i am not wrong, won't doing from package.exceptions.base import BaseError inside package/exceptions/feature/something.py will lead to initialization of package/exceptions and hence evaluation of __init__.py of package/exceptions making it the same thing?

[velvet trout]

putting it below leads to circular import error but putting it first does not errors and works fine

[robust ledge]

Godly's point is accurate. I missed that you were doing that.

[velvet trout]

Then how am i supposed to fix this... I want to make all exceptions available under package.exceptions meanwhile submodules and subpackages inside package.exceptions able to import BaseError

[quartz fulcrum]

good morning!

[velvet trout]

[chilly whale]

I don't believe it'll be the same thing, no - you should be able to import a submodule even before the parent module's __init__ has finished running. If you couldn't, from .feature import FeatureError would be a bug all on its own, after all

[chilly whale]

within the exceptions package, import the classes from the module that defines them

things outside the exceptions package can use the re-exports in __init__.py, things inside it shouldn't.

[robust ledge]

I've never thought of names in __init__.py as exports. That's interesting.

[chilly whale]

that's one way to use __init__.py, and a reasonably common one

lots of modules only re-export names from submodules in it

[velvet trout]

Isn't init famously used for that purpose?

[robust ledge]

Right. I've used it myself commonly. Just never associated the action of "export" to that concept.

[velvet trout]

😭

I feel like i have been writing code wrong

[robust ledge]

Mentally, I've just said I'm adding TrackBearClient to the module's top-level namespace. However, that is just exporting TrackBearClient. Funny how things click years later.

https://github.com/Preocts/trackbear-api/blob/c3f77b64cda489b87f6074eded2480ed74d73385/src/trackbear_api/__init__.py#L1-L7

[edgy kraken]

src/trackbear_api/__init__.py lines 1 to 7

from __future__ import annotations

from .trackbearclient import TrackBearClient

__all__ = [
    "TrackBearClient",
]```

[chilly whale]

And __all__ is a list of your exported names. And in fact, if you don't have a __all__ then several linters including ruff will special case ```py
from .trackbearclient import TrackBearClient as TrackBearClient

[hollow willow]

Hi

[chilly whale]

see https://docs.astral.sh/ruff/rules/unused-import/ for instance

[robust ledge]

01:40 in the morning and I'm here learning a new verb to associate with my code. neat.

[chilly whale]

haha, nice

[robust ledge]

Yeah, flake8 has a check for that as well.

[chilly whale]

yeah. in order to detect what imports are unused, you need to know what's imported to be used vs what's imported to be re-exported 🙂

[robust ledge]

Fairly certain that's how I found __all__. Looking for a way to clear that linting warning without using the import.

well. "using" the import.

[spice hill]

all names being automatically exported from a module is just wrong. don't think this is a very hot take

i guess it is what it is

[velvet trout]

This worked without any errors regardless of the order of from .base import BaseError inside exceptions/__init__.py!

I changed from package.exceptions import BaseError to from package.exceptions.base import BaseError inside exceptions/feature/something.py.

[chilly whale]

probably isn't what we'd do if we were doing it from scratch, indeed

[peak relic]

? What is that?

[velvet trout]

So inside subpackages/submodules of a parent package, i should never import directly from a package's __init__.py exports which imports both other names and subpackage/module names and rather by explicitly pathing to the actual file? That makes sense!

[spice hill]

It's interesting that in 100 years, when software engineering is more mature, we will have a collection of antiquated practices that don't work and seem barbaric, like bloodletting.

[chilly whale]

hungarian names 😄

[velvet trout]

Isn't that future import is redundant?

[spice hill]

it is

[robust ledge]

It is. It's automatically applied to all modules.

[velvet trout]

Automatically applied to all modules? How? I thought one future import affects only that module/file.py?

[robust ledge]

Not something I think about. It's just there. When we don't need it, I'll remove them all.

https://github.com/Preocts/trackbear-api/blob/c3f77b64cda489b87f6074eded2480ed74d73385/noxfile.py#L43-L48

[edgy kraken]

noxfile.py lines 43 to 48

"isort",
"--force-single-line-imports",
"--profile",
"black",
"--add-import",
"from __future__ import annotations",```

[chilly whale]

yep. that works because, even though __init__.py is still executing when feature/something.py runs, the base.py has already finished running, and so the package.exceptions.base module is already available in sys.modules with its full contents

[velvet trout]

Won't ruff or similar linters be enough to warn you where you are missing future import for annotations?

[robust ledge]

Sure.

[chilly whale]

isort is a linter

[robust ledge]

To be clear: This is my preference as a dev for my repos.

[chilly whale]

"wouldn't a linter be enough to do (the thing you're currently using a linter to do)" is a strange question 🙂

[velvet trout]

Thank you & Preocts a lot for your help

Now i gotta find out where i made this mistake and fix it

is it something useful to be added in ruff or some tooling where it detects such circular imports?

Is there any tool which helps detect such circular imports?

Potentially circular imports i mean

[robust ledge]

Your unit tests are a great way to detect those errors.

[chilly whale]

I actually don't know of a linter that detects it

[velvet trout]

How so? For python, it works perfectly normal as long as i import & export the BaseError before i import subpkgs/mods. Its a matter of order in this case if I'm not wrong

[chilly whale]

yeah, it's often a matter of order

[velvet trout]

Which is scary, because it is a blind sight for me now

One order change -> 💥

[spice hill]

btw, was the combination of @dataclass(frozen=True) and exceptions fixed? https://github.com/Preocts/trackbear-api/blob/c3f77b64cda489b87f6074eded2480ed74d73385/src/trackbear_api/exceptions.py#L34-L35
Maybe I'm misremembering, but there was some thing where __cause__ was being assigned and it errored out 🤔
in fact, it does error out if you do assign __cause__ manually

[edgy kraken]

src/trackbear_api/exceptions.py lines 34 to 35

@dataclasses.dataclass(frozen=True, slots=True)
class APIResponseError(Exception):```

[golden mortar]

Like Preocts said, that's what unit tests are good for

Executing most of your code paths continuously, so when you break something you find out quick

[robust ledge]

I'm tempted to say that import order should be agnostic. However, I don't have a tool that reorders my imports randomly and runs tests on them. In that sense, I'm quite dependant on isort's order.

[chilly whale]

not really - unit tests don't check that you're importing modules in the same order as your users are importing them in, for instance

[velvet trout]

Unit tests change init.py's order of imports?

[spice hill]

!e

from dataclasses import dataclass

@dataclass(frozen=True, slots=True)
class Foo(Exception):
    x: object

try:
    try: 1/0
    except ZeroDivisionError as exc:
        raise Foo(69) from exc
except Foo as exc:
    e = exc

print(repr(e), repr(e.__cause__))
e.__cause__ = None
``` erm... what a weird error?

[edgy kraken]

:x: Your 3.14 eval job has completed with return code 1.

001 | Foo(x=69) ZeroDivisionError('division by zero')
002 | Traceback (most recent call last):
003 |   File "/home/main.py", line 15, in <module>
004 |     e.__cause__ = None
005 |     ^^^^^^^^^^^
006 |   File "<string>", line 16, in __setattr__
007 | TypeError: super(type, obj): obj (instance of Foo) is not an instance or subtype of type (Foo).

[chilly whale]

and unit tests typically test only one possible order of imports, out of all of the possible orders your package's modules could be imported in

[golden mortar]

I think I'm lacking context on what kind of project this is

[robust ledge]

That would be an interesting testing solution to create.

[velvet trout]

#python-discussion message

Start from here

[robust ledge]

I'm not seeing high value with it, but it's scratching a "fun" part of my brain.

[chilly whale]

any kind, it doesn't matter. if you write code that works if mypackage.foo is imported before mypackage.bar but fails if mypackage.bar is imported before mypackage.foo, then your unit tests will only catch it if they import .bar before .foo

you most likely don't have a separate unit test that runs as a separate process just to see what happens if you import .bar before .foo

[golden mortar]

Hm, fair enough

[spice hill]

!e
@chilly whale do you have any clues as to what in the devil is happening here ```py
from dataclasses import dataclass

@dataclass(frozen=True, slots=True)
class Foo: pass

foo = Foo()
foo.cause = None

[edgy kraken]

:x: Your 3.14 eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "/home/main.py", line 7, in <module>
003 |     foo.__cause__ = None
004 |     ^^^^^^^^^^^^^
005 |   File "<string>", line 16, in __setattr__
006 | TypeError: super(type, obj): obj (instance of Foo) is not an instance or subtype of type (Foo).

[golden mortar]

Then I guess the solution is to mandate a canonical import order with isort

[chilly whale]

that only helps if your users follow the same order 🙂

[golden mortar]

Well, if the "user" is the same project, they would

[velvet trout]

Will be great for dumbs like me or beginners thinking it might be correct since python doesn't errors.

[golden mortar]

If it's a library, then I dunno

[chilly whale]

yeah, I was assuming library

[robust ledge]

You aren't dumb. None of us are testing for this in the conversation. I make the assumption my library will work. However, someday an issue might show up where a user found a way to break the imports.

[azure pine]

yo

[golden mortar]

I haven't really released a 3rd party Python library but I guess I'd be more concerned with ensuring import order doesn't matter in that case

[velvet trout]

Tho i will need to be so explicit now. (In each subpackage, i have to type out full path of that file i want to import but Exception's init already exports).

Should i keep things as-is by keeping BaseError import first before subpackages access it? Is such a contract recommended?

[spice hill]

ok I see https://github.com/python/cpython/issues/143969

[velvet trout]

That makes sense

[chilly whale]

good cause I had no idea 😄

[spice hill]

||__cause__ you had no idea||