python-discussion

[silent whale]

Yes amazon is indeed laying off people, but how can you be sure it's due to AI? Look 2021 and 2022 right off the pandemic how many people Amazon fired.

[ebon hollow]

"Cybersecurity software maker CrowdStrike said in May that it’s laying off 5% of its workforce or 500 employees, and directly attributed the cuts to AI."

[golden mortar]

And again, these are companies that have a vested interest in the idea that AI is good enough to replace human workforce, whether that's true or not.

[ebon hollow]

Your right about that. But that's not the only thing.

[silent whale]

Alright you got me there, I can't deny it then :)

[fossil steeple]

i think theory lemme pull up my notes

[golden mortar]

It's still not clear exactly how AI has allowed them to reduce their workforce.

[chilly whale]

why did they lay off 5% of their workers and not 100%?

[fossil steeple]

@dry yacht https://docs.google.com/document/d/1IyRlerYYLpIXZeMefvfLZHm_AP_S9nVSMw7fZpMcJLI/edit?tab=t.x3o3sfc9a20s#heading=h.nbiwng50na4d
the data structures tab

[golden mortar]

Lots of companies prematurely fired a lot of people and tried to replace them with AI, and then had to rehire them later when it turned out the AI was too unreliable.

[fiery yarrow]

ceo ain't gonna clean their own desk

[ebon hollow]

Well AI isn't good at some aspects. So Entry-Level Programmers get fired and I would say that now there's more competition and higher requirments

[golden mortar]

In my experience AI is not even remotely good enough to replace entry level programmers.

[ebon hollow]

Well I have no clue now, I see diversed opinions about it.

[fossil steeple]

do you think it will be? like is there any point in me going to go do like a cs degree if it's gonna be obselete by the time i finish it? (like 8 ish years time)

[golden mortar]

I don't think LLMs will ever be.

Maybe there will be some other technological revolution in the future, and I can't speak about that.

[harsh anchor]

8 years is a long time

[silent whale]

That's always been true... Just look at doctors from 1926 VS 2026
We even have to go to universities! Medicine is done, now that the patient can just buy pills at the pharmacy...

[slender urchin]

A good school will have adapted their program by then if needed

[fossil steeple]

3 more years in the irish education system plus 4 years in a degree

[golden mortar]

But the current transformer architecture I think has fundamental limitations that will ensure it's never more than a tool.

[ebon hollow]

Same, I still have time to change career paths.

[silent whale]

But at the endd of the day, you can still not do Computer Science if you think that's true. Like if yo u truly believe AI is going to replace coding, then don't do it!

[ebon hollow]

You got a great point there.

[fossil steeple]

yeah and i've heard that the job shortages are temporary but i could be wrong

[golden mortar]

Most likely job demand will ebb and flow over time.

[sullen dust]

how would it work if AI took over programmers jobs? like how would they continually run and update files wouldnt they still need human operation?

[fossil steeple]

i want to do it cos i like coding and computers in general
and i have like 20 copy pages filled with me trying to make logic gates in binary

[ebon hollow]

There still will be programmers working on it for example AI Engineers will be high in demand.

[golden mortar]

I mean, if we had AGI they'd be completely autonomous.

It'd be as if it was a person.

[sullen dust]

yeah lets hope AGI doesnt get implemented

isnt Elon Musk wokring on one?

[golden mortar]

I think going for a CS education is a completely rational choice.

[dull dune]

AGI is pure marketing, it's not worth talking about

[oblique spindle]

If AGI is implemented, humanity may not survive.

[dull dune]

that's a silly statement

[golden mortar]

There is no indication that anybody anywhere in the world has even the slightest clue what's needed to build an AGI.

Elon Musk least of all.

[sullen dust]

what was that optimus thing he was talking abt some time ago?

[golden mortar]

That guy talks a lot. It's not really worth taking seriously.

[sullen dust]

true but its entertaining

oh nvm its was complete bs lol i js searched it up

[oblique spindle]

If you need code which is not maintainable,scalable , doesn't respect the project structure and just works code LLMs are fine but the code debt they create will be overall loss to the company.

[ebon hollow]

Then how are the other companies doing it then.

[dull dune]

that's a short term thing. You have to have entry level positions to have more senior positions, medium term getting rid of entry level will result in lowering requirements

[golden mortar]

It doesn't really work for anything beyond the absolutely trivial.

[dull dune]

I encourage you to really look into stories about replacing developers with AI

[golden mortar]

Yeah, that's also a good point. Even if it was possible to replace juniors, which it isn't, it'd create a shortage of workforce in the long run.

[dull dune]

frankly, it's not happening

[fossil steeple]

isn't that what's happening now

[golden mortar]

No, it's not.

[dull dune]

tech was on a decade long hiring glut, and is correcting

[fossil steeple]

i thought a whole bunch of companies have stopped hiring juniors

[fiery yarrow]

bug bounties on open source projects are probably among the strongest damnations against llms right now. cURL is killing theirs because they're getting absolutely swarmed with hallucinated reports of CVEs that don't actually exist with code fixes that don't actually compile

[dull dune]

without LLMs, not much would have changed re layoffs

[golden mortar]

That's not because of AI.

[fossil steeple]

why is that now

[dull dune]

#python-discussion message

[fossil steeple]

so then consequently, in a few years time there should be a boom?

[oblique spindle]

Because it was a technology boom then a tech shortage then another boom now due to AI Misinformation there's gonna be another shortage then another boom due to this shortage...

[golden mortar]

That's because of post-COVID layoffs and rising interest rates reducing positions and releasing a bunch of more experienced developers into the job market.

[dull dune]

also, legitimately, while there may be good individuals in leadership and management, at any scale leadership and management are largely shit

[dry yacht]

i think theory lemme pull up my notes
Nah, this is "just" how to use it.
Theory would be like, for example for list that it's array-based and that operations like append(), pop(), or operator[] are guaranteed to run in (amortized) O(1) because of how it "exponentially" resizes itself.

These docs just show how you can use it in Python.

[dull dune]

not necessarily, tech as a whole has been over invested in for a while

[fossil steeple]

oh well

[golden mortar]

There may be. If the economy recovers and fewer people study CS, there may be way higher demand down the line.

[fossil steeple]

why is just in speech marks

[dry yacht]

Because that can be quite a lot in the beginning anyway

[fossil steeple]

good enough for me then

[fiery yarrow]

admittedly, this raises the question of the actual threat: it's not whether LLMs can replace programmers, it's whether LLM hawks can convince ignorant boards and HR that they can replace programmers. the former is probably impossible. the latter, not so much

[fossil steeple]

i heard a thing that ai is meant to be getting worse due to more ai slop could be wrong tho

[golden mortar]

That'll self-correct over time though.

[fiery yarrow]

granted, but it'll hurt pretty bad for a while

[oblique spindle]

I mean they should make something like a dislike system on GitHub, so If someone with a project with 32 stars disliked you, it'll be -32 or something so people don't submit obvious AI code.

[fossil steeple]

@golden mortar what do you work as

[golden mortar]

Software engineer

in fintech

[fossil steeple]

and just out of curiousity did u do a cs degree

[golden mortar]

Yes, I have an M. Sc in CS

[fossil steeple]

how many years is that now

[golden mortar]

I graduated in 2014

[ebon hollow]

In Belgium?

[golden mortar]

Nope

[fossil steeple]

no as in how long does it take u to do an m.sc

[golden mortar]

5 years

[fossil steeple]

plus the undergrad is 9

[ebon hollow]

Where is Funtech

I mean Fintech

[fossil steeple]

9 years jeez

[golden mortar]

financial technology

[golden mortar]

Huh?

[ebon hollow]

Oooh

[fiery yarrow]

do you mean like pre-university high school or something?

[charred python]

funtech sounds fun, though

[sullen dust]

damn i just got accepted for an interview for my post 16

[fossil steeple]

oh wiat maybe ur college system works diffrently is the m.sc a stand alone degree or did u have do smth before it

[fossil steeple]

in ireland you have to do an undergrad in smth and then u "add" to that

[golden mortar]

I think in some countries you have to do like a 3 year bachelor and then a 2 year master, but at my uni it was a single 5 year program

[rustic needle]

I got my graduatio 12 years late 😅 in 2025 ☺️

[granite wyvern]

When I did my BSc in CS it was a 3 year degree. Honours was an additional year. A masters was probably 5 all up then - BSc plus further work.

[ebon hollow]

Indeed

Can you still get hired without a degree even if you know everything?

[fossil steeple]

years so it's a masters in ireland which is 1 years plus the undergrad is 5 so yeah

[rustic needle]

In Sweden they changed it a while ago so thatnow you usually (always?) get a bachelor on your way to the M.Sc.

[granite wyvern]

Yes.

[golden mortar]

Ok, it wasn't like that in my time

[ebon hollow]

Easy Peezy

[rustic needle]

My time was 5 years as well

[ebon hollow]

I dont need 5 years

[rustic needle]

No bachelor

[granite wyvern]

Having a degree helps at the outset though - it effectively says "this person's spent time and learnt an expected spread of knwoeledge".

[robust ledge]

It's not easy to break through the crowd, but it is possible.

[ebon hollow]

Ill build an impressive portfolio and ill just take the exam of CS

[rustic needle]

How will the company assess that you know everything?

[ebon hollow]

Interview

[rustic needle]

Are they gonna interview you for 5 years? I think many companies will be looking at grades.

[charred python]

Some industries or companies that do government contract may require a degree. Other companies don't care (but without a degree, you'll need other experience, usually)

[ebon hollow]

I've seen people do it without

[ebon hollow]

but to get experience I need to get hired

[charred python]

right

[granite wyvern]

There's no "exam of CS". A degree requires attending and passing courses so that the university knows you've covered what they consider necessary. You can get extra credit for some things to avoid some courses, but that's usually when you switch degrees or come from another university where you've already done equivalent stuff.

[robust ledge]

I'm always curious about that. In the dozens of jobs I've had, I've never had my grades looked at. (to my knowledge)

[rustic needle]

If there is one person with a degree, and one without, I would probably pick the one with.

[granite wyvern]

There's stuff you learn in a degree you don't learn independently in the world unless you're interested in the theroy etc, and the theory does inform the way you think and gives you better tools for thinking.

[rustic needle]

Oh, I gave my grades to them if I remember correctly. But I was 0 and 1 year out from uni resp.

[ebon hollow]

Nope, You have to be enrolled but you don't have to attend courses. Professors don't check who came and who didn't you show up on the exam and make it and that's it

[granite wyvern]

It it true that attendance at lectures isn't checked. But attendance at tuorials sometimes is. You could show up at just the exam (and the practical assignments during the course). I don't like your odds.

[rustic needle]

In Sweden comp Sci you would have 4-5 exams per semester, some mandatory tasks/laborationsand and some project courses.

[rugged vapor]

hey guys

[rustic needle]

Of course you can skip the lectures. But to me they were a great (the best) way to learn. At least for mathematics.

[chilly whale]

sure, you don't technically have to attend courses, but it's very unlikely that you'll pass courses in later years without attending them

passing tests without learning the material only works for basics you already know, and falls apart once you need to learn new stuff

[dusk adder]

or just do what i do
learn the material at home and skip the lectures

[ebon hollow]

I'm already focused on learning coding; why should I enroll if I already know the material?

[harsh swallow]

Idk about Sweden, but i know that except for some degrees like medical, such thing was done to make degrees more compatible between EU countries.

[ebon hollow]

And also I'm saying if you know the material, if you don't then ofc you have to attend

[chilly whale]

CS doesn't teach you how to code. The stuff it teaches is tangential to learning to code, and teaching yourself to code won't teach you the stuff you learn in a CS degree.

[granite wyvern]

If that works for you, all to the good. FOr. lot of people it's helpful to be there; I at least learnt better by being there and writing notes as things went. Just watching/listening is less effective.

[ebon hollow]

Well Ill just learn what they learn right now

[dusk adder]

when i do go i try to participate
though one of my professors banned me from answering questions because i always give the answers too quickly .-.

[ebon hollow]

cuz I want to have multiple degrees without doing 10 years of attending classes

[granite wyvern]

Good. If that works, fine.

[dusk adder]

i dont like the public setting of universities too much

[granite wyvern]

Instant answers does reduce the engagement of the other students.

[chilly whale]

sure, obviously you can self-study all of the material required for a degree. Most people struggle to do it, though. Most people really have trouble forcing themselves to learn material that they don't find interesting without the pressure of grades and deadlines

[rustic needle]

Medical has been made more similar in EU as well (in Sweden we switched "AT "-> "BT" to become more similar)

[granite wyvern]

And answers always from one student also reduces their engagement.

[dusk adder]

i get it, but then again i dont see the point of going if i ruin it for others just by participating yknow?

[granite wyvern]

Then don't go.

[charred python]

just answer second.

[rustic needle]

That's reasonable, it's better to answer if you are UNSURE then if you are SURE. 🙂

[harsh swallow]

I meant like medical doctor is 5y and MSc or equivalent level, not BSc level (at uni, there's only PhD after that).

[golden mortar]

I'd usually wait a little and answer only if it seemed like no one else would

[steel whale]

or they can wait until their teacher calls them or sometimes not answer

[golden mortar]

Which was most of the time

[steel whale]

it's called being fair

[dusk adder]

what i can recommend for people in my position is that it might be better to ask questions rather than giving answers
good questions make professors happier than good answers for some reason

[granite wyvern]

Means you're thinking, and that they don't have to prise answers out of a class.

[rustic needle]

I'm not exactly sure what you mean. Medical Doctor here was 5.5 + 1.5 = 7 years, but they changed it to 6 years

[vestal scroll]

personally I don't see the benefit of the lecturer asking closed questions during lectures. much better for students to ask questions themselves instead

[dusk adder]

i think it also gives them an opportunity to naturally bring up subtleties that they would want to put as an aside

[harsh swallow]

For uni itself, it's 5y here and has no BSc level. That's just uni - not counting years of residency in hospital and specialisation after that.
I was adding to "now 5y programmes were split and there's BSc" thing that there are exceptions to the split even in EU

[steel whale]

6 or 7 years

😛

[chilly whale]

and the odds are very low that you'll find all of the material for the degree interesting. I would have never been able to self-teach systems architecture or linear algebra, for instance. I find them both incredibly boring.

[rustic needle]

yea I was a bit wrong, previously it was 5.5+1.5+5 years, now its 6+0.5+5

[harsh swallow]

Non-engineering degrees here are 3y for BSc + then 2y for MSc
Engineering is 3.5 + 1.5
Unified programme (no BSc in between) for e.g. medicine is 5y

[rustic needle]

I see, yes I was just considering technical science (engineering/math/computer/physics) when I made that remark.

[rustic needle]

Where are you?

[harsh swallow]

Poland

[rustic needle]

Cool, I did python programming for Ericpol at Ericsson, so visitted Krakow twice 😄 Cześć @harsh swallow

[undone wagon]

would people in 200 years be impressed with coders not using ai like we're impressed with mathematicians 200 years ago not using calculators

[charred tusk]

Depends on how useful AI gets

In its current state?

No.

[oblique spindle]

I mean I'm not impressed by mathematicians not using calculator

Like Obviously we didn't use calculators whole learning mathematics

[chilly whale]

and using LLMs while learning to code is like using a forklift while lifting weights at the gym

[charred tusk]

Well, OP didn't specify learning

[chilly whale]

you can't be a coder without learning to code, right?

[charred tusk]

Right

[ebon comet]

Proper maths people use equations and calcs are typically using numerical approximations which is different

[charred tusk]

Even if "AGI" does replace you, that just makes you a prompt-inator, not a developer yourself

[ebon comet]

I've been personally arguing with a prompt engineer about the quality of his code (since it's generating ginormous files etc) and lots of smells and he is convinced the toolchain doesn't need a good dev any more!

[harsh anchor]

"proper math" is just not very well defined

[ebon comet]

I don't disagree with you here; it's a vague idea

You can 'dress up' proper formulated mathematics in code and tensor arithmetic but it might still not be peer reviewed actual quality content and might even be pseudoscience written by a carried away philosopher

[finite rose]

Back then they only knew like the first twenty primes and couldn’t count higher than like, 37. (Exaggerated for effect)

[harsh anchor]

huh?

[ebon comet]

I'm referring to some messy code and ideas that a non dev friend has shared with me but doesn't want sharing publicly in a sarcy manner

[finite rose]

AI code reminds me of perl. It’s best for a single one time use then discarded.

[ebon comet]

I saw that 'pearl' of wisdom!

[charred python]

funny, one of the few times I use AI it is to make single-use perl code.

like to edit a bunch of config files or some other one time refactor.

[rustic needle]

People discarding AI like this, have you tried agents like Claude Code? (I haven't but I might soon)

[finite rose]

Theo has a recent vid reacting to an article about how code is now cheap, but software is still expensive.

[charred python]

I have a claude code subscription at work and a trying to give it an honest go.

[ebon comet]

I've heard that Claude Opus is not bad (I've tried Claude others; they make a mess of your code) - OpenAI seems almost OK

I tried google gemini as well; did not get on with it

[finite rose]

I have access to various AI agents through my IntelliJ Idea subscription

[rustic needle]

WE only have weaker non-claude models at work, so Ive tried claude vs those, but it's obvious that it's not working as well as it should be.

[ebon comet]

Some friends of mine had some fun with local AI generating little 10s vids. They used a 5090

I'm a poor man with only a 4060 🙁

[finite rose]

I didn’t understand what any of that meant except video is somehow involved 😆

[silver plover]

Good conversation but can we AI in #ot0-psvm’s-eternal-disapproval

[undone wagon]

old joke

[harsh swallow]

Old joke, doesn't work, and we don't like even "joke" malicious code here

[rustic needle]

I have a laptop without dedicated GFX :3

[granite wyvern]

Please don't use potentially destructive things as examples, even as a joke.

[marble mesa]

Ok sorry

[harsh swallow]

It always annoys me that those "jokes" use os.remove which literally can't remove directories, ever. At least be creative and verify your joke, making this mistake shows it's just the same "joke" copied over and over again...

[tall vector]

That's exacly why they're jokes, why they're fine, and why it's still silly that pydis removes them; because they cannot do actual harm

[tropic locust]

Is docs.python.org a good place to learn python?

[robust ledge]

Harmful or not, the intention does not jive with the community we want here.

[granite wyvern]

There's a tutorial there. I hear it's ok.

We mostly use the docs site for reference, when you need to look things up. Then it's extremely useful.

[cerulean ravine]

i;m tempted to say it could be instructive to understand the three (or more?) different reasons it won't work

[tall vector]

Then, frankly, I think the orwellian nature of that desired community is a stupid goal lol

[robust ledge]

The door is over there. 🤷‍♀️

[tall vector]

Classic Pydis, "your opinion is wrong, leave"

[granite wyvern]

They're not fine. Wait for an unwise person to "fix" your joke, and then bad stuff happens. They're just a bad idea all around.

[robust ledge]

You can read it how you like. I'm saying your opinion is valid, don't subject yourself to such stress.

[harsh anchor]

[cerulean ravine]

it's orwellian to stop people spreading malicious software?

[tall vector]

No, it's orwellian to stop them posting jokes which, as we have established, aren't actually harmful because they cannot possibly run successfully because they don't work

[chilly whale]

have... have you read 1984?

[robust ledge]

It's one of many places. You might also like realpython website, our !resources tag, and books like Automate the Boring Stuff or Byte of Python. All depends on what provides you with what you need.

[frosty oriole]

george orwell definetely wrote about this in his hit novel 1984

[granite wyvern]

Here's a gun. It has no bullets. it's just fine.

[tall vector]

I'm not against moderation, I'm against the misconstruing of jokes as potentially harmful when it takes a huge amount of effort to go from what was posted to code that actually can break stuff

[charred python]

We've always been at war with PHP.

[harsh swallow]

Depends on who pastes is, the quotes aren't always converted to smart quotes...
Backslashes will work here - just show a warning - because \W and \s are not valid escape sequences...
So it's only os.remove not handling folders - and only when that gets fixed, we'll see permission error.

[granite wyvern]

shutil.rmtree is hard work?

[tall vector]

Last I checked, a program running normally in userspace won't remove system32 even if you use the right method.

[granite wyvern]

And we always will be.

[chilly whale]

I'm confused how someone who has read the book could possibly think that telling people not to post commands that look harmful is "Orwellian"

[harsh anchor]

if someone knows about shutil.rmtree and wants to modify that program to use it...

[marble mesa]

Ok sorry jeez…

It wasn’t even potentially destructive then

[harsh swallow]

There were actually cases of that when there was no staff nor old regulars to cut it off - people just trying to impress each other with actual closer to working malicious code...

[tall vector]

I wasn't referring to a specific thing, I was referring to the general overbearing pydis approach to moderation in which everything is overscrutinised and responded to often disproportionaly, in my opinion

[chilly whale]

ok. That's not what "Orwellian" means.

[torn viper]

When type hinting class attributes, should I use the most concrete type possible (as I would for function return types) or the most abstract type possible (as I would for function arguments)?

[tall vector]

My bad, let me rephrase, "I believe the approach taken here is poor community management" instead then

[robust ledge]

Is the class an concrete class or an abstract class?

[harsh swallow]

Meanwhile just earlier today: I asked someone to open any terminal and they opened admin perms powershell without me asking for admin... I actually told them never to do that as they may break something even by accident because of those elevated perms

[frosty oriole]

i would say bring it up in #community-meta but...

[tall vector]

(I do appreciate calling out the 'orwellian' thing, I would rather be correct in my phrasing/anaologising, I just also don't want that to be a detractor from what I'm trying to say)

[torn viper]

Concrete

[robust ledge]

Then you'll want to make your annotations as accurate to the implementation as you comfortably can.

[torn viper]

It's really a choice between list and Iterable. But now that I think about it, I should probably be specific.
-# After all, if my principle is "as specific as possible" then why not "Any" for everything, which is not helpful.

[finite rose]

https://docs.python.org/3/tutorial/index.html

[chilly whale]

wrt. policing commands that appear to be malicious, there's two aspects that come into play. One is that this community has a ton of beginners, and beginners are known for pasting commands they don't understand - which is why we don't want people sharing things that would cause harm if executed. The other thing is that things that look harmful at a glance and require people to figure out whether or not they're actually harmful are a drain on our very limited moderation resources. It's harder for us to moderate if we need to treat looks-harmful-and-is differently from looks-harmful-but-isn't, and it's much easier to just say "don't share things that look harmful"

this is a better topic for #community-meta than for #python-discussion though.

[charred python]

But officer, it wasn't a real gun.

[chilly whale]

it's just a joke, bro

[finite rose]

😝

[lusty spoke]

wsg bad bitches
I want to add a toggleable overlay for a plugin I'm making but I want it to impact performance to the least degree possible, what python option should I use

[robust ledge]

Any is the opposite of specific, isn't it?

[tropic temple]

really depends on how you want to interact with the data. if all you need with the iterable is to iterate through it, then that would suffice. if you need other things in list, then list

[torn viper]

"As abstract as possible" is what I intended to say (I've been at this for awhile)

[robust ledge]

Ah, I see. Yeah, for that just don't use annotations.

[charred python]

Can you give us some examples of that class attributes you refer to? I can think of both cases (concrete and abstract) being valid.

[torn viper]

I mean, it's a dataclass so who am I to say how the data is used (what if slicling is something that is needed at some point)

[cerulean ravine]

you know what the data represents, no?

[zenith lintel]

sup

so you want a gui for it basically @lusty spoke

[torn viper]

Indeed, but I'm looking for more general advice (which in this case I guess is "whatever makes sense")

[finite rose]

Is “pydis” an overloaded term? Beyond a python disassembly library?

[robust ledge]

It also refers to this channel.

[zenith lintel]

isnt it python discord

[charred python]

@dataclass
class ShoppingCart:
    items: list[Item]

Here I use list because I expect my shopping cart dataclass to contain a concrete list, instead of something like an iterable that fetches something dynamically (or whatever)

But I can imagine other use-cases where you may want to be as abstract as possible. I think it depends.

[tropic temple]

pygen also

but it aint gen no more :(

[robust ledge]

#bringbackpygen

[half pewter]

Pepperidge Farm Remembers

[finite rose]

Ah. Someone above seemed to be using it as a “python ‘dis”, as in “disrespectful.

[robust ledge]

✨ Language ✨

[torn viper]

@dataclass(frozen=True, slots=True)
class FilterGroup:
    target: str
    filters: list[Filter]

    @classmethod
    def ofType(cls, target: type, filters: Iterable[Filter]) -> Self:
        return cls(target.__name__.lower(), list(filters))

-# Don't ask why I need to store class names. It's the culmination of a number of bad decisions that I'd rather just keep running with than making the objectively correct choice of fixing at this point.

Though, given that I'm literally converting to a list… it probably should be a list and not an iterable

[robust ledge]

I wouldn't use a list with a frozen class but that's just me. I don't know that python will stop mutation of FilterGroup.filters.

[torn viper]

Good point, should be a tuple then?

[robust ledge]

Does order matter?

[harsh anchor]

just don't mutate it

[robust ledge]

Well, a tuple would be the frozen alternative. A frozen set if Filter is hashable and order doesn't matter.

Now I'm curious, does a frozen dataclass allow mutation of a list attribute?

[torn viper]

I'd rather have the assurance that order is preserved (and not have to worry about what counts as a duplicate filter)

[nova grove]

is there any library out there that allows python to crosscompile to windows? I've been compiling with a windows VM so my friends can use my apps and its annoyin

[harsh anchor]

yes

[torn viper]

>>> @dataclass(frozen=True,slots=True)
... class Test:
...     x: list[int]
...
>>>
>>> z=Test([])
>>> z.x
[]
>>> z.x.append(5)
>>> z
Test(x=[5])
>>>

[edgy kraken]

Hey @torn viper!

Please edit your message to use a code block

Add a py after the three backticks.

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

This will result in the following:

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

[robust ledge]

That they do. My head, right now, does't like this but I'm also certain I've allowed this before in things I do.

Yup, I have. https://github.com/Preocts/trackbear-api/blob/3ab7cc8aaf67faa2251b5e0a17f33fcbecb27a89/src/trackbear_api/models.py#L291-L292

[edgy kraken]

src/trackbear_api/models.py lines 291 to 292

work_ids: list[int]
tag_ids: list[int]```

[tropic temple]

Huh i've never thought about using mutable things in frozen dataclasses. maybe because my only usecase of frozen=True is to make it hashable, which makes using list in it impossible

[robust ledge]

And yet, isn't that frozen dataclass hashable by nature?

[torn viper]

I don't exactly know how Python could be setup to disallow it (or more broadly, disallow mutating elements of properties of frozen dataclasses, especially given that you could nest those element arbitrarily deep in the dataclasses' properties)

-# Probably dataclass' instead of dataclasses'

[tropic temple]

hashable dataclasses need to be frozen, frozen dataclasses dont need to be hashable
-# iirc

[robust ledge]

It couldn't. It's chasing something better provided by a more strict language. I just find it neat that I've never stopped the think about it. Or, at least, I don't recall doing so.

[spark quiver]

hello everyone im looking to start learning python does anyone know where is a good place to start thank you

[robust ledge]

Interesting.

[brisk gazelle]

!resources

[edgy kraken]

Resources

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

[chilly whale]

!e You've seen this, right? ```py
t = ([1, 2, 3],)
try:
t[0] += [42]
except Exception as exc:
print(f"Caught exception: {exc}")
print(f"{t=}")

[edgy kraken]

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

001 | Caught exception: 'tuple' object does not support item assignment
002 | t=([1, 2, 3, 42],)

[torn viper]

Fair enough

[robust ledge]

Yup! Love that one as an example, actually.

[spark quiver]

thank you very much

[chilly whale]

one of my favorite Python weirdnesses

[charred tusk]

wait a second...

[torn viper]

To me it makes sense, since it's no different than in a langauge like Java writing:

final List<int> items = new List();

-# It's been awhile since I did Java, so I could have the grammar wrong there.

[robust ledge]

I like the idea of a frozen dataclass. Python gives me a "close enough" solution. I roll with it, and then get pointed to neat things like this in hindsight.

That and anything hashable can be used in sets and I use sets a lot.

Set operations make things go brrrr

[remote geode]

this is java?

[chilly whale]

t[0] += something behaves like t[0] = t[0].__iadd__(something)

[granite wyvern]

We need a deepfreeze, like deepcopy.

[torn viper]

Yes? At least close enough.

-# Replied to teh wrong thing

[chilly whale]

yes. and that's the explanation for this weirdness.

🙂

[robust ledge]

Maybe some sort of system to check on that permission as well. You know, to make sure if we are changing something that we have the right to do so.

[remote geode]

pretty sure given all the new frozendict discussion, this isn't possible?

[torn viper]

Yes? At least close enough.

[remote geode]

Also seems against python ethos

its just the wrong language if u want privacy like that

[robust ledge]

What is the proposed behavior of a frozendict when it holds a mutable and that mutable is mutated?

[frail scaffold]

Whats deepfreeze and deepcopy?

[torn viper]

I would assume it's mutated?

[robust ledge]

I do agree that at some point you cross over a boundary and the request of the system exceeds what it should be able to do.

[torn viper]

At least I'd assume it would be to dict as tuple is to list

[remote geode]

its muted, and if the frozendict contains something that isn't hashable then trying to hash the frozendict frails

[robust ledge]

I would guess that as well, currently. Haven't read the chatter on frozendicts

[spare raptor]

why are all objects heap allocated in python

[remote geode]

bc its interpreted?

[harsh anchor]

that is python's conceptual model

[spare raptor]

being interpreted doesnt stop a language from using the stack

[robust ledge]

deepfreeze doesn't exist. copy.deepcopy will make a copy of a mutable object (like a list or dict) and copies of all nested objects within it. Versus a shallow copy.

[remote geode]

i guess, but wouldn't objects going on the stack require them to be immutable?

I guess a fully slotted class could go on the stack

[spare raptor]

no

[rapid bridge]

Python can't do the kind of analysis required to guarantee an object allocated on the stack would actually have no references when the stack pops the frame

[remote geode]

if you can add arbitrary amount of size to an object, how can it safely go on the stack?

[rapid bridge]

And the internals don't allow relocation to try and mitigate this

[harsh anchor]

just don't deallocate it 📈

[spare raptor]

stack_ptr -= (whatever the size of your value)

[chilly whale]

by having it reference something heap allocated

[tropic temple]

i would assume any language that does reference counting couldn't put their stuff on the stack

[chilly whale]

at least not without some way to copy it to the heap after the fact

[rapid bridge]

Recently Python even moved parts of the call stack off of the conceptual assembly stack

[spare raptor]

that doesnt make sense.... reference counting could be for dynamic things like arrays, or values that need to stay alive until they are no longer used, and everything else stays on the stack

[rapid bridge]

Every object in Python stays alive until it's no longer used

[spare raptor]

yes i know

[peak valve]

Guys I code with python since 4 years and I have no clue of what language I should learn next. I was thinking something like c++, c or go. Anyone got any suggestions for me

[harsh anchor]

everything in any language should stay alive until it's not used 🥴

[robust ledge]

What do you want to build?

[rapid bridge]

Are you suggesting implicitly unboxing values?

[slender urchin]

Depends on what you want to make

[rapid bridge]

Pypy does do that afaik

[spare raptor]

but for example

x: int = 5
y: int = 10

what is stopping these from being on the stack?

[tropic temple]

there would need to be a lot of confusing pre processing to make sure every variable couldn't be that second case (values that need to stay alive until they are no longer used)

I guess they could just move it to the heap the second another reference points to the same thing, but at that point just put it all on the heap in the first place again :p

[craggy willow]

del x

[spare raptor]

ah

[peak valve]

Usually my projects consist of high requests per second. So concurrency and high perfomance is my goal for a language

[rapid bridge]

For local variables, pypy would probably unbox those. Maybe Pythons experimental JIT would as well

[spare raptor]

oh unbox like take them out of PyObject?

[rapid bridge]

There's a separate optimization for ints where their instances are reused rather than allocated anew every time.

[rapid bridge]

Yea

[chilly whale]

when you say "these", do you mean the integers or the variables?

[torn viper]

Haskell? It might not be something you end up using (I don't), but I found that it rather changed my perspective on a number of things (and for that, I think that learning it was worth it)

[rapid bridge]

Yea

[rustic needle]

But some "use" doesn't really mean used, but just hanging around forever. I'm looking at you, memoized methods 😒
(https://discuss.python.org/t/memoizing-methods-considered-harmful/24691)

[charred python]

That is "being used". Remembered for later.

[rapid bridge]

Yeah that one's kinda obvious

[chilly whale]

sure. It's also a memory leak in all but name. Remembering something you don't need to remember any more.

[rapid bridge]

Cache can't even use a weak dict

It just kinda has to work like that

[spare raptor]

like how js does it with like (not exactly)

struct Object { 
  ObjectType type;
  void* ptr;
  
  short small_int;
};

[rustic needle]

Why does this stop x from being on the stack? You don't need to make the used stackspace available immediately, do you?

[spare raptor]

stack space needs to be calculated at compile time

[rapid bridge]

Pypy is a JIT, so I think it'd just straight up generate code that works with machine words and gracefully degrades to bigints if things go poorly

[spare raptor]

this would break the entire reason the stack is fast, the reason the stack is fast to my knowledge is 1: it's a cache hit magnet, and 2nd, there are no holes in the stack

[spare raptor]

ah

[chilly whale]

do you know that there is only one copy of 5 and one copy of 10 in the program?

[rustic needle]

Is that related to weakref? Why couldn't it (or my own custom memoize decorator) use wekrefs?

[rapid bridge]

(to be clear, so would a real JS impl)

[spare raptor]

yes this would be at global scope in one file

[chilly whale]

no, that's not what I mean

[rapid bridge]

Because two distinct instances could be equal, and so should provide the same cached result

Even if the first was lost before the second one was even constructed

[chilly whale]

I mean that even if you do ```py
x = 10
y = 9 + 1
z = int("10")

[spare raptor]

I could guess that

java does similarly with like strings

[chilly whale]

!e ```py
x = 10
y = 9 + 1
z = int("10")
print(x is y is z)

[edgy kraken]

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

True

[spare raptor]

nice

[chilly whale]

which means that if the integers were put on the stack, the program would use more memory, not less, and have less cache locality, not more

[charred python]

I believe you but I have a hard time believing it. So if I were to do x = random.randint(0, sys.maxint) it has to find an existing object with the integer value (if it exists)?

[spare raptor]

really?
there would be a cpu preformance hit because it has to go to memory to reach that pointed value, no?

[rapid bridge]

If you consider the cache in a thumbnail service, keyed by say ("path", w, h), the tuple of a previous invocation would be long gone before the next invocation could happen given a weakdict (of course, functools.cache would be incorrect here)

[chilly whale]

no, this optimization only applies to integers in a small range, not the whole range of ints. -5 through 256 IIRC

[charred python]

ahh, ok 😌

[chilly whale]

it's the "small integer cache"

[rustic needle]

About 2: if you try to use the holes, sure. Otherwise i see no problem with del x just hiding x from the interpreted code but still occupying the space on the stack until it would have been removed as usual.

[spare raptor]

yea that makes sense i guess

[chilly whale]

the stack and the heap are both in memory, so it has to "go to memory" no matter what

[spare raptor]

ah

[silver plover]

Do you imagine a scenario where Python would ever interns strings more aggressively for similar purposes?

[chilly whale]

Python already interns strings pretty aggressively, and lets you choose to intern strings yourself if you want something even more aggressive

[spare raptor]

but if on the stack this could be optimized to each using a byte of memory perhaps instead of the 16 bytes PyObject does on 64 bit systems

[rustic needle]

Ah, with a custom memoize I guess that could be fine.

[chilly whale]

sure. Historically Python had two different integer types, one that was arbitrary size and one that was limited to the size of the C int type on the underlying platform. That was removed in Python 3 because the CPU cost of all of the checks and conversions far outweighed the memory savings.

[spare raptor]

they checked at runtime?

[silver plover]

Oh, I don't really see the current behavior as 'aggressive', more opportunistic/ low hanging fruit, as I understand it. I've never directly used intern... I know it's there, but somewhat wondered if it should be used more widely. Or perhaps it is in some libs I use, dunno.

[rapid bridge]

It's not possible to check anywhere but at runtime

[chilly whale]

yes - you need to, because those two different integer types could both be passed to some function like sum or math.sin or whatever

[rustic needle]

Wait. Hmm. What does this mean actually? Why is this a problem?

A cache doesn't necessarily have infinite size, so it can't be to guarantee the function+args doesn't get computed a second time.

[spare raptor]

maybe, but even using 8 bytes of memory (sizeof(long long)) is a win over 16 bytes of memory

[rapid bridge]

Functools.cache does, and a cache eviction strategy that's just "did the object live long enough" is probably rather ineffective

[chilly whale]

wait, it couldn't be 8 bytes of memory when you pass it to sum

[rustic needle]

The stack is no less in memory than any other variable. In most systems. No? (I'm not familiar with modern systems, nor mmu so I shouldn't make statements like this)

[chilly whale]

sum takes arbitrary Python objects. You need everything you pass to sum to be a PyObject* (or some similar discriminated union)

[spare raptor]

im saying if all integers on the stack used max int size

[chilly whale]

it wouldn't help, because it could still take something that's not an int

or an int that's larger than the limit of long long

[spare raptor]

so like 128 bit ints?

[chilly whale]

that works until someone wants to pass a 129 bit int

[spare raptor]

can python ints scale infinitely?

[chilly whale]

yes

[spare raptor]

ah

yea that would be difficult

to get py ints on the stack

[chilly whale]

even setting that aside, think of functions like print

[rustic needle]

Probably.

[rapid bridge]

You can assume the sum is given only integers within the scope, construct byte code that assumes this, and when it turns out it's wrong, rewrite the bytecode to do the more general setup

[chilly whale]

print takes completely arbitrary objects

[charred python]

I can finally count all the atoms in the observable universe?

[rapid bridge]

Graalpy and the new adaptive interpreter sort of do this

Tho I think they go general -> specific

[spare raptor]

hm

[spare raptor]

well so does printf in c

[chilly whale]

it's not really analogous, no

printf in C takes a format string that tells you how to interpret each of the things you've passed it

print in Python doesn't have a parameter like that, so print needs to discover for itself what type each argument is

[spare raptor]

hm

python types could provide optimizations

in the future maybe

[chilly whale]

in printf, the format string is the discriminator, in print the ob_type field of the PyObject structure is the discriminator

[spare raptor]

wow you know a lot about python

[chilly whale]

🙇‍♂️

[granite wyvern]

It really doesn't care much. obj.__str__ FTW.

[chilly whale]

yes, sure

[spare raptor]

how does it know what obj is though? I think that's what hes getting at

[chilly whale]

but even so, the point is that it couldn't just take a long long off the stack

[rapid bridge]

str is looked up on the ob_type

[chilly whale]

it needs to be passed something that tells it that the pointer it's been passed should be treated as a pointer to a long long and not to some other type of thing

[spare raptor]

yea

[chilly whale]

you can only get rid of the PyObject wrapper if you replace it with some other way to introspect the type

[spare raptor]

dynamic typing is difficult to optimize it seems

[spare raptor]

optimizations with types!!!
make the linting do stuff

[rapid bridge]

The idea would of course be to implement everything a second time for long long specifically, and wrap it back into PyObject when needed. It'd be pretty nightmarish in practice tho

Graalpy does work like this

[chilly whale]

either that, or have a new base type of PyObject with an extra field saying whether this thing is a PyObject or some optimized type

or jam that directly into PyObject

[spare raptor]

problem with that is that uses more memory for all objects, no?

[chilly whale]

not for all objects, no

[spare raptor]

right now PyObject is at the biggest it should be because 16 bytes is just enough to be copied in one instruction

[chilly whale]

I mean, very easy example, you could make the ob_type field store a sentinel that indicates that this is some optimized type

[spare raptor]

yes, and reinterpret the pointer as that optimized type

[chilly whale]

like, if ob_type == 0x1 then the ob_refcnt field isn't a refcount at all but the value of an integer object

you can do this, and it would save space, but it would probably cost far too much CPU by littering special cases everywhere

there is no ob_ptr field

[spare raptor]

oh

like whatever points to the underlying object

[chilly whale]

PyObject is 16 bytes, composed of exatly 2 fields, each 8 bytes - ob_type and ob_refcnt

[spare raptor]

then where is the actual value?

[chilly whale]

after the PyObject header

all objects start with the same common prefix, and the ob_type field dictates what comes after that prefix (and before it, actually, because the gc sticks some fields before it depending on the type)

[spare raptor]

ah

so pyobject is dynamic size

[chilly whale]

more or less...

the better way to think of it is that PyObject is a base class, that base class is 16 bytes, and subclasses of it might be larger

[spare raptor]

is PyObject a C++ class?

[chilly whale]

no, but its memory layout is very similar to one

[spare raptor]

so like

struct Point { 
  // PyObject header
  int x;
  int y;
};

[chilly whale]

https://github.com/python/cpython/blob/main/Include/cpython/longintrepr.h#L93-L101

[edgy kraken]

Include/cpython/longintrepr.h lines 93 to 101

typedef struct _PyLongValue {
    uintptr_t lv_tag; /* Number of digits, sign and flags */
    digit ob_digit[1];
} _PyLongValue;

struct _longobject {
    PyObject_HEAD
    _PyLongValue long_value;
};```

[chilly whale]

the memory layout of a Python int object is:

PyTypeObject *ob_type;
Py_ssize_t ob_refcnt;
uintptr_t lv_tag;
digit ob_digit[n];

[spare raptor]

so this would work?

[chilly whale]

yep

[spare raptor]

that's cool, but how are these put in arrays?

[chilly whale]

that's more or less how a tuple looks in memory

[spare raptor]

it seems like it would be very slow to iterate over them, because you have to figure out each element's type, then increment the pointer by that many bytes

is that why for loops are slow in python?

[chilly whale]

you don't usually put a PyObject itself in an array, you usually put a pointer to it in the array

[spare raptor]

oh

[dusk adder]

would you prefer it if python used curly braces instead of : and indent?

[spare raptor]

Bython

[granite wyvern]

!e
No. gah!

from __future__ import braces

[edgy kraken]

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

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

[spare raptor]

I heard python for loops are slow, is that true, and if so, why?

[dusk adder]

it's one of those things that seems to be most identified with python

[celest osprey]

because python is a stack vm so it pushes and pops and decref/increfs the refcount of each object in the for-loop

[fervent plank]

man

why do I find python harder than java 😭🙏

[celest osprey]

subjective but I think most people would say otherwise

[robust ledge]

Are you interested in what you're building in python?

[granite wyvern]

If you're coming from java, less is spelled out overtly in python.

[spare raptor]

please elaborate?

[dusk adder]

ive met several people who say that, and it seems fairly arbitrary

[spare raptor]

why?

[chilly whale]

like a Python list is basically ```c
PyTypeObject *ob_type;
Py_ssize_t ob_refcnt;
Py_ssize_t ob_size;
PyObject **ob_item;
Py_ssize_t allocated;

[spare raptor]

ah

[chilly whale]

the question is always "compared to what?"

[spare raptor]

not compared to anything, just anything about python's implementation that might cause preformance issues to occur in for loops specifically

[chilly whale]

if you're comparing to C loops, it's apples to oranges. C loops know that they're only dealing with integers. Python loops can deal with any type of object, so there's overhead involved in figuring out what types of objects they're dealing with

[celest osprey]

Python's interpreter is a stack machine, and it executes bytecode.
The interpreter reads the bytecodes, and some bytecodes push values onto the stack and some pop values onto the stack. Because of the pushing and popping, the stack owns a reference to the object and it has to increment the refcount and decref it when it is popped of the stack, so there is a lot more overhead. Not to mention the dynamic typechecking which adds evern more overhead

[spare raptor]

like for for i in [1,2,3,4,5]?

[chilly whale]

in Python you can do for line in file: or for elem in some_list: or for number in itertools.count():. That abstraction isn't free, and since Python is an interpreted language it doesn't know until runtime what type of object it's actually looping over

[spare raptor]

ah

so dynamic typing is difficult

[celest osprey]

i mean yeah it's slow ig

[spice hill]

that should hopefully improve once the JIT gets better

[celest osprey]

so to optimise python you got to assume types

and then fallback if that assumption fails

[spare raptor]

also the gil is being removed, what is it being replaced with?

[celest osprey]

it's not being replaced with anything, it just allows for free-threaded python, afaik

[spare raptor]

then how are race conditions avoided

[chilly whale]

the gil is one big lock. It's being replaced with lots of small locks and lots of atomic operations

[spare raptor]

ah

[spare raptor]

where do they put the locks?

[chilly whale]

everywhere 👀

[spare raptor]

like what gets wrapped in a mutex

[chilly whale]

but in particular, in free-threaded builds, every object contains a mutex for locking access to that object

[spare raptor]

ah

[chilly whale]

and there's other types of locks, too

[spare raptor]

so is that in like the pointer to the pyobject or the pyobject itself

[chilly whale]

in the pyobject itself, I believe

[spare raptor]

ah

[chilly whale]

https://github.com/python/cpython/blob/main/Include/object.h#L156-L167

[edgy kraken]

Include/object.h lines 156 to 167

struct _object {
    // ob_tid stores the thread id (or zero). It is also used by the GC and the
    // trashcan mechanism as a linked list pointer and by the GC to store the
    // computed "gc_refs" refcount.
    _Py_ALIGNED_DEF(_PyObject_MIN_ALIGNMENT, uintptr_t) ob_tid;
    uint16_t ob_flags;
    PyMutex ob_mutex;           // per-object lock
    uint8_t ob_gc_bits;         // gc-related state
    uint32_t ob_ref_local;      // local reference count
    Py_ssize_t ob_ref_shared;   // shared (atomic) reference count
    PyTypeObject *ob_type;
};```

[chilly whale]

yep

[spare raptor]

what is _object?

a gc object?

[chilly whale]

every object

[spare raptor]

ah

[chilly whale]

PyObject is a typedef of struct _object

[fervent plank]

im a complete beginner

I tried python first it was too hard, then I tried java and found it easy

[spare raptor]

oh

well, java makes things more clear

[fervent plank]

ppl told me python is simple

[harsh anchor]

it's about as easy as java

[spare raptor]

public static int add(int a, int b);

this tells you everything you need to know

def add(a: int, b: int) -> int

these types are optional in python, which means you sometimes just dont know

[fervent plank]

nah

I think that's the reason

[chilly whale]

one possibility is that you found Java easy because you had already learned some Python. You might have found Python easy if you started with Java and then moved to Python. You might just be discounting all of the stuff that's common to both languages and which you already learned while trying your first language

[robust ledge]

Usually is.

[chilly whale]

the biggest part of learning to program isn't learning any specific language, it's learning concepts - what's a function? what's a variable? what's a loop? how do I read this error message? - and a lot of that is common no matter what language you're using

[fervent plank]

yeah in college when we code with python it's in stuff like Webot which I'm not interested in and I don't understand, but when we code in java it's stuff small window applications, many console-based programs etc etc

so ig I associate python with hard stuff I don't understand

[frosty oriole]

which year are you?

[chilly whale]

it definitely is much easier to learn if you find what you're writing interesting

[fervent plank]

1

[fervent plank]

god I wish they teached us stuff we'd need in the future 😭🙏

[robust ledge]

They do. They teach you how to learn.

[fervent plank]

I don't get why this "webot" thing was even there in the first semester

[fervent plank]

okay

[frosty oriole]

you just need to get past the introductory stuff

[chilly whale]

hey, wait - I let this slip by earlier, but this is wrong. It does care, because the type is the thing that holds obj.__str__! You can't get the __str__ without the type (the PyObject points to the PyTypeObject, and the PyTypeObject contains the tp_str field). You're right that it doesn't have, like a switch over every possible type, but it instead indirects through a vtable owned by the type.

[fervent plank]

aight

okay thanks for making me realize chat

[chilly whale]

for what it's worth, Python and Java are very similar languages, in the grand scheme of things. If you get good with either one, you'll be able to pick up the other one pretty quickly.

[pallid garden]

-# really?!

[harsh anchor]

quite

[pallid garden]

ive never thought of python and java as "similar"

then again, i dont know java in-depth enough

[finite rose]

Java is based on C

[dry pike]

You could even say that Java is the father of Python

[rugged barn]

They honestly are

A lot of concepts are very transferable

Which I found incredibly nice

[pallid garden]

i mean...

ive never had to think of private/public in python

[robust ledge]

Except python is older, so that's weird.

[pallid garden]

or return type of main

[finite rose]

I think most Turing-complete languages are going to share many similarities

[pallid garden]

i dont think there's a concept of static in python?

[finite rose]

Static methods?

[pallid garden]

static class

[dry pike]

I didn't say that it would be accurate, only that you could say it.

[pallid garden]

or rather, static functions

[dry pike]

that's what @staticmethod is, no?

[robust ledge]

You could say it, but I'd expect a willy-wobbly timey-wimey explanation for the age difference.

[rugged barn]

!d staticmethod

[edgy kraken]

staticmethod

@staticmethod```
Transform a method into a static method.

A static method does not receive an implicit first argument. To declare a static method, use this idiom...

[finite rose]

A static class in Java is really just a scoping control

[pallid garden]

outside of a class

[rugged barn]

How would that look like?

[robust ledge]

Wouldn't that just be a function then?

[pallid garden]

-# yes ive just been listing each of the keywords from public static void main

[dry pike]

mods, get her

[rugged barn]

Mods, mods!

[pallid garden]

wait what

[rugged barn]

Also, minor rant

But I actively dislike the "right way" to write Javadocs

Because what do you mean, I'm supposed to write @implsec on the documentation

Teaching people how to implement stuff is weird

[finite rose]

Well in Java there’s no such thing as “outside a class.” There is no global space like in Python

[pallid garden]

correct

and so... i struggle to think of them as "similar"

there are so many differences between them

[rugged barn]

There are many similarities too

Depends on where you look

[pallid garden]

there's no list comprehension in java

[robust ledge]

Do you not normally write documentation on how to implement what you've designed?

[rugged barn]

I don't usually write them on the docstring

I usually do that on a separate document so I can easily share as markdown

[pallid garden]

java never had a GIL

etc, etc.

[autumn pelican]

never conform

[rugged barn]

I kinda get that Java is a lot more contract-bound and all that but, I don't find it great to tell people how to do things

But when in Rome, do as the Romans so
Whatever

[robust ledge]

The way I see it, even the function signature is telling the next-dev what to do. If it's any more complex than that, the instructions in the docstring only further helps the next-dev.

[wet fox]

once i got to concurrency in java i was out

ugly looking code

still my first lang and has a place in my heart

for a video game funny enough that wasn't minecraft

[finite rose]

A static method in Java is the same kind of thing as a static method in a Python class…… mostly.

[wet fox]

for a private server 👿

[dry yacht]

Stream.parallel() or the ThreadPoolExecutor make it pretty easy tbh

[wet fox]

i just dont get completable future chaining enough to not make it super nested

[finite rose]

But Python gives you three scopes in a class, instance, class, and static. Java only has instance and static. I’m tbh I still don’t really understand the difference between a static and class method in Python, other than you cannot refer to a static class method via an instance reference, only via the class name reference, like Foo.my_static_method()

[autumn pelican]

document like you are always working with a team, your team member might just be you coming back 6 months later trying to figure out what you wrote

[spice hill]

@staticmethod could arguably be removed from python with no major loss (besides breaking existing code)

[jagged belfry]

!e ```py
class A:
@staticmethod
def b():
print(1)

a = A()
a.b()

[edgy kraken]

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

1

[finite rose]

Well I was lied to! See, even the one thing I thought I knew I didn’t know

So what’s the difference then?

[spice hill]

@classmethods accept a cls parameter

[dry pike]

classmethod gets the current subclass passed

[jagged belfry]

Static methods are methods that are organized with a class, and are passed nothing by default.
Class methods are methods that operate on the class, and are passed the class as cls by default.
Instance methods are methods that operate on the instance, and are passed the instance as self by default.

[finite rose]

Not to confuse the issue, but a “static” method in Java aligns more closely to what you’d call a class method in Python

[charred tusk]

I don’t really get why static methods need to exist

[finite rose]

static methods in Java know what Class they belong to

[charred tusk]

What what

[spice hill]

Exactly.
https://mail.python.org/pipermail/python-ideas/2016-July/041189.html

Honestly, staticmethod was something of a mistake -- I was trying to do something like Java class methods but once it was released I found what was really needed was classmethod. But it was too late to get rid of staticmethod.

actually, maybe not quite...

Can you override a static method in a subclas in Java?

[finite rose]

So static methods in Python seem to just be a namespace for methods.

[steel whale]

does anyone have an article about how the L3 cache works?

[charred tusk]

On your CPU?

[steel whale]

the general workflow of most L3 caches

[jagged belfry]

Pretty much. A way to put a function in a class without changing it

[finite rose]

No, you can hide them though by redeclaring a static method with the same name. However, they are not inherited and there’s no runtime polymorphism. The method to be executed gets bound at compile time, not run time, unlike with instance methods.

[dry pike]

what happens if you call the Java static method from a subclass? it still refers to the superclass?

[dry yacht]

You probably mean the "shared cache", no? That's typically the L3 cache nowadays, but it doesn't have to be the L3 one

[finite rose]

And that’s generally considered bad practice in Java, whereas in Python (and I think JavaScript?) it’s a useful feature.

[dry yacht]

It depends on the static type of the object

[steel whale]

i don't see it being referred to as the shared cache

[rugged barn]

Static just belongs to the namespace while classmethods are class aware

you arguably can do static with classmethods anyway but like
eh

classmethods are much more powerful

[finite rose]

It depends on the type of the reference variable you use. If you have a static method() in both class Parent and Sub,

[dry yacht]

@dry pike Java has two types for each variable: The static type and the dynamic type
If you do:

A variable = new B();  // assuming B inherits from A
variable.func();
```then the static type of `variable` is `A`, but the dynamic type of `variable` is `B`, so if `A` has a `static` function `func()` then it would call `A.func()`, even if `B` overloads it

[dry pike]

but I mean, within the functin

because someone said that the function gets a reference to the type

[rugged barn]

the only like, actual valid reason may be semantics

[dry yacht]

Well, you know which class you're currently writing code in, no?
I assume this is what they meant (obviously then there's also reflection, but I won't get into that)

[timid ember]

dude how do you read the pyrr documentation

[finite rose]

It’s bad form to call a static method from an instance variable because of the possible confusion. So you would call Foo.aStaticMethod() from an instance method in the Foo class and never this.aStaticMethod() and then you will never be surprised.

[timid ember]

i had to go to the source code to find what the parameters of a matrix creation function expect

[spice hill]

!e
That's the difference between static methods in Java and class methods in Python then. You can do this in Python: ```py
import math

class Point:
def init(self, x, y):
self.x = x
self.y = y
@classmethod
def from_xy(cls, x, y):
return cls(x, y)
@classmethod
def from_polar(cls, dist, angle):
return cls.from_xy(math.cos(angle) * dist, math.sin(angle) * dist)

class LoggedPoint(Point):
@classmethod
def from_xy(cls, x, y):
print(f"creating a point with {x=} and {y=}")
return super().from_xy(x, y)

p = LoggedPoint.from_polar(425.63012111456584, 0.1628311844604171)
print(p.x, p.y)

[edgy kraken]

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

001 | creating a point with x=420.0 and y=69.0
002 | 420.0 69.0

[finite rose]

Yes, class methods can be overridden in subclasses. It’s nice. You can probably only do it in Python because it’s a dynamic language. Java resolves a lot of things at compile time so they are fixed in stone at runtime, whereas Python can happily redefine your foo:str variable to be foo:int at runtime. Now, the fact that you can do something like this doesn’t mean you should do something like this.

[robust ledge]

Substitution principle was respected in their example though.

[spice hill]

LSP is kind of a handwavy thing

[finite rose]

It’s an aspirational goal, not a roadblock

Is a square really a subclass of a rectangle? It depends.

[spice hill]

Whether a subclass method is compatible with the superclass method depends on what promises the superclass makes. If the code has no docstrings or comments, one is often left to speculate what the intended promises were

[finite rose]

a square has a restriction imposed on it that rectangle doesn’t have. Then again, a square can just be a degenerate case of a rectangle where width = height

[jagged belfry]

I guess here's an example of where staticmethod is helpful:

class Shape:
  @staticmethod
  def compute_area():
    raise NotImplemented

  def area(self):
    raise NotImplemented


class Rect(Shape):
  def __init__(self, width: int, height: int):
    self.width = width
    self.height = height


  @staticmethod
  def compute_area(width: int, height: int) -> int:
    return width * height

  def area(self):
    return self.compute_area(self.width, self.height)

data = Rect(5, 5)
if data.area() > Rect.compute_area(5, 4):
    pass

[steel whale]

i'm having fun with SIMD

[spice hill]

What's the point of having a compute_area in the parent?

[jagged belfry]

To raise if the subclass is improperly implemented. Could be an ABC but I didn't bother

[steel whale]

Collecting numpy
  Downloading numpy-2.4.1.tar.gz (20.7 MB)
     ━╸━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.8/20.7 MB 18.5 kB/s eta 0:17:56
  WARNING: Connection timed out while downloading.
  WARNING: Attempting to resume incomplete download (786 kB/20.7 MB, attempt 1)
  Resuming download numpy-2.4.1.tar.gz (786 kB/20.7 MB)
     ━╸━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.8/20.7 MB ? eta -:--:--```
what the fuck is my ISP on

[granite wyvern]

Yes. I was missing the cpython internals side of this discussion and the "how much info do you need" from that point of view. I was more going: print itself doesn't really care about which type, as long as it can indirect throuugh it to the __str__ method. Which is largely why I then shut up 🙂

[granite wyvern]

Fair enough. It's less.. dynamic. Things are quite explicitly spelled out.

[jagged belfry]

18.5 kB/s
What year is it

It's 1995. I'm getting 18.5kb/s on a download over my isp via dial-up
It's 2026. I'm getting 18.5kb/s on a download over my isp

[spice hill]

148 kb/s in 1995 was probably very good

[spice hill]

!e

from datetime import datetime, UTC
print(datetime.now(UTC).year)

[edgy kraken]

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

2026

[steel whale]

2026 and im still geting these speeds over ethernet

they genuinely have something against me icl

[granite wyvern]

We used to see that throughput on ethernet when one side was half duplex and the other full duplex and they were thus mutually misconfigured. A recipe for packet loss (and thus appalling TCP retry behaviour).

Ah, the modern pleasures of gigabit and full duplex and no crossover cables.

18kbps was pretty good for dial up. ISTR that trailblazer modems would achieve that, and not much else. At one point, anyway.

[dry pike]

wait you can get 11Gbps in some places now?

oh god nvm, 50Gbps

that means you could (theoretically) download 1TB of data in 160 seconds

[chilly whale]

Python and Java are both high level object oriented garbage collected imperative procedural algol-derived languages where objects are passed by and assigned by reference. And both are compiled to bytecode and then interpreted by a virtual machine. They're very similar in the grand scheme of things. Both of them are much more like each other than something like Fortran or even something like C

[spice hill]

yeah, the big difference is static and dynamic typing

otherwise, pretty similar

[charred tusk]

Enterprise gear is over 100 now
I didn’t think anything over 10 had trickled down to consumers yet though

[dry pike]

yeah this is consumer, Ziply Fiber that I found online

I assumed enterprise stuff was much higher since a while ago

[charred tusk]

Can I just like, live, in the data center?
All I need is a sleeping bag
I just want the Ethernet cable

[granite wyvern]

You will want earplugs or ANC headphones.

[charred tusk]

That’s okay

[granite wyvern]

And a warm coat or something.

[charred tusk]

I’ve got fur

Now what about this thing labeled “halon”??

[granite wyvern]

Better than CO2 extinguishers.

[ionic bluff]

certain spots are not so bad in a dc.

[charred tusk]

Do I inhale or exhale first?

[granite wyvern]

Best to inhale before they go off.

[ionic bluff]

having to stand in a cold aisle does suck after a few minutes.

[dry pike]

first exhale, then exhale again

then forgot how to inhale for 10 seconds

then inhale

[jagged belfry]

Yep. If you have a good local link, ethernet is faster than most ssds.

[dry pike]

what if we put fiber in da PC

[jagged belfry]

you can do that

[dry pike]

lemme guess it's so expensive that no one makes those

[jagged belfry]

50$

[dry pike]

?

[jagged belfry]

Amazon has a bundle: PCIE card to SFP, SFP transceiver, 3.8m cable for 48$

of course, you need a router, and probably a longer cable, but like

[dry pike]

why would you need a router

[jagged belfry]

Well, because this replaces ethernet?

[dry pike]

it's not replacing ethernet wdym

fiber within the PC, not to the PC

[jagged belfry]

oh

[granite wyvern]

It's a NIC. Replaces an ethernet NIC.

[jagged belfry]

Yeah, and a fiber cable, replaces the etehrnet cable

[granite wyvern]

I went looking for "optical sata" and didn't get what I wanted.

[jagged belfry]

Fiber within the pc isn't worth it. Too many turns and you're already electrically so close the translation between light and electric isn't worth it

[granite wyvern]

Unless it's all optical.

[dry pike]

does it introduce some latency?

wow, optical computer

[granite wyvern]

COnversion will.

[jagged belfry]

Someday there will be functional optical transistors on a mass scale, but that day is not today

[dry pike]

that's what I meant

[dry pike]

do you think we'll survive that long?

[jagged belfry]

Probably? Like, I'm gonna be real, if we die it'll be self-inflicted. Modern crops and fertilizer techniques aren't going away and could resurrect us even after most apocalypse scenarios. Despite the global economic chaos, "Everyone is gonna die" isn't particularly realistic. It's hard to kill an entire species. To clarify, if the US set up a relatively small array of modern nuclear reactors, it could reverse the flow of the mississippi river and desalinate all the water needed.

As for individuals, idk. Depends. A fucked up political state can last a few centuries, but the internet has kind of changed everything (and with the advent of mesh and ham networks, isn't going away ever)

[dry pike]

between 150 and 200 species go extinct every day.

[jagged belfry]

And there are trillions of species on the planet. As bad of an ecosystem disaster as it is, there is a big difference between a disaster and obliteration.

You can think of it in terms of nines of uptime/reliability

[dry pike]

I don't know what "nines" means

[silent whale]

I had to search if this is true, and it is... this is depressing

[jagged belfry]

99.9% (3 nines) uptime is 8 hours of downtime per year

99.99% (4 nines) is 52 minutes of downtime per year

[granite wyvern]

which is 3 9s.

[jagged belfry]

99.9999% (6 nines) is 32 seconds of downtime yearly

[fiery yarrow]

uptime.is moment

[jagged belfry]

It's estimated there are about 8 million species on earth. 99.9999% of 8 million is 8 species lost per year

[dry pike]

I think a puzzle site I play on has about 3 nines scheduled

[jagged belfry]

Now there's a lot of impact variance. For example, there's a lot of species of bugs and bacteria that are not as striking as losing big cats or rare birds.

But to drag this back to humanity

[dry pike]

justice for small cats

[jagged belfry]

At one point in history, it's estimated there were about 1300 humans left alive.

[dry pike]

the world was so close to healing

[jagged belfry]

To get there from today, 99.9999999% of humanity would need to die

nine nines. That's just intrinsically difficult. Not much in reality can be that thorough. so, no, I don't think there's a concern there in a "wipe out" moment. Same with regular cats and regular dogs.

[silent whale]

Just learned that 99% up time equals to 3d and 16h cool

[fiery yarrow]

catastrophic events to large-scale environments can occur when a single species disappears

[silent whale]

I'm not a biologist and I don't understand anything about biology. Why are you bringing those numbers? What is your point? Is it that humanity is close to extinction?

[granite wyvern]

This all came from NWJ asking if we would live longer enough to see all optical computers.

[jagged belfry]

No. It's that humanity is very far from it.

It also relates to optical computation as it exists today

A modern CPU has a bit error rate somewhere south of 1/10^12

[dry pike]

apparently quantum entanglement is not instant

[jagged belfry]

or 99.9999999999% accuracy

If you want to compete, then optical transistors also must reach that error rate, which is (IIRC) one of the main struggles today with that and quantum chips

[silent whale]

99.9999999999% (12 nines) is 0.32 milliseconds of downtime per year

[dry pike]

sorry, that was the formation of the entanglement

[jagged belfry]

My blog is six nines, roughly

discord.. significantly less

github less than discord

[dry pike]

how much work did it take to get your blog to 69s?

[half pewter]

coming dangerously close to six sevens

[jagged belfry]

Not much. A stable framework, nginx, and then hosting it on a vps instead of a cloud service.

[fiery yarrow]

micro, not milli

32 millis is 9 nines

[tropic temple]

if your downtime is close to six sevens you should be reevaluating your choices

[fiery yarrow]

the 3d printing website thingiverse is infamous for being constantly broken. there's a joke their uptime isn't 5 nines, but 9 fives

[dry pike]

What if we do hexadecimal

4 Fs

[fiery yarrow]

huh, not usually the case where one can pull an "i saw that" on nedbat

[ebon comet]

Better sites: https://github.com/oernster/3D-printing-info/blob/main/stls/stl-sources.md

[lusty spoke]

yurrrr

[pallid garden]

good morning (again)

[ashen cipher]

my website is up for 9 zeros

well depending on what website wer'e tlaking abuot

[fiery yarrow]

that's too high. make it 12 zeros

[ashen cipher]

i mean it needs to be up for a bit

10 zeros sounds more correct

[pallid garden]

if my website is hosted on say, cloudflare, and cloudflare goes down, how could i switch over to aws programmatically?

[harsh anchor]

what do you mean by "hosted"

[pallid garden]

change dns record?

[pallid garden]

say i deployed a static page to cloudflare

just a simple html

[pallid garden]

my dns provider is also cloudflare, how would i change my dns record if cloudflare went down

[harsh anchor]

you can't

[robust ledge]

Are you looking for a DNS failover solution?

[pallid garden]

huh, i never heard of that

let me look into it

[pallid garden]

wait, does this mean i have a dns setting that automatically switch over to a secondary host if the primary host goes down?

[robust ledge]

More or less. Load balancing.

[pallid garden]

or does this mean when the dns goes down, a secondary dns takes over

[robust ledge]

I'm confused by "dns goes down".

[pallid garden]

so what happens when cloudflare goes down?

like last year

[robust ledge]

Half of the states are no longer online.

[pallid garden]

my website worked just fine, it was on aws

[robust ledge]

You want a multi-cloud solution?

[pallid garden]

im not sure what the definition is

i guess if you want a service to survive these kind of events, you basically need to be the nameserver?

[robust ledge]

It sounds like you want load balancing (or whatever name said service calls it). Mutli-cloud just means you're using more than a single cloud provider. To your "what if cloudflare goes down"; there are 300+ cloudflare datacenters. Not all of them go down at once. If all of cloudflare disappeared, nobody would care your website was down. Least of all, you.

[pallid garden]

i can imagine having my website hosted on vercel (and by proxy, aws), cloudflare and on-prem, and then just having the dns failover to the secondary and tertiary host

but if the dns nameserver goes down theres basically nothing you can do?

[robust ledge]

The DNS host that can change the route needs to be somewhere.

[pallid garden]

it's not possible to have my records on both the google nameservers and the cloudflare nameservers, right?

If all of cloudflare disappeared, nobody would care your website was down. Least of all, you.
this is just an hypothetical scenario

[robust ledge]

Hypothetically, that is a freaking doomsday situation.

[pallid garden]

im not actually going to spend the engineering effort to ensure my little site survives internet catastrophes

this is all just academic discussion

[robust ledge]

I don't understand networking enough to answer this correctly.

[pallid garden]

i dont even know whether my question even makes sense

turns out it is possible, but it's only for enterprise customers

at least according to an answer in here: https://community.cloudflare.com/t/can-i-use-cloudflare-when-domains-nameservers-are-pointed-to-google-domains/345284

[robust ledge]

🤷‍♀️ I think I know what you're asking. But take the AWS us-east-1 dynamodb issues that took down a chunk of the internet for a day. The takeaway for corps there wasn't that they needed a multi-cloud solution. They just needed to use what they had more correctly. ((multi-region in the case of aws))

So maybe work on being resilient to failure in your cloud of choice. Once those patterns are understood you can tackle the increased difficulty of a multi-cloud solution.

[pallid garden]

my surface-level understanding of nameservers goes like this: if i request for say example.com, and their nameservers is cloudflare, my computer first gets directed to cloudflare dns to ask for the IP address, and then it gets directed to that IP address where the site is hosted

so what if the cloudflare dns goes down?

suppose i need this service to be like crazy resilient

[robust ledge]

The world cries.

[granite wyvern]

Anyone dependent only on them is stuck.
Most people have a few.

[pallid garden]

have a few?

[granite wyvern]

Eg 1.1.1.1 (cloudflare) 8.8.4.4 (google) and etc

[pallid garden]

it's possible to have your records on multiple dns nameservers?

[granite wyvern]

I'm talking about the upstream nameservers a client like your laptop might use. On UNIX, the /etc/resolv.conf file.

[pallid garden]

well, my primary is cloudflare, my secondary is quad9

[granite wyvern]

There's 2 sides to this: who your client/laptop asks, and who hosts the zone file (the names/ips in some domain like example.com).

[pallid garden]

so if cloudflare goes down, will i still be able to access my site through quad9, which i have registered on cloudflare?

[robust ledge]

Define "goes down".

[granite wyvern]

So my laptop here's using:

search wauk
nameserver 172.16.3.1
nameserver 1.1.1.1

at present. The 172 address is my home firewall/dns/dhcp server.
The 1.1.1.1 is cloudflare.

[pallid garden]

cloudflare dns servers borked

[granite wyvern]

Zones are usually hosted on multiple machines/services also.

[pallid garden]

i dont actually understand zones yet

i should probably look into it

[granite wyvern]

It's the collection of records for a domain.

[pallid garden]

so i can host it on multiple services?

i can use both cloudflare dns AND route 53?

oh wow https://superuser.com/a/1725125/551781

turns out yes

and this is quite commonplace

[granite wyvern]

Pretty sure. The parent zones have NS records for the machines which can answer questions about your zone.

[dry pike]

[pallid garden]

i see

today i learnt

[granite wyvern]

NS == nameserver

[robust ledge]

Nobody mentioned rust. Go back to waiting for your cargo build /j

[granite wyvern]

So when you talk to a public service like 1.1.1.1 it in turn talks to other DNS servers to get your answers. There's lots of caching.

[pallid garden]

cant open discord when rustc is running

OOMs

[granite wyvern]

More swap.

[robust ledge]

Just buy more memory.

[granite wyvern]

Or fast disc.

[pallid garden]

in this economy???

[dry pike]

I don't even write rust

[robust ledge]

I know, right . It hurt so much.

[dry pike]

also, rust is always on topic

also, it's just a fun emoji

[pallid garden]

@mods is this true?

[dry pike]

mods will say it's not, but that's because of their programming

deep down they know

[pallid garden]

TIL route 53 is called route 53 because dns port is 53

[robust ledge]

Just wait until you learn why boto3 is called boto.

[pallid garden]

idk, why is boto3 called boto

[granite wyvern]

Why is GNU called GNU?

[pallid garden]

bottle dolphin?

OHHHHHHHHHH AMAZON

haha

[robust ledge]

🐬

[granite wyvern]

Can you enlighten this ignorant user here? (me)

[pallid garden]

boto is apparently a species of bottle dolphin that lives in the amazon river, which is a play on amazon web service

[granite wyvern]

Ah, thank you.

[crystal scarab]

i used python for something that has made my life easier, and it is amazing (im very new to coding)

[barren sentinel]

Any thoughts for making this faster? I want to try removing a directory, run a command, then put it back if it fails. Currently I’m using shutil.move + tempdir.TemporaryDirectory, but it’s very slow for large folders. Any ideas for speeding it up?

[crystal scarab]

like 10x faster then how fast used to do it

just felt like saying that here

[pallid garden]

rename the folder to a temporary name?

[orchid pier]

Python too hard bro

[robust ledge]

Sweet! What'd ya do?

[pallid garden]

wait does shutil already do that

[barren sentinel]

Hm, that might work? My main concern is it wouldn’t if the folder name doesn’t matter to the program being run (in this case it’s ty)

[pallid garden]

i see

hmmm

idk whether it would work but i would try adding a pattern with https://docs.astral.sh/ty/exclusions/#excluding-files

maybe ignore all **.temp.**?

[crystal scarab]

i work at a car dealership, and i have to make facebook ads (no i did not make an auto fb ad maker and am not advertising 😭) but it just scrapes info from my companys website using selinium, then i have a little qt gui that comes up and it fills in the answers for the fb prompts, and it also downloads the listings images so theryre easy access and i experimented with threading for the first time, and it made the scrapping go from 2 min to 20 sec, and it felt like i waas einstein, and i made the gui really pretty took me like 2 days, all around went from taking 20 minutes to making a fb add to 2 min, i just have copy and paste boxes for the writable stuff

its not automated just makes life easier

[pallid garden]

🎉

[robust ledge]

That's kick ass, actually.

[barren sentinel]

Ooh, that is an idea. Similar would also be like adding a command line config for forced ignoring a file, though that would be hard to persist

[pallid garden]

oh right

that might work

[crystal scarab]

i dont have img perms

😭

how do you get img perms

[barren sentinel]

Either of these would also make it easier to re-run, since it things go wrong it’s quite annoying having to unzip a 1 gig zip again

[crystal scarab]

its made my life so much easier and i am so happy and proud of myself

[pallid garden]

bribe the mods to make you a staff

[barren sentinel]

You can’t, at least for this channel. You can post images in any of the off topics or in the help post channels though (though also please don’t dox yourself)

[crystal scarab]

i was just gonna show img of my programs gui

[pallid garden]

you can send in #ot0-psvm’s-eternal-disapproval and just link the message here

[crystal scarab]

thankyou i put them in there

[barren sentinel]

Gaming, what ui framework did you use?

[crystal scarab]

pyqt5

[barren sentinel]

Hm, I might have to push that up my list of ones to look at from my project

[crystal scarab]

emojis make everything look better😎

[barren sentinel]

(For those in the know I still haven’t pinned down what to use for my regex visualizer ;-;)

[crystal scarab]

it auto opens my dealerships page, aswell as my fb, and it saves the cookies for my fb, and also has a feature to delete those cookies its amazing

[pallid garden]

my eyebrows raised instinctively when i heard that

[crystal scarab]

i just wasnt about to format my own images and spend another hour or two, aligning everything easier to just type it out 😭

so emojis on top

[pallid garden]

i guess im too boring a person haha

[barren sentinel]

Though I did have an idea for making a backend-agnostic easily transpilable gui framework, does this sound completely insane or semi reasonable?

It would be immediate mode, based around only percent-based or absolute positions (so no flex boxes) layouting outside-in

[crystal scarab]

i also just sent a photo of what it looks like when i input a url

the cycle took 14 secounds to pull all the data

[pallid garden]

how many pages?

[barren sentinel]

And use an imgui-style draw list for everything in the backend made of primitive draws/text draws for the backend agnosticism

[fathom vortex]

A GUI framework without Flex boxes would be kinda useless tho?

[crystal scarab]

it pulles one listing, i downloaded 20 images, ran all the info in open ai to get a description, and pulled all the web data for the info to fill in things, it could prob be faster but im happy with it for now

[pallid garden]

i see

i guess it's way faster to code it like this

[crystal scarab]

500 ish lines

im also very new to coding

i started learning python maybe 4-5 weeks ago

[pallid garden]

oh wow that's impressive

[barren sentinel]

That’s one of the main things I’ve been thinking about with this, since I really like immediate mode, but that makes flex boxes basically impossible easily, so my idea is to have inside out layouting using only absolute positions or percentage based sizes which should cover my needs, since I don’t see myself needing flex boxes for most of my uis

[fathom vortex]

If it works for ya, cant say much about it

I have written a layout engine and yeah, flex boxes make things tons complicated

[crystal scarab]

my job said i was alowwed to spend time doing anything that would help me sell more cars, i have been spending atleast 8-12 hours a day learning to code for the last 35 ish days 💀

[barren sentinel]

Immediate or retained?

[crystal scarab]

my brain hurts beyond belife

[white knot]

Wow am learning new patterns from copilot suggestions

[dry pike]

hope they aren't antipatterns

[fathom vortex]

Don't use AI for that, you might learn antipatterns

Writing code like an llm is not desirable skill to have

[white knot]

Ah but they are quite easy when i think about it.

Like its inserting to a db. I thought of dumping everything at once or one by one

But it suggest to make it into chunks which is better for network calls

[white knot]

But you could filter out and learn the good things that it gives right

[dry pike]

how would you know that they're good?

[fathom vortex]

You dont know whats good as a beginner and ai loves to hallucinate and gaslight

[white knot]

ah like some sorta feeling.
For example here making many network calls to supabase db was bad right

[dry pike]

but why would middle ground be better?

[fathom vortex]

That feeling is called intuition and you get that from experience

[white knot]

Yea thats correct. Not believing it entirely

[fathom vortex]

Ideally dont really use AI for learning, take stackoverflow or other resources

[white knot]

Like it process in batches. So like number of times the loop runs becomes less

[fathom vortex]

Batching isnt always faster

[dry pike]

by that logic, wouldn't running it only once be faster?

[white knot]

But i have a json with like 20k lines from which am inserting. Would that work like one http request can have that much stuff

[dry pike]

that's probably quite within range of default limits of the web server you are interacting with

[white knot]

Oh

I thought supabase would block me

[dry pike]

you think they would block one large request and not a bunch of chunks of the same data?

[white knot]

Yea

[dry pike]

the chunks take more resources to deal with

[white knot]

So i have to time.sleep it ?

[dry pike]

for what reason?

if you can do it all in one request, do that

[white knot]

Ok I will try. But this what copilot suggested

[granite wyvern]

os.rename ? should be instant if you just move things sideways.

[dry pike]

but you had no reason to trust it

[white knot]

😅

[barren sentinel]

Is that not the same thing move does?

[granite wyvern]

Well I'd have thought so, but yousaid it was slow.

A cross device move involves a recursive tree copy, which would be slow. But on the same fs? Just a rename of the top dir3ectory should be immediate.

[barren sentinel]

I don’t that that matters for this, but that is actually a good thought, since in one of my setups the default tempdir behavior would make the rename cross-drive

[granite wyvern]

You can tell TemporaryDirectory to put the directory in the same directory as the thing it will be replacing. If you're doing what it sounds like you're doing.

[barren sentinel]

But then it would still be discoverable

[granite wyvern]

With eg listdir? Yes. But it will be unreadable IIRC (at least by people who aren't you).

[barren sentinel]

By the program being run (ty in this case)

[granite wyvern]

It's not clear to me what your use case is. I've got an atomic_filename(target_filename,.....) context decorator I use to make files atomicly. Make a temp file, write to it, rename to the target file is nothing goes wrong. And I've a similar thing for directries.

[granite wyvern]

If the temp file isn't named like a file ty cares about, it shoud ignore it.

Sorry, temp dir. But the same applies.

I don't know ty's habits, but most things won't bother examineing dot names (.foo)

[barren sentinel]

Hm, would module discovery still happen if the file was renamed

[granite wyvern]

Again, I don't really know what you're doing.

If you need to import stuff, having the right dirs in sys.path does the trick. If you're doing something else, provide details.

[marsh fulcrum]

Hello

[barren sentinel]

Minimizer, trying to remove as much code as possible before the whatever issue in ty stops happening

[marsh fulcrum]

I am a Python script developer

[granite wyvern]

So:

• move something away, try ty
• repeat until happy?