The Nineteenth Byte

emanresu A

01:14

We just released Vyxal 42.0 with a huge amount of new features

7

Dannyu NDos

Woah

Is that version system akin to GNOME?

emanresu A

(check the date lol)

Dannyu NDos

Dang it.

lyxal

01:28

@emanresuA my name is Lyxal golfing and I approve of this message

Someone

woah

I haven't even thought of a prancc yet

Unrelated String

01:54

i'm going to prank everyone by actually doing anything at all

they;ll never see it coming

DLosc

I also have a language announcement. After years of losing golfing contests, I've decided it's time to give up on my principles and bring Pip into the modern golf meta. To that end, I'm adding compressed string syntax and switching to the log-bytes scoring system of Fig. The combined language will be known as "Pig."

Bubbler

02:24

Pig must be the name of a lang that is Python but with builtin names in Pig Latin

DLosc

02:37

orfay iway inway angeray(enlay(amesnay)):
    intpray(iway + 1, amesnay[iway])

I hate it already

Unrelated String

think of some way to pig latinize the structure of the syntax too

function calls are postfix

operators are prefix

iway inway ((amesnay)enlay): orfay
    (+ iway 1, [iway]amesnay])intpray

maybe the for keyword should delimit the end of the entire block lmao

DLosc

But if you move something to the end, you have to add -ay to it. So would it be orfayway?

Unrelated String

ahahah

emanresu A

Operator precedence depends on whether its name starts with a vowel or not

Dannyu NDos

02:53

Pig latin reminds me of Korean 도깨비말.

Care to suggest an example input?

→ Cabare tubu subugebest aban ibixabampble ibinpubut?

...Yeah, not quite compatible to English phonotactics.

l4m2

tio.run/…

Why arguments isn't sparse?

att

03:32

probably because it's not an array?

Sandbox Posts

05:49

0

A: Sandbox for Proposed Challenges

Vending Machine Simulator code-golfinteractiverandom Alternate title: Linear Logic Simulator Objective Write an interactive program that asks you to design a vending machine, and then simulate the vending machine. The fuss is, the vending machine is designed for simulation of linear logic. Design...

Dannyu NDos

Help

l4m2

07:37

CMC How high can you reach in Minecraft given n oak planks? You're in a 1x1, height=inf hole, so e.g. slab can be used to increase 1b height

emanresu A

09:19

@l4m2 If n ≥ 8 (= crafting table + 3 ladders) you can in theory climb arbitrarily high, although doing so requires dreamluck

l4m2

@emanresuA Recycling ladder?

emanresu A

Picking up a ladder while standing on the one above yes

But the chance that you'll pick it up is small

Rydwolf Programs

13:55

LMAO so there's this AP test review site called Fiveable (5 is the highest score you can get), and they just sent out an email that they're rebranding to "fourable"

witj this link: fourable.me

What a way to be reminded it's April fools lol

Simd

15:33

I have code that uses divide and conquer. It works on a 2d numpy array. It splits the array into two halves, solves the problem for each half and then merges the result . I want to speed it up using parallelism by solving both halves in parallel recursively. Is that plausible in python?

This is really a hole in my coding skills

pxeger

It's possible but because of the GIL you likely won't get much of a speedup

unless you use multiprocessing (multiple separate Python processes) instead of just multithreading, which would allow proper parallelism

but that has higher overhead and is more difficult to use

Simd

Hmm... Can numba or cython release the GIL?

pxeger

or unless the majority of your time is spent in numpy functions which release the GIL

numba and cython are alternative runtimes and they don't even have a GIL, I believe

Simd

The computation in each half is currently in numba wrapped numpy code

pxeger

ah I see what you mean

I have no idea whether numba releases the GIL, you could do a test to find out

Simd

15:39

"Numba will release the GIL when entering such a compiled function if you passed nogil=True . Code running with the GIL released runs concurrently with other threads executing Python or Numba code (either the same compiled function, or another one), allowing you to take advantage of multi-core systems"

This might be too hard for me sadly. At least I don't know what to do

pxeger

running code in multiple threads isn't too hard

start with docs.python.org/3/library/threading.html

Simd

@pxeger I am worried about the overheads

pxeger

how long does your code take to run?

Simd

@pxeger about 1 second. The problem is that I run it thousands of times

pxeger

the overhead of starting threads won't be more than a few ms so you'll still see most of the benefit

Simd

15:44

Those thousands of times can't be done in parallel. So I want to speed up the 1 second

pxeger

alternatively if you use a threadpool (admittedly more complex) that should reduce the overhead

Simd

@pxeger I think I would need help with that

pxeger

you could try ChatGPT, it's pretty good at Python

yours doesn't seem like too tricky a use-case

Simd

@pxeger interesting! My experience is that it always gives incorrect code :)

I can describe the problem/solution if you have the time?

pxeger

import concurrent.futures

with concurrent.futures.ThreadPoolExecutor(max_workers=2) as executor:
    for i in range(1000):
        futures = executor.submit(conquer, left_half(data[i])), executor.submit(conquer, right_half(data[i]))
        results = concurrent.futures.wait(futures).done
        output[i] = combine(results)

something along those lines would work I think

Simd

15:55

Thanks. What do you think the overhead would be like?

I love anything with the word future in it :)

Or I could wait for python 3.13 :)

The Empty String Photographer

TIL: You can use your mobile phone as a security key.

DLosc

@Simd I assume you start all your Python programs with from __future__ import barry_as_FLUFL, then?

Simd

16:13

@DLosc I will now!

@TheEmptyStringPhotographer ???

DLosc

(TIL that everything that's ever been in __future__ is still in there, even if it has since been included in the core language. So you can also do stuff like from __future__ import print_function, unicode_literals, with_statement if you want.)

Simd

16:28

@DLosc that is pretty poor

DLosc

I imagine it's done that way so that old code that imports stuff from __future__ won't break when the future arrives

Simd

right but given that everything else will break I am not sure it makes sense

bpa.st/RNUA is the code. The two lines that do the calls I want to parallelize are:

d_left = get_max(cusum_A, left_half, k, level_for_first_column=end_levels[0])
d_right = get_max(cusum_A, right_half, k, level_for_first_column=end_levels[1])[:, ::-1]

@pxeger ^

Rydwolf Programs