CS people what do you call // ?

[ehproque@neopaquita.es]

@khleedril @futurebird I'm a quantum physicist, let's call it 1

[futurebird@sauropods.win]

@khleedril

You and the whole fifth grade.

[ehproque@neopaquita.es]

@khleedril @futurebird ℏ=h, where is your God now

[dalias@hachyderm.io]

@futurebird Comment?

*ducks*

[llewelly@sauropods.win]

@futurebird in all seriousnessness, the use of "floor divide" implies the result is rounded towards negative infinity, in contrast with "trunc divide" , which implies rounding toward zero, an important distinction when one operand turns out to be negative. Now here's a question: in your favorite languages, which way does integer divide behave? Does the language definition even require a particular behavior? (C90 and C99 for example call it "implementation defined". )

[jhominal@mastodon.social]

@martok this operator has returned an int since it was added in Python 2.2, so this seems weird?

[dalias@hachyderm.io]

@futurebird Jokes aside, is it a floor or a towards-zero like the awful C "division" operator? That would affect proper naming a lot.

[david_chisnall@infosec.exchange]

@futurebird

I can't read this and not interpret // as a start-of-line-comment marker. So this line is just 16. And I call 16 an integer.

[martok@mstdn.social]

@jhominal Only if both arguments are int. If either is float, the result is float even if its value is by definition int-ish.

I still think this violates the Principle of Least Astonishment, that's why I just don't use it anymore. int(a/b) is not that much more typing.

[yonder@spacey.space]

@ehproque

I still like to...

#define PI 3.14159265358979

it feels artisan. bespoke. handwoven

@futurebird

[jonathanhogg@mastodon.social]

@futurebird Years ago I got offered "show your Python skills" quiz on LinkedIn and the answer to this question was given as "integer division". I was so furious I put in a complaint. It is, of course, "floor division". And Python-style modulus is the One True '%' operator –C's remainder can get in the bin.

[futurebird@sauropods.win]

@yonder @ehproque

no no no THIS is "hand crafted" ... only the finest pi made of the most dubious loops.

import random
import math

inside=0.0
resolution=60000
for i in range(resolution):
x,y = random.random(),random.random()
if math.sqrt(x*x+y*y)<1:
inside+=1

pi=4*inside/resolution

print(pi)

[futurebird@sauropods.win]

@yonder @ehproque

importing math feels very goofy. I should write a square root finder to avoid that. How do those go? It's like newton's method I think.

[futurebird@sauropods.win]

@yonder @ehproque

And now that I think about it, we don't need random either, just make it an grid, if it's fine enough it should be OK, right?

[darkling@mstdn.social]

@futurebird @yonder @ehproque Newton's method converges wonderfully quickly, so it's often a good place to start. I believe that CORDIC algorithms can also do square roots (although I haven't seen or derived one myself) -- those can be nicer for implementation if you don't have an FPU (or are building one), because they tend to be all about adds and bit-shifts and not much else.

CORDIC only generates one digit per iteration, though, whereas Newton doubles the precision each time.

[burnitdown@beige.party]

@ehproque @futurebird i need more pi

[gbargoud@masto.nyc]

@yonder @ehproque @futurebird

Can random.random() be stuck in a constexpr to do that at compile time or did someone rightfully shoot down the idea of compile time nonsererminism?

[ehproque@neopaquita.es]

@burnitdown @futurebird I'm out, but here's some cake

[futurebird@sauropods.win]

We don't need any random numbers to be THIS terrible.

inside=0.0
resolution=100
for x in range(resolution):
x=x/resolution
for y in range(resolution):
y=y/resolution

if x*x+y*y<1:
inside+=1

pi=4*inside/(resolution*resolution)

print(pi)

[c0dec0dec0de@hachyderm.io]

@llewelly @futurebird floor can also be defined as going toward zero.

Programmers are amazing at not knowing the foundations of things. I worked somewhere that was absolutely confounded by numpy’s use of statistical rounding (x.50 goes to the even value not up, to avoid biasing your data in a systematic manner). I, freshly graduated from college and having taken a stats course, knew what it was immediately.