The Nineteenth Byte

10:00 AM

I can't wait for Unicode to be a full image standard.

Well, we really only need 256 intensity selectors, then we can combine red, green, blue, and transparent (space?) symbols with the appropriate intensity selector to recreate rgba. Combined with newlines, a text file can now represent a 2D image. Add form feeds, and you can have animations too!

We can then use this system to render all other symbols we'd ever need. Finally, the dream will come true!

pxeger

@Adám we don't even need intensity selectors - just use the 256 Braille characters which can be interpreted as binary numbers!

Adám

True.

@pxeger So how would you represent a dark-skinned cis-male having an almost-white boy and a medium-tan neutral-gender child with a generic-skin trans-female?

pxeger

well to be honest I would argue that doesn't need to be in Unicode

Adám

If I were in charge of Unicode, I would definitely prefer no (or at least less) emoji over lots of combinations with ZWJs.

pxeger

10:07 AM

You can just put a dark-skinned block, then a cis symbol, then a male symbol, then an almost-white-skinned block, then a child symbol, then a male ..., and let the human work out what it means

Adám

@pxeger Sure, but sticking in the ZWJs ensures that the big tech companies can boast about being inclusive.

And as a bonus, it also makes the cost of entering the market so high that big tech doesn't need to fear any newcomers.

Think of the kids!

A username

Hi

pxeger

⢒⠴⠄⡦⣶⡎⠄⣶⡶⢦⠴⠄⣮⢦⠶⣆⣶⢶⢦⠄⣶⢮⡎⠄⡶⢦⣮⠄⢪⡶⢖⣆⣶⠦⢦⠄⣶⡮⢦⡎⠶⣶⡎⠦⣎

(that's Braille-Unicode-Binary for "I, for one, welcome our new Unicode overlords")

Adám

I, for one, appreciate rendering of bit octets as letters.

pxeger

CMC: Braille-Unicode-Binary (reference implementation: Try it online!)

Adám

10:12 AM

@pxeger dupe?

pxeger

oh

:(

A username

The best way of saying thanks

pxeger

missed opportunity to say "1 byte taken of Neil"

3

A username

That's horrible

pxeger

~~Neil had better not help me golf too many times, or he'll have died of cannibalisation~~

3

A username

10:15 AM

This is worrying

Remind me not to help anyone golf

Neil

@RedwolfPrograms that's just data:text/html,%3Ctextarea%3E but then again I would describe Notepad as a text editor

@cairdcoinheringaahing well you've only got yourself to blame

pxeger

@Neil May I introduce you to t.pxeger.com/xs.html? It's a mildly better version of data:text/html

Neil

10:31 AM

data:text/html is golfier though

@Ausername after reading my answer he golfed 112&2/ to 2/56&

StackMeter

11:02 AM

@Adám yes

why does Jelly/05AB1E use so many builtins

there are 256, and most of them are literally so far-fetched: just use a library

Adám

They are golfing langs!

StackMeter

yes

but why some things like halving, when divison exists

how is halving going to matter

and the decrement and increment operators

just use 1+

A username

Eclipses are cool

StackMeter

they are

it's like the sun was swallowed

A username

11:17 AM

Or the moon, in this case

StackMeter

yes

send a pic over

A username

Sorry cant

but is cool

StackMeter

ok

send link then

@ᴅᴀɴɪᴇʟッ who is this

@xash welcome to the nineteenth byte

A username

Just google 'lunar eclipse'. Something recent should show up

Or ask Lyxal.

lyxal

Can confirm

A username

11:23 AM

Southern hemisphere gang

lyxal

StackMeter

lol

lyxal

Taken a few minutes ago from my backyard

StackMeter

I'm well into the northern hemisphere

and it cold out here

lyxal

@Ausername is the moon red for you too?

I mean it probably is

But is it as red as it is here?

StackMeter

11:24 AM

bruh

the moon is grey

its a piece of rock

lyxal

Moon looks red to me

A username

Right now, no, it is totally invisible.

Peak of eclipse.

15 minutes ago it was.

StackMeter

@lyxal then you must be mistaken

pxeger

Nice OPPO phone you got there Lyxal

(please be careful when uploading images from your phone to the internet; they can contain metadata best kept private, especially location/coordinates)

A username

Wait what?

pxeger

11:27 AM

exiftool can reveal said metadata

ExifTool Version Number         : 12.26
File Name                       : PqKl0.jpg
Directory                       : .
File Size                       : 153 KiB
File Modification Date/Time     : 2021:05:26 12:23:13+01:00
File Access Date/Time           : 2021:05:26 12:24:39+01:00
File Inode Change Date/Time     : 2021:05:26 12:24:14+01:00
File Permissions                : -rw-r-----
File Type                       : JPEG
File Type Extension             : jpg
MIME Type                       : image/jpeg

A username

I knew screenshots contained embedded XML, but...

pxeger

do they?

A username

Here's a screenshot, open it in a text editor.

StackMeter

@pxeger I no longer trust images

pxeger

<x:xmpmeta xmlns:x="adobe:ns:meta/" x:xmptk="XMP Core 6.0.0">
   <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
      <rdf:Description rdf:about=""
            xmlns:exif="http://ns.adobe.com/exif/1.0/">
         <exif:PixelYDimension>638</exif:PixelYDimension>
         <exif:PixelXDimension>762</exif:PixelXDimension>
         <exif:UserComment>Screenshot</exif:UserComment>
      </rdf:Description>
   </rdf:RDF>
</x:xmpmeta>

StackMeter

11:31 AM

images are banned henceforth

yes

Maybe

y e s

####  #   # #####       #   # #   #   #   #####
#   # #   #   #         #   # #   #  # #    #
####  #   #   #         #   # ##### #####   #
#   # #   #   #         # # # #   # #   #   #
####   ###    #          ###  #   # #   #   #

  #   ####   ###  #   # #####         #    ####  ###  ##### #####         #    ###  #####  ###
 # #  #   # #   # #   #   #          # #  #     #   #   #     #          # #  #   #   #   #   #
##### ####  #   # #   #   #         #####  ###  #       #     #         ##### ####    #      #
#   # #   # #   # #   #   #         #   #     # #   #   #     #         #   # #  #    #     #
#   # ####   ###   ###    #         #   # ####   ###  ##### #####       #   # #   #   #     #

Sorry if that wraps weirdly...

?

StackMeter

11:36 AM

dammit it wrapped weirdly

I give up

lyxal

@pxeger jokes on you it's not an oppo phone

StackMeter

nope - still bad

pxeger

@lyxal OnePlus? RealMe? (owned by oppo)

A username

@lyxal Is it an Ephone?

lyxal

@pxeger realme :)

StackMeter

11:40 AM

no I don't see the image

images have had an EXCEPTION

Neil

@StackMeter so ++ and -- should never have been in C?

StackMeter

11:59 AM

@Neil yeah, just use +1

A username

What about C++?

And C--?

StackMeter

same byte count

@Ausername C+1, C-1

A username

Renaming the languages...

StackMeter

Yes

Neil

actually C++ is ((C+=1)-1)

Adám

12:06 PM

@StackMeter I'd be surprised how often I wish that APL had increment and decrement functions. Jelly is based on J, and J is based on APL, but J does have increment, decrement, double, halve, square, squareroot, …

Neil

12:17 PM

CMC: Score a Blackjack/Pontoon hand. Examples: "AA" -> 12, "AT" -> 21, "JJ" -> 20, "2468" -> 20, "3579" -> 24, "A357" -> 16, "A234" -> 20. Input can be a list of any 13 integers or ASCII code points between 50 and 205 inclusive e.g. 2-9TJQKA

Aaron Miller

12:32 PM

So close...

caird coinheringaahing

@Bubbler ROs can set up feeds using the Feeds user, mods have to change it to be a specific user such as NMP

hyper-neutrino

do we need our feeds back?

lyxal

Temporarily

caird coinheringaahing

12:52 PM

I think it'd probably be best to restore them, at least until Redwolf is sure he's fixed the current issues

May 22 at 21:32, by caird coinheringaahing

Feed links: New Main Posts. New Meta Posts. New Sandbox Posts

hyper-neutrino

1:13 PM

→ 43 messages moved to Trash

→ 8 messages moved to Trash

on second thought it might be confusing without the name of the feed indicating where the post came from

@cairdcoinheringaahing I don't suppose you saved the feed avatar URLs anywhere?

pxeger

...

oops, that's Redwolf's bot's

Anush

I want to optimize a discrete function. Are there good libraries for this sort of thing? (accessible from python?)

pxeger

@hyper-neutrino voilà

hyper-neutrino

:D thank you

pxeger

@Anush optimise how? for performance/speed?

Anush

1:22 PM

it's a function I want to minimize

pxeger

oh as in find the minimal value of? SageMath can probably do it

Anush

it's just not defined in an easy way. I just have a function I call and it returns a value

pxeger

brute force?

Anush

@pxeger yes.. smart brute force I am hoping for

pxeger

min(range(-10000, 10000), key=mystery_function)?

Anush

1:23 PM

it's in 1000 dimensions :)

pxeger

uhh

Anush

I can do a greedy hill climb

just looking for anything smarter

pxeger

machine learning?

Anush

go on...

how do you use machine learning for this sort of optimization?

pxeger

they tend to be good at smart optimisation of multi-dimensional black-box functions

@Anush I dunno, I've never used it :)

Anush

1:25 PM

:)

StackMeter

@Anush why

Anush

@StackMeter ?

I need more words :)

hyper-neutrino

→ 4 messages moved to Trash

caird coinheringaahing

Man Feeds are buggy :/

StackMeter

@Anush why 1000 dimensions

Anush

1:37 PM

that's just the problem

StackMeter

so you don't know why you need 1000 dimensions?

hyper-neutrino

yeah sometimes feeds doesn't post shit when we want and other times it posts 43 sandbox posts at once

i'm just glad it doesn't get rate limited so I can just trash them all at once

Anush

@StackMeter I do but describing the whole problem will take a long time and bore you

StackMeter

I have time

a lot of time

caird coinheringaahing

Redwolf's gonna wake up to some fun :P

StackMeter

1:42 PM

@cairdcoinheringaahing why

hyper-neutrino

@StackMeter his bots are ded again

StackMeter

did he get demoted

lel

caird coinheringaahing

It looks like both Redwolf and the bots died at around midnight Texas time, so they've likely been asleep as we've been discussing removing the bots

hyper-neutrino

if his internet isn't reliable enough i can discuss hosting it on my server with him; i doubt it takes any resources

StackMeter

@cairdcoinheringaahing :58129017 :58129017 :58129017 does this work

caird coinheringaahing

1:45 PM

ಠ_ಠ My desktop died yesterday, so I'm transferring everything across to my laptop. Apparently, even when you have browser sync turned on, Tampermonkey doesn't do anything, so I need to reinstall all 24 of my userscripts

hyper-neutrino

oof :/

caird coinheringaahing

@StackMeter It only pings once and please test chat in the Sandbox

StackMeter

ok

i did not know

there's a chat sandbox?

hyper-neutrino

yep. but please don't ping people just for testing (unless they agree to it)

StackMeter

ok

hyper-neutrino

1:49 PM

and definitely do not play with flags

StackMeter

don't worry I only pinged the bots

@hyper-neutrino I've made one flag in my life

hyper-neutrino

russian bot is not a bot

wait nvm maybe they are

nvm their math profile says they are one. thought it was just a user with a weird name based on their rep but it looks like they just got lucky with an answer getting a lot of upvotes

StackMeter

ok

do not worry

I have more than 5 braincells

because 172 > 5

Anush

:)

StackMeter

@Anush are you ready to tell me why you need 1000 dimensions

Anush

2:01 PM

sorry.. it's too long

it's not exactly 1000

StackMeter

I can wait

@Anush what is it then

Anush

hmm.. the number depends on the version of the problem

here is a different but related problem.

Imagine a have a long list of letters and I want to encode each one as either 0 or 1

for any encoding I can compute the (Shannon) entropy in the usual way as - p_0 log(p_0) - p_1 log(p_1). How do I choose the encoding that maximizes that value?

@StackMeter please ask if anything is unclear

caird coinheringaahing

I don't think you can, right? There's an infinite number of ways to encode an arbitrary letter as either 1/0, so there's no way to find the optimal max encoding

Anush

@cairdcoinheringaahing if there are 26 letters there are 2^26 different possible encodings

pxeger

@cairdcoinheringaahing what do you mean by "There's an infinite number of ways to encode an arbitrary letter as either 1/0"

because the number of letters is finite

Anush

2:13 PM

each letter either gets mapped to a 0 or a 1

so you could just look at all 2^26 possibilities. But if I allow capitals too then that gets painful

caird coinheringaahing

For example "Is it a consonant or a vowel?", "Does it have an odd number of 'ends' or an even number?" etc.

Anush

@cairdcoinheringaahing we are just mapping each letter to a 0 or a 1

caird coinheringaahing

You can choose literally any criteria to map a given letter to 1 or 0 though

hyper-neutrino

right but only 2^26 distinct mappings exist

Anush

so the question is how to maximise the entropy of the mapped letters

hyper-neutrino

2:16 PM

~~brute force all 2^26~~

caird coinheringaahing

@hyper-neutrino Ugh, I hate this kind of stuff ಠ_ಠ

hyper-neutrino

:)

Anush

in such a way that you could try to do it if you allow capitals too so we have 52 letters

@cairdcoinheringaahing sorry :(

caird coinheringaahing

2^26 is only 67108864, that's well within forcing range

@Anush No, it's on me, not you :)

Anush

yes.. 2^52 is not good though

hyper-neutrino

2:17 PM

wait that's actually smaller than i expected

Anush

let's assume we have lower and upper case letters

caird coinheringaahing

4503599627370496 (2^52) is less forcible :P

Anush

indeed :)

I wonder if this can be formulated as some standard problem that I could then plug into a tool to optimize

If anyone is interested I could provide some starting code in python to play with

and data :)

StackMeter

You want to choose the value that maximises p0log(p0)

oh wait it's all negative?

Anush

@StackMeter I want to choose the mapping that maximises - p_0 log(p_0) - p_1 log(p_1).

note the minus signs

StackMeter

2:26 PM

then you want to make p_0 and p_1 as even as possible, no?

Anush

where p_0 is just the proportion of times 0 occurs and likewise p_1 is the proportion of times 1 occurs (i.e. 1-p_0)

@StackMeter yes exactly

StackMeter

so you try and make the sum of the numbers of letters on the p_0 side equal to that of the p_1 side

Anush

sum of the frequencies of the letters

you are unlikely to be able to make them exactly equal but as close as possible

so yes

how can you do that?

StackMeter

well I think there's a challenge based on just that

making two sublists from a list as even as possible

though a tip - putting e, t, h, a, i and o on one side likely isn't correct with any real, long piece of text

Anush

if you could help me make the challenge that would be great

hyper-neutrino

2:30 PM

this problem is basically just given the frequencies of each letter, find a subset summing to as close to 0.5 as possible, right?

Anush

@hyper-neutrino yes

obviously you can extend to more groups than 2

user

@RedwolfPrograms Yay

Anush

hmm.. en.wikipedia.org/wiki/Largest_differencing_method

that looks fun to implement

StackMeter

@user lol

user

It's unlikely any of you actually like it, but whatever

@StackMeter ?

Wezl

2:33 PM

@user it's just too big to fit in my small little brain

user

:(

Wezl

I'm sure it has some inspiring things I could distill but I don't know where to start

StackMeter

@user I "lol"'d because you responded to a 16 hour old message 4 pages back in the transcript

user

My Scala detector beeped :P

caird coinheringaahing

CMC: Given a lowercase character (one of abcdefghijklmnopqrstuvwxyz), output how many "ends" it has (using DejaVu Sans Mono font). For example, a has 1, k has 4, r has 2. This is the full mapping. In weird edge cases such as i, go by the values in that link

Oct 9 '17 at 18:54, by Martin Ender

I felt a great disturbance in the Force...

StackMeter

2:38 PM

@cairdcoinheringaahing so did i

user

@StackMeter btw you may want to reduce your starring rate just a little. We always have nonsense on the starboard, but I at least prefer to see slightly higher-quality nonsense :P

StackMeter

ok

user

Also, whenever one of my messages gets starred, it makes me realize it sounds stupid :/

Wezl

^

caird coinheringaahing

@user So long as people aren't starring loads of messages, I think its generally ok to star anything you want. I'll cancel multiple single stars that happen on a group of messages, but if someone wants to drop the occasional star and no one agrees, no harm done

StackMeter

2:42 PM

unfortunately, I have been unable to reproduce this issue, therefore I'm marking it as statusno-repro

caird coinheringaahing

They leave the starboard the fastest anyway

user

@cairdcoinheringaahing I know, and I'm not saying it's a huge annoyance or that StackMeter must stop starring stuff, I just wanted to suggest they might want to slow down (apologies @StackMeter if it came out the wrong way)

caird coinheringaahing

That said, I don't think stars should be used as a "I find that funny" reaction, rather they should be "I think others will want to see this". Those can overlap, but aren't always the same thing :)

StackMeter

it's ok

user

Heh, I posted a Scala-themed KotH in the Sandbox a while after Redwolf mentioned Scala

@UnrelatedString It's not really partial application, it's basically treating the method as a field and then treating the field access as a postfix operator, so (1+) would be (1.+). No one uses postfix operators anyway, they're considered a mistake

StackMeter

2:48 PM

@user postfix is a sin

user

Amen

Wezl

*disagrees*

infix is a lot worse

user

Mixing them is bad

I guess if you have a stack-based language, postfix is alright

Wezl

@user *is making a language with only postfix and infix*

hyper-neutrino

postfix is the only operator type that really makes sense to mix with stack operations

though i don't get why you'd have postfix (or any) ops in a stack based language; IMO it's not really postfix it just looks like that in principle

StackMeter

2:50 PM

@Wezl Infix is normal langs, postfix is stack-based langs and prefix is dumb

hyper-neutrino

lisp: am I a joke to you

user

@hyper-neutrino In a concatenative language, it's not really postfix, but you could have a stack-based language with infix, postfix, prefix, all of that

hyper-neutrino

although lisp isn't even really prefix

Wezl

the problem with infix is it's less expressive. In prefix and postfix, operator operator means something but in most infix languages it's an error

user

And for good reason

hyper-neutrino

2:51 PM

@user true

StackMeter

@hyper-neutrino me: yes

Adám

@Wezl *cough*

Wezl

@Wezl so composition is much worse, except trains fix this

StackMeter

@Wezl why would you need that

Wezl

@Adám *wink*

user

2:55 PM

@StackMeter Composition can be extremely useful, but there's nothing wrong with having to do f.g instead of f g

@user To be clear, I'm talking about Scala here. I have no idea how useful or idiomatic it is in other languages

Wezl

there is something seriously wrong about using parentheses for grouping, and there's a reason stack-based languages don't need it

user

You can have a stack-based language that uses parentheses, right?

Wezl

yes, but what do you mean by "uses", because forth uses them for comments for example

user

Ooh, Agda's mixfix looks both awesome and a pain

@StackMeter May I ask why? (unless you were just making a joke)

Wezl

another huge problem is that a b c has to be parsed differently if b is an infix operator, which I'm fixing by having a b c use b postfixly(?) and a b: c use b infixly(?)

pxeger

3:02 PM

I was considering using something like those "mixfix" operators in my practical programming language (which is not and likely will never be public). I didn't know about them in Agda so it's nice to see some prior art

user

Couldn't you just not use prefix and postfix with infix

@pxeger I feel like those would be amazing for embedded DSLs, but I'm not so sure it'd be too helpful for practical purposes

Wezl

forth supports mixfix operators, and lisp, and red/rebol

user

Interesting, I thought Forth was just concatenative

Wezl

you can have basically macros that take tokens instead of a pre-parsed list

in lisp and red/rebol, mixfix operators ==== macros

pxeger

My language supports arbitrary custom operators using characters from the set !%^&*-=+:~./<>?|, and \;,@$[]{}()#' are reserved for other purposes. Is it a bad idea to use backticks for custom literals (à la Ruby %-literals or JS's backtick strings), or should I add some special syntax for double-quoted strings (because otherwise I've run out of ASCII characters)?

I know some people here don't like backticks, and I'm not a massive fan because they're really hard to use in Markdown, but do you think it's that bad a problem?

Wezl

3:07 PM

backtick-strings are fine (IMO)

pxeger

But then there's also things like heredocs, or Zig's multiline string literal syntax using backslashes, so...?

Or strings with arbitrary delimiters like Swift's #" "# (I think it's also in C++ and C#)

maybe I want those

Redwolf Programs

Morning everyone!

hyper-neutrino

hey there @RedwolfPrograms

Redwolf Programs

NP/SP down :|

hyper-neutrino

yep, we've already restored the feeds

Redwolf Programs

3:14 PM

I think my internet went down

user

Give them some soup and rest if they’re feeling down

Redwolf Programs

Right as I went to sleep I noticed I lost wifi, but I was too tired to investigate

hyper-neutrino

should I run the bots on my server?

other than when the service needs to update, it is pretty reliable

pxeger

they should definitely be run on a server

Redwolf Programs

It's not an issue with wifi reliability. This is the first time this has happened I think.

Just bad timing.

@pxeger They are

pxeger

3:16 PM

so why does it matter that you lost internet at home

Redwolf Programs

Because my server's at my home

hyper-neutrino

right but also you like, sleep, i don't, so if anything bad happens i can fix it (/hj)

pxeger

well I meant run it on a proper server in a datacenter with reliable internet

Redwolf Programs

I have reliable internet I just don't write reliable code lol

Like I said, this is the first time this has happened

In something like 2-3 years

Okay, I think I see the problem

The bots noticed they closed, but never attempted a reconnect

Odd, NP/SP never noticed they disconnected, but the socket to watch New Posts did

Okay, bots are back online and (hopefully) fixed

hyper-neutrino

well in any case, should we maybe leave the bots in your room to test some more? seems like they're not ready for prod yet

Redwolf Programs

3:23 PM

Seemed to be a weird issue with JS's promises

If we've got the feeds running already, I guess I could go ahead and move these back for a little while then move them back to here whenever they're ready

I doubt there'll be any more issues, but I doubted there would be any more issues last night lol

pxeger

try unplugging the internet and plugging it back in again to check they can reconnect properly?

Redwolf Programs

I have internet now of course

It was like a two minute outage

pxeger

but I mean, in case it happens again, to check your reconnecting code works

Redwolf Programs

I know one of the sockets did reconnect properly, so I just copied over its code

pxeger

oh

Redwolf Programs

3:28 PM

I see y'all had a lot of fun trying to get the feeds working again

pxeger

uhuh

hyper-neutrino

eh. i knew it'd happen

any time a room with feeds is unfrozen or a feed is added there is total chaos. as we all know, SE chat works perfectly with no issues or weird things at all

pxeger

important information: a Beans emoji is a candidate to be added in Unicode 14

Redwolf Programs

Exciting

Can't wait until 2022 though

@Neil data:text/html,<textarea> is awful. data:text/html,<html contenteditable> is much better as an entry level text editor.

Then you can add scripts and styles until you regret everything

pxeger

data:text/html,<html contenteditable>this is a surprisingly usable text editor, thanks redwolf

Redwolf Programs

3:34 PM

data:text/html,<html contenteditable><body style="font-family:monospace;font-size:10px;line-height:1.25;white-space:pre-wrap;word-break:break-word">

for a better one

pxeger

data:text/html,<html contenteditable><body style="font-family:monospace;font-size:10px;line-height:1.25;white-space:pre-wrap;word-break:break-word">this is far too small

Redwolf Programs

Then make it 11px

or 12px

You can even use ctrl + b or ctrl + i for bold and italics

Transcript for