@ngn btw, are you ready to make ngn/k's sources available?

because, well, being able to try something offline isn't bad :-)

@EriktheOutgolfer Can't you try it even if it is closed-source?

@Adám it's not even that currently

it's just a README file

ngn has only given the sources to TIO, and now only the executable is there (he also asked Dennis to delete the sources after pulling)...hm, what if I download that?

it's not like TIO itself is closed-source

@EriktheOutgolfer Exactly.

@Adám the problem I see is that I'm using Windows :-)

@EriktheOutgolfer I bet it is Linux-only anyway.

Ubuntu is currently on hiatus until June 26th, so I can upgrade it to 18.04.1 by then :P

@Adám I don't think that sources are Linux-only...

hm, I might be able to convert an ELF to an EXE, only if I find such a tool online because I won't bother doing it by hand

@Adám I can't find the /opt directory here :(

@EriktheOutgolfer It doesn't work like that.

@Adám yeah I suspected so, tried both pinned repos

@EriktheOutgolfer Here is part 1 and part 2 of the ngn/k executable in hex.

@Adám hm, let's see if the halves are correct...

(not that I don't trust you :P)

@Adám um, what is the output format there?

@EriktheOutgolfer Hex. Every pair of two characters forms a byte.

@Adám what do the whitespace chars represent?

@EriktheOutgolfer Nothing, just a quirk of the TIO wrapper.

@EriktheOutgolfer I don't have a plan really

my code is written in what most people here would consider a very bad style

@ngn ngn/c

@ngn doesn't mean it can't be compiled ;)

@Adám nah, it's "ngn style" :P

@Adám just plain c

@EriktheOutgolfer it's only linux and freebsd, 64-bit

(to be honest, I haven't tested on freebsd lately - I should)

@ngn There is no license listed at gitlab.com/n9n/k. What is the copyright status?

@Adám there's nothing there, it's just a homepage so Dennis has something to link to

@ngn I know, but I'm asking if we're allowed to copy and redistribute the executable.

@Adám haha :) at least have the decency to use k

@ngn hey, Adám might not know/be comfortable to use k...

@ngn Why don't 4.5E4#1:"/opt/k-ngn/k" and 4.5E4_1:"/opt/k-ngn/k" work?

@ngn Ah, 45000 works. No exponential format in K?

@Adám technically no - you can use it only "for a quick test and for golfing purposes" as mentioned in my homepage, but of course I wouldn't mind giving access to people who show genuine interest in k

@Adám yes, 64-bit ints

@ngn And you still don't have floats, do you?

I'm happy to keep chatting, but shouldn't we move this conversation to the orchard?

@ngn ↰ I guess, but it isn't in my power to do so.

@ngn oh, almighty Dennis, it looks like a little off-topic conversation sprouted in here, but we don't have the power you have, so can you please do us that one little favor of moving messages, to the great APL orchard?

@Adám you can't reply to another chat room...

@EriktheOutgolfer you didn't "@" him

sure, but "almighty @Dennis" sounds weird (<-- @ in this message)

@Adám regarding floats: I started adding some support for them but I hit a wall with printing a float

I'm not smart enough to do it properly

also, in an unrelated on-topic note, I finally got the ELF file, yay

@ngn Maybe print the fractional part after integer part, separated by a period. Just thinking.

@EriktheOutgolfer Uh oh, ngn will come after you. Was it really that hard to recombine and translate the hex to binary file?

@EriktheOutgolfer for what

also yeah that^^

@ASCII-only ngn/k

@Adám I came home from yet another day @ finals :P

@Adám no, wrong target! he should come after Dennis if he dares ;)

@EriktheOutgolfer my team of laywers will be watching you - the moment you start using ngn/k for anything but quick testing and golfing, a swat team will break in and arrest you :)

@ngn do you really think I want to reverse-engineer the ELF?

cat /opt/k-ngn/k | xxd | head -n 1792
# cat /opt/k-ngn/k | xxd | tail -n +1793 | head -n 1792
# cat /opt/k-ngn/k | xxd | tail -n +3585

you know I'm not really known for intentionally violating licenses

@EriktheOutgolfer not really, it was a joke

@EriktheOutgolfer yes

@ngn obviously

@EriktheOutgolfer Something like:

in←(⎕NEW⊂'Clipboard').Text
⍞←'Copy next part and press Enter'
in,←(⎕NEW⊂'Clipboard').Text⊣⍞
⎕IO←0
(⎕UCS hex⍳(part1,part2)∩hex←16↑⎕D,⎕A) ⎕NAPPEND 'k' ⎕NCREATE 0

@Adám it's not as easy as it sounds

@ngn Maybe not, but I suspect others have done it before you.

@Adám wow why would I do that when I can just click ngn's link (or write up my own code, it's not that hard), remove the 0x part, and then xxd -r the hex?

@Adám I've looked at some ftoa() impls and they are rather complicated

I'd like to achieve two things: first - simplicity

second - I'd like the output for a float to contain enough information for the value to be reconstructed exactly

@EriktheOutgolfer Well, you do need transfer via the clipboard to a file, no?

@ngn As in bit-wise exactly?

@Adám just Ctrl+A, Ctrl+C, Ctrl+V :P

it's not really that large

@EriktheOutgolfer Of course. I was just showing how you can automate it and give it a nice "interface".

@Adám yes, if you take the output as a string and execute it, you should get back the value you started with

@ngn That may be impossible. (Without special notation to denote the binary remainder)

no fuzzy nonsense like ⎕pp or ⎕ct

@Adám not impossible, just hard

with every digit you print you get closer to the actual value on the real line, so after enough digits it becomes unambiguous

@ngn Well, actually, every binary number has an exactly decimal representation, no?

@Adám yes, possibly very long

@Adám Which parts of the conversation should I move?

@Dennis I think everything from chat.stackexchange.com/transcript/message/45004407 to chat.stackexchange.com/transcript/message/45005293 relates to K/APL more than to TIO.

← 67 messages moved from talk.tryitonline.net

@Dennis chat.stackexchange.com/transcript/message/45004984 has its context here now.

mini-poll: if 2 2 2\6 is 1 1 0 (the last 3 digits of the binary representation of 6), what would you expect 2 2 2\6 7 5 to be?

⎕←2 2 2⊤6 7 5

@Cowsquack
1 1 1
1 1 0
0 1 1

ngn, ^

@Cowsquack so, the columns are the binary representations

I think it would be better if it was transposed?

⎕←⍉2 2 2⊤6 7 5

@J.Sallé
1 1 0
1 1 1
1 0 1

Seems more natural to me

actually nvm, I thought it did it the other way around, yeah transposed is what I would prefer

J also does it the transposed way

   2 2 2#:6 7 5
1 1 0
1 1 1
1 0 1

yup

that makes more sense to me too

do you allow non-ascii variable names in your k?

@Cowsquack no

@Cowsquack though I'm ready to stray off the path if there are good reasons :)

related question: if 2 2 2\5 are the last 3 binary digits, (,2)\5 is the last binary digit, what would you expect of 2\5?

for example, I would reeealy like J much more if it did ^^^

@ngn the same as (,2)\5

because right now, I always have to refer to J's vocabulary for finding out which symbol does which

@Cowsquack utf8 decoding costs some complexity and some performance, but there's no reason why ascii-compatible code pages shouldn't be allowed

@ngn probably the whole binary representation of the right argument?

@J.Sallé that's the answer I was hoping for :)

Because as of now you'd have to know how many digits the binary representation has, right?

it comes up too often in golfing

yup

@J.Sallé right

@ngn Yeah, I'm with J.Sallé, extend the scalar as necessary.

@ngn what is the difference between passing in non-ascii chars and passing in multiple ascii chars?

@Cowsquack ascii is 7-bit (0-127), non-ascii chars need to be encoded somehow, usually as utf8 which uses multiple bytes with a set 8-th bit (codepoints 128-255) to represent a single char

@Cowsquack the simplest thing I could do to support unicode is to treat any char beyond U+007F as a letter - then I wouldn't need to do any utf8 decoding

of course, unicode has many non-letters and that may be confusing...

@ngn so this avoids the costs to complexity and performance?

@Cowsquack yeah, k is supposed to be simple and fast

@ngn what do you mean?

@Cowsquack Unicode contains many "characters" that aren't "letters".

for example in J, I would be able to do ⊂←<, but ⊂ is not a letter

@Cowsquack example: "LΛ" is two codepoints (U+004C,U+039B) but three bytes (0x4cce9b); in order to determine if the second character is really a letter, I would have to decode the bytes

but if I treat any byte >128 as a letter, I wouldn't need to decode - from k's point of view LΛ would look like a 3-letters variable name

ah okay

.oO( Just pretend you're java and embrace UTF-16 )

@Ven k wasting 2 bytes where 1 would do? no way :)

well, really, you're only wasting one of those 2 bytes :P

@Ven rephrase: "k wasting 1 byte where 0 would do?" :)

@Cowsquack well, I suppose you want to treat ⊂⊂ as two separate identifiers

but that's hard - it requires decoding and knowledge about where the letters are in unicode

yes... okay, I think I see what you are getting at now

@ngn No, we really need to look at whether a character is a letter or not, and only letters should combine with neighbours.

@Adám that's what I did in ngn/apl long ago, but I'm reluctant to do in ngn/k

yet another "encode" question: what should 2\7 8 return?

@ngn You mean [[1,1,1],[1,0,0,0]] or [[1,1,1],[1,0,0,0]]?

as 2\ means "as many digits as necessary", 2\7 is 1 1 1 and 2\8 is 1 0 0 0 - they have different lengths, so I could return (1 1 1;1 0 0 0) or pad with leading zeros: (0 1 1 1;1 0 0 0)

@ngn Backslash + Backtick = Hell

@ngn I suppose it is easier to pad until rectangular than to remove leading 0s, so I say ragged.

@Adám it's not so easy to pad

@ngn ↓↑ ;-)

@Adám that would pad with trailing 0s

k doesn't have mix (↑)

@ngn I think adding leading 0s until rectangular is the "standard"?

@J.Sallé Yeah

@ngn No, that would pad with trailing. you need ⌽↓↑⌽

But if making a ragged array would be easier and wouldn't affect whoever's doin' the golfs, go for it.

@Adám still wrong but doesn't matter, I see what you mean

@ngn Why is that wrong? Reverse along last axis, mind you!

@Adám you need each

⌽↓⌽ is a fairly weird smiley.

@J.Sallé I think I'll go for rectangular to avoid nasty surprises, even if it's slightly less efficient

k heresy :)

@ngn sounds good

@Ven looks kinda creepy

@Ven cats have that sort of eyes

⌽↓⌽, not so creepy

(and not so, eh, smiley)