The APL Orchard

rak1507

00:20

@LdBeth I wonder what it's for

LdBeth

01:34

@rak1507 It's something you'd have on a lisp machine, parsing various kinds of time/date text formats

rabbitgrowth

07:01

Trying to make {⍸2=+⌿0=∘.|⍨⍳⍵} tacit. Managed to turn it into ⍸2=(+⌿0=∘.|⍨∘⍳), and then ⍸2=∘(+⌿)0=∘.|⍨∘⍳. But is there a way to get rid of the parens altogether?

Fawn Locke

No, I don't think so

You can use an old golfer's trick

⍸2=1⊥0=∘.|⍨∘⍳

1⊥ is summation here and continues the forks :)

rabbitgrowth

Ah lovely, thank you

I was assuming it was always possible to write a train parens-free

"Decode in base 1" - that's quite clever :)

Fawn Locke

07:21

Or ⍸2=0+.=∘.|⍨⍤⍳

∘.|⍨⍤⍳ and ∘.|⍨∘⍳ are identical semantically (when monadic), but i find the former nicer to think about.

rabbitgrowth

What do you mean by "nicer to think about"? Aren't they exactly the same in the monadic case?

I've been going to ∘ by default because it looks simpler visually (which is probably not the best reason to choose something).

I'm not sure if I'm thinking about this in the proper way, but f⍤g feels like it's doing something simpler in the dyadic case: apply g to the arguments, then apply f to the result of that.

Whereas if you were to describe f∘g in words, it would be more of a mouthful.

Fawn Locke

07:41

Back, ⍤ is "atop" in my brain, f (g X). So using ∘ for it feels wrong, I prefer to think about it as a lapse of judgement on Dyalog's part

∘ is a nicer symbol for atop anyway, but It'll have nice symmetry with before once that's added (⍛)

Well Dyadically they are different, f (X g Y) vs X f (g Y)

The former being ⍤ and more atop-y, hence the association

Dyalog should've made monadic ∘ Y f (g Y) imo

rabbitgrowth

Would monadic ⍥ be (g Y) f g Y then?

Fawn Locke

Probably not, I haven't thought about what it should be. I think ⍥ is atop monadically in BQN too - which feels like a mistake

But there isn't much it can be

I very rarely see f⍥g⍨ Y

rabbitgrowth

08:02

Anyway, I think I'm starting to grok tacit. Sometimes I can immediately recognize that I want a ⍥, etc, although it can still take me some time to figure out where to put what exactly.

where to put what? what to put where?

Anyway, yeah. The progress feels nice.

Fawn Locke

Yes, I forgot to say that. You've picked it up extremely quickly

rabbitgrowth

I've been learning on and off for more than a year actually. This is what got me interested: reddit.com/r/adventofcode/comments/k8zdx3/…

Thought it was garbled text at first :)

When I found out it wasn't, I was immediately hooked.

(I think it was mostly because of the ⎕s.)

Fawn Locke

Oh, I see. Learning tacit is unfortunate, you can't use it casually because you do not understand it, and you don't understand it because you can't use it, once you get over the initial curve the growth of learning becomes exponential

rabbitgrowth

Think that's true of many other things too.

Fawn Locke

Yeah, I think this is APL's only feature that's like that though

rabbitgrowth

08:17

It felt like a bunch of arbitrary rules at first, and seeing examples like 5(+,-)2 in the APL Wiki helped - you can read it as "plus and minus". That answered the "why" for me.

I should add that ]box on -trains=tree has been extremely helpful. Also it looks pretty ;)

Fawn Locke

Yeah, I like tacit because it expresses some ideas very naturally. And reveals patterns in your code

rabbitgrowth

I don't think I would otherwise have thought of programming as building a tree of functions. It really is beautiful.

Richard

08:51

@ZtA0 using )COPY you can append two workspaces. Or maybe I misunderstood your question

Brian BED

09:28

@PyGamer0 OOooo one more thing. i want support for quaternions :D

PyGamer0

@BrianBED uh would that be useful

might as well add symbolic algebra to APL lol

there must be a SymAPL

Brian BED

09:48

right now i really want to add a way to get my quardinates in my game to be imaginary but imaginary numbers with 3 axies don't even have neat properties so i guess i'll settle for trigonometry :/

i do wanna make my quardinates for my 2d version be imaginary tho. that'd be cool

PyGamer0

i should actually try making something in APL for once

Brian BED

yee its super fun. i don't wanna go back to python :)
if i ever try a new lang im prob gonna try rust but honestly if i can get good at doing stuff in apl then i'd rather do it apl

PyGamer0

@PyGamer0 the question of course; is WHAT should i make lol

Brian BED

@PyGamer0 i started with making some of my old python starter projects in apl, and then i just let my imagine carry me beyond!

xpqz

10:11

Here's a good one for APL: make a sliding tiles puzzle solver (en.wikipedia.org/wiki/Sliding_puzzle)

PyGamer0

hm i dont think my 2 brain cells can solve that lol

xpqz

It looks easy, but isn't.

PyGamer0

0 2 1
3 4 7
6 8 5

ill say ^ is the puzzle

where 0 denotes the empty piece

xpqz

      state←¯1+4 4⍴?⍨16
      state
┌→──────────┐
↓ 0 15  5  8│
│14 11  2  7│
│ 4  3  1  6│
│ 9 12 13 10│
└~──────────┘

      moves state
┌→────────────────────────────┐
│ ┌→──────────┐ ┌→──────────┐ │
│ ↓15  0  5  8│ ↓14 15  5  8│ │
│ │14 11  2  7│ │ 0 11  2  7│ │
│ │ 4  3  1  6│ │ 4  3  1  6│ │
│ │ 9 12 13 10│ │ 9 12 13 10│ │
│ └~──────────┘ └~──────────┘ │
└∊────────────────────────────┘

PyGamer0

1 2 3
4 5 6
7 8 0

^ and that is the solved version

@xpqz ok so does that output all possible moves?

xpqz

10:23

Yeah.

That can be done nicely in APL.

Solving it though...

PyGamer0

@xpqz my brain has stopped working already

Kamila Szewczyk

why?

it's just manhattan distance of permutations.

PyGamer0

manhattan distance?

xpqz

That's a decent stab that will get to to 4x4.

Taxicab geometry

A taxicab geometry is a form of geometry in which the usual distance function or metric of Euclidean geometry is replaced by a new metric in which the distance between two points is the sum of the absolute differences of their Cartesian coordinates. The taxicab metric is also known as rectilinear distance, L1 distance, L1 distance or ℓ 1 {\displaystyle \ell _{1}} norm (see Lp space), snake distance, city block distance, Manhattan distance or Manhattan length, with corresponding variations in the name...

Brian BED

just ask for input wasd and then find where 0 is with 0⍳board and then if the input is "a"
then flip 0 and the number on the left... well in my head sounds easy but idk. you also need to catch errors because of like... if you use "d" to move the right of 0 youd get an error

oh wait... ahhh a solver for it

xpqz

10:28

Was thinking more of an automated solver giving you the sequence of moves.

Brian BED

right that sounds harder

PyGamer0

how do i generate all permutations of an array?

Kamila Szewczyk

many possibilities!

the most convenient one is dfns.pmat

xpqz

For 5x5 manhattan distance isn't sufficient.

PyGamer0

how do i use dfns library on tryAPL.org?

rak1507

10:35

@xpqz here's my bad attempt {⍵~⍨⍥⊆∪{⊖⍉⍵}\{0::⍵⋄⌽@((0 0)(0 1)+⍸0=⍵)⊢⍵}¨{⌽⍉⍵}\4⍴⊂⍵}

Fawn Locke

Only some are available, the workspace on that site is only 512kbs which won't be sufficient for that amount of permutations, anyway

rak1507

@PyGamer0 you don't

xpqz

@rak1507 My ungolfed mess was:

moves←{
     d←(0 1)(1 0)(0 ¯1)(¯1 0)      ⍝ Move offsets
     m←⍵
     {⌽@i ⍵⊢m}¨(,⍳⍴m)∩d+⊂i←⊃⍸0=m
 }

PyGamer0

let me try making a permutations generator first

rak1507

@xpqz yeah (,⍳⍴m)∩ nice

Kamila Szewczyk

10:37

surprised nobody used stencil

rak1507

I tried but it wasn't nice

Brian BED

i wonder if it'd be neater for ○ to mean tou instead of pi. like a full sircle radius is ○2 with a radius of 1, and ○ is a full sircle not a half one... ¯\_(ツ)_/¯

Kamila Szewczyk

pi is useful in a lot of formulas

i guess we'd be writing ○.5 instead

PyGamer0

⎕IO←0
Permute←{⍵[?⍨≢⍵]}
Permutations←{x←⊂⍵⋄x{a←Permute⍵⋄a∊⍺:⍺∇⍵⋄⍺,⊂a}←⍵⋄x}

^ i really dont know what i was thinking here

Brian BED

12:04

{¯3○÷/⍵}1 1
this is a way to get the radians of a point with an x and y... but i have no idea how to avoid the test case 1 0 where it gets a "cant devide by 0" error. i cant find any way of fixing it except something like this (⎕io is 0): `{0=1⊃⍵:¯3○÷/1E¯12,⍨⊃⍵ ⋄ ¯3○÷/⍵}1 0` but it's incredibly ugly

Kamila Szewczyk

what you really want is a two argument arctangent.

sometimes called atan2

Brian BED

yea thats what {¯3○÷/⍵} is

i looked here for the formula math.stackexchange.com/questions/1201337/…

Kamila Szewczyk

that's not what atan2 is.

because it fails for one of the defined cases

you could use the inner product formula the author has showed too.

Brian BED

oh so how do i do atan2?

Kamila Szewczyk

APL doesn't have it as a built-in. you must special case it according to the following definition:

Brian BED

12:17

aw

well thanks

Kamila Szewczyk

you could use vectors though

Brian BED

yea havent learned that yet

Kamila Szewczyk

using the inner product formula

Brian BED

ok wait ill google a bit. ill prob figure it out

Kamila Szewczyk

it's given in the post you linked.

Brian BED

12:18

i did also think about using 12○xJy

which works ¯\_(ツ)_/¯

Kamila Szewczyk

that's a smart idea

Brian BED

:D

Kamila Szewczyk

phase of the complex number is defined in terms of arctangent

Brian BED

yep

{12○+/1 0J1×⍵}1 1
tada!!

Kamila Szewczyk

:)

⋄{¯9 ¯11+.○⍵}3 3

TryAPL

12:22

@KamilaSzewczyk 3J3

Brian BED

u sure that makes sense?
{12○+/1 0J1×⍵}3 3
0.7853981634

{12○1 0J1+.×⍵}3 3
this works too

Kamila Szewczyk

⋄57.2958×12○3J3

TryAPL

@KamilaSzewczyk 45.00001609

Kamila Szewczyk

seems to be correct

Brian BED

13:53

dtr←○180÷⍨⊢
this is a way to get degrees to radians...
now i reeaaalllyyy want a radians to degrees again and i wanted to do `dtr⍣¯1` but if the input is 0 it gets a devide by 0 error...
i tried `dtr←○0.005555555556×⊢` and the 0.0055 is just ÷180, and i was expecting it wouldn't get a devide by 0 because it doesn't devide but it still gets the same error. any smart idea how to fix this or should i just make a seprate function "rtd" which has the exception? again... reallyyyyy inelegent :(

dzaima

@BrianBED huh, that's weird

dzaima/APL handles it just fine

it's (○⍣¯1) 0 that fails

Fawn Locke

Side note: dtr←○÷∘180

dzaima

you can do dtr←(180÷⍨○1)×⊢ but that's way less cool

xpqz

@dzaima dyalog bug?

dzaima

probably

Brian BED

14:02

@dzaima thanks so much. its way better than my exception idea

Brian BED

16:39

https://imgur.com/K26AhCp
have you ever pressed alt+p in explorer? pretty neat :)

it works for pdfs, word documents, and lots more :o

LdBeth

19:19

I usually do “type” in terminal :D

LdBeth

19:55

⋄ -/0 ⍴1

TryAPL

@LdBeth 0

LdBeth

well it is surprising, because subtraction has no identity

It is probably because -n equal to 0-n

Transcript for