The APL Orchard

eyepatch

04:17

I'm struggling to define a lambda to work over n-wise reduce. I'm looking to have a window of 3 elements which reduces over an input array and return the minimum of (left - middle),(right-middle). Something like 3{(⍺-⊃⍵)⌊((⊃⌽⍵)-⊃⍵)}/0,a,0
I can't seem to figure out how to work with n-wise reduce with 3 elements.

dzaima

@eyepatch 3 f/arrOf3Elements is always gonna be equal to f/arr, so i don't quite understand what you want to do

eyepatch

⎕←3,/0 9 6 1 6 2 0
Returns the window, instead of a window, I want to apply a lambda. I suppose I can always apply afterwards.

Dyalog APL

@eyepatch
SYNTAX ERROR

dzaima

@eyepatch yeah, n-wise reduce reduces the window by f, at no point do you see an n-element list unless you specifically make it

⎕←3⊢/1 2 3 4 5 6

Dyalog APL

@dzaima
3 4 5 6

eyepatch

04:32

⎕←3⊣/1 2 3 4 5 6

Dyalog APL

@eyepatch
1 2 3 4

eyepatch

Where did my 2 3 4 5 go?

dzaima

to explain in another way, 3 f/A B C D E is equal to (A f B f C) (B f C f D) (C f D f E)

eyepatch

in any case: {x y z←⍵⋄(x-y)⌊(z-y)}¨3,/0 9 6 1 6 2 0 works, but seems a little unnecessary.

Oh. That's helpful.

I didn't get that before.

dzaima

@eyepatch yeah, i've wanted some built-in for f¨N,/ many times before

eyepatch

04:35

That seems the natural definition of 3 f/A B C D E

Also tremendously useful in the domain.

dzaima

agreed

hmm, i could make n f\ be f¨n,/ in my APL

@eyepatch oh, {x y z←⍵⋄(x-y)⌊(z-y)} ↓⍉↑3,/ A is like 20x faster and more APLy

now i want n f\ to be ↓⍉↑n f/ :|

eyepatch

04:57

Sure. That makes sense.

I was playing around with solving HackerRank problems in APL and then translating to C++.

e.g. gcc.godbolt.org/z/G5sPCa

dzaima

@dzaima actually no, that doesn't make sense, what I want is to have n someSymbol A to be ↓⍉↑ n,/ - no operator

Adám

@eyepatch If you want to access the actual windows, use f⌺3 but beware that the first and last application of f will have "shards" (the first/last 2 elements padded to 3)

eyepatch

I was doing padding myself. :)

Adám

⎕←{⊂⍵}⌺3⊢1 2 3 4 5 6

Dyalog APL

@Adám
┌─────┬─────┬─────┬─────┬─────┬─────┐
│0 1 2│1 2 3│2 3 4│3 4 5│4 5 6│5 6 0│
└─────┴─────┴─────┴─────┴─────┴─────┘

dzaima

05:07

@Adám and also beware that even with padding allowing the following to work just fine, f⌺3 ⊢ 1 2 errors.

Adám

@dzaima That was a bug that has been fixed in 17.1, due very soon.

(I still think that an even shorter ⍵ should be allowed, and that it should just error if trying to access ⍺ in such a case.)

dzaima

@Adám so e.g. ({⊂⍵}⌺3) ,1 would be ⊂0 1 0?

Adám

@dzaima Yes. And {⊂⍵}⌺3⊢⍬ would be 0⍴⊂0 0 0

ngn

@eyepatch that's 2⌊∘-/2-/

dzaima

@ngn slower than my faster one though

Adám

05:15

@Quintec Welcome back! I guess APL is just very mnemonic.

ngn

@dzaima but simpler

eyepatch

Not sure it's quite right since it's the min(x-y,z-y)

dzaima

@ngn i'd agree with "shorter", but imo it's overly complex & less understandable

eyepatch

Or rather, it's supposed to be that, but appears to be min(x-y,y-z) instead

dzaima

@eyepatch the ∘- fixes that

Adám

08:52

@dzaima ∘- kind of looks like an eyepatch ○.∘-

dzaima

13:55

@dzaima and why not an extra 9x faster

J. Sallé

@dzaima using an extra ¨, even!

dzaima

and why not another 40% less :D

J. Sallé

Nice!

Cut them ¨s down

dzaima

and why not a bit more, though now the texting starts to wary a lot

Torsten Grust

14:35

{x y z←⍵⋄(x-y)⌊(z-y)}¨3,/ A is the Haskell version.

dzaima

it's also the slowest.

Adám

@TorstenGrust Just for comparison, how does that look in Haskell?

dzaima

and 3 {x y z←⍵⋄(x-y)⌊(z-y)}\ is the in-dev dzaima/APL version

Torsten Grust

@Adám map (\[x,y,z] -> min (x-y) (y-z)) . chunksOf 3

Adám

@TorstenGrust I see, yes, that looks similar. What does \ do?

Torsten Grust

14:46

`\` is the lambda

Adám

@TorstenGrust haha, markdown is "awesome"

Torsten Grust

Urgh ;-)

Adám

@TorstenGrust \

Torsten Grust

Thanks, I'll make copies of this for future uses.

Adám

@TorstenGrust Uh, can you see my markdown? It is ``\``

Torsten Grust

14:49

Thanks, I should know better. I create all of my lecture material and notes in Markdown. For years now.

In Markdeep, to be precise.

casual-effects.com/markdeep

I can highly recommend taking a look at it.

dzaima

@Adám another SE quirk is that the message history is available in the transcript (quickly openable with a shift-middle-click om the arrow on a message), which easily allows seeing markdown without even abusing the links!

Torsten Grust

This is practically all we use in my group to create written material.

Adám

@dzaima Ah, cool.

Torsten Grust

I also like the comprehension style: [ min (x-y) (y-z) | [x,y,z] <- chunksOf 3 A ]

Adám

@dzaima I don't seem to need Shift. However, I can click>"history" on other peoples msgs, but not on my own.

Torsten Grust

14:56

{ e(x) | x ∊ X, p(x) } ≡ [ e x | x <- X, p x]

dzaima

@Adám for me on chrome the shift immediately focuses the opened tab, whereas just middle-click makes me scroll to it

Adám

@dzaima Ah, for me too on FF.

dzaima

@Adám about the 2nd part, what do you mean? unedited messages don't show history in live chat, and i can see history for every message in the transcript

Adám

@dzaima Oh, you're right. Never mind.

Torsten Grust

15:28

Oh, looks like I'll travel to Denmark in September?!

Adám

@TorstenGrust You won?

Torsten Grust

It says "Phase II – Winning Professional Entrant" on the web page. I haven't heard anything yet, though.

Adám

@TorstenGrust I just asked the supreme judge. It is indeed so. Congratulations. See you there!

Torsten Grust

I am floored, really.

Thanks, @Adam.

J. Sallé

@TorstenGrust ⌊Torsten! Congratulations!

Torsten Grust

15:34

:-)

Adám

@J.Sallé ⌊?

APL Riddle: 1+@3+⌊/¨2⍬⍬

J. Sallé

@Adám He did say he was floor ed

Adám

@J.Sallé Oh, I totally missed that msg.

J. Sallé

@Adám it was a bad joke anyways hahahahahahah

Adám

@J.Sallé Can you solve my riddle?

J. Sallé

15:42

@Adám I don't think so? All I can say is that ⌊/⍬ seems to yield the largest integer?

Adám

@J.Sallé Yes, "infinity".

Torsten Grust

"To infinity and beyond"

Adám

@TorstenGrust Yes. Nice.

J. Sallé

Ah, I see now

That's very clever

Adám

@J.Sallé Wait, what? You only see now, after you figured it out‽

J. Sallé

15:47

@Adám I hadn't figured it out >.>

Delphi is currently consuming most of my ~soul~ attention

Ah I seem to have forgotten how markdown works, too

I blame Delphi

ngn

16:08

@TorstenGrust grats! are you going to publish your solutions?

Adám

@ngn Are you going to publish yours? Btw, you could easily have won if you had not gotten so many minus points for readability and quality of comments ;-)

ngn

@Adám minus?! :P

J. Sallé

@Adám Did he golf the comments as well? heheheheheh

ngn

@Adám well, a losing solution isn't as interesting for publishing, so no

@J.Sallé what is a comment? :)

Adám

@J.Sallé what comments?

J. Sallé

16:17

Comments were 0 bytes then, I guess >.>

Adám

@J.Sallé Do comments byte count include the ⍝s?

J. Sallé

@Adám I don't know; are ⍝s considered comments by the interpreter?

Adám

@J.Sallé It has its own token type:

⎕←(3⌷⍤1⊢201⌶⍬)⌷⍨⊂(2⌷⍤1⊢201⌶⍬)⍳⊃200⌶,⊂'⍝c'

Dyalog APL

@Adám
┌─────────┬────────────┐
│SCOL_LAMP│MINI_COMMENT│
└─────────┴────────────┘

dzaima

@J.Sallé in dzaima/APL comments weren't even a token, but now i've got a boolean to enable extra tokens so that it could be used for syntax highlighting too, in which case ⍝abc is a single token

Adám

16:29

@dzaima Oh, I just asked the interpreter, and it also says ⍝abc is a single token. I guess it is only for syntax colouring purposes that it distinguishes the lamp and the text. CC:@J.Sallé Fair enough, I think.

J. Sallé

@Adám Then I think lamps should be counted towards comment byte counts

Adám

@J.Sallé In either case, ngn's comment byte count is 0.

J. Sallé

Hah

ngn

@Adám 0 comments, all of excellent quality :)

Adám

@ngn At least you had no typos in your comments and no descriptions of the code were wrong.

dzaima

23:02

finally decided that dzaima/APL should be pulled so fancy stuff yay

Adám

@dzaima I think I get all but the last one. Dyadic ⍎ is "apply"? And ` arrayifies a function?

dzaima

May 24 at 9:03, by Adám

@Ven @dzaima Wouldn't it be neater to have `f mean the arrayified f and extend ⍎ to allow an arrayified function as left argument (and in that case allow any right argument) so you can say (2⊃`f`g`h)⍎Y to do g Y

Adám

@dzaima Oh wow, that's a while ago.

dzaima

was indeed. i had completely forgotten about it until i looked at the commit history :p

Adám

What am I going to do about APLcart when 18.0 comes out? I'll need to revise everything to consider using ⍥Over, ⍤Atop, and ⍨Constant…

@dzaima ⍩ is not a synonym of ᐵ.

dzaima

23:10

@Adám yeah, i'm lazy

@Adám huh, didn't know about A⍨, but it does seem pretty useful in trains

@dzaima introduces problems in SBCS too, as i don't want to waste a char for having 2 equal things, but that'd mean needing to complicate my encoding

Adám

@dzaima But you can handle Unicode, so only one edition needs to be in the SBCS.

@dzaima Absolutely. ⌈/*2⍨ is much easier on the eye than 2*⍨⌈/

@dzaima Oh, and all three are confirmed now

dzaima

@Adám this is my current encoder, and that doesn't support multiple chars mapping to the same thing

Adám

@dzaima So? Just don't handle one of them. Or maybe I'm not understanding the issue. Are you using SBCS internally?

dzaima

@Adám does that mean that ⍢ is gonna have to wait?

Adám

@dzaima Yes, ⍢ and ⍫ are probably for 19.0 as they have rather large internal implications.

dzaima

23:16

@Adám i could just make it error on encoding ⍩, but that's just not nice

Adám

@dzaima But hey, maybe if we do three at a time, we can get ⍛ too. It makes tacit so much nicer too.

Maybe when we add ⍥k for Depth, we'll go all the way and add A¨ and A⍤k and A⍥k

I even think A⍢ should be allowed, so you can do A⍢⊃ instead of A@1

H.PWiz

I think having A¨ over A⍨¨ isn't worth it. Although A¨ probably coudn't stand for much else

Adám

@H.PWiz For me, it is mostly a matter of consistency. I like the idea of arrays acting like functions that return themselves.

I intrigues me.

dzaima

@H.PWiz stuff like 0¨ instead of 0⍴⍨⍴ is cool though, and 0⍨¨ is way less obvious

Adám

Not to mention that if you have to write 0⍨¨ you might almost just write 0⊣¨ now.

Transcript for