The APL Orchard

1:02 AM

what's the shortcut in )ed to align the comments? (if there is one)

4 hours later…

5:28 AM

@rak1507 I don't think there is one by default. Configure the keyboard shortcut for action <AC>:

6:26 AM

I have two arrays with the same dimensions, call them A and B, and a third boolean array, C. I want to create an output array, where the values from come from either A or B depending on whether C is true of false. So for example, if A←1 2 3 4 ⋄ B←11 22 33 44 ⋄ C←0 0 1 1, I want the result to be 1 2 33 44`.

What is the nicest way to do this?

@EliasMårtenson The first thing that comes to mind is (A,[0.5]B)[↓(1+C),⍪⍳≢C]

      A←1 2 3 4 ⋄ B←11 22 33 44 ⋄ C←0 0 1 1 ⋄ (A,[0.5]B)[↓(1+C),⍪⍳≢C]
1 2 33 44

I did this for a one-dimensional array: 0 0⍉(A,[0.5]B)[;C]

haven't played around with dyadic transpose yet

The problem is that I need a generic solution.

@Razetime It's the first time I used it. :-)

generic, meaning?

6:35 AM

That said, KAP doesn't support the 0 0 form, so I still need to find a better way.

@Razetime One that works regardless of dimensionality of the arrays.

That said, I guess I could do: (⍴A) ⍴ 0 0⍉((,A),[0.5],B)[;,C]

@EliasMårtenson (C/⍥,B)@{C}A

OK, that exposed me to @ and ⍥. Please allow me to read up on them first.

Yet another need for Under (⍢): B⊣⍢(C/⍥,⊢)A Try it online!

@Adám aaaa so smart

This goes into APLcart now.

6:48 AM

yes please

@Adám I can't find documentation for ⍢ in the dyalog documentation.

Because it hasn't been implemented yet.

I see. :-)

Is there a writeup on it somewhere?

⍢ is basically a generalisation of the @f and @A

@EliasMårtenson BQN

If I had been quicker, I'd have protested adding @ back in 16.0 and instead argued that we should add the more general form.

@EliasMårtenson dzaima/APL :)

7:01 AM

@EliasMårtenson Btw, GNU APL has an ad-hoc syntax for this, namely A⊢[C]B but I do not recommend adding such obscenities to the language.

@dzaima Huh, I didn't know about that.

@EliasMårtenson C⊃¨A,⍥⊂¨B works too, of course, but super-inefficient.

C⌷⍤0 1⊢A,⍤0⊢B is better.

@Adám about % at all or that it works on high rank arrays? (i've mentioned % somewhat recently, but i had no clue if i had actually bothered to implement it for high rank)

at all

7:20 AM

@Adám Actually, I considered adding something like that (as a library call, not a language extension)

So, say I have an array A, and it's got elements 1 2 3 4 5 6

and I have a function f

I want to get 1,(1 f 2),2,(2 f 3),3,(3 f 4),4,(4 f 5) and so on

@Adám Actually, this one would be efficient in KAP, if I understand the code correctly.

@dzaima it'd be a good idea to remove a ton of NYIs in the repo

some dzaima/APL perf comparison

@Razetime ⊃,/2f{⍺(⊣,⍺⍺,⊢)⍵}/A

7:27 AM

side note: this isn't for golfing

@Razetime s/NYI/NGBI/g (not gonna be implemented)? :P

@dzaima oh come on there's so many things implemented that are marked NYI

like ? is marked NYI

oh you mean in docs/chars.txt

yes indeed

yeah that's just plain outdated all around

7:29 AM

that's what i generally refer to when trying something

ok i'll try to update that

@dzaima And in Dyalog.

@Adám ⊃,/2(⊣,f,⊢)/A simpler?

@Razetime Yes, of course. D'oh.

@Adám i'm quite happy with C⊃¨A,⍥⊂¨B being 2x faster in dzaima/APL

(as dzaima/APL would be, and of course is slower for everything else)

7:34 AM

@Adám this gives 1,(1 f 2),2,2... , I don't want repeated elements

@dzaima (that's for 10000 item arrays. for larger ones they're more similar, but dzaima/APL still can be like 30% faster)

@Razetime ⊃,/2(⊣,f)/A then.

hm nice

@Adám but this eats the last element

7:51 AM

@Razetime Sure, but trivial to append.

aghh

No, that's an invalid train ;-)

not in dzaima/APL :)

8:21 AM

@dzaima pushed

2 hours later…

user41805

10:06 AM

@Adalynn welcome back. yeah for sure you didn't miss out on any changes to se chat

10:46 AM

What's a good way to zero the lower triangular of a (square, boolean) matrix?

you could do a boolean and with ∘.<⍨⍳≢m

RGS

@xpqz Just remember that ∘.f with f one of < ≤ ≥ > is your friend when you want ... triangular matrix ....

@xpqz aplcart.info?q=extract%20upper ― all add some synonyms.

RGS

(∘.<⍨∘⍳∘≢×⊢)

{⍵∧∘.<⍨⍳≢⍵} more readable imo

10:53 AM

@RGS beautiful

@rak1507 That only works on Booleans, though. (But I assume you meant to prefer a dfn.)

@Adám he specified that it was a boolean matrix

RGS

Coding with generality is always (?) a good thing.

Sure, just that the APLcart entry (and RGS's equivalent) is more general.

fair enough

10:57 AM

@Adám I did search dilligently on APL cart, but in a glass-half-empty way: zero triangular etc. But I think it would be nice if this showed on just 'triangular' as every other operation on triangulars seem to do.

@xpqz It does show on "triangular":

So it does; obscured from my vision by not thinking of it as "extracting the upper" but "wiping the lower" :?

My problem, not APLCart's.

I've added "zero out lower" now.

4 hours later…

2:48 PM

Ok, more array gymnastics: given a boolean rank 2 matrix A and Ys Xs, zero all elements on row Ys and col Xs in A, except A[Ys;Xs], and leaving all other elements as is. So given:

3 3⍴1 0 1 0 0 1 1 1 1

and 2 2, we should get

3 3⍴1 0 0 0 0 0 0 0 1

(⎕IO←0)

A chess rook, basically

{x y←⍵ ⋄ m←⍺ ⋄ (2⌷m)←0 ⋄ (2⌷[2]m)←0 ⋄ (⍵⌷m)←⍵⌷⍺ ⋄ m} horrible solution

{⍺∧1≠+/⍵=⍤1↑∘.,⍨⍳≢⍺}

@rak1507 You're on fire today. Nice.

having exams soon magically makes me productive in everything that isn't doing revision

4 hours later…

7:30 PM

fun

oh this was you

oh, I've discovered that before

well, this is the reverse

I've added a note about this to the issue report. We'll see what happens.

7:54 PM

      r ← ⍤0 1
      +r⍨ 1 2 3
2 3 4
3 4 5
4 5 6
      r
┌─┬───┐
│⍤│0 1│
└─┴───┘
      +r⍨ 1 2 3
0 0 0
0 0 0
0 0 0

now this was unexpected

Same kind of thing, I suspect.

I assumed it would error but it's interesting that it doesn't

      r←⍤0 1
      +r⍨2 3
4 5
5 6
      +r⍨4
8
      r
  0 1
┌─┘
⍤
      +r⍨2 3
0 0
0 0
      +r⍨4
VALUE ERROR
      +r⍨4
       ∧

weird

Again, it is the display that does it:

      ]box off
Was ON
      r←⍤0 1
      +r⍨2 3
4 5
5 6
      +r⍨4
8
      r
⍤ 0 1
      +r⍨2 3
4 5
5 6
      +r⍨4
8

8:02 PM

bc why wouldn't outputting have side effects lol

Dec 10 '20 at 11:40, by Adám

@rak1507 It is bug 17228 in the interpreter: When the interpreter needs to call a callback function to print something, it should create a ⎕ORed clone of the item to print and pass that, but instead it does an in-place ⎕ORing and so a side effect of printing with a callback is that the item that was printed changes from being a function to be a ⎕OR.