The APL Orchard

AviFS

2:06 AM

Is there a way to map across ⍺, rather than ⍵?

For instance, 1 3 2↑¨⍳5 would give:

1 1 2 3 1 2

The closest I can come in this particular case is 1 3 2↑¨(⍳5)(⍳5)(⍳5), but perhaps one can do better?

dzaima

1 3 2↑¨⊂⍳5

¨ maps across both arguments always, so if you want one to be constant, just ⊂ it

AviFS

The best I have rn is 1 2 3{⍺↑¨(≢⍺)⍴⊂⍵}⍳3

Oh wait, sorry dzaima, let me read

@dzaima Woah, no way, thanks!

This isn't even the slightest bit familiar

Is it time to reread the Legrand book?

I'm not sure I totally get it, though. Mind elaborating on how the constant bit works

hyper-neutrino

so let's break down the individual components

AviFS

Thanks, please

hyper-neutrino

so the right side ⍳5 you obviously get what that means

that returns an array, which is a vector (1-dimensional) of dimensions 5

⎕←⍳5

TryAPL

2:15 AM

@hyper-neutrino 1 2 3 4 5

AviFS

Wait! We do have a bot working?!

hyper-neutrino

if you then enclose it with ⊂, it packs the array into a single-cell array

so now you have a single-cell array containing a vector

⎕←⊂⍳5

TryAPL

@hyper-neutrino
┌─────────┐
│1 2 3 4 5│
└─────────┘

AviFS

Right

hyper-neutrino

@AviFS yep, as of a few days ago :)

AviFS

2:15 AM

The constant bit is new to me, though

@hyper-neutrino Awesome!!

hyper-neutrino

when dzaima says constant they just mean if you want one side to not loop

so you're looping on both sides between 1 3 2 and 1 2 3 4 5 initially

AviFS

Are you really?

hyper-neutrino

but now you're looping between both sides of 1 3 2 and ⊂ 1 2 3 4 5

yes, by default it loops through both sides

AviFS

That makes sense in principle

I hadn't realized it mapped over both though

hyper-neutrino

⎕←1 2 3{⊂⍺,⍵}¨1 2 3 4 5

AviFS

2:16 AM

Trying to make sense of that

TryAPL

@hyper-neutrino
LENGTH ERROR
      ⎕←1 2 3{⊂⍺,⍵}¨1 2 3 4 5
                ∧

AviFS

It always looked to me like it was mapping only over ⍵

hyper-neutrino

⎕←5 4 3 2 1{⊂⍺,⍵}¨1 2 3 4 5

TryAPL

@hyper-neutrino
┌─────┬─────┬─────┬─────┬─────┐
│┌───┐│┌───┐│┌───┐│┌───┐│┌───┐│
││5 1│││4 2│││3 3│││2 4│││1 5││
│└───┘│└───┘│└───┘│└───┘│└───┘│
└─────┴─────┴─────┴─────┴─────┘

hyper-neutrino

@AviFS right - most things you've worked with probably had a single-cell or scalar ⍺

AviFS

2:17 AM

@hyper-neutrino hmmm, that's true. super interesting

@hyper-neutrino ah, that makes a lot of sense

thanks for explaining that part

hyper-neutrino

glad it makes sense :D

AviFS

Gosh, I'd never realized

This makes it easy to see for me, too

⎕←1 2 3↑¨3 2 1

TryAPL

@AviFS
┌─┬───┬─────┐
│3│2 0│1 0 0│
└─┴───┴─────┘

hyper-neutrino

take what i say with a grain of salt tho - i'm pretty new to APL so stuff i say may be totally wrong and just happen to work out in specific situations

AviFS

no, it's super helpful and makes a lot of sense

Question then:

dzaima

2:20 AM

@hyper-neutrino you're on point :)

hyper-neutrino

woo \o/

AviFS

Can you only map then if ⍵=1, ⍺=1 or ⍺=⍵

dzaima

@hyper-neutrino (except that that's a bit excessive and could just be ⎕←5 4 3 2 1{⍺ ⍵}¨1 2 3 4 5)

TryAPL

@dzaima
┌───┬───┬───┬───┬───┐
│5 1│4 2│3 3│2 4│1 5│
└───┴───┴───┴───┴───┘

hyper-neutrino

oh

yeah that

AviFS

2:21 AM

haha, i did notice that

@AviFS My sense is there are some that work so long as one is a multiple of another

dzaima

(how does the wiki page of scalar extension not mention ⊂ ಠ_ಠ)

AviFS

And some might work even if not multiple. Just duplicate and cut off

hyper-neutrino

i believe so yeah - if either is a scalar or singleton or otherwise they need the same length

AviFS

Or is this always the case: only map then if ⍵=1, ⍺=1 or ⍺=⍵?

@hyper-neutrino well it's good enough for now, thanks a bunch!

dzaima

@AviFS you can map when either side has rank 0, or one has 1=×/⍴it, or both have equal shape

(note that the case of 1=×/⍴arg is a historical wart and shouldn't be relied on)

AviFS

2:24 AM

@dzaima fancy fancy

Can you give an example of thing with such shape please?

⎕←×/⍴(1)(1)(1)

TryAPL

@AviFS 3

AviFS

is the only thing I can think of, which doesn't work

dzaima

@AviFS more fancy: you can map when ⍺(∨⍥(×/⍴)∨≡⍥⍴)⍵

AviFS

arghhhhhh

dzaima

@AviFS ⎕←⍴ 1 1 1⍴1

TryAPL

2:26 AM

@dzaima 1 1 1

AviFS

more fancy: you can map when

⍺(∨⍥(×/⍴)∨≡⍥⍴)⍵⍝∨<?⍴∨≠*⍴○⍺⍳⍤⍥⍣⎕⍟⍋⍲⍫!⍟⊖⍉⍋⍱⍟⍲!∩⌈∩⍀|⍺⍝⎕∘*⍵○∊⍴↓∨?¯<≠>≤⍴∆⍳~↓⍺_⊤⎕’∘⌈⊤⊤∩∘⊤⍺’⍵↑↓○?∨○

@dzaima ah, thanks

dzaima

@AviFS (but better just ignore that part and just use ⊂)

AviFS

haha, of course

also, golfy question:

CMC: Which I'll cheat and steal for my submission

Remove the first element after every nonzero element. (The element to be removed is guaranteed to be a zero.)

So: 0 0 0 1 0 2 0 2 0 3 0 3 0 2 0 2 0 1 0 0 0 0

Becomes:

0 0 0 1 2 2 3 3 2 2 1 0 0 0

My current way is a bit sketchy:

⎕←{⍵/⍨1,≤∘0¨2-/⍵} 0 0 0 1 0 2 0 2 0 3 0 3 0 2 0 2 0 1 0 0 0 0

TryAPL

@AviFS 0 0 0 1 2 2 3 3 2 2 1 0 0 0

AviFS

3:03 AM

@AviFS To clarify, this was the golfing question!

hyper-neutrino

is this just implement a stack / backspace an array

in which case what if you input 1 2 3 4 0 0

dzaima

@AviFS in that example it'd work to just remove every second element

hyper-neutrino

is it this problem

AviFS

@dzaima No, only remove zeros after nonzeros

Not at all, where are you guys getting the stack bit?

hyper-neutrino

oh wait, no it isn't - i misread

AviFS

3:16 AM

haha, no worries

dzaima

oh, i also misread a bit

⎕←{⍵⌿⍨¯1↓1,0=⍵}0 0 0 1 0 2 0 2 0 3 0 3 0 2 0 2 0 1 0 0 0 0

TryAPL

@dzaima 0 0 0 1 2 2 3 3 2 2 1 0 0 0

hyper-neutrino

is input non-negative?

dzaima

BQN

AviFS

@hyper-neutrino Yup!

dzaima

3:21 AM

@dzaima if i understand correctly, {⍵⌿⍨¯1⌽0=⍵} also works

AviFS

I got it down to this, just struggling to trainify the left end:

(w/⍨1,2≤/⊢)w

Where w is the array

dzaima

(⊢(/⍨)1,2≤/⊢)w should work

AviFS

Defn might be shorter:

⎕←{⍵/⍨1,2≤/⍵}0 0 0 1 0 2 0 2 0 3 0 3 0 2 0 2 0 1 0 0 0 0

TryAPL

@AviFS 0 0 0 1 2 2 3 3 2 2 1 0 0 0

dzaima

yeah, trains lose a lot when / is involved

AviFS

3:26 AM

Darn, I forgot about the / bit with trains

You're right

But actually, they'd be the same even without

⎕← ≢¨'{w/⍨1,2≤/⍵}' '(⊢(/⍨)1,2≤/⊢)'

TryAPL

@AviFS 11 13

AviFS

Without parens req for /, they'd both be 11

So still no better

@dzaima that's also 11

AviFS

5:15 AM

@dzaima @hyper-neutrino Thanks to your APL help, I've outgolfed both of the other two APL submissions so far! Thanks again!

hyper-neutrino

nice! two things though - 1. i discourage "beats other answers" notes because they can become false later and it's not really necessary and 2. don't comment under other posts saying you outgolfed them

i might not mind, but people might find that really annoying lol

AviFS

@hyper-neutrino Uh oh true, I'll definitely remove point 1

hyper-neutrino

i'll give it a shot but i doubt i can outgolf lol

AviFS

But for 2, I'm only doing it because I've seen others do it

It's the first time I do and I'm happy to delete

But I'm almost positive it was Adám, which feels trustworthy

hyper-neutrino

fair enough ¯\_(ツ)_/¯

AviFS

5:19 AM

And on second thought, tbh I don't intend to leave the "Shorter than the other APL answers so far!" for long

Nor the "Will add explanation soon!" for that matter :P

The idea was really anything but to show off

I just figured there's 3 APL answers now...

Out of just 16 answers

I'm figuring people don't care to see another one, so I added that to get some attention, haha

hyper-neutrino

that's fair enough :p again it's not really a big deal anyway

AviFS

I figured people might not upvote the third submission in a language they've seen several

And it's not really enough of a win for people to notice

They're all in the 60s, haha

hyper-neutrino

eh, it's more likely to not get upvotes due to being too old of a challenge that people just skip over :P

AviFS

True...

Sh*t, I didn't realize it was 7 freaking years old

hyper-neutrino

even if someone posts an answer that loses to an earlier post i'll still upvote it if i found it interesting to look at or good by some metric

AviFS

5:22 AM

Of course, so would I

I don't know how interesting it is, tbh

It's just shorter

Thanks for the upvote, I'm guessing that was you, haha

hyper-neutrino

it was, i forgot to the first time :P

AviFS

how could you!

@Adám Searched Google for your bounty post...

Let's go!

hyper-neutrino

question - what's a good way to partition an array at specific indices?

e.g. I want 1 4 6 f 1 2 3 4 5 6 7 8 to give (1 2 3) (4 5) (6 7 8)

or 1 0 0 1 0 1 0 0 f 1 2 3 4 5 6 7 8

Razetime

@hyper-neutrino dyadic ⊂?

apl.wiki/partition

hyper-neutrino

oh ಠ_ಠ i do not know how i missed that

Razetime

5:34 AM

@AviFS gonna be difficult to outgolf this without a different approach

AviFS

You think?

I feel like some minimal things could be improved

There's a lot of parentheses and ¨

I'd be surprised if there weren't a few bytes that could be shaved off

See!! You did find something, razetime

hyper-neutrino

how do i turn an array of vectors into a matrix? ideally it should pad if jagged

disclose didn't work, as i (somewhat) anticipated

or whatever ⊃ is called

AviFS

↑

hyper-neutrino

oh lol

AviFS

⎕←↑(1 2 3)(4 5 6)(7 8 9)

hyper-neutrino

5:44 AM

ty

TryAPL

@AviFS
1 2 3
4 5 6
7 8 9

AviFS

It's called mix!

Don't ask my why

hyper-neutrino

ah, good to know, thanks! yeah idk what to search for, lol

AviFS

And it'll conveniently pad if not rectangular

⎕←↑(6 4) 5 3

TryAPL

@AviFS
6 4
5 0
3 0

AviFS

5:45 AM

@hyper-neutrino for sure! thank you for all your help

and for giving me a question I could answer, haha

hyper-neutrino

:P

hyper-neutrino

6:05 AM

⎕←⊂ 2 2 ⍴ ⍳ 4

TryAPL

@hyper-neutrino
┌───┐
│1 2│
│3 4│
└───┘

hyper-neutrino

⎕←⊂¨2 2 ⍴ ⍳ 4

TryAPL

@hyper-neutrino
1 2
3 4

hyper-neutrino

is there any way to like, enclose each col (or row)?

Razetime

probably using rank

hyper-neutrino

6:14 AM

i think this is a bit of an xy problem lol. i'll try to figure out a non weird way of doing this and ask if i have a more specific question

Razetime

      ⊂⍤1⊢2 2 ⍴ ⍳ 4
┌───┬───┐
│1 2│3 4│
└───┴───┘

hyper-neutrino

ah

i see. ty

Razetime

rank is really really good for multidimensional arrays

AviFS

7:20 AM

@hyper-neutrino Unless I'm missing something, this is just the opposite of mix from before!

Flip ↑ and you get ↓

⎕←↓2 2 ⍴ ⍳ 4

TryAPL

@AviFS
┌───┬───┐
│1 2│3 4│
└───┴───┘

AviFS

And if you want to do by column instead of by row, you can do:

⎕←↓[1]2 2 ⍴ ⍳ 4

TryAPL

@AviFS
┌───┬───┐
│1 3│2 4│
└───┴───┘

AviFS

It defaults to ↓[2]2 2⍴⍳4 in this case

@Razetime This work too!

Razetime

7:37 AM

yeah I should have used that

RGS

12:32 PM

@rak1507 @Razetime are you doing the bioinformatics contest final now?

(now = in 28min)

rak1507

yea will you?

RGS

Yeah

Whatever I don't do in the next ~6/8h, I won't do at all

Tomorrow I won't have time for it

Razetime

@RGS probably will

I'm awake rn

RGS

Right, cuz it's later for you.

Razetime

I'll probably lose about 8 hours sleeping

If I can think through it well I'll probably stay up a bit

RGS

12:41 PM

But wait, is there any scoring based on time? I thought it's just points gained

Razetime

i am sure there's time based scoring on this one

it's also time limited

RGS

Sure, Dyalog's competition is also time limited but there's no scoring based on time

xpqz

@RGS I've made quite a few changes to the book, making consistency changes based on your suggestions. Let me know if you would prefer a fresh cut (I'm happy either way).

RGS

@xpqz fresh cut, please

Sorry I haven't posted any revisions lately :'(

xpqz

Coming right up.

@RGS It's a lot of words, and you have a day job -- I'm happy with anything.

Razetime

12:48 PM

@RGS i'd say there's a big difference between a few months and 24 hours :)

RGS

It still is limited 🙃

Both 1 and 45464623455634572546246 are finite integers, although I agree the latter is bigger.

xpqz

@RGS Wait, are you a mathematician?? :)

Razetime

so in a 24 hour period it does make more sense to score based on how quick answers come in

RGS

@Razetime oh yeah, absolutely! it makes definitely more sense, I just didn't think they did that. For some reason I thought the next tasks are going to have such granular scoring that timing didn't matter.

Razetime

or they're just going to be absolute pain

maybe if they're hard enough they won't have to worry about time based scoring

RGS

1:02 PM

yeah

Well, good luck

rak1507

1:14 PM

lol wtf is this

problems so hard

RGS

yeah

and no perfect scoring

rak1507

I don't even understand problem 1

where do 101101 and 010001 come from

RGS

The point is that you don't have enough information to know how to fill in the ?s, so you have to infer from whatever you can what those ? should be

And my interpretation for the sample I/O, is that they say

well, for the first person, if they had strings 101101 and 010001 then their genotype would be 111102, and that matches with ?11?0?, so that's our guess

My understanding is that it is our job to figure out what are the most likely binary strings or smth like that

oh darn, for the life of me, I cannot understand P2 :P

Razetime

1:35 PM

i can kinda get p2 but it;s not obvious how the solving will go

RGS

Well, now I understand the question as well, but I also have no idea how to attempt to solve it :P

Razetime

based on point value superspreaders is probably the easiest

RGS

One thing that bothers me is that there is a mathematically correct answer for P3, but they score based on simulations they did and not on the actual correct value?

Or do you think they did the simulations with the actual correct answer?

Razetime

1:51 PM

it seems that they have done it with the correct answer

cause people are scoring 100% on it

Razetime

2:34 PM

yay i figured out Q1

rak1507

2:48 PM

nice

Razetime