The APL Orchard

« first day (1708 days earlier) ← previous day next day → last day (1197 days later) »

Razetime

09:18

Wong Jia Hau

@Razetime thanks

5 hours later…

Sohang Chopra

13:57

Hi everyone! It's my first time here - I am a beginner in APL.

I answered a question on Code Golf Stack Exchange in Dyalog APL. I am posting the link to my answer below - please comment on my answer if you have any suggestions for improving it / reducing character count.

My answer link : https://codegolf.stackexchange.com/a/235768/107187

xpqz

@SohangChopra Can you explain how you run your code, or perhaps post a TIO link? tio.run/#

You can also run one-liners right here in the chat: ⎕←⌽'Hello world'

TryAPL

@xpqz dlrow olleH

Elias Mårtenson

@SohangChopra {⍺[⍵]} can be better written as ⌷⍨

That'll save you 4 chars.

Sohang Chopra

14:32

@xpqz I am running the code on tryapl.org.

LdBeth

@SohangChopra and (?2⍴⍨≢) can be (?2⍨¨), new feature in Dyalog 18

Sohang Chopra

@LdBeth I tried your code, and it works - but I don't understand how it works. Can you please explain?

@EliasMårtenson Thanks for your suggestion, I'll edit my answer accordingly.

LdBeth

15:00

@SohangChopra in 18 dyadic function ⍨ is just a short hand for {⍺}

and seems they also have optimization so N⍨¨ is as fast as N⍴⍨≢

dzaima

@LdBeth {⍺⍺} you mean

rabbitgrowth

15:43

Sorry if this is a dumb question, but I found this expression from https://www.jsoftware.com/papers/50/50_36.htm which calculates the first ten Fibonacci numbers:

fib2←{⊃+\∘⌽⍣⍵⍳2}

fib2¨ ⍳10
0 1 1 2 3 5 8 13 21 34

Am I correct to assume that this is not very efficient, since the `+\∘⌽` operation is run `+/⍳10` or 45 times?

Oops, looks like I messed up the formatting there...

dzaima

SE chat doesn't allow markdown in multiline messages

but yeah, fib2¨⍳n is O(n^2)

rabbitgrowth

Ah I see

Is there a better way to do this?

dzaima

you can go all iterative, but it's not too pretty

Master Quiz

15:58

In f@g⊢A can g be applied to A to obtain just a boolean or can it return a vector of indices?

dzaima

@MasterQuiz if g is a function, it must return a boolean vector

Master Quiz

@SohangChopra I've quickly wrote a 17 character solution: 1∘⎕C@((⍳∊?↑?⍨)∘≢) (I think it should be correct). Is it true that you can't erase the parentheses after @?

Master Quiz

16:16

@Adám Why did you use I-beam instead of {?2:⍵⋄1⎕C⍵}¨

@MasterQuiz 16 characters: 1∘⎕C@((⍳∊?∘?)∘≢)

rabbitgrowth

I think 1∘⎕C@(?2⍴⍨≢) works too?

xpqz

@rabbitgrowth Here's one:

Fib ← { ⍝ Tail-recursive Fibonacci.
    ⍺ ← 0 1
    ⍵=0: ⊃⍺
    (1↓⍺,+/⍺)∇⍵-1
}

dzaima

@xpqz the point was to calculate the first n fibonacci numbers. The original is O(n) for only one too.

xpqz

16:34

@dzaima Right. I offer {r←⍬⋄_←{{p←+\⌽⍵⋄r,←⊃p⋄p}⍣⍵⊢0 1}⍵⋄r}

⎕←{r←⍬⋄_←{{p←+\⌽⍵⋄r,←⊃p⋄p}⍣⍵⊢0 1}⍵⋄r}10

TryAPL

@xpqz 1 1 2 3 5 8 13 21 34 55

dzaima

and that's what I meant with not too pretty all iterative

xpqz

You're basically describing all APL code I compose :)

Master Quiz

@rabbitgrowth I prefer to work with IO set to 1, so it would be 1∘⎕C@(1-⍨(?2⍴⍨≢)), but yes

rabbitgrowth

17:01

Ah right, I shouldn't have assumed ⎕IO←0. Sorry!

Thanks @xpqz! For some reason it didn't occur to me that you could use ⍣ with a function with side effects. It appears that your solution is about twice as fast. This is really interesting.

xpqz

There is a lot of detailed fib-stuff in dfns, which you may have seen already.

rak1507

@SohangChopra ⊢⎕C¨⍨¯1*∘?2-∊∘⎕A is 16

⎕IO←0 of course

ah missed the @ stuff above

~~oh they don't work anyway with keeping capitalised letters the same~~ nvm it does you can just uppercase random letters and it'll work, clever

1 hour later…

Master Quiz

18:19

@rak1507 Isn't this same as ⊢⎕C¨⍨¯1*∘?2⍨¨

rak1507

19:05

No, it doesn't keep capitals capitalised

Leo

19:26

@xpqz Thanks, I'll have a look.

Adám

@rak1507 {⍺⍴,⌿1+(,⍵)⊤⍺?⍥(×/)⍵} then. (⎕IO←1 for APLcart.)

rak1507

fair enough...

I feel like it should probably be ⎕IO independent if possible but sticking with one makes sense too

Adám

@MasterQuiz Where?

@rak1507 One feature I'd like would be a checkbox/switch/parameter/whatever to flip all entries between ⎕IO←0 and ⎕IO←1, but obviously, that is a major job, and makes the ongoing job of maintenance/additions harder.

rak1507

yeah that's overkill probably, lot of effort for not much gain, normally pretty easy to adapt something to be ⎕IO independent unless it uses particular tricks

Adám

… which one or two entries do; something like getting indices and padding with zeros.

rak1507

19:41

yep

Master Quiz

@Adám Here but I just realised that this answer is 4 years old... Also there's a typo in "Rquires".

Adám

@SohangChopra If you see lots of ¨ then there's a good chance you can "factor them out": {(?2)⌷⍵,1⎕C⍵}¨

Master Quiz

@Adám It becomes also a lot more readable

Adám