The APL Orchard

« first day (1041 days earlier) ← previous day next day → last day (1910 days later) »

Sherlock9

02:44

@AntonioNikishaev @TessellatingHeckler Thank you both!

14 hours later…

Adám

17:13

Announcement: DuckDuckGo now has an !aplcart bang; you can search directly on APLcart by entering !aplcart with your query. This also works in your URL bar if you use DDG as your browser search engine.

ngn

i like that

Antonio Nikishaev

17:59

How would you do «f⍣g but giving intermediate results»?

Adám

yesterday, by Adám

Q: Repeated function application in APL

Given the integers p, q and n I would like to generate a vector of length n with the elements (⌊p ÷ q) (⌊(⌊p ÷ q) ÷ q) ... In other words I want to construct a vector where the i:th element is {⌊⍵ ÷ q} applied i times on p. How can I do that in APL?

apl

Antonio Nikishaev

@Adám I don't want "vector of 𝑛", I want «until g»

Adám

@AntonioNikishaev Ah. Sorry.

ngn

@AntonioNikishaev dyalog doesn't have scan-fixedpoint, so you must work around it with something like {⍵,⊂f⊃⌽⍵}⍣{g⊃⌽⍵}⊂ or recursion

Adám

@AntonioNikishaev @ngn Maybe I should add {⍺←⊢ ⋄ r⊣⍺ ⍺⍺{r,∘⊂←⍺ ⍺⍺ ⍵}⍣⍵⍵⊃r←⊂⍵} to APLcart?

Antonio Nikishaev

18:16

@Adám Thanks. Yes! I tried to search "iterate", "unfold"

ngn

@Adám that might scare the noobs :)

Adám

@ngn Do you have a better (shorter) general definition?

ngn

better lobby for a scanny ⍣

Antonio Nikishaev

@Adám BTW do you have an idea why is chat.stackexchange.com/transcript/message/52727538#52727538 ?

Adám

@ngn Like ⍣(⍳10) and ⍣(⍳∞)?

ngn

18:18

old relevant conversation

Adám

@AntonioNikishaev Can you give me a full TIO repro?

@AntonioNikishaev What does "unfold" mean?

ngn

@Adám i don't know, either something like that or a separate squiggle. in k they are just / and \. obviously, apl can't do that because it doesn't know the valence of functions in advance to tell apart "reduce" from "converge", and "scan" from "converge-iterate" (or whatever fancy name it has now)

Antonio Nikishaev

@Adám tio.run/##SyzI0U2pTMzJT9dNzkksLs5M/v@ob2pm/…

Adám

18:34

@AntonioNikishaev aplcart.info/?q=iterate#

@AntonioNikishaev Thanks, I'll have a look.

Antonio Nikishaev

@Adám like in Haskell, e. g. unfoldr (\x -> if x>0 then Just (-x, xdiv2) else Nothing) 42 is [-42,-21,-10,-5,-2,-1]. Here function returns (a) what elem goes to the result (-42), (b) the next state (xdiv2 aka {⌊⍵÷2}), (c) whether to stop (Just(v,state) or Nothing). Shouldn't be necessary that complex, but the idea is this.

ngn

18:48

⎕←{⍵>0:(-⍵),∇⌊⍵÷2⋄⍬}42

Dyalog APL

@ngn
¯42 ¯21 ¯10 ¯5 ¯2 ¯1

ngn

@AntonioNikishaev ^

Antonio Nikishaev

@ngn yes... but why would I ever want a recursion? ;)

ngn

@AntonioNikishaev i can do it without recursion in k

1_*'{0<*|x}{(-x),-2!x:*|x}\42

Antonio Nikishaev

well... -{2/x}\42 in k7. but we're talking about apl

ngn

18:58

that's not isomorphic to the haskell code above

k5/k6 could do -(-2!)\42 too

in apl: -{⍵,⌊2÷⍨⊃⌽⍵}⍣{0≥⊃⌽⍺}42 or -{∪⍵,⌊2÷⍨⊃⌽⍵}⍣≡42

dzaima

@AntonioNikishaev hint at the behavior

Antonio Nikishaev

@ngn that -x is not necessary, just a (simple) way to do unfoldr (\x -> if cond x then Just (x, next x) else Nothing) a₀ would be nice [But that's just me, and I'm only "using" APL for 2 weeks, so maybe I'm completely wrong]

dzaima

@dzaima more direct hint

ngn

@AntonioNikishaev afaik there's no simpler generic way to do unfoldr than the ones i mentioned above

and k's \ is just a special case of unfoldr in which the results (a) are the same as what's fed into the next iteration (b)

Antonio Nikishaev

19:20

@dzaima thanks. That (almost) explains it. I specifically checked ⎕drs, but the docs for ⎕dr don't say anything about it trying to auto-compact.

1 hour later…

dzaima

20:37

think i'm finally going to try to somewhat optimize my line parsing code. though it's fun, it's annoying to have to write everything performance-requesting in train form :/

Adám

20:49

@AntonioNikishaev I'll make sure we fix that. Thanks!

→ 10 messages moved to Trash

2 hours later…