The APL Orchard

« first day (1003 days earlier) ← previous day next day → last day (1948 days later) »

Bubbler

00:08

I guess I could work from binding strength, but having one would still be handy

1 hour later…

Bubbler

01:18

TFW APL wins over Jelly.

2 hours later…

Sherlock9

03:29

@Adám It's happened again! No trees for trains!

I tried the commands again and the output handling system is there, ]rows etc does print dots, and boxing still works on arrays

4 hours later…

Adám

07:36

@Sherlock9 Can you save your session and send it to me or support@?

Sherlock9

That's a good question. Where in RIDE is the save session button?

Adám

@Sherlock9 There isn't one: aplcart.info/?q=save%20session%20file#

Sherlock9

08:14

Now I have to hunt where it saved the file

Found it!

Adám

08:33

@Sherlock9 Yeah, but generally, posting people's full email addresses in plain-text is a bad idea.

Bubbler

09:30

So it happens that both of my top two APL answers have the ⊥

Adám

@Bubbler Yeah, ⊥ is awesome. I just wish it had 2 as default left argument.

Bubbler

Looks like many people agree with that :)

Sherlock9

09:50

@Adám Ah sorry. Are you able to delete my message as room owner?

And I've sent you the session

Adám

@Sherlock9 No, only mods can. I've asked Mego to fix it. Flagging is probably overkill, as my spam filter works quite well :-)

dzaima

10:06

but i dont want to optimize bitvec concat

dzaima

10:17

@dzaima while at it, in what order should (⎕←1 ⋄ ⎕←2 ⋄ ⎕←3) output the numbers (aka in what order are the expressions evaluated)?

also, should (1⋄-⋄3) 2 error or give 1 ¯2 3?

Adám

@dzaima ¯2 2

dzaima

@Adám so just ignore the ⋄s when parens contain both functions & arrays?

Adám

@dzaima No, (1⋄-⋄3) 2 is like ((1)(-)(3)) 2

dzaima

@Adám but still have function arrays or not that too?

Adám

@dzaima I don't really see how the function arrays can work generally. Just use (f⊂⍛,g⍮h) if that's what you mean.

dzaima

10:28

@Adám that's so messy/unreadable & not-obvious-at-all though

Adám

(⊂⍤f,⊂⍤g,⊂⍤h) looks more consistent.

I don't see how arrays of functions can work if you don't have a way to turn functions into arrays, like J's gerunds, K's f[] and Dyalog's ⎕OR'f'

dzaima

@Adám what i'm suggesting is that (f⋄g⋄h) is exactly (f⊂⍛,g⍮h), and also i still haven't removed `

Adám

@dzaima Sure, but how do you nounify a single function? How or when are the functions applied as opposed to being used as data? Etc. etc.

dzaima

@Adám again, i'm suggesting (f⋄g⋄h) to be just an alias to an already representable thing, and we've had the discussion about noun functions before and this is what it ended up as

Adám

@dzaima How does (x⋄y⋄z) know what to do?

dzaima

10:37

@Adám it depends on the types of the variables, which, when known, will be obvious

Adám

@dzaima And thus you were asking about a mixed type thingy.

dzaima

@Adám exactly, i was asking if mixed types should be acceptable or error

2 hours later…

ngn

12:51

@Adám what is "K's f[]"?

Adám

@ngn I thought that was how you nounified a verb: (f[];g[];h[])

ngn

@Adám no, you just write (f;g;h). f[] would call f with :: as argument.

Adám

@ngn Oh. OK then dzaima's syntax much resembles K's.

ngn

@Adám s/nounified/nominalized/ (blame Latin)

also, f g h are already nouns :) almost all identifiers are nouns. "almost" because k has (yuck) reserved words

Sherlock9

13:44

Oh this is turning out to be an interesting problem: What's a good way to write the following function

powermod←{
    a b mod ← ⍵
    r ← 1
    {
        r ← mod | r × a
    }⍣b
}

Because that's obviously wrong. The inner dfn won't be called. But I'm not sure how to get mod|a*b without running into too-large numbers

And {a b mod←⍵⋄1{mod|⍺×⍵}/b⍴a}299 792 458 doesn't work either

ngn

@Sherlock9 if b can be large, you should do repeated squaring to make it run faster. it's easy to implement recursively or with 2⊥⍣¯1 and reduction

@Sherlock9 recursive: powermod←{a b m←⍵⋄b=0:1⋄r←m|×⍨∇a(⌊b÷2)m⋄2|b:m|a×r⋄r}

Bubbler

14:00

Isn't there an idiom for it?

ngn

non-recursive: powermod←{a b m←⍵⋄b←⌽2⊥⍣¯1⊢b⋄(m|×)/b/{⍵,m|×⍨⊃⌽⍵}⍣(¯1+≢b)⊢a}

@Bubbler i don't see it in their idiom list

Bubbler

14:15

I just wondered because J has one

ngn

j is a lot more mathsy

Bubbler

14:32

Not golfed: Try it online!

Bubbler

14:48

32 bytes

ngn

@Bubbler this is ok for golfing, but ⍺×⍵*2 risks an overflow by doing more than one multiplication without mod

Sherlock9

15:27

@ngn Yeah, the point was I was trying to make an exponentiation algorithm (mod m) that worked for large enough a and b that it would normally overflow. Thanks a bunch :D

Adám

Announcement: First Dyalog '19 videos now available!

4 hours later…