Good news: Preparation of FinnAPL idiom list for inclusion in APLcart is now over half way done.

Thanks @ngn @Adám @dzaima for the help with the scan thing

If APL evaluated 2 3 4 + 2 null 4 as 4 null 8, what would be the argument for not returning 2 6 null as the result of 2 3 4 + 2 3?

@PaulMansour because giving errors is good, there are a couple orders of magnitude less times that'd be useful (from the already incredibly few for nulls) and a couple orders of magnitude more cases when that'd be explicitly very very horrible

(a completely pointless point, but what about 1 2 3 + ,2? In Dyalog, because history, that vectorizes the two, but it'd be ambiguous now with null extending)

Agreed. I assume by "incredibly few" you mean the few primitives that would return null given non-null arguments (like 0 divided 0)?

But then every primitive has to handle nulls in some way, many of which will not lead to errors, because that is the point. No?

@PaulMansour I meant in general useful, I see nulls as being pointless, and besides some performance edge-cases, I don't think they should exist (as a global setting, at least. opt-in would be kind of okay-ish, but APL really isn't suited for that).

Got it.

Please let's not have a setting for it!

@Adám, can you remind me of how to ⎕S for multiple strings? (if it's possible, I think it is)

@J.Sallé just have an array of strings as the left operand?

@dzaima I thought that was it too, but it's not exactly what I need, unless I'm doing something wrong

@J.Sallé So what do you need?

⎕←'you' '\w+'⎕S'\u0' '&'⊢'can you remind me'

@Adám
┌───┬───┬──────┬──┐
│can│YOU│remind│me│
└───┴───┴──────┴──┘

Let me see if I can illustrate it: If I do something like 'a' 'b'⎕S'&'⊢'abc a b c', it returns 'ab' instead of 'abc' 'a' 'b'

I need the latter rather than the former

@J.Sallé You need to ravel each. Remember that 'a' 'b'≡'ab'

@DyalogAPL This is one of those cases where my " syntax would be nice.

Ah, I knew there was a trick I was forgetting

@Adám oh, you've thought of it too?

or am i forgetting that i got it from you. it's been a while since i got here

@dzaima My idea is that "abc" is ⊂,'abc' while "abc" "def" is 'abc' 'def' and "a" "b" is ,¨'ab'. I'm not sure if dzaima/APL agrees.

It doesn't solve all issues, but it does take care of a large swath of them.

@Adám ah, my idea is that "anything" is ,'that same anything' (except escaping - maybe " could also allow for traditional escape sequences?)

@dzaima Yeah, I thought of the escaping too, and that might be good, but it is orthogonal, and so not my focus.

@dzaima Hm, ,'that same thing' — is that the same thing as mine?

@Adám you have an additional ⊂ after that

@dzaima Right, I have ⊆ after that.

@Adám I don't like that that breaks how strands work (what would ("ab") "cd" be?)

So any time you have "something", you have consistent depth of 2.

@dzaima Yeah, I thought about that too. That one in particular is no problem, though; it is (⊂'ab')(⊂'cd')

@Adám or you could have constant depth 1, and not break into strands

@dzaima No, because I wanted to solve the oft-repeated issue: 'list 'of' 'words'⍳'words' with "list" "of" "words"⍳"words"

@Adám imo that's just a problem of how ⍳ works and that in APL having a built-in for ⍳∘⊂ would be unheard of. it also continues teaching the wrong fact that arrOfE ⍳ E for any E works (while that'd only be true for "-strings and simple scalars)

@dzaima Right. ⊆⍵ is actually slightly problematic too, as it teaches that …f⊆… solves your problems, but what if I need the depth to be at least 3?

i should mention that I don't like ⊆ either

@dzaima My fault. I insisted it be added :-(

For a while, I had a campaign for that as my email signature.

well it does solve a problem, but my thinking is that the problem should've been solved deeper down (but that sadly probably isn't possible without breaking everything)

I think I'm reiventing the wheel multiple times over, but I want to finish this answer before I post it here for you guys to see, so I have another question: Is there a way to guarantee ⌸'s ⍺ to always be in the same order? Is grading the argument the only way?

@J.Sallé I usually prepend the ordered elements to ⍵, then subtract one from the counts/remove the first index.

Ah, that's clever actually hahahah

didn't think it could be solved so simply, but it makes sense

@J.Sallé Iverson actually originally had two editions of the function (it wasn't an operator yet), one that maintained order, and one that sorted.

Smart, Mr. Iverson! I wonder why it didn't catch, though ⍨

I don't think your solution is gonna work for me though. ASCII art challenges are a pain in this case

I'll just sort the result of ⎕S, I guess

Actually I can't do that either because ascii art again ugh

@user9791370 Hello. Interested in APL?