@pmikkelsen Hah! Your "APL9 from Outer Space" talk is exactly what rekindled my long-standing intention to look at Plan 9 more closely.

@pmikkelsen I sort of agree. Ideally, we'd have a native build, but running an entire Linux kernel just for Dyalog seems worse than running it through some emulation layer.

I don't quite get what's going on with this: ↑0⍴⊂(↑L)+↑R. This paper is claiming that primitive scalar dyadic functions are defined to give the above result on empty input.

I just don't quite get the reason for Mix.

Aw. Kind of sad that ÷¨⍬ gives a divide-by-zero error.

Why shouldn't it be? Accessing elements from ⍬ gives you 0

@B.Wilson are you sure it's a mix? Could be in ⎕ML←2 where ↑ is first which makes more sense

@SilasPoulson Yeah, I understand the mechanics of the current behavior, but it's a definitional/design question more than anything. What should ÷¨⍬ return? If your problem domain is positive reals, but the input data just happens to be empty, it's sad to have to special case the base case.

@SilasPoulson Ah! Thank you. I was beginning to lose my sanity a bit.

Suppose you could set []div to 1 preventing the error but feel it makes sense to error. You're asking what's the reciprocal of each of my no things.

@SilasPoulson it should really be ⍬ as well I think, ¨ on empty arrays should return an empty array of the appropriate type imo, APL just determines the type by evaluating on the prototype as there isn't a type system