@nathanrogers If you want to see the current uses of Key, you should probably read e.cd. If you want to see how I used Key with Depth Vectors, then you can read the Publications section of the README for the "Key to a Data Parallel Compiler." I have an extended version of that paper, but I don't recall whether I have posted a link that one or not.

@nathanrogers I almost certainly have a bug in page 5, but I certainly don't know what it is.

@nathanrogers Are you still having trouble with getting Co-dfns installed or compiled?

Interesting challenge: starting with a depth vector representing a tree where each element is listed in the vector according to a depth-first, pre-order traversal, what is a nice, short way to compute the table showing the distance from each node to every other node in the tree, where distance is the number of edges in the shortest path connecting any given two nodes.

Example Depth vector:

0 1 2 3 1 2 3 3 4 1 2 3 4 5 6 5 5 6 3 4 5 6 5 5 6 3 4

And corresponding distance matrix:

0 1 2 3 1 2 3 3  4 1 2 3 4 5  6 5 5  6 3 4 5  6 5 5  6 3 4
1 0 1 2 2 3 4 4  5 2 3 4 5 6  7 6 6  7 4 5 6  7 6 6  7 4 5
2 1 0 1 3 4 5 5  6 3 4 5 6 7  8 7 7  8 5 6 7  8 7 7  8 5 6
3 2 1 0 4 5 6 6  7 4 5 6 7 8  9 8 8  9 6 7 8  9 8 8  9 6 7
1 2 3 4 0 1 2 2  3 2 3 4 5 6  7 6 6  7 4 5 6  7 6 6  7 4 5
2 3 4 5 1 0 1 1  2 3 4 5 6 7  8 7 7  8 5 6 7  8 7 7  8 5 6
3 4 5 6 2 1 0 2  3 4 5 6 7 8  9 8 8  9 6 7 8  9 8 8  9 6 7
3 4 5 6 2 1 2 0  1 4 5 6 7 8  9 8 8  9 6 7 8  9 8 8  9 6 7
4 5 6 7 3 2 3 1  0 5 6 7 8 9 10 9 9 10 7 8 9 10 9 9 10 7 8

@ngn @pierre @AttilaVrabecz Any idea what version of K this is? {s[x]#(,//y)@<s[x]_sv(s _vs!*/s:^y)x}

@Adám looks like kona/k3

@arcfide 26 bytes: ∘.(≢∪~∩)⍨⊢↑¨∘↓(⌈⍀∘↑-↑¨⍳∘≢)

@arcfide {∘.(≢∪~∩)⍨∪¨⌊\¨,⍨\⌽⍋⍋⌽⍵}

@ngn ^ can probably be even shorter

Same byte count but looks nicer is {∘.(≢∪~∩)⍨(⌊∘⊃∪⊢)\⌽⍋⍋⌽⍵}

Is ⌽⍋⍋⌽⍵ the breadth-first, post-order traversal? (I'm only guessing the name)

Hmm, looks like it is a breadth-first traversal, and I guessed wrong what pre-order meant

@H.PWiz oh, 24 bytes, nice! :)

@ngn Thanks. I'd post it if I knew how to use Kona on TIO.

@Adám which challenge is it a solution to?

@ngn ⍉

@Adám how do i call it? {...}[order;nestedarray]?

@ngn No idea, I guess that or with reversed args.

100000>1.797693135E30

I'm getting domain error

@Adám

⎕←100000>1.797693135E30

@H.PWiz
0

⋄⎕IO←0⋄⎕DIV←1⋄⎕←100000>1.797693135E30

@nathanrogers
0

X(

in Dyalog this is giving me Domain Error

⎕←100000>1.797693135E308

@nathanrogers 308, not 30!

Sorry

⌊/ (0=11○x)/x←1J3 2J3

why doe sthis return 1.797.....

⎕←⌊/⍬

@H.PWiz
1.797693135E308

@nathanrogers Because 0=11○x is 0 so x gets compressed to ⍬ and then ^

so how can I only get the smallest of a list or ⍬?

@nathanrogers That's a bit inconsistent.

I want the least from a list where the least is a non imaginary number

that is very consistent

if there isn't one, then there isn't one

I'm slightly surprised that you can't compare with ⌊/⍬ though

@nathanrogers When reducing over an empty vector, APL will attempt at returning the identity element. The identity element is an argument which leaves the other argument unchanged. For min ⍺⌊⍵ it is the greatest representable number.

@H.PWiz It is because of rounding. the 10-digit precision display rounds up beyond max-float, which in turn is an invalid input when ⎕FR=645.

how would I express "the min of all items that are not imaginary"?

⍞←100000>⌊/⍬

@Adám 0

@nathanrogers Like you did, but there aren't any in x!

i know that

but 1.77...E308 is also not in my list

⎕←⌊/⍬ ⋄ ⎕pp←34 ⋄ ⎕←⌊/⍬

so (⌊/(0=11○x)/x) returning 1.777.... doens't make any sense

@ngn
1.797693135E308
1.7976931348623157E308

I know, so how do I express what I want?

I'm not asking how is ⌊ implemented

I'm asking how to get the result I'm after

@nathanrogers What do you want there are none? Really ⍬‽

nothing

if there are none, then its an empty list

obv

@nathanrogers There is no such thing as "nothing" in APL.

⍬

nil, nothing , nada

@nathanrogers That is something; a length-0 list.

@nathanrogers Well, there is ⎕NULL, but I wouldn't recommend it.

which is what it should be

empty

@nathanrogers (∨/b)/⌊/(b←0=11○x)/x←1J3 2J3 would do it. Or ⊃b/⌊/(b←0=11○x)/x←1J3 2J3 if you want to golf.

@DyalogAPL this looks like a bug in dyalog

any arith op with 1.797693135E308 causes a domain error

@ngn Why? .48 rounds to .5

on the other hand 1.7976931348623157E308 works

@ngn Yes, that number is beyond the range on a 64-bit float.

@ngn That's less.

I think 1.797693135E308 should at least error, before doing anything with it

@Adám ^

@H.PWiz It does, no?

⎕←1.797693135E308

⎕←1.797693135E308

@H.PWiz
1.797693135E308

@ngn
1.797693135E308

lol :)

lol :D

We allow you to input and handle DECFs in 64-bit mode, but not do math with them:

⍞←1E1000

@Adám 1E1000

⎕←2 3⍴1E1000 1E2000

@Adám
1E1000 1E2000 1E1000
1E2000 1E1000 1E2000

⎕←1+2 3⍴1E1000 1E2000

@Adám
DOMAIN ERROR

@H.PWiz hahah, you're a couple milliseconds faster than me :)

I have a question

tio.run/##fVXBbttGEL3rK/…

this is a primitive ray caster

@nathanrogers Whoa.

tio.run/…

this function is giving me problems

the csting function in particular

there's a bug in the cast function to dtermine which color to show, and I've been staring at it for 2 days

and it's making me nuts

@nathanrogers Wait, does it run in normal Dyalog APL (not compiled)?

yes

but you need codfns image

to display the result

I believe the sphere and plane detections are working like they're supposed to, I've reviewed them several times, but the sphere is coming back as two colors

@nathanrogers OK, but when running in normal Dyalog, can you at least inspect the result and verify that it is as expected?

it's a very large result

800x600x3

@Adám ugh. so ⌊/⍬ prints as something that not only doesn't evaluate to its exact self, but causes errors when you do arith on it... that's why i want ngn/k floats to round-trip

at least the output of the Ray function

@ngn Welcome to float land!

I don't know how to do an isolated test of "cast"

blob:imgur.com/bdf59022-309e-4cb6-85ae-e3d59a9b9ce2

imgur.com/gQpdZdZ

there we go. you can see the sphere should be red, but the top section is green

something is wrong in that cast function expression to determine the color index of the nearest distance found

@nathanrogers Maybe scale the whole thing down to 16×12 and hand-check?

good idea

didn't think of that :P

Oh

I'm never filtering out the nonzero vlaues

rather, I'm never filtering out the less than zero values

ah! but that gets rid of the plane X(

⎕FR←1287⋄⎕PP←34⋄⎕←⌊/⍬

@DyalogAPL Hey!

⎕←⌊/⍬⊣⎕FR←1287⊣⎕PP←34

@Bubbler
9.999999999999999999999999999999999E6144

Interesting

@Bubbler You can also use a leading ⋄ to have multiple statements but not output the result of the first:

⋄ a←2 ⋄ b←3 ⋄ ⎕←a+b

@Adám
5

Oh thanks