the k tree

« first day (404 days earlier) ← previous day next day → last day (1131 days later) »

richie

12:55

rng:&&0,

chrispsn

13:11

@richie nice

'Where' is eerily common in short, fast solutions

@Traws i like this if we don't assume & or # are available

although it's not quite right for the zero case

Traws

13:34

@chrispsn yes, it should be {-1+\x(1,)/`i$()}

chrispsn

@Traws @richie - added, thanks. If take exists then it looks like icsa's ({&x#1}) is a little faster, but i love that it only needs to use where, and list literal syntax gist.github.com/chrispsn/…

also worth noting k7 (but not yet k9?) deferred execution of !x - not sure how this whole thing should account for such cases

2 hours later…

richie

15:49

Another option for length is {y+1}/

richie

16:13

Scratch that that code is so wrong it's not even funny haha

coltim

Assuming the count of a scalar is 1, would 0+/{1}'(), work? (at least in k7)

richie

Yeah that should work. The idea I wanted to express was 0{y;x+1}/

ngn

@coltim in k9 / is eachright and there's no way to give an initial value for reduce

but this could work: {y;x+1}/0,

coltim

16:29

ooh, that's pretty clever!

ngn

not really. it's just start with 0, go through the elements and add 1 for each.

the only tricky bit is that you must force the function to be dyadic

coltim

I guess I expect / to "use" the input data, vs. it being a for loop on the count

(in other languages it'd be the difference between a for loop and a forEach loop)

richie

16:57

Here's where in k4/k7

w:{-1+,/l+!#l:~-1=!:',/,x}

Assumes ! Exists

3 hours later…

chrispsn

20:29

It feels like a +\ soln could work for where

Or an index into the original array

chrispsn

21:22

oK has: {,/(0|x)#'!#x} github.com/JohnEarnest/ok/blob/gh-pages/oK.js#L213

which is what 'k tech tree' has listed already

chrispsn

21:36

looking for something that doesn't involve nested arrays

ngn

21:51

@chrispsn {+/~(+\x)>\!+/x|:0} ?

still uses nested though :( what was that thing with step functions?

chrispsn

 (`step 0 2 3!"abc")@!4
"aabc"

ngn

thanks

chrispsn

@ngn at least there's no catenation

ngn

22:10

seems it should work with step functions, but i'm running into bugs

{1+(`step(+\x)!!#x)@!+/x|:0}

@chrispsn {1+(`step(-1,+\x)!-1,!#x)@!+/x:0|$[x~*x;,;]x}

JohnE

I've been tinkering with something you folks might find interesting. A compiler from a K subset to GLSL fragment shaders: beyondloom.com/tools/specialk.html

ngn

@JohnE cool :) dyalog:co-dfns = k:.. co-lambdas?

JohnE

co-dfns aspires to solve the much more challenging task of general computation on a GPU. Special-k is just a nice concise k-flavored syntax and sugar for writing fragment shaders that draw pretty animations on the screen.

on the plus side, though, since I'm only using GLSL ES in WebGL it should theoretically run on just about anything that can handle a modern-ish browser. Some people have said it works in Chrome on Android

I might integrate this thing with iKe eventually. You could do some really crazy shit if you could define uniforms and have both a shader and general oK program communicating with it

ngn

22:33

@JohnE i thought drawing pixels in iKe was strange. vector graphics and shaders should be more in the spirit of array languages.

JohnE

well, when I first wrote iKe webGL was a lot less mature

ngn

@JohnE what about canvas?

JohnE

iKe uses canvas. And it does have a vector drawing primitive

but the way you draw things in a shader program is completely inside-out from that perspective

« first day (404 days earlier) ← previous day next day → last day (1131 days later) »

all rooms