@ktye can splice combine existing indices in addition to inserting stuff?

@coltim hmm, I guess kinda, in that you need to explicitly provide the value to replace with

e.g. x:!6;i: 1 2; ?[x;i;+/x i]

@ngn I found another segfault: `"\"

@ngn also, another generic-empty-list edge case, but should $() return ""?

@coltim ok does this: x:!6;i:1 3; ?[x;i;-1] -> 0 -1 3 4 5

github.com/JohnEarnest/ok/blob/gh-pages/docs/…

down to 19... there's probably some simpler equivalent symmetries that can make this shorter

++,'/(+|+|';)@'+2^'

@coltim what would be better? ()?

"feel cheated" because i (like everyone?) assumed it was implemented natively, and in k7 it was just a k string

@ngn but you use k-stringgs as well for other primitives

@ktye now i do, yes

should i add it? it doesn't clash with anything else

@ngn if you like k6 (more important: ok-doc) compatibility, why not?

@ktye wait, did k6 have that??

@ngn idn, but oK

?[x;I;[f;]y] insert (from kparc)

reluctantly adding a todo ..

maybe k9 plans it too: there's a line left blank between cond and amend,dmend.

'splice' feels like a cut extension - maybe it should go on triadic _?

that's for the gods to decide

@ngn well your cut is different to k9 cut

splice being on ? could be vestigial - you have an opportunity to fix it

@chrispsn that doesn't bother me. for the basic set of primitives i stick to k6.

@chrispsn wasn't cut "^" in k9?

@ngn yeah. I think splice should be triadic ^ in k9, if it gets added

Interesting library tree.readthedocs.io/en/latest/api.html

Via news.ycombinator.com/item?id=25886682

f:{                                      /x:dowker notation, y:crossing signs
 N::2*n:#x                               /n:number of crossings, N:twice n
 s:1-2*!n#2                              /s:all possible sets of smoothing signs for the crossings (n*(2^n) matrix of -1 1)
 S:{(x,'1+|x;0 1+\:x:(0<x)|:/y,-1+x|-x)} /S:smoothe the crossing numbered x,y in dowker notation (x can be negative)
 a:N!(x S'2*!n)@'/:+~y=s                 /generate all possible smoothed knots (results consist of neighbour pairs that form sets of disjoint unknots)

@rak1507 how's ^this?

whoa, amazing, I'll have to examine it more later

lots of words, but i'm afraid some areas might still be unclear

maybe but the fact each bit is broken down is great

@ngn right now that's what it does. it came up when trying to use ` 0: strList, which doesn't like () but works fine with ""'s

@coltim doesn't it?

@ngn I guess if the only value is (), was thinking like this (replacing the "" with ())

@coltim but 0: is right to complain that it's not given a list of strings. how did you obtain the ()?

often this comes down to my lack of prototypical elements in empty lists

@ngn I think it came from $:'ing a list of symbols

@ngn but why does it work with just () then?

@coltim () is empty, so we can consider it a list of strings without any strings :)

@ngn I see... somewhat related, but did the behavior of &() change recently?

@coltim yes, it's @chrispsn's idea for deep where

the motivation is to make &boolmatrix work like &boolvector - return the coords of the 1s

@ngn ooh. is part of that it returning ,!0 rather than !0 (like &!0 does)?

this extends to non-booleans, to more than 2 dims, and to ragged arrays.. and i'm not sure we've got all the details right yet

@coltim yes, unfortunately there's a discontinuity at 1 dimension - &vec returns a plain (not: enlisted) vector, while &ndimarray returns a list of n vectors

original discussion in the apl room

@ngn ooh, so () is standing in for the latter case? (I think I came across some old APL stuff talking about this sort of nuance)

@coltim apl (dyalog, actually) has had ⍸vec only for a couple of years iirc, and ⍸anyarray only since the last version

@ktye hmm, I guess this is slightly different than the k7 version? (if my understanding is correct, the k7 version can take just a single index, and the oK version with a function feels like a @ variant?)

@ngn the problem with ⍳ i was talking about is that ⍳,3 returns 0 1 2 instead of the mathematically correct (,0)(,1)(,2), for backwards-compatibility reasons related to the treatment of singletons as if they are scalars

@ngn oh I meant the "how do you represent higher-dimensional arrays when you have typed empty arrays" thing

@coltim well, in ngn/k currently you can't :(

i want to fix that but there are more important fires to extinguish right now - arithmetic and "find"

@ngn yeh, I don't recall there being an elegant solution (I think it was more like "empty arrays are messy" or something)

@coltim the traditional way is to store a "prototype" - one hidden element in the empty array, used only to determine the structure of what's missing

my favourite way to explain prototypes:

@ngn ah yes, the empty array jokes =P

56849027: I really like being able to apply a function to the values specified by the indices; I've run into that when golfing

emptiness has a color; absence a form

@JohnE is that a joke? explain pls?

@coltim to expand on this, it would be great to be able to do ?[x;i;10/0^x@i:0 1+*&~0>':x] instead of @[x;t-1;(0^x t)+10*]_t:1+*&~0>':x. i.e. remove the values from x at i (the latter of which could be multiple indices), apply the monadic f to those values (as a single arg to the monadic f), and stitch the results (which may be one or several) back in at the first index in i

@ngn more like poetry I guess

@coltim (f could also be a value)

@ktye yeh, sorta like the inverse of (...)#x

(in that example I was using list_int to drop a specific index from the list)

@rak1507 do you know what these mean: positive/negative smoothing, laurent polynomial, unknot?

first two, no, unknot, yes, but I can always google the terms

the unknot + trefoil are about the only things I remember from a short introductory thing about knot theory I went to

that and having to do lots of colouring stuff :(

@rak1507 "smoothing" is from the recursive definition of bracket polynomial

the classic algorithm goes like this: we choose a crossing (the first blue circle), we smoothe it in two different ways (the other 2 blue circles), compute the bracket polynomials for the resulting knots, multiply one by A and the other by A^-1 (this depends on the crossing sign) and sum them

being an array programmer, i decided not to do recursion, but to process all crossings at once :)

this might have been a mistake - it would have been easier to explain if my solution followed the algorithm from the definition

maybe to explain but not necessarily to implement/port

not sure if it would have been a shorter or longer solution

how does K handle recursion out of interest?

@rak1507 it handles it well :)

does it have an equivalent of ∇?

in ngn/k o is a special variable name like apl's ∇

ah

not a symbol? :(

is z a special variable name too?

we're short on symbols

I heard unicode has some ;)

@rak1507 i think not but .z is a special namespace

Ah

because I think I had an issue when I tried to do z:something, but maybe that was just me

officially k doesn't support unicode. my support for unicode is an own extension.

@rak1507 oh, it could have been because x y z are default argument names, like ⍺ ⍵

ahh yeah that's it

that makes a lot of sense

so if you don't have an explicit arg list (like {[a;b;c;d;e] ..}), and you mention z, the function automatically becomes triadic

yep

unlike apl, k functions know their valence

in what way do APL functions not know?

@rak1507 {2×⍵} - is this a monadic or dyadic function?

oh wow, I have never noticed that if you do 1{⍵}2 it uses both arguments!

the answer is: it's neither. it's ambivalent. it can be applied monadically or dyadically

And in K you can only apply a function one way?

when you give it fewer args, it forms a projection (a.k.a. currying or partial application)

when you give it more, it signals a rank error (rank is a concept that applies to functions, as well as arrays)

Yeah that makes things clear now because when I tried to use z as a variable name what I got back must have been the representation of a curried function

if it looked something like {..}[..], probably

yeah I think it was something like that

there are exceptions, for instance f/ is ambivalent in k

and some primitive verbs can take up to 4 args, notably @ and .