@coltim in k9, ? flips dict domain and range

Like find but on the entire data structure instead of specific elements

@coltim how would =`a`b`c!0 1 0 work?

@coltim _sym may be suitable for "delete global variable" too. both are rare.

@chrispsn flips as in swaps them? why would anyone..

@ngn if calcualted correctly, 25kB are only field names longer that 2 chars. some are keywords.. does minification fail because that would change the interface? maybe it could still be done with a test case run in a loop by renaming each of the symbols.

@ngn it comes up a fair bit! For example take targets key intersection for dicts so swap can help target values

@chrispsn interesting. I was a bit surprised that ?: generally isn't implemented on dicts

@ngn I must be going crazy, I thought I saw things like =0 1 2 3 4!"abcda" returning "abcda"!(,0;,1;,2;,3;,4) (which struck me as a bit weird)

@ngn yeh so far the only real usecase I've come across for _sym lowercasing stuff is doing type-specific stuff where `c~_@x is a bit nicer than |/`c`C=@x

@coltim i might have broken stuff while i was playing with "find". i still don't know what to do with rank-sensitivity

@coltim ok, done

@ngn I tracked it down. it was the opposite, i.e. the keys having duplicates but the values being unique

codegolf.stackexchange.com/a/171410/95516 @ngn would it be possible for you to make an explanation for this answer?

@rak1507 ha! explaining that would be very challenging :) that answer is from more than a year ago, it's like new code to me. i'll try

thanks, no worries if you can't, I can try and decipher it

@rak1507 have you done any knot theory before? or at least something related to topology?

Nope... a tiny tiny amount but really nothing of any substance

@rak1507 btw, when i first saw that challenge it was not very clearly explained, had only 2 tests, and even they were incorrect

lmao, yeah I saw that mentioned in the chat

but i liked it so much, i spent a week trying to understand knots and polynomials and stuff, i prepared some tests, the community reopened it, and i added a 500 rep bounty, but it expired without anyone solving it (the brainflak solution came later), so i posted this as proof that it's possible to solve

I spent some time researching on this codegolf.stackexchange.com/questions/197129/…, trying to find some sort of invariant but I didn't really get far

But I did find that challenge, and it looks like that produces an invariant of some sort

So that could potentially be used

@rak1507 it doesn't. the bracket polynomial is not an invariant. there are other polynomials that are.

It says '(Again, most knot theorists compute the bracket polynomial as an intermediate product on their way to computing the Jones polynomial, which is invariant across all transformations, but we will not be doing that.)'

so the bracket polynomial can be used to compute a different one maybe

And then that'll work

yes

i would suggest reading wikipedia instead of trying to come up with an invariant polynomial by yourself. it took humanity centuries to come up with good invariant polynomials :)

I was going to use the Jones polynomial and calculate that from the bracket one maybe

@rak1507 did you learn some k already?

not enough to decipher your solution yet

I guess I could look online for other ways of calculating it but I haven't found many resources

but i don't have to explain how primitives and basic idioms works, right?

e.g. !2#n being "odometer"

no

great

what does 'syms oom' mean?

@rak1507 my impl uses a predetermined amount of memory dedicated to symbols. "oom" means out of memory - shown when the limit is exceeded.

makes sense

is there a need to raise the limit?

nope

cmc: given a list of even-length int lists, pad them with 0s so they are centred, e.g. (1 2;3 4 5 6 7 8;9 1 2 3) -> (0 0 1 2 0 0;3 4 5 6 7 8;0 9 1 2 3 0)

cmc?

chat mini-challenge (stackexchange jargon)

@ngn r:{y(n/(!n)-2\x-n:#y)};{m r'(m:#'x)_'(,(|/#'x)#0),'x}

ktye.github.io/…

your link is broken

why does it break? because no http:?

it only works with ktye because of i\ div and i/ mod

noob question: is there a flip adverb (I think that's the right term) like ⍨ in APL?

    ktye.github.io/…

@rak1507 no, not in general

damn

and there's no rotate?!

@rak1507 maybe ngn has it.

@rak1507 rotate:{x_y,x#y}

where's my ⌽ :(

|

ah, you mean the dyadic version? you see, k has less primitives in general. but is rotate needed that often? even if, it's easy to define.

@ngn I think this works

@rak1507 NYI in golf.k =P

@rak1507 {,/|(0,x)_y}

@ngn that works for positive rotations. an obvious overload would be to handle negative integers

what absolute mad person decided % would be a good symbol for division!

@rak1507 iverson. % is the one that looks closest to ÷

is there a 'statement separator' ?

@rak1507 ;

@rak1507 what else should it mean?

thanks

@ktye modulo

I guess it makes a bit of sense

@rak1507 "flip" in k is usually a synonym for "transpose", monadic +

@coltim well done. i'm sure you can shorten that

@ngn you use idiv for 3%2? is that new?

@ktye original k returns floats, but ngn/k has been returning ints for a long time

@rak1507 just so you know: in the original k6 mod and intdiv are y!x and (-y)!x respectively. in k9: y\x and y/x (yes, args are swapped even for div)

@coltim i have 25

@ngn i get golf.k error message here: ngn.bitbucket.io/k/…

@ktye leading spaces mean continuation lines

there's an invisible \l golf.k before the other lines :)

@ngn hmm. what's the point of continuation lines?

@coltim sometimes a lambda can't fit on a line

f:{something
   something}

@ngn like separate statements? why can't it just scan until it hits a "}"?

I know some earlier k's had implicit ";"s at the end of lines

@coltim maybe it can if the file is parsed in one go, but it seems easier and probably more efficient to grab a group of lines, parse them, execute them, forget the asts and anything associated with that, and move on to the next group

@coltim all ks have that, afaik. in k code, ";" and "\n" are treated the same way

@ngn ya learn something new every day. it seems to be (another) difference between q and k

i parse everything. that may only be a problem with huge generated programs.

@ngn not anymore for k9

@chrispsn what's the new way?

@ngn mod div

uh.. i should've guessed

@chrispsn but the args are still swapped

@ngn I dunno how, but 24 bytes

@coltim nice! my 25 bytes were {p,'x,'p:&'-2!-m-|/m:#'x}

oh hmm, -1 with |x,0 instead of 0,x,0

@ngn hmm so -n!x is intdiv too?

@coltim yes, heritage from k6

@ngn folds arms and here I thought % being intdiv was okay

@ngn bug?

nyi error:
-6$'(1 2;3 4 5 6 7 8;9 1 2 3)

I came up with {p,'x,'p:{x#0}'{x%2}@-l-|/l:#'x}

I gave up with the rotation

@chrispsn if -6$1 2 is nyi, so should be that one, no?

best i got in k9 so far is ++(2 div M-L)_'(-M:|/L:#'x)#'x

@ngn oh. for some reason i thought -6$1 2 was ok

@chrispsn for now (-n)$ works only with strings

and might pad with 0N in ngn/k anyway

I miss ⍨

@rak1507 you're still under the influence of apl. breathe! :)

i find it easier to read bracketed stuff than commuted verbs

only one char more

but if you want to do it multiple times, you end up with lots of nesting and it gets confusing/ugly

{x#0}'{x%2}stuff vs ((stuff)%2)#0

I don't really like the second one

would rather do 0#⍨2%⍨stuff

@rak1507 most primitives in k (and maybe apl too?) were designed so the left arg is more likely to be a term, and the right arg a longer expression

yeah that makes sense

% and # are exceptions I guess

-2!x is x div 2 (in k6)

it seems more likely to me to do expression div 2 than something div expression

@ngn ah cool

@rak1507 instead of x#0 you can use "where" (the original from which apl took ⍸⍵): &x

I saw that

that was clever

Another thing is the use of ' a lot, is that fine in K unlike ¨ in APL?

@rak1507 yes, eaches (') and flips (i.e. transposes: +) can do any equivalent of ¨ and ⍤

sometimes you may need eachleft (\:) and eachright (/:) too because k has no enclosed scalars

and the performance is fine vs trying to do something that works on a whole array? (if you can even do that)

i mean, there's no equivalent of apl's ⊂0 1, only ,⊂0 1 - the latter being k's ,0 1

@rak1507 building off of @chrispsn's comment, this could also be #[%[stuff;2];0]

I forgot you could use [] to apply to a function

@rak1507 k works better on leading axes, so we prefer wide rather than tall matrices, and this preference is reflected in some of the primitives, e.g. odometer !2 3 4. per-row overhead is negligible for wide matrices

- 2 in k9... {x@'-(2 div M-L)-\:!M:|/L:#'x}

@ngn alright cool

{x@'((L-M)%2)+\:!M:|/L:#'x} ...

@rak1507 the same principle from APL applies - if you each a function, it needs to either be called few times, or do a lot of computation. {1+x}'!10 is still very much bad

ok I think this is my cleanest 23 bytes

@dzaima right yeah, but it seems less bad than in APL

or you should go to less effort to try and not use it maybe

@coltim wow. and {|x,0}/ -> (|0,)/

@rak1507 I think each being more is just a consequence of no rank and embracing vectors. k does have the advantage of static parsing though, which does speed up execution of scalar code

oh yeah that's true

interesting stuff

@ngn isn't this equivalent to 0|/ in older K and 0|/: in newer? (max over, with 0 as initial argument)

@beagle3 / here is "converge" "n times do", because the function is monadic and there's an int on the left

Oh, I see. I only saw the latest, didn't follow the entire discussion ..... apologies.

@beagle3 no problem. the relevant link is here

something to impress @rak1507 with (i hope) - in k n monad/x is like apl's f⍣n⊢x. can you guess what n monad\x is? :)

iterate/dzaima APL character that I can't be bothered to type?

@rak1507 yes :)

:D

@dzaima a "T" with 2 dots on top, if i remember correctly?

@ngn yep, ⍡

yeah that one

note the beautiful symmetry between / and \ :)

it is nice I have to admit

how do you do windowed reduction in K?

oh I think I saw it in the oK docs

@rak1507 in k6 there is n'a which is like n,/a in apl, and then you can each on that. idk about k9.

Huh, why is ': a thing then?

@rak1507 also: ints\ encode; ints/ decode; str\ split; str/ join

@rak1507 iirc it changed between k5 and k6 so i kept both

@ngn is there an APL /, replicate?

@rak1507 we don't do replicate. we do at-where

now that is cool

@rak1507 these overloads are gramatically monadic operators binding to the left arg, this forms a derived function that is monadically applied to the right argument. k was running out of symbols.

although @ indexing seems almost the wrong way round for the name, "array at indices"

@rak1507 "[items from] array at [positions] indices"

@rak1507 normally you don't need it. a@x is a x. it's only needed if a monadic verb follows, which would be interpreted dyadically otherwise. or use (a x)&..

@ktye aha makes sense

I like how arrays seem syntactically the same as functions that return what is at an index

@rak1507 yes, one of arthur's greatest achievements is the unification of syntax for array indexing, function application, and dict lookup. few languages do that.

it's certainly interesting, I don't know if I like it or would prefer it over different things but it is definitely an interesting idea

@rak1507 I like it too, but it's less so in k9; f#L is filter if f is a function, but L#L is intersection; not sure if there are any other places where the equivalence is lost.

to me function application and array indexing/dict lookup are two very different ideas so unifying the two seems a bit odd

@beagle3 well, A#B is 'filter if in A from B'

so it's sort of similar

@rak1507 yes, but one of the charming (for me) things about the prior unification is that you can replace a function with a cached version of it (array or dict), and in k4 it just worked . Now it only mostly works. There are other "almost equiv but not quite" things; e.g. if you have a list L and a dictionary version of it d:(!#L)!L (! is iota, # is rho).

yeah that's a nice point (the caching)

They would be interchangeable for most things, but !d gives !#L and !L is not the same thing

@rak1507 python and js do those 2, but not function application. so it's not entirely odd.

yeah, arrays are like dicts but with indices as the key (conceptually) so that makes sense that those two are similar, but functions seem entirely different

"associative array" is another term for "dictionary"

all three of those (arrays, dicts, funcs) map values from a domain to a range

@rak1507 Mostly in the sense that functions are "read only". K bridges the gap between them in other ways as well - for example if f[x;y;z] is a function and a[i;j;k] is an element in a 3-dimensional array which yield the same values, then projections like f[x;;z] and a[i;;k] are as-compatible as the original f and a elements.

@rak1507 if you squint, you can imagine an array being a function that takes indices instead of arguments

@ngn this basically

@coltim yeah that's what I mentioned earlier, it does sort of make sense, I like it!

i must disappear, sorry

np

@beagle3 hmm, I hadn't considered that downside