@chrispsn what is the @: in the middle doing?

@ktye modified assignment - like i:i@i . There's probably a way to simplify it

@chrispsn interesting. how is the operation i@i called, if i is a permutation index? and what is i@i@i@i.. doing?

ngn.bitbucket.io/k/#eJzLtNI1tzfnylSIwYcAyzAK4w==

A shuffle?

I do wonder if dicts could make the code cleaner too

@ktye permute? :)

or self-permute? (a composition of a permutation with itself)

@coltim @ngn btw this can now be {(x?x)=!#x}#

@coltim but yes, as far as i can tell the find-based approach is way faster than the non-find one

@chrispsn maybe not cleaner, but: {<!/(x;i)@\:&~@[;0;0]~':x@:i:<x}

last one: {(@[;0;:;1]@~~':x@i)@<i:<x}#

obvious simplification {@[(~':x@i)@<i:<x;0;:;0]}_

i like this layout since it makes ordinal's role clearer

here's an awful one

*'-1_{x^*x}\

fwiw here's BQN's 'mark firsts' github.com/mlochbaum/BQN/blob/…

@chrispsn I wonder if f#x could just be syntax sugar for x@&f x under the hood or something. it appears to be a decent bit slower

@coltim it also matches keys#dict/i#dict

@coltim heh, clever

@coltim oops: {!((0#x)!()),x!x}

@coltim how about {!,/x!'x}

@chrispsn quite close to working, provided the integer overloads of ! are handled. guess there is also preserving the typing of empty lists

also in some k's x!y make-dict requires x and y to be lists

@coltim how about {!!/0,x!x}

@Traws ooh wow, not sure how it does it but it does handle those edge cases. I don't understand why but it doesn't seem like to like symbols =|

@coltim ah yeah, it's because 0 mod does not work as an identity with symbols.. but {! ::'x!x} seems to work

@Traws ok, handles symbols and also nested lists (so long as there are no scalars)

I guess the "top-level" case of ?scalar is not how ? distinct works as it requires a list

What does like do in K7?

@Razetime not 100% sure but it's likely similar to k4's

hm

I'll add the K4 example for now

@coltim the previous form could be just {!0!x!x}.. anyway the idea works by applying identity to the items of the dictionary, which deletes the duplicate keys. but I guess I'm abusing a ngn/k mechanic, I don't know if this behavior is intended (@ngn). it does not work in other Ks

@Traws not really intended, it must be a side effect of how i implemented arithmetic operations on dicts

@ngn I also got this oom error

"what should dict0,dict1 do" is an interesting question

@ngn "upsert" seems to be the generally accepted version and seems sorta related to how the arithmetic ops work

@ngn i catenate both, key and values

@ktye but that could create a dict with duplicate keys

even when the args don't have duplicate keys

yes, it does.

@ktye it shouldn't cost too much to redirect x,y (where both x and y are dicts) to @[x;!y;:;. y]

@Traws fixed, thanks

@ngn it doesnt work for me. your @[..] is also inconsistent: ngn.bitbucket.io/k/#eJxziDZUMLLWMba2sjaJ5XIA8yAcAE18BdE=

@ngn I think I prefer to append the keys that don't collide and leave the rest intact, because this looks very strange