@scrawl here is oK's implementation of dict,dict: github.com/JohnEarnest/ok/blob/gh-pages/oK.js#L117

duplicate x, iterate over the keys of y setting x[yk]:yv

the confusion you're getting is probably because I iterate over keys, and use dget() to retrieve the keys, so both times I get the first binding for the value

it would be possible to tweak this to iterate over key/value pairs but if you're using denormal dictionaries on purpose it's sort of a "play stupid games, win stupid prizes" scenario

an argument could potentially be made that the make-dict operator ought to squash denormal dictionaries together into proper dictionaries; it's a more expensive operation, though

@coltim how's this? {$[x;!,/(,'x)!'x;x]} tests

i note there's no grade or sort in k tech tree yet

kelas had a nice quicksort implementation here

@ktye fixed, thanks

@Traws @scrawl think of it as an amendment - each key-value pair from the second dictionary overwrites the first seen matching key and its corresponding value in the first dictionary

@chrispsn it doesn't seem to like nested lists (b/c of a length error that can be worked around with {$[x;(,'x)!','x;x]}) but I think there's a segfault (@ngn) when razing a bunch of nested dicts of different sizes/ranks or something, e.g. ,/((,,7)!,,7;(,,7)!,,7;(,(1 2 2;2 3 3))!,(1 2 2;2 3 3))

@coltim smaller repro: (`a!0),(,,`b)!1

this probably has something to do with the rank sensitivity of find

interesting that k4 just throws a 'type error when joining dictionaries of different ranks (or at least on that specific example)

(I imagine joining dictionaries of different ranks falls into the "play stupid games, win stupid prizes" category)

@coltim is the "upsert" you mentioned the current behavior in k4 and ngn/k or is it something else?

@Traws both - (k4/q doc link)

@chrispsn I think this is cute

@ngn oh, hang on.. this is just my stupidity. i've fogotten to let the error through in the function that handles dict amendments

@coltim works now, though the result may look a little strange

@ngn hmm, 'idx on the ,/?

@coltim yeah, but idk why the web version doesn't print it

@ngn I had to do a Ctrl+F5 to get it to show up

ah.. :)

thanks

@ngn, what do you think of eval working recursively on a parse-tree ?

@Traws . is more like apply than eval

@Traws does any other k impl do that?

@ngn k7

@ngn . string tho

@Traws strange. k9 doesn't.

@coltim right, . string is more like eval (or parse+eval)

having some mechanism to eval a nested parse tree would be fun

to me it seems more useful.. but of course I'm not considering the implementation details and performance impact

@coltim yep, because we have `p to get the parse tree but don't have the counterpart of evaluating it

it could be also a symbol like `p if not .

hmm what if .str was just parse, and .non-string-list was eval?

i have tokenize: `t"str", parse: `p toks or `p"str" and value: ."str" or .code. the result of `t is a flat list of tokens, and of `p a flat list of code points.