@ksi all syscall args go in registers, so their types don't really matter as long as they can fit in 64 bits

@ksi interesting. i wonder what's so different about close(). maybe i should switch to syscall wrappers too at some point

j's way of exposing the tokenizer to the user is an interesting idea: news.ycombinator.com/item?id=22725596

I wonder how much that could be extended to e.g. replace symbols with words (disambiguating overloads), or with non-ascii symbols (to achieve some "idealized" version of the syntax that could be copied/pasted)

@ktye In dyalog, you can also display a function as a tree that is very useful (see dfns.dyalog.com/n_dft.htm), and Marshall Lochbaum wrote a script to do the same thing in J (code.jsoftware.com/wiki/User:Marshall_Lochbaum/Formal_Parser). This is very useful, but it's in fact one of the things I don't like about tacit expressions: representing them in a linear way is far from optimal. I do not have the same problem with K's xyz functions.

That said, I find +/%# very cool

@yiyus you mean the tacit form (which is not in k)?

@yiyus display a function as a tree - k has `p@ for that

yes, i mean tacit verbs like in dyalog and j

specially trains with forks and hooks, obviously i do not see any problem with tacit verbs like sum:+/

@ngn: "if i start making up too much syntax of my own, i fear nobody would be interested." on the contrary - why would anyone use a k clone when they can use the real deal? exploring new variations on array languages should be encouraged. one of the most exciting projects over the last decade is enchilada (see here: enchiladacode.nl) the author, robbert van dalen, took ideas from k and joy.

robbert has a successor to enchilada - "spread". here: github.com/odipar/spread

@bakul: e.g. dictionaries are atomic.

@bakul: might as well add that limiting dictionary keys to names often forces the programmer to fall back (sometimes in mature applications) when the keys are extended to non-names.

@bakul: also: nice to see you again!