@ngn any comments on this? https://chat.stackexchange.com/transcript/message/53485307#53485307 your current UNQ has worst case complexity for sorted input, while it could be much faster than the general case.
i wish we had sane intellectual property laws (we = jurisdictions where the berne convention applies) and geniuses like atw weren't incentivized to be dicks
that looks like an improvement! in i2 i did grade-up within unq: https://github.com/ktye/i/blob/master/_/i2/k.go#L1408 i also thought about idom detection for ^?x to split the algorithms on k-level. you don't use ^x like k7 i guess. Also a kimplementation with each-prior.. could be used.
@chrispsn would be interesting to know how many + are addition and how many flip. i'm thinking about using +x for abs (because of complex numbers). doing the analysis based on the parse tree may be more complicated (to distunguish monadic and dyadic), given you have multiple k dialects as input.
nsl.com (stevan apter's) can be all found in nsl.com/k - no need to view-source; and curl/wget can mirror it. However, be aware it is a mix of k versions (including a few of Apter's own versions).