News for hackers. Stuff that matters.
Re:BQN: An APL Variant from Marshall Lochbaum by mncharity on Saturday August 15, 2020 @10:15PM (#24174281)
A python or julia implementation, integrated with normal numpy, might enable low-barrier exploration and incremental adoption?
Re:BQN: An APL Variant from Marshall Lochbaum by mlochbaum on Saturday August 15, 2020 @10:37PM (#24174415)
This is a nice idea. I'd thought about trying to run BQN on top of an APL (difficult because existing APLs' boxed or nested models are at odds with BQN's based "like a normal programming language" array model), but trying the more mainstream APL-adjacent langauges hadn't occurred to me. numpy at least looks promising.
Re:BQN: An APL Variant from Marshall Lochbaum by mncharity on Monday August 17, 2020 @05:23PM (#24192319)
A story of "It's a python package, a DSL for numpy" could reduce several barriers to exploration and incremental adoption. DSL's let one use the familiar host language, but for that one little bit where you'd be willing to learn a new and better way of doing things.

A further step in that direction, might be to have a second, more familiarly pythonic interface. Sort of a Q for K. But with minimized cognitive distance from a pythonic ideal, numpy api, and BQN. Attempting to create a low-barrier space encompassing them, for learning and trade offs. Though I'd have to think about what that might look like. Semantics and vocabulary might be learned separately from syntax. Perhaps someone might use BQN-native syntax for prototyping, and then dump it for release as company-acceptable "just using a normal python library" verbose pythonic code. A counter argument is the APL and pythonic ideals for code are rather different, so it might be worth using python and numpy simply as infrastructure, without worrying about chasing hearts and minds.