the k tree

« first day (775 days earlier) ← previous day next day → last day (760 days later) »

4:55 AM

@rak1507 it's always amazing how many people seem to read about APLs, recognize some good ideas, and then crawl back to the hole they came from convinced that they extracted everything of value

more fun for us I guess

5 hours later…

rak1507

9:55 AM

@JohnE yeah, especially when they regularly use things like numpy which are crap compared to any proper array language

dzaima

@rak1507 the noteworthy difference is that "regularly" isn't "always", and then being similar to the other non-array-y code is much more important

but then again, there are things like SQL and regex which are practically DSLs embedded in languages

rak1507

I agree, but I was mostly talking about how people think there's no point in using anything from APL other than ⌈ and ⌊ but then go on to solve problems in a way which would be much nicer in APL

dzaima

10:16 AM

the big problem with arraylangs is that they don't necessarily solve any single problem. It's hard to consider them as full-fledged languages because they (supposedly) suck at managing large programs, data structures and control flow, and generally have bad performance when you aren't doing array-y things, but also really aren't useful as DSLs as you'd often need to go back & forth with the data being processed, at which point fancy array functions (like numpy) are just better

rak1507

10:27 AM

they definitely have issues with doing things that aren't array-y in nature, but for programs that do things that are array-y, like a lot of things on that blog for example, they're (imo) always much better, and it's a shame that people dismiss them and prefer to use other things

I'm not saying that people should use an APL for everything

dzaima

@rak1507 but it's hard to use it for only for some things. You rarely write a production program that just computes whether a given square is a magic square. And even for that you'd need some UI, and APL already flops on that

rak1507

@dzaima integration is definitely a problem, UI I don't think is though, depends on what the task is and the required output format

shakti can be called from nodejs and python which is cool, shame about the licensing requirements or whatever though

dzaima

regex and SQL nicely set their boundaries on what they do and what they don't. APL tries to do everything, but is only really good at array-y stuff

rak1507

agreed

4 hours later…

beagle3

2:35 PM

In my opinion it's more of a SEP-field. There's 4 or so very-different-paradigms that I'm familiar with: The dominant Algol world (C, C++, Java, Perl, Pascal, Basic, Python...); The meta-world of Forth and Lisp (and dylan, and scheme and ....); The array world (APL/J/K/...) and the logic world (Prolog, and friends, and in some ways SQL). A new language in the same world gets a "fair" evaluation. A language from an unfamiliar world is mostly ignored as fringe/crazy/weird regardless of merit.

JohnE

2:55 PM

The way I see it, APLs have two serious problems: commercial licensing, which curtails the kind of hobbyist/enthusiast tinkering which organically feeds innovation and new applications, and the fact that the maintainers of existing APLs stick to a purview which is traditionally suitable for APL. In my body of work I think I have produced some compelling proofs-of-concept that K is much more generally applicable as a language than the databases-and-CLI-apps cubbyhole it is shoved into

but to expand beyond those confines the languages need "batteries included" for stuff like graphics and audio IO, more general networking support, etc

dyalog and J seem the furthest along in that direction, but dyalog is commercial and J is, well... honestly it's just not easy to use

beagle3

@JohnE I have huge appreciation for your K work that I'm familiar with (iKe, oK and special K). But showing it to people produced the same SEP-field response that I get from showing "plain K" - which is why I suspect (lack of ) familiarity is a much bigger problem than applicability to a specific problem.

2 hours later…

JohnE

4:36 PM

sure, but these things feed into one another. A perception that something is suited only for a narrow domain breeds disinterest if you're outside that domain. A perception that something is general makes it appealing to a wider audience, with a wider set of needs and applications

1 hour later…