/g/ - Technology

Lisp, fags say 4th, niggerfaggots say neither

what happened to this board?

>low effort thread

>fucking retarded op trying to fit in

nice infix syntax there pleb

>>10790

it has always been dead

I'm bored.

>>10367

Forth has some nice features. The idea of organizing my code into "screens" of 80x26 characters is comfy. ThinkingForth or whichever book I was reading explains the theory of structuring a project into components, such that a component is comprised of a few screens of well-factored words. I could get to like that.

What I can't like about Forth is the speed. Forth will always be slower than C, because of its design. I was tempted to think that because it had that weird execution paradigm and was so fucking hard to use that it must be like a goddamn lightsaber if I stopped being such a microbrain and just learned to use it properly. A sophisticated weapon from a more civilized age or whatever. But, no; that's apparently a huge disadvantage. From what I've read, modern CPUs, with their branch prediction and whatnot, do not play well with Forth's "threaded" style of execution.

Not that I could find any recent benches, but in the '90s gforth was something like 6x C. Or at least, that's true of gforth. There are modern variants that are presumably less retarded if you want to drop $1k for a license to whatever give-a-shit dialect. But if you're doing that, it's probably for some legitimately useful purpose where the *size* of the code is critical, and not because you like dicking around with exotic read: shitty languages.

As for Lisp, SBCL benches around 3x C, IIRC. You can also have the compiler automatically vectorize arrays if you don't mind being explicit about types. I don't want to fuck around with memory-managed VM languages anymore, but SBCL can probably do whatever you need, short of producing an executable that isn't upwards of 30MB.

neither. neither are practical (i mean still much better than using mainstream PL). SML

>>10844

>$1k for a license to whatever

iForth's 100 USD. The optimizer's amazing and the package is good but it has like JVM-tier startup delays, and you can't fix that because they don't give you the source of the kernel (even after you pay). After the one-time payment, updates are free (but rare).

SwiftForth's 400 USD, it has a very simple but extensible native-code optimizer, and you get all the source. After the 400 USD payment you can pay 100 USD on subsequent years to get updates for a year. There's almost no reason to do this, just look at https://www.forth.com/swiftforth/revs/ , but you can skip years and re-up whenever.

But even after you consider these options, if you want anyone else to use your software ever, and not just accept compiled executables, then you need your apps to be compatible with gforth anyway. Which is an incredible pain.

conclusion: use APL instead. J's open source and really, really good. And although APLs have a much higher "this is unreadable holy shit" shock to the uninitiated, they--unlike Forth--become easier to read with skill, and when you come across someone else's APL you will--unlike with someone else's Forth--be able to understand it and get use out of it, rather than gag a bit before trashing and rewriting the thing.

Not lisp.

>>11247

I installed Factor yesterday! I'm gonna make it, fam!

>>11249

I was half-remembering VFS Forth, which goes for 379 USD, and 1232 if you want the "pro" version. According to its own benchmarks, it's pretty good:

>Test time including overhead (ms) VFX iForth SF3 SPF4 W32FS gff

>Time (ratio to VFX): 1.00 1.32 2.36 1.27 1.82 3.90

But yeah, unless I find myself dabbling in robotics, I don't think I have any legitimate reason to mess around with Forth and its Brainfuckishness. Even then, I'm skeptical I shouldn't just use FASM for that sort of thing.

J was the next thing I looked at in my pursuit of a useful yet esoteric language. Seems pretty solid.