2022-06-28 01:17:12 <lilo_booter > hi all - i mentioned my rpn experiments with rust previously - took it a bit further and now have it running with wasm and xtermjs - http://geminidev.hopto.org/hettie/~pi/rrpn/ (enter help to see current dictionary)
2022-06-28 09:42:08 <lopata      > what's the difference btwn an eek and an erd ?
2022-06-28 09:45:30 <lopata      > 01 < g e^e^k  vs  < n e^r^d
2022-06-28 10:24:51 <josuah      > a few days ago, I asked about advises about fun Forth things here
2022-06-28 10:24:54 <josuah      > for publishing an article
2022-06-28 10:25:02 <josuah      > and the article is out: The Gopher Times 5
2022-06-28 10:25:50 <josuah      > - gopher://bitreich.org/1/tgtimes/
2022-06-28 10:25:52 <josuah      > - https://gopherproxy.net/bitreich.org/1/tgtimes/
2022-06-28 10:26:11 <dlowe       > Not gopher:// ?
2022-06-28 10:26:34 <boru        > Get with the times. Everyone's using gemini now.
2022-06-28 10:26:42 <dlowe       > lol that'll teach me not to read the backscroll fully
2022-06-28 10:27:32 <josuah      > boru: the topic of Gemini is very welcome here too
2022-06-28 10:28:16 <josuah      > dlowe: that was for whoever not having a gopher client at hand
2022-06-28 10:28:31 <josuah      > curl supports gopher:// since long :) and netcat too
2022-06-28 10:28:50 <josuah      > (just printf '%s\r\n' "$path" | the query in it)
2022-06-28 10:29:03 <josuah      > I suppose openssl s_client can do the same regarding gemini://
2022-06-28 10:31:36 <boru        > Should be fine, yeah.
2022-06-28 11:10:34 <dlowe       > got me to look at retroforth
2022-06-28 11:11:07 <lispmacs[work]> several gemini clients also support gopher, at least lagrange and elpher, anyway
2022-06-28 11:27:18 <lagash      > Nyxt supports both - I should know, I helped contribute! :)
2022-06-28 11:45:19 <dlowe       > retroforth's uses of anonymous functions (quotes) are pretty inspired. `[ 'true s:put ] if` is much prettier than IF..THEN metacompilation hackery
2022-06-28 11:47:51 <dlowe       > same goes for `[ ... ] repeat` and `[ ... ] while` - just simplicity itself
2022-06-28 11:49:41 <remexre     > yeah, I'm exclusively doing those too
2022-06-28 11:50:32 <remexre     > https://git.sr.ht/~remexre/rtf/tree/trunk/item/src/prelude.fth#L107 (and L160-192)
2022-06-28 11:58:24 <KipIngram   > Looks like there are numerous ways to avoid that IF ... THEN hackery.  I don't have those in my system either, but I don't do anonymous functions either.
2022-06-28 11:58:33 <KipIngram   > I do it all with conditional returns.
2022-06-28 11:58:44 <KipIngram   > And conditional recursion.
2022-06-28 11:59:07 <KipIngram   > And I'm convinced its considerably more "compact" in terms of overall code.
2022-06-28 12:01:09 <dlowe       > I could be convinced of this
2022-06-28 12:06:42 <eris[m]     > Hmm, i disagree that quotes are simpler
2022-06-28 12:08:01 <eris[m]     > implementationwise, that is
2022-06-28 12:13:47 <remexre     > I'm pretty sure simple { and } words are not significantly more complicated than IF and THEN, in terms of the metaprogramming involved
2022-06-28 12:13:55 <remexre     > (thinking of a DTC implementation, at least)
2022-06-28 12:14:50 <remexre     > : { ['] JMP COMPILE, HERE 0 , ; IMMEDIATE
2022-06-28 12:15:08 <remexre     > : } HERE OVER ! CELL + LITERAL ; IMMEDIATE
2022-06-28 12:15:18 <remexre     > something like that?
2022-06-28 12:15:27 <remexre     > (my actual implementations are more complicated and allocate)
2022-06-28 12:16:06 <dlowe       > if you wanted to make the compilation easy, you add an unresolved jump, compile the quote as usual, then compile the address literal, then resolve the initial jump
2022-06-28 12:16:16 <dlowe       > not the most efficient thing
2022-06-28 12:18:48 <remexre     > ya, that's what I did above
2022-06-28 12:19:16 <remexre     > tho what i did above doesn't quite work ig
2022-06-28 12:19:39 <remexre     > forgot the code field
2022-06-28 12:21:15 <dlowe       > it'd be better during runtime to be able to compile the quote recursively in its own area to avoid the jump but then you've ruined the simplicity
2022-06-28 12:21:55 <remexre     > ya, that's what i do in my real implementation
2022-06-28 12:22:11 <dlowe       > how do you handle nested quotes?
2022-06-28 12:22:35 <remexre     > compiler has dynamic memory allocation
2022-06-28 12:22:56 <dlowe       > yeah, that's a far cry from the usual bump allocation of most forths
2022-06-28 12:25:10 <remexre     > yeah, I'm doing something closer in complexity to qbe or llvm than to most forths; this forth is meant as a compile target as much as a user-facing language
2022-06-28 12:27:17 <remexre     > (and I haven't ever implemented a lot of "classical" compiler optimizations before, and kinda wanna try building some DSLs for them that compile to this forth; especially a small datalog engine)
2022-06-28 12:27:37 <dlowe       > nice :)
2022-06-28 14:41:59 <KipIngram   > eris: I've never done quotes, so I can't really comment.  It doesn't actually "feel" like it would be simple, though.
2022-06-28 14:42:58 <KipIngram   > dlowe: And yes, to the extent I have thought of it it seemed like putting it inline and jumping over it would be the simplest path to it.
2022-06-28 14:43:07 <KipIngram   > But slightly less desirable, as you noted.
2022-06-28 14:43:51 <KipIngram   > I guess it's probably not "too" non-simple, done that way.
2022-06-28 14:44:14 <KipIngram   > And compared with the usual IF THEN, maybe about the same?
2022-06-28 14:44:57 <KipIngram   > I think the conditional returns are even simpler - they're just ordinary words.  No extra hoop jumping at all to build stuff that way.
2022-06-28 14:45:04 <eris[m]     > inlining it and jumping over prevents certain things
2022-06-28 14:45:15 <eris[m]     > mm, true
2022-06-28 14:45:31 <KipIngram   > Conditional recursion takes just a tiny bit of extra diddling.
2022-06-28 14:46:16 <dlowe       > it's just a conditional branch back to the beginning
2022-06-28 14:46:22 <KipIngram   > You at least have to go find your most recently defined word.
2022-06-28 14:46:38 <KipIngram   > Yes, sure.  But you do have to go look that up.
2022-06-28 14:47:03 <KipIngram   > But any implementation of tail recursion requires that too.
2022-06-28 14:47:26 <dlowe       > Do you?  Just store the pointer to the head during compilation and use that
2022-06-28 14:48:08 <KipIngram   > Fair enough, but that's "extra."  Beyond standard compilation I mean.  It's barely any extra, I agree, but it is.
2022-06-28 14:48:28 <dlowe       > okay. I couldn't remember if that was already stored or not
2022-06-28 14:48:30 <KipIngram   > I'm picking nits.
2022-06-28 14:48:41 <dlowe       > there are a lot of little variables used during compilation like HERE
2022-06-28 14:48:48 <KipIngram   > Well, it's pointed to by CURRENT.  Or something related to it is.
2022-06-28 14:48:54 <KipIngram   > It's easy peasy to do.
2022-06-28 14:48:56 <KipIngram   > No doubt
2022-06-28 14:49:20 <KipIngram   > You have to have that info somewhere in order to be able to properly extend the linked list.
2022-06-28 16:42:18 <dlowe       > KipIngram: how do you do a multiple dispatch under that model?  Like I have something that can be four values and I want to do a different thing for each
2022-06-28 16:42:49 <KipIngram   > I'm not quite sure what you mean - you mean like a switch statement or something?
2022-06-28 16:42:53 <dlowe       > yeah
2022-06-28 16:43:44 <dlowe       > I'm not challenging the idea just trying to understand how it would work in practice.
2022-06-28 16:44:10 <KipIngram   > Yeah, I don't actually have a word for that, but if I were to add one it would take an integer argument and would just jump forward that many cells.  Most likely I'd factor that structure into a word of its own, and then would have the words in the option list do a double return.
2022-06-28 16:44:24 <KipIngram   > : foo switch bar bam boom ;
2022-06-28 16:44:43 <KipIngram   > bar, bam, and boom would double return, so bar and bam wouldn't fall through to the following options.
2022-06-28 16:45:08 <KipIngram   > swtich would just skip n cells.
2022-06-28 16:45:41 <KipIngram   > But there's not really an easy way that I've seen to do that with *clauses* of codes instead of single words of code.
2022-06-28 16:45:53 <KipIngram   > quotations might help with that, actually.
2022-06-28 16:46:17 <KipIngram   > Generally rather than try to define chunks of code anonymously, I just name them and have done with it.
2022-06-28 16:46:58 <KipIngram   > But I'd usually use my .: word to define them, which means I could later drop all the names from the dictionary, after I'd compiled the stuff that used them.
2022-06-28 16:47:20 <KipIngram   > .: does the same as :, except it sets a bit in the header.
2022-06-28 16:47:30 <KipIngram   > .wipe scans the dictionary and unlinks anything with that bit set.
2022-06-28 16:48:11 <KipIngram   > Forth encourages a lot of factoring, so I think it should offer an easy way to de-clutter the namespace.
2022-06-28 16:48:44 <dlowe       > I think overlays are the traditional forth mechanism for that
2022-06-28 16:48:57 <KipIngram   > Having the names around later, even if they're not in the lists, is nice for decompilation.
2022-06-28 16:51:02 <KipIngram   > Overlays would move the code in and out too, right?
2022-06-28 16:51:29 <KipIngram   > A good vocabularly mechanism helps too - you can sequester those "helper words" in a vocabulary.
2022-06-28 16:51:48 <dlowe       > sure, but you're talking about moving 1k of code in and out :)
2022-06-28 16:52:14 <KipIngram   > Sure - I'm generally ok with clever use of disk-resident stuff.
2022-06-28 16:52:34 <dlowe       > I've been going through the old Starflight source and it uses overlays a lot
2022-06-28 16:52:41 <KipIngram   > Forth's mechanism for that is fairly nice, actually.
2022-06-28 16:52:57 <KipIngram   > I tried really hard to write my  BLOCK word so that the "already resident" code path is very very fast.
2022-06-28 16:53:04 <KipIngram   > As fast as I could make it.
2022-06-28 16:53:12 <KipIngram   > And that part is a primitive.
2022-06-28 16:53:31 <KipIngram   > And I have a  LOT of disk buffers.
2022-06-28 16:53:47 <KipIngram   > Seems like a good way to use our modern plethora of RAM, at least in a desktop environment.
2022-06-28 16:54:43 <dlowe       > I was amused at the rant in Starting Forth against the terrible trend of storing code in named files of different sizes
2022-06-28 16:54:51 <KipIngram   > I needed to do a mod 255 operation in that fast part of BLOCK, so I went and found a slick way to do that that's faster than using %.
2022-06-28 16:55:24 <dlowe       > just bit-and by #ff, no?
2022-06-28 16:55:32 <KipIngram   > And yes, I could have made it be 256 instead, but that didn't work as well with my overall planned RAM management.
2022-06-28 16:55:45 <dlowe       > oh, 255. hm.
2022-06-28 16:56:05 <KipIngram   > It's certainly not as fast as mod 256 would be - you can hardly beat that.
2022-06-28 16:56:21 <KipIngram   > But it's 3-4 times faster than using idiv.
2022-06-28 16:57:17 <KipIngram   > But I'm keeping those 4k disk buffers in a 1 MB RAM page, and I needed that last 4k for something else related.
2022-06-28 16:57:41 <KipIngram   > I've got 255 4k buffers, and then 255 "buffer records" that keep up with what's in the buffers, their stale status, etc.
2022-06-28 16:57:50 <KipIngram   > When it's all said and done I waste 16 bytes of the 1MB page.
2022-06-28 16:58:13 <KipIngram   > I'm planning a bi-level RAM management scheme, where the "big blocks" are 1 MB.
2022-06-28 16:58:32 <KipIngram   > That's just going to be a fixed block size allocator.
2022-06-28 16:58:45 <KipIngram   > The level below that, though, will use ropes to issue out blocks of various smaller sizes.
2022-06-28 16:59:09 <KipIngram   > One of you guys put me onto ropes a few months ago.
2022-06-28 16:59:43 <dlowe       > yeah, ropes are very cool
2022-06-28 18:58:26 <pointfree   > Hello!
2022-06-28 18:58:34 <eris[m]     > hallo!
2022-06-28 19:35:17 <KipIngram   > Hi pointfree!