00:00:03 pikhq_: do you mean threads as continuations? 00:00:16 either is theoretically possible, but the second seems more likely 00:00:32 I said what I meant. 00:00:43 It's not really threads. 00:00:44 It's coroutines. 00:00:46 wow 00:00:53 where did it get the continuation library from? 00:00:58 Please. 00:00:59 He wrote it. 00:01:01 "Continuations as threads" is a really stupid way to say 00:01:04 "Coroutines with stack pointer". 00:01:11 And misleading. 00:01:22 (also, for an emulator, continuations sound like a really good way to do savestates) 00:01:36 in Underlambda, you can serialise continuations to stdout if you want to 00:01:44 and later read them in and just pick up off from there 00:01:53 That's pretty close to what the savestate support is, IIRC. 00:02:03 Well. 00:02:06 High-level wise. 00:02:15 It's in C++, so low-level, it's a fuckton of painful serialisation. 00:02:30 And then deserialisation. 00:03:10 how many languages have specified that all interps must be able to do an (interp-specific) serialisation/deserialisation of any subset of the entire program state, including a continuation? 00:04:04 Interp-specific? Lame 00:04:10 It should be the program format 00:04:20 that's one possible option 00:04:32 but that would ruin a lot of possible optimizations, I think 00:04:50 program format with an extra pair of parens is what derla uses, IIRC 00:04:51 but I'm not sure 00:05:28 Fuck optimisations, platonically perfect language 00:06:17 perhaps have two serialisation options 00:06:26 an optimised one, and a portable one 00:06:35 or just require all interps to be able to do it either way 00:06:47 :t forkIO 00:06:48 Not in scope: `forkIO' 00:06:53 and have an option where they just reformat serializations rather than load them 00:07:19 @hoogle fix 00:07:19 Data.Function fix :: (a -> a) -> a 00:07:20 Control.Monad.Fix fix :: (a -> a) -> a 00:07:20 module Control.Monad.Fix 00:07:31 ais523: Surely you can encapsulate a loading of a serialisation and then serialise that program? 00:07:32 Subprograms. 00:07:46 I don't think fix is actually in Control.Monad because I have it imported in ghci and it was not in scope. 00:08:01 CakeProphet: it's in Control.Monad.Fix, if you look more carefully 00:08:03 It says Control.Monad.Fix. 00:08:13 oh hey look at that it does. 00:08:20 elliott: hmm, you could write a small program to do it 00:08:25 :m + Control.Monad.Fix 00:08:27 ... 00:08:41 especially as deserialization never immediately runs what it deserializes 00:08:44 obviously I am not mentally exhausted right now. 00:08:44 ais523: how, if loading a serialisation ajumps into it? 00:08:45 ah 00:09:02 you can add an extra ^ to jump into it, as always 00:09:12 Surely you want to clear the stack first 00:09:21 (clear) dip call 00:09:24 or provide it with an argument 00:09:34 @hoogle IO a -> IO () 00:09:35 Foreign.Marshal.Error void :: IO a -> IO () 00:09:35 Network withSocketsDo :: IO a -> IO a 00:09:35 Control.Exception block :: IO a -> IO a 00:09:41 actually, continuations clear the stack themselves 00:09:44 it's a simple use of ultra 00:09:50 ultra + pop clears the stack 00:09:51 Oh, it's not just a function? 00:09:52 Gross. 00:09:54 hmmm, I thought void was in Control.Monad as well. 00:10:06 elliott: it's not a function, because continuations don't return 00:10:14 ais523: Yes they do, when the program executes 00:10:16 erm 00:10:17 you can make function-like not-quite-continuations pretty easily, though 00:10:18 ais523: Yes they do, when the program terminates 00:10:25 They return to the OS 00:10:33 what, really? continuations are like goto, not like gosub 00:10:34 No reason why an underlambda program can't be the OS to another 00:10:38 all things must eventually return. 00:10:40 oh, I see what you mean there 00:10:42 except for those that don't. 00:10:46 ais523: When a program using continuations exits, does your computer explode? 00:10:50 that's interesting and I should think about it 00:10:59 elliott: no, the process quits 00:11:01 ais523: Or, alternatively: How does a CPS program terminate? 00:11:04 and that's a different sort of level from other things 00:11:08 A CPS program /takes a continuation as an argument/. 00:11:19 That's the OS. 00:11:21 Or your shell, or whatever. 00:11:29 the thing is, Underlambda isn't CPS, and doesn't assume things written in it are 00:11:30 Not being able to pass your own is a needless limitation. 00:11:42 ais523: No, but call/cc is equivalent with CPS. 00:12:16 it'd be easy enough to make continuations that were return-to-argument-on-exit rather than quit-on-exit 00:12:21 Underlambda is implemented mostly in itself 00:12:29 CakeProphet: it is, but it's a new addition 00:12:37 if you have an Underlambda interp that doesn't do continuations, you can change it to one that does with a string replace 00:12:52 ais523: Yeah, but it's gross to have continuations that you can't control the exit behaviour of in the core anyway. 00:13:15 Being able to write an "Underlambda OS" in Underlambda would IMO be very much in line with its goals. 00:13:20 it would 00:13:20 (Without writing your own interpreter.) 00:13:25 I'm reconsidering my design a bit at the moment 00:13:26 oerjan: apparently too new for my version of ghci 00:13:31 ais523: But thankfully, they don't actually need to take a continuation. 00:13:34 They just need to return. 00:13:35 in particular, what exactly c should do 00:13:44 Any program that used continuations like it did before should work, unless it has dead code at the end. 00:14:06 I'm trying to work out what would happen if you just lopped off the Q at the end 00:14:23 hmmm okay so.... 00:14:29 ais523: well, you should maintain the stack 00:14:31 fix (void . forkIO) -- fork bomb yes? 00:14:32 at least IMO 00:14:41 so here's the basic operation: #XCx|=> X(X(x))x 00:14:43 ais523: in fact, a continuation might want to push the stack at the end of execution 00:14:51 hmm 00:14:52 what's X? 00:14:54 oh, hmm 00:14:57 anything 00:15:02 that's the rewrite rule that implements C 00:15:02 can you use special notations for your variables? 00:15:09 rather than mixing them with operations :P 00:15:26 elliott is probably my favourite boy name [/random] 00:15:27 which doesn't form continuations, but which is the basic operation that lets you implement them in user code 00:15:35 ok, umm, I have no idea how that rewrite rule works 00:15:35 tiffany: spelt like that? 00:15:40 idk 00:15:46 let me use square brackets 00:15:51 CakeProphet: the parent halts too quickly to call it that, i think 00:16:03 (also, # and | should be interpreted as ^ and $ in regexen) 00:16:15 #[X]C[x]|[X]([X]([x]))[x] 00:16:21 it's probably clearer without the brackets 00:16:45 START history 'C' rest END = history (history (rest)) rest 00:16:51 no, not history 00:16:53 stack 00:16:53 ais523: Why does the history matter? 00:16:54 Oh. 00:17:07 START stack 'C' future END = stack (stack (future)) future 00:17:14 ais523: Why not (stack future)? 00:17:15 yep 00:17:19 So you can determine the rest of program? 00:17:25 because that would automatically run future if you tried to evaluate it 00:17:28 Right. 00:17:48 fix (forever . forkIO) -- fork bomb yes? 00:17:59 ais523: That seems correct to me, then. When future runs off the end of the program, the continuation returns, if you called it. IMPORTANT NOTE: Any "exit program" functions have to be careful not to specify that they exit the interpreter, just the current program. 00:18:00 not entirely sure how forever :: m a -> m b 00:18:05 CakeProphet: better 00:18:16 "current program"? ouch 00:18:19 I guess b is just bottom? 00:18:24 CakeProphet: because it never returns anything, it can return any type :P 00:18:24 Well, it's not as bad as it sounds. 00:18:28 ah right. 00:18:31 ais523: The "future function" should be ((future)@), where @ basically means "in a new context". 00:18:31 I mean, such a concept doesn't exist 00:18:36 @src forever 00:18:37 Source not found. I've seen penguins that can type better than that. 00:18:37 unless you can somehow work it out using rewrite rules 00:18:42 oerjan: so then the fix is completely unecessary here. :P 00:18:44 forever = fix (>>) 00:18:44 there's no "in a new context", you manufacture the new context by hand 00:18:49 using dip, etc 00:18:57 ais523: I'm telling you to introduce one; all it does is run the program with a fresh, empty stack, then push the stack at the end as a single element. 00:19:06 CakeProphet: um no 00:19:07 The only relevant effect being that any "exit" instructions are defined to only halt the current program. 00:19:15 oerjan: oh right nevermind. lol 00:19:16 ais523: It's not the idea of a current program that's inelegant here, it's an exit instruction. 00:19:23 if that's meant to be implemented in the interp directly, it'd increase the complexity of the language by about 1000% 00:19:24 You can't just halt the program forever at any point in a more pure model. 00:19:36 ais523: I'm hinting slightly that you should have no exit program instruction. 00:19:42 yep, I get it 00:20:08 because if you envision an underlambda program in a higher context, "exit program" turns out to look something like "halt computer" 00:20:12 which is insane 00:20:26 oh, I changed the definition of C, it seems 00:20:32 it doesn't give you the whole stack, just the future 00:20:42 ah, so you have to copy the stack yourself? 00:20:43 makes sense 00:20:55 Cx => (x)x is reasonable, after all 00:21:02 as is 00:21:02 as grabbing the whole stack can be done as (~aa(n:^)~*(*)*^)n:^ 00:21:05 * oerjan notes that the e function was the only function that wasn't implemented in terms of itself in his unlambda self-interpreter 00:21:11 sQx = s(s)x 00:21:12 ais523: ouch :P 00:21:22 oerjan: what does e do again? 00:21:23 elliott: that's commonly abbreviated to A 00:21:25 so it could support what elliott said 00:21:33 oh, I see 00:21:35 elliott: exit 00:21:36 elliott: exit the program 00:21:36 exit 00:21:43 what does @ do again? 00:21:45 in unlambda 00:21:47 I don't expect people to type out (~aa(n:^)~*(*)*^)n:^ every time, that's precisely why you can write A instead 00:21:50 read character 00:21:56 right 00:21:57 especially as it's a useful and commonly used operation 00:22:14 OTOH, the fact that it can be implemented in terms of more basic instructions is good as it means interps in BF or whatever don't need to implement A themselves 00:22:21 apply the argument to i or v, according to whether it was successful 00:22:27 ais523: hmm, I'm wondering how an underlambda interpreter would do preemptive multitasking 00:22:37 the opposite of A, in a sense, is ('A_~)_~('^_^A)_~'^_ 00:22:56 ais523: you can't just divide programs into the first N instructions and the rest and selectively evaluate them, because (a) IO (not a huge problem), but (b) those N instructions could loop forever, or whatever 00:23:00 and lol, ~_~ '^_^ 00:23:07 -!- AndGregor has quit (Ping timeout: 245 seconds). 00:23:10 A grabs the stack into a single stack element; that operation, I, runs a program in a new context 00:23:32 so really, you'd want to be able to infra a continuation and have things just work 00:23:54 yep 00:24:18 ais523: hmm, if you _do_ want exit, there's an obvious way 00:24:20 so the problem with removing the quit from the end of a continuation, is that it'd go back to the last place you called a continuation, not the last I 00:24:35 #@x| ==> (empty string) 00:24:41 ais523: you just have to redefine |, somehow 00:24:58 well, the | at the end is one of the most sacred parts of Underlambda 00:25:00 as is the # at the start 00:25:08 they are non-negotiable, and changing them will make m4 cry 00:25:42 ...underlambda is written in m4? 00:25:49 oerjan: it's written in everything 00:25:59 the only way m4 can do input is the equivalent of C's #include 00:26:14 ais523: well, what I meant was basically 00:26:16 so in order to read an Underlambda program, you change the quote marks to # and | and then include it 00:26:20 you just have to redefine how much the exit instruction erases 00:26:24 yep 00:26:32 by having an operation (foo)L that runs foo "with the | after it" 00:26:37 as in, (foo)L -> foo| 00:26:58 so, my current definition of Q as a preprocessor rule is Q/(-)S(((!(a(:^)*):^)):)~^ 00:27:05 (the (-)S is there for debugging, I think) 00:27:16 and it's actually obliterating the rest of the program 00:27:28 /but/, all I need to do in order to make "new context" an instruction is to add an extra element in the moand 00:27:29 *monad 00:27:40 what is monadic about it 00:27:46 (you know the "you could have invented monads" thing? I did, several times) 00:27:50 we need to ban the word monad forever :( 00:28:00 well, take a rule like a/(-)S((a(^!^)a*a(:(^!^)a~*a~(^~!^)a~*a*~^)*(^!^))((e%)S))~^ 00:28:14 basically, each command is translated into what amounts to a monad action 00:28:19 -!- AndGregor has joined. 00:28:23 where the monad is similar to Cont 00:28:32 but it allows two forms of execution: running the program, and grabbing a copy of the program 00:28:35 fair enough 00:29:29 ais523: maybe Underlambda is the language that will finally work to do an esoteric OS in... designed to be implementable anywhere, persistable state, defined in terms of a very simple core, and now perhaps with the ability to run subprograms? 00:29:33 oh, and the ((e%)S) stuff happens when you try to run a command that needs arguments on an empty stack 00:29:41 I dunno how easy it'd be to implement in assembly with no memory management, though 00:29:47 elliott: yes, although I can't think of a way to do preemptive multitasking offhand 00:29:56 ais523: yeah, that's what I was thinking about 00:30:03 how easy do you think an implementation on the bare metal would be, out of interest? 00:30:07 the problem with Underlambda is not implementing it, but implementing it efficiently 00:30:19 an implementation that works but has exponential performance is pretty trivial 00:30:22 heh 00:30:41 ais523: hmm, preemptive multitasking seems like it must be a primitive, or else it requires Greenspunning the language _itself_ 00:30:49 Greenspunning? 00:30:58 ...tenth rule? 00:31:03 I think you can do it with preprocessor rules, anyway 00:31:06 elliott: aha 00:31:15 proof: you can't do "run N instructions", because of arbitrarily complicated loops within; so you need to split the loops 00:31:20 but then you have to replicate the control structure used 00:31:22 why don't you just implement Lisp /with/ Underlambda preprocessor rules? 00:31:30 mumble mumble, it requires a self-interpreter, Q.E.D. 00:31:34 ais523: I meant underlambda in underlambda 00:31:52 elliott: oh, you do that all the time, that's what the rewrite rules /are/ 00:31:57 and the reason that the core is so simple 00:32:09 so that you only have a relatively small number of commands to metacircularly redefine every time 00:32:36 e.g. the language doesn't require interps to have error-checking like checking for an empty stack, but you can bolt it on after the fact if you like 00:33:07 I fear Underlambda's a little too ambitious, I keep restarting it and trying to do it right this time 00:33:14 I'm already on version 2.0 after abandoning version 1.0 00:33:50 I have a naive implementation of core+preprocessor that I'm using to check if naive implementations of core+preprocessor are enough 00:34:15 written in rather outdated Perl (as in, it's trying to work around the lack of a language feature that has since been added) 00:42:07 -!- elliott has quit (Ping timeout: 250 seconds). 00:42:58 -!- ais523 has quit (Remote host closed the connection). 00:44:41 -!- Zuu has quit (Ping timeout: 260 seconds). 00:48:37 -!- ive has quit (Ping timeout: 250 seconds). 00:49:26 -!- elliott has joined. 01:07:24 -!- AndGregor has quit (Ping timeout: 258 seconds). 01:14:27 -!- AndGregor has joined. 01:25:33 Sigh 01:25:41 fizzie: Can I blame you for SOCK? 01:25:45 I need to blame someone 01:37:58 What the hell is OOBINLINE, anyway 01:38:05 Some weird protocol thing, looks like 01:47:24 -!- copumpkin has quit (Ping timeout: 240 seconds). 01:55:24 -!- monqy has quit (*.net *.split). 01:55:24 -!- sebbu has quit (*.net *.split). 01:55:24 -!- MDude has quit (*.net *.split). 01:55:25 -!- myndzi has quit (*.net *.split). 01:56:41 -!- augur has quit (Remote host closed the connection). 01:57:01 -!- yorick has quit (Ping timeout: 260 seconds). 01:57:08 -!- yorick has joined. 01:57:26 -!- monqy has joined. 01:57:26 -!- sebbu has joined. 01:57:26 -!- MDude has joined. 01:57:26 -!- myndzi has joined. 02:10:51 -!- copumpkin has joined. 02:15:46 -!- pikhq has joined. 02:16:02 -!- pikhq_ has quit (Ping timeout: 258 seconds). 02:23:13 olsner: Hey, I just realised something 02:23:25 olsner: There's already a type-safe Haskell assembler 02:23:25 -!- AndGregor has quit (Quit: Bye). 02:23:57 Harpy 02:24:13 I dunno if it can write the machine code to disk though 02:37:14 Deewiant: Wrt SOCK, should I do input validation on popping or after everything? 02:37:19 Please tell me it's the reasonable latter 02:40:21 -!- HackEgo has quit (Read error: Connection reset by peer). 02:40:59 -!- HackEgo has joined. 02:41:39 -!- derrik has joined. 02:42:09 -!- derrik has quit (Client Quit). 02:46:38 -!- augur has joined. 03:00:37 fork fork fork 03:00:40 forkity fork fork 03:01:16 do fork while forking forks forever true. 03:01:26 what 03:02:26 I am flabbergasted that I cannot talk about forks and be appreciated for my intellectual contribution. 03:02:59 me too 03:03:19 hi 03:03:53 fork monqy while monqy 03:03:58 hi 03:04:42 -!- BeholdMyGlory has quit (Ping timeout: 240 seconds). 03:07:19 -!- BeholdMyGlory has joined. 03:07:35 while (fork()); 03:09:35 Nooooooooooooooooooo 03:10:05 can someone say the digit one please 03:10:06 thanks 03:11:47 elliott: look in the topic 03:11:54 o thanks 03:12:00 there is an extra space in the topic 03:12:04 before two of the |s 03:12:05 fix plz 03:12:07 it upsets me 03:12:39 -!- Guardric has quit (Quit: Leaving). 03:12:42 -!- CakeProphet has quit (Ping timeout: 252 seconds). 03:13:24 sheesh 03:14:02 someone said there was a way to get irssi to put the topic in the command line for editing, but i've forgotten what it was 03:14:50 Ctrl+T? 03:14:52 just a guess 03:14:55 -!- oerjan has set topic: It is the 90s and there is time for the requirements of supervision and control of transmissible spongiform encephalopathies, also an Esolang event @ Hel/Finland on 3.10.2011: https://wiki.helsinki.fi/display/lambda/esoteeriset+ohjelmointikielet | I think pointers are considerably more useful than lambda calculus | 12345678 | http://codu.org/logs/_esoteric/. 03:15:07 now tab worked, i'm sure it didn't last time... 03:15:16 as in /topic? 03:15:28 Q: How do I easily edit existing topic? 03:15:28 A: /TOPIC 03:15:28 --http://irssi.org/documentation/faq 03:16:04 no, as in /topic ... and now it _stopped_ working again 03:16:08 -!- MDude has changed nick to MSleep. 03:16:56 hm, tab completing something else made /topic tab work again 03:18:23 heh 03:18:38 BAD: R received wrong data and/or misplaced it 03:18:38 BAD: R didn't push 6 03:18:38 BAD: trying to close socket with K reflected 03:18:39 nooooo 03:20:20 well thatsbdumb 03:20:42 i didnt realize i couldnt do /top 03:21:41 quintopia: yep, that's precisely what makes it _stop_ working :( 03:21:44 Deewiant: Hey, Mycology lies 03:22:40 after i do that, typing /topic explicitly no longer works until i complete something else 03:22:48 oerjan: i know. lets hack it and fix it! 03:23:33 ... 03:23:53 quintopia: /tope^W works XD 03:24:30 elliott: ? 03:24:31 so it does 03:24:47 i wonder if there is a patch for this already 03:24:54 have you asked in #irssi? 03:24:57 oh, so does /top 03:25:01 oerjan: re oerjan: i know. lets hack it and fix it! 03:25:31 elliott: yeah, quintopia is so naive 03:25:55 argh this is such a bad crap 03:26:03 i know its hard to hack the pile of spaghetti called irssi 03:26:21 but i think if i had a day, i could figure it out 03:26:25 quintopia: that's not the naive part 03:26:39 what is? 03:26:48 i'm sure elliott can explain 03:26:57 no, only zzo can 03:26:59 * oerjan whistles innocently 03:27:08 ask zzo 03:27:09 i... doubt zzo38 could 03:27:19 yes he could. quintopia just wouldn't get it. 03:27:45 well then, if i wouldnt get it, then i probably shouldnt care 03:30:00 argh what the FUCk is wrong with this 03:30:02 ogkfhlmkjhlkhpv;g 03:30:28 ok maybe you're meant to... pad it ou? 03:30:29 t 03:30:30 no 03:30:34 hmmm 03:30:47 jasiodjdfoidjgoidsf 03:31:05 hi 03:31:29 GOOD: sent "Foobar" with W 03:31:29 GOOD: W pushed 6 03:31:29 GOOD: closed socket with K 03:31:29 "9\"+55<" 03:31:30 okay that's 03:31:32 not right 03:31:58 oh my cology 03:31:58 OH GOD I' MSUCH AN IDIOT.... no wait no im not 03:32:08 yeah no im not an idiot 03:32:09 yeah you are 03:32:10 but what 03:32:14 copumpkin: :'( 03:32:15 <--- crying 03:32:17 okay no 03:32:26 sorry the crying doesn't stop for anyone 03:33:12 ok i'm reading wrong 03:33:44 i want to bite this laptop for not working 03:33:55 rational adult behaviour 03:35:01 i hear biting things is a common disease among horses 03:35:12 maybe im a hors 03:37:07 "< >:#,\n | goe\n#:<\"BA\n\">:#,_\nlects\"\n\"+_v#!\n0\"+<>:\n00#vp" 03:37:07 ok th- OH WHAT IS 03:37:08 WHAT IF 03:37:12 FUCKIN' 03:37:12 VECTOR 03:37:13 SHIT 03:37:20 popVecOffset 03:37:21 BAMBLES 03:37:37 that should really be a type error somehow. 03:38:26 also mööse. would MÖÖS be a good fingerprint name? 03:38:43 not four bytes 03:38:53 i think 03:38:56 i mean in latin-1, of course 03:39:02 ah. 03:39:06 nobody uses latin-1, oerjan 03:39:46 it's close to irssi's default CPsomething fallback. at least it contains those characters. 03:40:22 Windows-1252? 03:40:36 cp1252 03:40:38 hmm, STRN doesn't look... so hard 03:40:49 "cp1252" is ambiguous. 03:41:02 Multiple vendors have used "code pages". 03:41:04 well it's what my irssi setting has. 03:41:20 And the registered name for the charset is Windows-1252. 03:41:24 in any case, it's latin-1 + some extras 03:41:35 It's not a strict superset of Latin-1. 03:42:08 It's a superset of the printable subset of Latin-1. 03:42:09 -!- ive has joined. 03:42:15 whatever. everyone recommended i set my irc fallback to it, anyway. 03:42:34 pikhq: well duh, it's not like there's extra room 03:43:15 Well, presumably you IRC primarily in English and Norwegian. 03:43:31 pikhq: i had it recommended in this channel. 03:44:05 Both of which, IIRC, have Windows-1252 as the primary legacy charset. 03:44:26 it allows me to see some messages from people not properly utf-8. 03:44:47 -!- CakeProphet has joined. 03:44:48 -!- CakeProphet has quit (Changing host). 03:44:48 -!- CakeProphet has joined. 03:45:02 Yeah. 03:45:18