00:00:00 --- log: started forth/05.09.16 00:09:00 --- quit: sproingie (Remote closed the connection) 00:10:06 --- quit: Quartus (Remote closed the connection) 00:10:26 --- join: Quartus (n=trailer@ansuz.pair.com) joined #forth 00:15:19 --- quit: Quartus (Remote closed the connection) 00:24:44 --- join: Quartus (n=trailer@ansuz.pair.com) joined #forth 00:29:17 --- quit: Quartus (Remote closed the connection) 00:32:25 --- join: Quartus (n=trailer@ansuz.pair.com) joined #forth 00:37:29 --- join: amca (n=plump@as-bri-3-4.ozonline.com.au) joined #forth 01:05:04 Hello 01:29:27 --- join: aum (n=aum@60-234-156-82.bitstream.orcon.net.nz) joined #forth 01:30:07 --- part: aum left #forth 02:22:54 --- quit: amca (Read error: 113 (No route to host)) 02:22:59 --- quit: Quartus (Remote closed the connection) 02:23:32 --- join: Quartus (n=trailer@ansuz.pair.com) joined #forth 02:46:49 --- nick: Raystm2_ -> Raystm2 05:45:34 --- join: folajimi (n=folajimi@shell3.sea5.speakeasy.net) joined #forth 05:45:55 --- quit: folajimi (Client Quit) 05:53:00 --- join: aardvark (n=folajimi@shell3.sea5.speakeasy.net) joined #forth 06:00:17 --- nick: aardvark -> aardvarx 06:01:22 Morning, all. 06:03:17 --- join: tathi (n=josh@pdpc/supporter/bronze/tathi) joined #forth 06:10:41 --- join: madwork (n=foo@derby.metrics.com) joined #forth 06:27:17 --- join: ccfg (i=ccfg@dsl-roigw3de0.dial.inet.fi) joined #forth 06:27:19 morning 06:29:04 --- nick: docl_ -> docl 06:31:16 docl, morning. 06:33:37 hi aardvarx, what's up? 06:34:10 docl, have you ever used isForth? 06:35:37 Good morning, Men! 06:36:49 Morning, Ray_work . 06:36:54 no, I don't think I have tried it yet 06:37:00 k 06:38:10 hi Ray_work 06:40:45 * docl downloads isforth 06:45:19 aardvarx: got questions about isForth? 06:46:41 --- quit: ccfg (Read error: 110 (Connection timed out)) 06:47:52 tathi, I was just wondering who would use it (for what purpose). 06:52:51 ah. 06:53:12 Also, how can I find the quotient of two values declared using a #define statement without using the division sign (i.e. '/')? 06:53:28 I know it's OT, but I'm having issues with it right now. 06:54:05 err...the division sign is how you normally do division, what's wrong with that? 06:54:15 I assume you're talking about in C? 06:57:12 tathi, yes. 06:57:16 and...I440r doesn't like C or the normal C/unix libraries much, and considers direct-threading to be "the only true forth" implementation model. 06:57:30 I gather he couldn't find a tool that fit those criteria, so he wrote IsForth. 06:57:52 tathi, I do concede that it seems silly, but my experience with numerical analysis has influenced me to avoid the operation where possible. 06:58:17 What do you mean by "direct threading"? 07:00:07 basically each colon definition is a list of addresses of the words that it calls 07:00:32 http://en.wikipedia.org/wiki/Threaded_code 07:00:49 Thanks. 07:01:08 http://www.complang.tuwien.ac.at/forth/threaded-code.html 07:01:24 this one describes it more as Forth people usually see it 07:04:34 tathi, do you have an example of actual threaded code to illustrate what the Wikipedia article is referring to? 07:05:06 hmm... 07:06:29 No worries if you don't have any examples. 07:06:54 well, words in IsForth look something like this: 07:07:14 : ' defined 0= ?missing ; \ in forth 07:07:27 tick: 07:07:31 call nest 07:07:36 dd defined, zequals 07:07:40 dd qmissing 07:07:41 dd exit 07:07:56 if that helps... 07:09:04 'nest' is a routine that pushes the address onto the return stack, and then loops through the rest of the definition 07:09:22 grab address, branch to it, repeat 07:09:34 You mean all the dd* instructions? 07:09:45 oh, sorry. 07:10:03 yeah, that's an assembly directive, short for "define doubleword" or something. 07:10:05 I'm new to forth. 07:10:17 okay. 07:10:29 it outputs a 32-bit value for each argument. 07:10:49 err...everything starting with 'tick:' there was assembly language. 07:11:13 so the rest of the instructions are subroutines? 07:11:50 yeah, they would either look much like this one (i.e. colon definitions; call nest; dd ) 07:11:55 or just be straight assembly language. 07:17:54 --- join: PoppaVic (n=pete@0-1pool74-238.nas24.chicago4.il.us.da.qwest.net) joined #forth 07:18:10 Mornin' 07:18:14 yeah...I'm not explaining that very well. 07:18:18 Morning 07:18:44 How goes it? What am I missin'? ;-) 07:18:52 Basically, in direct threading, you have a bunch of addresses that point straight to actual machine code. 07:19:04 yah 07:19:13 With indirect threading, the addresses point to a header which points to the machine code, and may also have other data. 07:19:48 "to a header" (aka. "or some structure") 07:19:55 yeah 07:20:48 I think Gforth info-file and the FIG site cover the variants prettywell 07:21:28 now, don't forget "call-threaded", either 07:21:59 * PoppaVic thinks he recalls F-pc was "call-threaded" 07:22:19 also called "subroutine threaded" -- instead of addresses, you have actual "call" instructions. 07:22:26 right 07:22:33 takes up more space than direct threading (on many processors) 07:22:50 and also, surprisingly enough, is often slower than direct threading. 07:22:58 or, at least you jump into the addr which has a "call foo" - but can be anything thereafter. 07:23:06 right 07:23:35 Actually, I think the "slower than direct threading" thing is an artifact of x86's stupid "instruction-cache snooping the data cache" design. 07:23:46 likely 07:24:27 This is more than I can comprehend at this point :P 07:24:44 yes, it takes awhile - and maybe some drawings 07:25:26 Yeah, I thought there was a good web page describing it somewhere -- but I can't find it. 07:25:39 You're right though, the gforth info page isn't bad... 07:25:55 perhaps somewhere along the ring - forth info is spread all over between hell & gone. 07:26:36 aardvarx: info page for gforth under Engine and then Threading if you want to try a look there. 07:27:00 will do. 07:27:03 thanks/ 07:30:10 I think it's interesting that so many forths experiment with so many methodologies. Always seemed to me a tribute to the essential flexibility. 07:30:52 yeah. 07:31:19 Don't you think this is a case of "too many cooks in the kitchen?" 07:31:35 --- join: JasonWoof (n=jason@pdpc/supporter/student/Herkamire) joined #forth 07:31:35 --- mode: ChanServ set +o JasonWoof 07:31:36 and having (thanks to Quartus) been taking a closer look at the ANS standard recently, I have to say it's impressive how many options they manage to allow. 07:31:50 while still providing predictable behaviour so you can write portable programs 07:32:31 aardvarx: maybe.. But, I believe I can unequivocally point at linux for the worst offender. 07:33:07 --- join: sproingie (i=foobar@64-121-2-59.c3-0.sfrn-ubr8.sfrn.ca.cable.rcn.com) joined #forth 07:34:15 tathi: hmm. Perhaps. I think it's more a tribute to their organizational-skills. I'm never going to go nuts over ANS (why the hell does this always look like 'anus' to me? ;-), but it never hurts to have some starting-point. 07:34:49 So, no more ANS-heckling, PoppaVic ;) 07:35:24 dude, it's just that I get upset when folks rant over it, or ansi-C, or posix. Pretending that "the work is done". 07:35:28 yeah. I'm just saying, I always thought of ANS as specifying "you have to do things this way", but... 07:35:57 I'm coming to think of it as more that "however your thing actually works, you should provide an overlay with these semantics" 07:35:58 tathi: sure, it's more like guidelines and "to be 'compliant' behave like thus.." 07:36:05 right 07:36:44 Mind you, I've not written any executable test-code for this beast, but it seems likely we can sorta' embrace ANS 07:36:50 ..or areas of it, anyway 07:37:32 I just always thought that some of the things I like to do (like separate 'forth'/'macro' dictionaries) were incompatible with the spec, but now I think I could create an ANS layer over top of that model pretty easily. 07:37:33 What folks always need to recall is you can get a paradigm-shift that invalidates part of "the standard" 07:37:38 (for those who want it) 07:37:44 exactly 07:37:55 forth also ANS 07:39:57 or, in your case: "only retro also ANS" ;-) 07:41:19 PoppaVic, I gues when you put it in that context, I do agree that having people running around saying "IT IS DONE" is not helpful. 07:41:33 *guess* 07:42:20 sure. ANd, then we get the lispers in ##C: basically they want to throw out the baby with the bathwater, and we are all supposed to switch to lisp. 07:42:42 (or python, or whatever the hell they want to advocate) 07:43:13 language-wise, i can see their point completely 07:43:19 The *other* side of the coin is justifying learning 3...N languages, because "each language has something special" 07:43:28 but as a platform ... i can't think of a language with more runtime policy than common lisp 07:43:36 except perhaps java, and perhaps not even that 07:43:45 sproingie: "runtime policy"? 07:44:04 yeah, as in stuff the runtime library has to do 07:44:08 ahh 07:44:08 with C, it's pretty minimal 07:44:12 right 07:44:23 on the flip side, it's also stuff you're stuck with the runtime library doing 07:44:33 How about Forth? 07:44:34 I know they often debate wild, CS-issues about lisp-dev 07:44:52 forth basically has no runtime policy 07:44:53 I guess it's academic at this point... 07:44:56 aardvarx: well, look at the variations 07:45:22 brb - nature screams... 07:45:27 or safer to say that as an interpreter most of the time, forth has the policy of its interpreter 07:46:38 ok, gotta restart X to clear some droppings it left permanently on the screen. brb 07:46:54 --- quit: sproingie (Remote closed the connection) 07:50:20 Yeah, the way you use "policy", I suppose I'll buy that 07:51:48 tathi: the weird thing is, the wording in ANS makes it sound like it's telling you how to implement your system 07:51:59 Right. Exactly. 07:52:04 Well, that is bad. 07:52:33 aardvarx: I believe it's an artifact of std-docs. They can't just "suggest" in the doc. 07:52:35 That is the easiest way to alienate people like PoppaVic. 07:53:09 aardvarx: it's worse when adherents pretend those docs to be written in stone-tables with that semantic-view. 07:53:47 ..because THEN, instead of being "damned useful", the docs (and the fanatics) end up as a straitjacket. 07:54:12 JasonWoof: yeah, I've been looking at it more closely over the last couple of days, and I think they're actually just specifying semantics. 07:54:17 * aardvarx scoffs. 07:54:24 Who needs acolytes anyway? 07:54:46 It is kind of "how it should work", but I think I could overlay it on my little forth if I wanted. 07:54:56 Morning all. 07:55:01 Quartus, greetings. 07:55:02 hi Q ;-) 07:55:15 yeah, tathi I'm pretty sure it could be a layer or voc 07:55:22 I mean, I might have to replace some of my words with ANS versions, but I _could_ provide an ANS layer if someone wanted it (I think) 07:56:41 PoppaVic, it can. There's a "nearly-ANS' layer for F-PC, for instance. One of the few discrepancies there is that the : is its own interpretation loop; in Standard Forth that's not how it works. Unless you're doing that, it should be eminently possible to write a Standard compability layer. 07:57:03 --- join: sproingie (i=foobar@64-121-2-59.c3-0.sfrn-ubr8.sfrn.ca.cable.rcn.com) joined #forth 07:57:04 Quartus: ok, I thought so. 07:58:32 The Standard isn't an 'only this' specification; it's a 'must be at least this'. You can extend a system significantly and still be compliant. Quartus Forth has inline words, for instance; that doesn't in any way affect its Standard compliance. 07:58:54 Yep. I can envision that. 07:59:11 In some ways, it's superior to this "posix standard" noise. 07:59:36 Quartus, that definitely sounds much better to me :) 07:59:38 Further to your point about the speed of subroutine threading vs. direct threading, PoppaVic, as soon as you add even trivial inlining, you become appreciably faster than direct threading. 07:59:54 Quartus: of course. 08:00:13 Are you sure about that? 08:00:19 Yes. 08:00:25 Quartus: I imagine that 'see' will throw the whole mess up, however. 08:00:40 speed of direct vs subroutine has a bit to do with architecture 08:01:02 just about any modern arch does better with subroutine 08:01:03 aardvarx: sure. It's akin to writing nice, neat macros in C - that can expand to a lot of code right in-place. 08:01:18 It's not as bad as all that; decompilation is diassembly in a native-code system, but it's not hard to re-map short inlined sequences back to Forth words. That's what I do in Quartus Forth. It's not 100%, but it suffices. 08:01:28 I'm trying to reconcile that with what I am reading from Wikipedia... 08:01:37 Quartus: interesting 08:01:55 wikipedia's full of shit in a lot of areas. quartus is in a position to know the reality :) 08:02:41 That's true. 08:02:42 otoh, I always figure functions are always easy to find and read. And, I'm willing to take a bit of a speed or size hit for readability - trusting that another, finer-grained tool can make shit smaller or faster. 08:03:25 PoppaVic, I don't use SEE as a programming tool. If I suspect the compiler is buggy, I might, I suppose. 08:03:44 There are 38 'must be inlined' words in Quartus Forth presently; 24 of those are two-byte sequences that map directly back to Forth words. 08:03:47 Quartus: oh, it's sometimes handy to 'see' how some similar word did its job. 08:04:07 PoppaVic, yes, and you can still do that. 08:04:36 Quartus: for an interactive "development environment", 'see' is handy. Once you are prepared to distro some app, We prolly can live w/o this sorta' decompiling-info 08:04:38 And because it's a Standard system, you could always run a direct-threaded Standard system on the desktop and use a one-to-one SEE to your heart's content. 08:04:48 exactly 08:05:03 And then move your source back to a native-code environment when you want it to run at speed. 08:05:11 right! 08:05:54 It's really not difficult to read the SEE output in a native-code system, though. It does mean knowing assembler, but if you're working at that level you pretty much need to anyway. 08:06:16 And, at least in Quartus Forth, it doesn't use a lot of esoteric assembler. 08:06:39 Mostly moves and branches. 08:06:51 sure, but again: we are definitely speaking about "traditional do-all forth" versus "we know we can live in the former and recompile far faster/better" 08:07:45 Quartus, "esoteric assembler"? 08:08:09 Seldom-used or tricky assembler instructions. 08:08:39 aardvarx: bordering on "undocumented uncodes" or "hacks" 08:09:26 To me, native-code generation is the obvious choice. It prevents old-school Forth code modification (where values are stored directly into the thread), but I have yet to see an example of that technique that was the only way to achieve the desired result, or the least confusing of the available options. 08:09:27 PoppaVic, now that's just plain nasty! 08:09:57 If thread == subroutine, why not just say so? 08:10:09 Right, Q. There _could_ be a benefit, say... In cross-compiling to some embedded-device. 08:10:09 It doesn't, aardvarx, as such. 08:10:41 aardvarx: the definition of "thread" has changed significanly over the decades of micros 08:10:59 PoppaVic, I'd do native-code in an embedded environment too, but with less agressive inlining. 08:11:12 ..and most forth-docs are guilty of using that very term at least 2 ways. 08:11:32 Q: certainly, because space would be at a premium. 08:11:44 Right. But the speed advantages are huge. 08:11:51 absotively 08:12:12 Look at CONSTANT in Quartus Forth. 5 constant foo is exactly equivalent to : foo 5 ; inline 08:12:13 PoppaVic, what constitutes a "thread" and does it have any relevance to "multithreaded" applications? 08:12:29 This is one of the reasons I like the idea of this bytecode vm under-forth/over-C 08:13:09 aardvarx: in forth, they can mean a LL of some type, or a "task" or a real "thread", or almost any other perversion. 08:13:24 LL? 08:13:29 linked-lists 08:13:38 Ahhhh!!! 08:13:39 tathi: I wonder if you could write two versions of the words that are incompatible with ans. Have a sort of vecabulary or something. and it starts out with the ans versions, and if you want the cool ones, you have to do "set nocompatible" or something at the top 08:13:46 such as when they speak of word-threads in vocs 08:13:51 So that's what amca was trying to say yesterday!! 08:13:55 aardvarx, a thread, Forth-style, is a list of subroutine addresses. 08:14:24 Quartus: yeah, and I try to speak of LOL (lists of lists) and Lists 08:14:34 Although the _list_ itself is not a subroutine, right? 08:14:40 right 08:14:42 'multi-threading' refers to a form of multitasking, and is something else. 08:14:53 Thanks. 08:15:16 aardvarx: usually, the 'list' is either nodes-of-something, or we mean "steps to execute in a colon-word" 08:15:40 In fact each subroutine (word) in a threaded Forth system is either a primitive (machine code), or a list of word addresses. 08:15:48 exactly 08:16:18 and, bear in mind that "word addresses" is why they threw in that whacky XT term 08:16:31 This is how it's implemented, and depending on the exact mechanism it's either called 'direct' or 'indirect' or 'subroutine' threading. 08:16:47 In subroutine threading, you don't need what used to be called an 'inner interpreter' loop; the CPU does that for you. 08:17:07 And when you add optimization such as inlining to a subroutine-threaded system, you get what's called a 'native-code' system. 08:17:21 * PoppaVic enjoys watching the analysis and remembers the Old Days ;-) 08:17:26 Quartus, are you saying that they "are calling a rose by a different name"? 08:17:37 No; each one is a different flower. 08:17:41 k 08:17:50 the commonality is "flower" 08:18:08 They smell different though, right? 08:18:27 yep, and they can be internally quite dissimilar 08:18:33 Conceptually, a word is still built up from other words in a native-code system, just as it would be in a 79-vintage FIG system. : star 42 emit ; does the same thing, looks the same. 08:19:22 What still creeps me out is how Forth "compiles itself" 08:19:29 heh 08:19:30 thread in forth has the same connotation of thread in multithreading, but "threading" doesn't automatically mean "multithreading" 08:19:35 I find that Forth implementors are so proud of their conceptual grasp of the implementation method, that they're terribly anxious to 'expose the plumbing', as it were. Virtually every Forth primer opens the kimono in the first couple of chapters, which I think confuses the hell out of newcomers. 08:19:35 Sort of like "an uncreated creator..." 08:19:42 Metacompiling forth is very rare 08:20:00 Quartus: I think I'd agree 08:20:01 Quartus: yep. even starting forth does this. i think the sad reason is because there isn't much else to show by that point 08:20:18 sproingie: well, remember the age of that book 08:20:23 Quartus: you can only print lines and boxes of stars for so long before you wonder what it's good for. then you get to do it yourself 08:20:34 Hell, as Scott Adams once said, they don't just open the kimono, they do naked jumping-jacks in front of the window. 08:20:46 lol 08:21:02 if bigforth weren't such a buggy festering pile, it'd make a good tutorial, what with the 3d turtle 08:21:18 bigforth??? 08:21:20 I'd not wish Bigforth on other than an enemy 08:21:21 That's new. 08:21:29 aardvarx: ignore it 08:21:33 okay. 08:21:34 lol 08:21:37 bigforth's fast, reasonably complete, and got lots of neat features 08:21:42 Yet you won't find a book on learning C that talks about how symbol lists are implemented in the compiler. Or what the intermediate code generation looks like. 08:21:43 it also features a million ways to crash 08:22:07 Yeah, a great selling point. 08:22:12 Quartus: I agree, and - generally speaking - even forthers can avoid that for many moons 08:22:54 aardvarx, metacompilation isn't so hard to imagine. For instance, new versions of GCC are recompiled using GCC; no different. 08:23:09 What forth needs is a book with a decent number of examples that - as I often tell new C'ers - emulate handy programs. 08:23:15 * sproingie still isn't convinced how well forth syntax scales up to large scale control when the flow is very nonlinear 08:23:44 Quartus, I guess that's what happens with the LFS project. 08:23:45 sproingie, why do you think it wouldn't scale? 08:23:53 aardvarx - LFS? 08:23:59 linear flow, it's beautiful. : wash load fill agitate drain spin fill agitate drain spin end ; 08:24:01 sproingie: as near as I can tell, this is due to fanatics, haqueers, and the general issue of the best forths are for-pay. 08:24:12 Yet, I get the willies just thinking about life before the first compiler. 08:24:23 Quartus, LFS =LinuxFromScratch. 08:24:33 Ah. 08:24:34 nonlinear flow with branches ... well, reading an "if" in forth still tends to be pretty yick 08:24:41 You compile gcc from a host. 08:24:46 aardvarx: "adventures are the nonterrifying horrors others suffer" 08:24:59 sproingie, if it's a complex task I'd use a finite-state table, but then I'd do that in any language. Nobody likes nested conditionals. 08:25:06 However, the question remains: where the first compiler came from? 08:25:16 nope 08:25:21 I know where they came from 08:25:21 yep, if/then is an awful construct for actual complex state flow 08:25:34 It's like that chicken/egg routine... 08:25:38 aardvarx, the first compiler was a human being. Creating the first compiler program was a natural progression, to save the human all that work. 08:25:39 aardvarx: it took godlike patience and a lot of toggle-flipping. 08:25:42 "Who's on first" 08:25:43 nondeterminism, backtracking, polymorphism, that's where it's at 08:25:54 blech 08:26:02 * aardvarx shudders. 08:26:23 unfortunately, forth will give you the tools to assemble such tools, but no implementation of it 08:26:28 sproingie, what the hell is a nondeterministic program? Random-number generator strapped to an I/O port? 08:27:18 I go to a great deal of trouble to make 'deterministic' programs, I don't need help going the other direction. :) 08:27:30 Quartus: read up on "amb", lot of lisp and scheme books cover it 08:27:58 I belive a lot of it is academic and buzzwords 08:28:44 one semi-contrived pseudocode example of using it might be something like: session := verify_credentials(amb(ticket,smartcard,password,fail)) 08:29:06 Ah, ok. They're using 'nondeterministic' to mean something other than nondeterministic. 08:29:08 PoppaVic, I guess you'll be fine as long as you don't spew that sort of heresy in #lisp :P 08:29:10 hrm...looks like it means much the same thing as "functional" programming language. 08:29:32 aardvarx: I try to not "spew" and I've never bothered the #lisp'ers 08:29:33 Quartus: it's nondeterministic in the sense of an NFA -- there's still a programed path through it, but it's just not apparent from the text of the program 08:29:41 and yes, amb is deeply rooted in FP 08:29:43 specifying result conditions rather than how you get there. 08:30:21 err... s/result conditions/desired results/ 08:30:26 tathi: that's perhaps the most succinct explanation of declarative i've seen 08:30:34 amb is declarative conditionals, basically 08:30:41 tathi: as in a prototype? 08:30:43 Certainly achievable in Forth. 08:30:55 Quartus: everything is achievable in forth if you write it yourself 08:31:03 often it doesn't even take much code 08:31:11 Right. Easily achievable, I mean. 08:31:12 right, hence: I believe forth is never really developed. 08:31:24 but writing it yourself means debugging it yourself 08:31:37 ..Forth is a lot like a kid that never does more than toddle-around in diapers. 08:31:48 debugging a nondet or backtracking mechanism is wickedly hard 08:31:51 Forth/Forths 08:32:20 PoppaVic: it does more than toddle around, it does running backflips. but you gotta train it first 08:32:24 laboriously 08:32:46 right - almost like their "parents" prefer to keep them in diapers. 08:32:59 well, part of the reasons for this are obvious 08:33:16 1) There is no good, "standard" way to use libs 08:33:25 2) Everything is source, or not at all 08:33:57 I don't find it laborious. But I share source, and I re-use other people's wheels whenever possible. 08:34:03 yeppers 08:35:03 I got the feeling that "Bigforth" was supposed to cure the above, but just falls on its nose. 08:35:26 ..maybe it was just a vehicle to market a "for pay" system. 08:35:37 If you want esoteric programming features, you may be well-served to use a language expressly designed around them. Prolog, for instance, is probably useful to somebody somewhere -- but you wouldn't write a calculator app in it. 08:35:53 Quartus: actually prolog is exquisite for calculator type apps 08:36:14 Quartus: bad choice of example ... but yah i get the idea 08:36:20 Ok, that just illustrates my ignorance of Prolog -- let's say 'video game', then. 08:36:46 you might write the AI in prolog and possibly the menu. certainly you wouldn't write a sprite engine in prolog 08:36:57 Sure. It's not a general-purpose system. 08:37:00 you could, but the code wouldn't really use any of prolog's functionality 08:37:53 windows uses prolog to resolve dependencies in the network driver stack 08:38:08 it actually embeds a little prolog engine 08:39:34 unification, there's another exotic control mechanism i hadn't thought of 08:40:16 might be a little harder in forth, what with having a stack instead of bindings 08:40:50 I consider Forth an ideal base for algorithmic experimentation; when used well (short definitions, in-code testing) it debugs better than anything I've ever used. 08:40:58 ahhhh, shit. I just realized a use for "user variables". Sunnabitch. 08:41:42 * PoppaVic bangs his skull off the wall 08:41:50 my previous retroforth project is stuck because of the lack of exception handling words 08:42:15 I hate exceptions, I really do 08:42:16 i could use a global control flag and check it everywhere, but that's just nasty 08:42:33 i have a utf-8 decoder. it runs across a malformed sequence 08:42:39 what am i supposed to do, exit the program? 08:42:49 I've never seen a decent, sensible, readable exception-system 08:43:07 PoppaVic: you haven't looked very hard 08:43:10 How about Java? 08:43:10 I never see much of an issue in scream & die 08:43:17 aardvarx: like I said ;-) 08:43:23 yeah that's robust software 08:43:26 Hmmm. 08:43:28 PoppaVic, have you looked at Standard CATCH and THROW? 08:43:34 make the entire system die because of a malformed sequence 08:43:43 I thought it was the best thing going in terms of exceptions... 08:44:07 aardvarx: java's exception handling is bog-standard try/except, and frankly it ain't that fun 08:45:04 common lisp has fabulous exception handling 08:45:39 i could make it return a "success" or "fail" flag on the stack, but frankly that's just disgusting 08:46:05 with rf, i could screw with the carry flag, but that doesn't propogate well, anything could reset it 08:46:13 * aardvarx needs to read up on the need for exception handling anyway... 08:46:23 i just gave one example 08:46:50 decoding a utf-8 stream, come across an invalid sequence. communicate this to some caller that may be some arbitrary distance back 08:47:01 and allow it to recover 08:47:19 As opposed to just exit()-ing ? 08:47:38 yes, as opposed to having the entire operating environment exit 08:47:59 Fair enough. 08:48:00 invalid utf-8 happens every time you cat a binary file or even a slightly malformed textfile 08:48:12 I thought there was more to it than that for exceptions... 08:48:27 that's a simple usage of exceptions 08:48:33 Anyway, that is still nice to know. 08:48:38 sorry, back. *sigh* 08:48:42 * aardvarx heads out for food. 08:48:49 but the idea is to communicate an error state to a caller that may be some arbitrary number of levels back in the call stack 08:48:55 and allow it to recover from it 08:49:16 Q: no, all I've seen is code-bodies and ideas that really bug me. 08:49:18 recover != fix, right? 08:49:34 recover == do-something and resume 08:49:41 recover as in take some alternate action 08:49:49 and either exit gracefully or skip over 08:49:55 * aardvarx blows raspberries. 08:49:56 basically, recover as in not die horribly 08:50:03 I generally dislike the idea But, I'm beginning to think its personal. 08:50:09 only exception system i know of that can really "fix and retry" is common lisp 08:50:25 That's *exactly* what I am concerned about. 08:50:27 others have to fake it 08:50:35 That's just lame. 08:50:51 lispers would say so 08:51:13 "Fake it 'til you make it" sounds so wrong in this context... 08:51:13 exception handling means a lot of things. throw/catch is minimal 08:51:28 I *WILL* admit that these 'exceptions' might be preferable to every func testing returns/errno 08:51:28 YMMV, right? 08:51:42 PoppaVic: that's the purpose of exceptions 08:51:46 Minimal, but adequate in the vast majority of cases. 08:51:52 yup 08:51:54 bbiab... 08:52:02 but, as a rule: most exceptions-fanatics just don't WANT to consider tests. 08:52:13 (or so it seems) 08:52:14 tests and exceptions are orthogonal 08:52:48 exceptions are not for catching errors in code, they're for catching exceptional conditions that you expect to happen anyway 08:52:53 stuff outside your control 08:53:28 like I said, I've looked for a clean C-ish or forthish example, and nothing I've seen makes me want to embrace it. 08:53:29 there is a lot of java out there that DOES paper over poorly specified code with exceptions 08:53:42 but this doesn't malign exceptions, just the state of the engineering art in java 08:53:57 exceptions aren't to be embraced 08:54:07 I'd love to add it to the bytecode-vm, but I'm not sure what I'd want the guts to do - that isn't some noxious mess 08:54:08 they should be the exception, not the rule :) 08:54:32 anyway, i gotta get to work 08:54:35 * sproingie waves 08:56:33 See ya! 08:56:37 Quartus: I started to review dpans and throw/catch - looks like days of reading and study, so I'll have to pass for now. 08:56:51 THROW and CATCH aren't too tough; minutes, not days. 08:57:22 no, days. I'd have to integrate the concepts myself, and into the still-evolving vm 08:57:22 CATCH is like an EXECUTE that returns a flag. 0 if there weren't any THROWn exceptions, or the THROW code if one was thrown. 08:58:15 THROW unwinds the context (stacks, primarily) back to the most recent level of CATCH. 08:58:19 where doth the handlers reset the machine? 08:58:33 what do you mean? 08:59:19 I presume a handler will either repair the issue, or warmstart, etc. So it either has to get back to the thrower/tryer or restart at some-point. 08:59:31 It's all in how you build it. 08:59:40 yah, so also is my thought 09:00:12 --- quit: tathi ("out biking") 09:00:22 : foo ... ( some error that can't be recovered from at this level ) -n throw ; : bar ... ['] foo catch if ... perform recovery then ... ; 09:00:49 geezus keerist 09:01:12 In this case, say foo was sending characters to an output device. If it can't, it throws an error up to bar; bar does whatever is required, and then (possibly) retries foo. 09:01:46 Here's my issues... 09:01:55 1) I can envision a handler list; 09:02:11 2) "foo" throws ex#4; 09:02:30 3) we need to CALL upon the handling; 09:02:37 What does 3) mean? 09:03:07 3(just like any other opcode, colon-word) 09:03:15 what do you mean 'upon the handling'? 09:03:20 ahh 09:03:29