00:08:17 _Pb [n=_Pb@75.131.194.186] has joined #scheme 00:08:59 -!- karlw [n=karl@adsl-99-157-202-134.dsl.pltn13.sbcglobal.net] has quit [Read error: 60 (Operation timed out)] 00:15:51 Might tasty hash stacks, Mrs McDunnough 00:23:54 jao [n=jao@214.Red-88-6-174.staticIP.rima-tde.net] has joined #scheme 00:24:16 -!- kenpp [n=kenpp@84.92.70.37] has quit [Success] 00:31:09 jengle [n=jengle@69.0.90.45] has joined #scheme 00:31:23 Zhivago, are you still around? 00:31:59 i guess i didn't need to ask that. 00:32:29 Hash stacks are one thing; a stack with hash is another, and a stash of hash is yet another. 00:33:15 well, I asked a question about grouping unrelated procedures into a header. zhivago replied that he doesn't do that. is it still generally common to save a useful procedure in its own file, ie. the filename "usefulproc.c" 00:33:41 I'm only curious because I want to know best how to store code for later use. 00:40:57 hash stash dude 00:42:49 :O 00:43:15 Charlie 00:43:21 hash stash Charlie 00:43:46 I'll have your hash; I love it! 00:44:18 Bombs, rope, lamp oil? 00:46:35 minion: memo for arcfide: Re your "Are they now?" reply to synx -- you should know better, that's a good nail he hit, and if you also attended Mitchfest, then this is what Danvy mentioned in one of his presentations (or maybe it was someone else who did). 00:46:36 Remembered. I'll tell arcfide when he/she/it next speaks. 00:49:48 What, nobody wants a nice stack with hash? I had the stack this morning, though without hash. 00:49:59 incubot: foo (+ 1 2) 00:50:00 Error: unbound variable: foo 00:50:03 michaelw, please explain how for a general procedure call, an extra procedure argument as a continuation can be `optimized away' to be just like multiple return values. 00:50:55 incubot: #|foo|# (+ 1 2) 00:50:56 3 00:52:00 incubot: #|foo|# (display 1) (newline) (display 2) 00:52:00 2# 00:53:18 *jcowan* still longingly craves the perfect way of providing extensible lexical syntax. 00:53:22 -!- jengle [n=jengle@69.0.90.45] has quit ["Leaving"] 00:54:47 jcowan: strings + macros. 00:55:02 Please explicate, eli. 00:55:23 (foo "arbitrary syntax here") 00:55:48 This is a syntax-rules world: taking apart strings is not an option. 00:56:16 And that's not lexical syntax, it's syntax: it only works when doing Scheme, not when reading arbitrary sexprs. 00:56:31 By "lexical syntax" I assumed you mean "concrete syntax". 00:57:13 No, I meant things like '#(...)' and "#e1.27" and "#s(cooby doo)" 00:57:32 That is what I mean by "concrete syntax". 00:58:22 Okay. So *the reader* is supposed to notice that (foo "foobarfoo") is lexical syntax? 00:58:37 I mean, when reading any datum whatsoever? 01:00:08 Macros work because they're expanded before evaluation -- if you don't have macros but you do have `eval' you can fake it by delaying evaluation. 01:00:15 If you want "reader macros" to work you need to run them before you construct the AST (aka `read'), which means that they cannot be plugged on top of the usual functions/macros system just like macros can't be. 01:00:17 A string version of `eval' can fake this too, and by using macros things are a little better. 01:03:02 I follow you up to the last point. 01:03:31 When READ reads something, it doesn't evaluate anything. The only way to make it do so is readtables, which means eval-when. 01:03:40 And I do so hate eval-when. 01:04:20 No, that doesn't mean EVAL-WHEN. It means that you need some mechanism of identifying a readtable at compile-time -- but it doesn't phase-mixed side effects, i.e. EVAL-WHEN. 01:04:57 Even if you have a readtable, the procedures it contains must be applied by the reader, which implicates runtime. 01:05:47 Anyhow, a full readtable is overkill. I only need extensible syntax, not arbitrarily redefinable syntax. 01:08:31 I don't know how readtables are related to anything here. 01:10:31 Because I want to be able to load a lexical-level module to change the interpretation of #-constructs, at least. 01:10:51 An (immutable) value isn't really first class until it has a datum syntax, and in the Lambda Order .... 01:16:32 And the essence of lexical syntax is that it doesn't just work when processing Scheme, but in all cases. 01:32:42 bombshelter13__ [n=bombshel@76-10-149-209.dsl.teksavvy.com] has joined #scheme 01:33:08 -!- bombshelter13_ [n=bombshel@76-10-149-209.dsl.teksavvy.com] has quit [Read error: 113 (No route to host)] 01:35:15 -!- sepult [n=user@xdsl-87-78-72-90.netcologne.de] has quit ["ERC Version 5.3 (IRC client for Emacs)"] 01:37:30 -!- Axioplase_ is now known as Axioplase 01:39:36 QinGW [n=wangqing@203.86.89.226] has joined #scheme 01:40:43 What is this sudden interest in standardizing first-class environments for Scheme? That's just stupid. 01:43:00 Time Suck^W^WTom Lord, that's all. 01:47:33 saccade [n=saccade@65-78-24-131.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has joined #scheme 01:49:30 -!- bombshelter13__ [n=bombshel@76-10-149-209.dsl.teksavvy.com] has quit [Read error: 113 (No route to host)] 01:51:14 I read "Time Lord". If only the Doctor could tell us what the next Scheme standard was like, we'd save a lot of time. 01:51:32 (*useless whovian interventio*) 01:51:54 We'd still argue. 01:52:05 I mean, you have to be pretty @#$* important in the Lisp community to have a one-letter signature, no? 01:52:26 Just like people now say R6RS was a mistake, they'll say R7RS _will_ _be_ a mistake and work to change the future. 01:53:48 *jcowan* busily builds wicker men against that day. 01:54:09 jcowan: err, are you talking of my signature? 01:54:20 Tom Lord's. 01:54:33 "-t" 01:54:48 ha. It never caught my eye. 01:55:06 Shouldn't that particular letter be reserved for the current active maintainer of T? 01:56:20 If there were one, perhaps. 01:56:29 *jcowan* tries to find an email from Axioplase, fails. 01:57:50 So what is your signature, then? 01:57:58 newline? 02:01:48 jcowan: "P!" 02:02:04 jcowan: but it's not the "-- \n" kind of signature though. 02:02:13 Neither is his. 02:02:14 Thanks. 02:02:24 Axioplase: the rest room is over there -> 02:03:16 -!- saccade [n=saccade@65-78-24-131.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com] has quit ["This computer has gone to sleep"] 02:03:32 foof: I beg your pardon? 02:04:12 it's a pun, Axioplase. 02:04:30 the name of P in English = "pee" = euphemism for "piss" 02:04:44 "rest room" being a euphemism for W.C., which is a euphemism for shitstool. 02:05:04 jcowan: "load a lexical-level module to change the interpretation of #-constructs" doesn't make any sense at all if you're talking about concrete syntax. 02:05:27 -!- blackened` [n=blackene@ip-89-102-28-224.karneval.cz] has quit [Remote closed the connection] 02:05:53 oh. Well, I was not going to grasp this pun. 02:06:01 IOW, I wish to be able to load an array module which suddenly allows #nA to create a literal array. 02:06:46 Since #A has no meaning in standard Scheme, the lexical syntax (as defined in ch. 7 of R5RS) must be changed. 02:07:15 However, I have no itch to change the meaning of "(car 32)" 02:07:21 rcy [n=rcy@S010600179a26267b.wk.shawcable.net] has joined #scheme 02:07:45 jcowan: If that's what you want, then fix the concrete syntaxes you want to use, read them as S-exprs like (#%hash 3 A ), and leave the rest of the work for macros and/or functions. 02:08:19 What happens to '(a b c (#%hash 3 A ))? 02:08:41 -!- luz [n=davids@189.122.90.116] has quit ["Client exiting"] 02:09:49 jcowan: Nothing. You need to deal with that too, of course. 02:10:21 How? If READ reads this in, it should return a list containing three symbols and an array, not three symbols and another list. 02:10:46 That's what I mean by not being quite first-class if you don't have a datum syntax. 02:11:05 "need to deal with that" = "define quote to do so". 02:11:44 I don't see how first-class-ness is related. 02:11:51 That won't do it. You have to redefine the behavior of READ directly. 02:12:25 To be first-class is to have most-favored-datatype status, the status of pairs or numbers or vectors. 02:12:59 Redefining quote doesn't help when you are reading sexprs from a socket and interpreting them as commands, e.g. 02:13:26 Your `quote' example is a good indication for the sense you're not making by asking for a lexically scoped reader syntax. 02:13:38 You want it to work in quote -- and there is no lexical syntax there. 02:13:58 You want it to also work with `read', and there is certainly no lexical scope there too. 02:14:16 (s/lexical syntax/lexical scope/ on the previous one too.) 02:17:13 peter_12 [n=peter_12@S0106000f66bc42b9.vn.shawcable.net] has joined #scheme 02:17:35 *eli* is still waiting for some clarification... 02:18:31 With all those flames about small/big Schemes, I wonder if we shouldn't just allow implementers/users of currently "small" schemes do their blabla on Thing One, and the others on Thing Two. At least, that should limit the fruitless debates. They can do whatever for Thing One, I'm just interested in Thing Two (which I want to be competitive and practical to use from scratch for most projects) 02:18:47 OmniMancer [n=OmniManc@202.36.179.65] has joined #scheme 02:18:57 You underestimate the capacity of Schemers for fruitless debates under *any* circumstances, Axioplase. 02:19:59 is (define (name . args) body) how the equivalent of &rest parameters are done in scheme? 02:20:04 Yes, OmniMancer. 02:20:28 Riastradh: I'm glad I'm not here when they have to decide of the color of the walls from the kitchen. 02:20:30 Axioplase: Many disagree for the need for two separate standards, and regardless the standards are related - thing two is a superset of thing one. 02:20:31 is there anything like keyword args? 02:21:05 foof: No, it's not a strict superset -- at least not according to what they said. 02:21:08 Not built-in or conventional, OmniMancer. 02:21:17 OmniMancer: You can implement a macro to do that, see let-keywords in Gauche. 02:21:36 eli: They said the reverse, that thing one is a subset of thing two. 02:21:59 foof: I'd define thing two first. And then cut off things (and probably patch a few others) until I get something that looks like R4RS. 02:23:46 foof: I didn't get the point of a strict subset -- and the only clarification they gave in this regard (and pretty much the only concrete point) is that the big version can require unicode and the little one not. So it's only a "subset" in a particular and possibly not too useful sense. 02:24:14 -!- jeapostrophe [n=jay@69.169.141.110.provo.static.broadweavenetworks.net] has quit [] 02:24:45 It's certainly not a subset in the sense of simply limiting the number of available procedures, because there will be semantic differences. 02:26:01 steiger_ [n=steiger@189.105.53.164] has joined #scheme 02:26:02 Right, which is why "subset" here is misleading. If you write a program that works with unicode-less (r5rs-style) strings, then switching to a unicode language may very well break it. 02:26:06 But if you leave some things unspecified in small scheme (like how much of the numeric tower is implemented, and what the range of characters are), then large scheme can mostly be defined as a backwards-compatible extension. 02:26:14 hey guys 02:26:33 i wanna paste some code for you, where's the best place? 02:26:36 lisppaste: url 02:26:36 To use the lisppaste bot, visit http://paste.lisp.org/new/scheme and enter your paste. 02:26:38 Doesn't work with strings. 02:26:42 steiger_, see the lisppaste URI. 02:26:47 ty 02:28:04 steiger pasted "untitled" at http://paste.lisp.org/display/86685 02:28:45 so, i can make no sense of the 2 last lines 02:29:02 why does it evaluates to this (shared ...) thing? 02:32:08 QinGW1 [n=wangqing@203.86.89.226] has joined #scheme 02:33:49 steiger_: cyclic list 02:34:17 sepult [n=user@xdsl-87-78-72-90.netcologne.de] has joined #scheme 02:34:36 -!- underspecified [n=eric@softbank220043052007.bbtec.net] has quit [] 02:34:59 It says "I call -1- the list (a a), and then return the list (-1- -1-)" 02:35:07 jeapostrophe [n=jay@69.169.141.110.provo.static.broadweavenetworks.net] has joined #scheme 02:35:21 Err, not cyclic here in fact. Just that it contains twice the *same* pair. 02:35:49 It's like (let ((one (list 'a 'a))) (list one one)) 02:35:56 Axioplase: yeah, i understood that. but why: 02:36:47 -!- rouslan [n=Rouslan@unaffiliated/rouslan] has quit [Read error: 60 (Operation timed out)] 02:37:36 Axioplase: .. but why does it contain the same list "object"? how can i make it evaluate to copies of the list? 02:38:16 they are the same list because you put the same list in both slots 02:39:17 (define (list-copy l) (reverse (foldl cons null l))) 02:39:21 OmniMancer: if i do (bm-aux 2 'a) it creates (list 'a 'a) 02:39:53 rudybot: eval (eq? 'a 'a) 02:39:54 synx: your "http://paste.lisp.org/display/86023,1/raw" sandbox is ready 02:39:54 synx: ; Value: #t 02:40:02 GeoffSK [n=chatzill@c210-49-251-180.kelvn1.qld.optusnet.com.au] has joined #scheme 02:40:04 rudybot: eval (eq? (list 'a 'a) (list 'a 'a)) 02:40:04 synx: ; Value: #f 02:40:27 oh, i see, synx 02:41:09 rudybot: eval (shared ((one (list 'a 'a))) (eq? one one)) 02:41:09 synx: error: reference to an identifier before its definition: shared in module: 'page 02:41:16 bah 02:41:19 rudybot: eval (let ((one (list 'a 'a))) (eq? one one)) 02:41:20 synx: ; Value: #t 02:41:31 it evaluates to (shared) to represent the lists are equal, and it is not necessary for the symbol 'a, since (eq? 'a 'a) is true 02:41:34 got it 02:44:10 -!- _Pb [n=_Pb@75.131.194.186] has quit ["Leaving"] 02:47:37 steiger_: shared seems to be no more than LET. 02:47:52 jengle [n=jengle@69.0.90.45] has joined #scheme 02:47:53 -!- dmoerner [n=dmr@89-151.res.pomona.edu] has quit ["Leaving"] 02:49:25 so, it creates a list with the same object twice (same value, same pointer. same everything). So it has to be EQ?ual. 02:49:59 i see 02:50:15 i'm doing a different-objects version 02:50:31 ty Axioplase , synx , OmniMancer 02:50:59 -!- QinGW [n=wangqing@203.86.89.226] has quit [No route to host] 02:53:49 tjafk2 [n=timj@e176199186.adsl.alicedsl.de] has joined #scheme 02:53:54 to install a PLaneT package I can do (require (planet dherman/json:1:2/json)) but that seems like a crazy bit of code to have in an actual application. In the application it would seem to make more sense to have (require json). 02:54:39 Which json? 02:55:30 eli, by "lexical syntax" I mean what R5RS means by it, which is only indirectly, if at all, connected with notions of lexical scope. 02:56:00 dmoerner [n=dmr@89-151.res.pomona.edu] has joined #scheme 02:56:05 -!- timchen119 is now known as nasloc__ 02:56:19 Hmm, my bad: it actually says "lexical structure". Anyhow, that's what I mean. 03:00:00 np steiger 03:01:25 peter_12: ideally it should just be (require (planet dherman/json:1)) or (planet dherman/json:1/submodule) for some submodule. 03:01:50 peter_12: The idea is major:minor, where minor versions change, but have the same interface (so won't break your code (but may be buggy)) 03:02:21 so author/package:major:minor/submodule is the full syntax, but just author/package:major should be sufficient for most applications and packages. 03:03:04 dherman is a dummy head though, and doesn't rename "json.ss" to "main.ss" so you need to add a /json at the end. 03:03:32 -!- OmniMancer [n=OmniManc@202.36.179.65] has left #scheme 03:04:35 (require (planet "json.ss" ("dherman" "json.plt" 1))) should work too if you want something in lists not weird syntax. 03:05:55 -!- jengle [n=jengle@69.0.90.45] has quit ["Leaving"] 03:08:58 -!- tjafk1 [n=timj@e176221231.adsl.alicedsl.de] has quit [Read error: 110 (Connection timed out)] 03:09:12 ...heh. I don't think the issue was about the syntax, synx. I think the issue was about relying on some remote database across the intertubes. 03:09:52 peter_12: You can use (planet dherman/json) to get whatever is the recent version -- but that's considered bad style since new versions can break the API. 03:10:29 Axioplase: `shared' is more like `letrec' that can work with values too, for example (shared ([a (cons 1 a)]) ...) 03:11:24 synx: No, dherman is not a dummyhead -- you cannot have a toplevel main.ss that makes sense of a (planet dherman). 03:12:29 jcowan: That's not good enough when you want to talk about concrete syntax extensions -- because rnrs has the advantage of using a fixed concrete syntax. 03:12:37 s/advantage/"advantage"/ 03:13:13 Quite. 03:13:44 -!- jeapostrophe [n=jay@69.169.141.110.provo.static.broadweavenetworks.net] has quit [] 03:13:45 So how can we move from a fixed to an extensible (but not necessarily redefinable) syntax without giving up the actual advantages? 03:14:07 You can't do that in any obvious way. 03:14:39 Like you said, the thing I suggested works only with considerable effort of redefining how `quote' and `read' work. 03:14:59 If you don't want that, then you have no way of extending it like you want. 03:15:49 bombshelter13_ [n=bombshel@76-10-149-209.dsl.teksavvy.com] has joined #scheme 03:16:06 The PLT `#reader' thing gives you support for per-expression custom readers, localizing the extension in a way that avoids the `eval-when' nightmare, but it looks like this is more than what you want. 03:16:44 rouslan [n=Rouslan@pool-71-161-75-11.cncdnh.east.myfairpoint.net] has joined #scheme 03:17:29 BTW, at least some of the SC members share the opinion that concrete syntax extensions should be the next logical step for "Scheme" to tackle. 03:18:03 But given the mess that this seems to be headed at, I doubt that anything like that can even be discussed. 03:20:22 eli: I meant (planet dherman/json) not (planet dherman/json/json) as is required by his package. 03:22:49 synx: OK, in that case you (or peter_12) should ask him for it, I'm sure that he'll either add it or explain why he doesn't want to do so. 03:23:17 Redefining quote is neither necessary nor sufficient, because read doesn't know what quote is. 03:23:42 jcowan: I did say "`quote' *and* `read'" 03:24:44 Changing read alone is necessary and sufficient. Hence the connection to readtables. 03:25:01 But it has to be a version of read that can invoke arbitrary procedures, unlike normal Scheme read. 03:25:07 Like SRFI-10, but less limited. 03:25:20 (a) Only if `read' is also used to read your code; (b) you wanted an extensible construct. 03:25:24 That's all fine. The trouble with SRFI-10 is that you have to do a *run time* thing to affect *lexical structure*. 03:25:43 (a) yes, (b) that *is* what I want, unless we are talking past each other again. 03:26:05 eli: I think it's an aesthetic thing really. I don't seriously think dherman is wrong for calling it "json.ss" not "main.ss". In fact, planet packages have a way to specify which file is the main one. It's just ignored at the moment... 03:26:27 There is no way to talk about a "runtime" without it being the runtime that happens at read-time. 03:27:23 synx: Well, IIRC, the convention used to be foo/foo.ss as the main file, and that changed with v4. My guess is that he's still using the v3 convention for some reason. 03:29:47 oh, well my bad then 03:31:27 Nothing bad, really -- it's a valid point, which I missed since I didn't see the original require line. 03:34:32 eli: Right. But if we are willing to accept only certain special cases, perhaps we can avoid requiring full access to every procedure at read time. 03:35:29 Seems that you have some specific meaning for "extensible" which is not the meaning I use. 03:35:46 underspecified [n=eric@walnut.naist.jp] has joined #scheme 03:36:57 By extensible, I mean that existing concrete syntax may not be altered, but new concrete syntax may be recognized at read time. 03:38:56 If the "new concrete syntax" includes arbitrary constructs then the problem is still there, and no way to get better than things like the solutions I suggested, or hacks like srfi 10. 03:39:41 QinGW [n=wangqing@203.86.89.226] has joined #scheme 03:42:40 -!- sepult [n=user@xdsl-87-78-72-90.netcologne.de] has quit [Read error: 104 (Connection reset by peer)] 03:43:01 -!- peter_12 [n=peter_12@S0106000f66bc42b9.vn.shawcable.net] has quit [] 03:44:17 -!- MichaelRaskin [n=MichaelR@195.91.224.225] has quit [Remote closed the connection] 03:45:16 peter_12 [n=peter_12@S0106000f66bc42b9.vn.shawcable.net] has joined #scheme 03:45:47 "jcowan: Which json?" system json the one that is installed. Just like which modules are installed in mzscheme/collects 03:46:50 synx: "but just author/package:major should be sufficient for most applications and packages" Do you mean it works or you would like that it works? 03:46:59 what if two people make a system json, peter_12? 03:47:18 synx: two people cannot make a system json 03:47:27 the system admin does that 03:47:44 What if you want to use two json libraries? 03:48:05 -!- dmoerner [n=dmr@89-151.res.pomona.edu] has quit ["Leaving"] 03:48:07 Or what if you have two projects that both use a different "logging" package? 03:48:21 logging1 and logging2? arbitrary and vague. 03:50:15 synx: This fixes problems. I understand why it works the way it does. However, it also creates equally bad or worse problems. 03:50:16 What if someone makes a smtp server, but leaves it stagnant, and someone else wants to do it better? 03:50:28 So... that's why you have author/package and not just package. 03:50:45 and :major is convenient because then you don't have to make a whole different package every time you change the interface. 03:50:47 peter_12, what's the problem you're trying to solve by writing (require json) rather than (require ...planet mumble json fnord...)? 03:51:58 I thought perhaps you wanted to distribute some code that would be independent of the remote database called PLaneT all the way across the intertubes, but I suppose that's not the case. 03:51:59 Riastradh: It is a good question. I don't think I have a solid reason except that it makes me worry. 03:52:06 Worry about what...? 03:52:28 I'm so worried about...I'm so worried about...the baggage retrieval system they've got at Heathrow...! 03:52:31 That package management becomes part of my application 03:53:07 that seems like a dangerous convolution 03:53:26 If module A uses C v0.1 and B loads C v0.2 and module C of any version can only be loaded once then that screws up everything 03:53:51 Who says only one version can be loaded at once? They're different modules. 03:54:05 But in any case, there are two separate issues here. 03:54:09 zbigniew: hey. I just tried scrmable.scm from your site. It fails with `Error: unbound variable: start.' Is it a known bug? (I've never used csi thouroughly, and just need to check that the package I'm using is properly working) 03:54:13 Riastradh: that is true for many modules (functional ones especially) but some modules wouldn't work that way. 03:54:18 module C v0.1 and module C v0.2 can both be loaded simultaneously... in theory at least. That might lead to trickiness if global side effects in module though. 03:54:20 One is that PLaneT serves as a naming authority. The other is that PLaneT is a component of a (very simple) distributed system. 03:54:23 dmoerner [n=dmr@89-151.res.pomona.edu] has joined #scheme 03:54:46 Like if you're using two database modules that both open a database file twice... 03:55:00 synx: Sure. System things like that. 03:55:08 I sure hope that I don't run into systems that work like that, synx. 03:55:24 luckily modules with side effects like that are vanishingly rare and would be ridiculous to upload on planet. 03:55:46 I think the point for me is that the PLaneT system is complex and doesn't solve all the problems so I'd rather take simple 03:56:19 synx: I don't think they would be ridiculous at all. 03:56:24 peter_12, I believe you can always take a PLaneT archive, extract it, and bundle it with your own source code. 03:56:34 -!- QinGW1 [n=wangqing@203.86.89.226] has quit [Success] 03:56:39 Riastradh: I think I will be doing that. 03:57:03 Provided, of course, that the archive from PLaneT be available under a free software licence compatible with that under which you intend to release your own source code. 03:57:06 Here's a problem peter_12. component A uses module v0.1, component B uses module v0.2, but you can only specify (require module) in your source code, using "planet-install module-v0.2" on the command line to install a package. 03:57:10 I know CPAN has its problems but I like how that works better in general and I am no Perl fan. 03:57:31 So you can never use component A and B together, since you can't specify which version of the module each component is designed to use. 03:57:49 How does CPAN work differently, peter_12, that enables it not to exhibit problems that you're concerned about? 03:58:20 if you install module v0.1, component B breaks. If you install module v0.2, component A breaks. 03:58:46 Riastradh: No. CPAN has different problems. It is up to the system admin to make sure the installed/available system modules can work together. That seems like the right location for the burden. 03:59:10 Or if you're using apt, component A blocks component B from being installed, thus leaving you hanging on the new version of various software, because one stupid developer won't update to use the new libraries. 03:59:16 Requiring the involvement of a system administrator makes it very hard to ensure that software works the same on two different systems. 03:59:20 But CPAN is very simple and PLaneT is much more complex (allowing for multiple simultaneous versions installed) 03:59:44 Riastradh: that would be a strike against Debian and that system is very well integrated 04:00:12 No, peter_12. It is not up to the system administrator to ensure that the dependencies of this or that Debian package are available. The Debian package system does that for the system administrator 04:00:16 If you say (require json) you have no idea what code you just required. No idea what version it is, what interface, who wrote it, nothing. 04:00:16 . 04:00:57 Riastradh: The important part is the Debian community/package management/system admins handle all of that. An application programmer does not. 04:00:57 If you said (require (planet dherman/json:1)) you know exactly what code you just required. Who wrote it, what interface. Add a :2 and you know what version. You know what it's going to do. 04:01:05 However, sensible packagers make sure that different versions of libraries are in different places in the file system. 04:01:13 Alas, not all packagers are sensible. 04:02:26 synx when you do (require scheme/ffi) do you worry about which version you get? 04:02:58 peter_12, if you are an application programmer, I certainly hope that you are aware of what libraries your application depends on, and what versions work, &c. PLaneT formalizes a naming authority for these libraries. The Debian package system does, too, I believe, and if you wanted to write a Debian package for your application, you would name the libraries that it depends on just like with PLaneT. 04:03:05 synx: You get the scheme/ffi that works with the current version of mzscheme you are using. 04:03:13 peter_12: Those are under greater scrutiny. Anyway, yes that does sometimes cause problems. 04:03:44 The Debian package system *can*. But nobody compels people to submit different versions as different packages, and indeed this is often not done. 04:03:54 When chicken 3 went to chicken 4, it lost all database support. To this day they haven't managed to adjust their FFI to enable the use of any databases. 04:04:09 Not the FFI. 04:04:16 Nobody's done the donkey work of porting them, that's all. 04:04:25 synx: so stay with chicken 3 until chicken 4 is ready for your use 04:05:18 right, so I'm stuck that way. If I had specified a version for the FFI though, I could use the old database modules until the new ones were updated. 04:05:28 and still use chicken 4 04:06:07 The issue's not in the FFI. 04:06:16 Compatibility is hard. 04:06:16 synx: Chicken 4 supports postgresql, gdbm and tokyocabinet. 04:06:25 (Cue Barbie.) 04:06:36 PLT's FFI is quite stable, so it hasn't had any of that sort of crisis in my memory at least. But (require scheme/ffi) in PLT does carry the danger that if the FFI changes, a lot of things will be instantly broken until they get updated. 04:06:37 What more do you want? 04:06:54 foof: er, sorry? I didn't think they had postgresql ported yet. 04:06:59 wtf is tokyocabinet 04:07:20 It's where Godzilla stores his address book and account balances. 04:07:31 tokyocabinet is a much superior alternative to gdbm 04:07:51 no SQLite? 04:07:53 so that means gdbm doesn't really count, huh 04:08:10 ... 04:08:16 HUH 04:08:40 gdbm counts if you have an application that uses gdbm... 04:08:43 Unless you're already using gdbm... 04:09:29 Oh I remember. I was making a scheme MUD in chicken since they started using module support. But it didn't support sqlite, so I ended up sticking with PLT. 04:09:40 Riastradh: "Compatibility is hard. (Cue Barbie.)" It is true that compatibility is hard. That is pretty well known to be a big problem and unsolved in a definitive way. 04:10:12 kilimanjaro [n=kilimanj@70.116.95.163] has joined #scheme 04:11:26 Suppose dherman changes something so you need a new initialization variable in his json module. Without module versioning, every program that uses json now breaks. 04:11:42 synx: It's a pretty far stretch from "not having ported sqlite yet" to "lost all database support." 04:11:51 With module versioning, every program that uses his json package remains fully functional, and if it's a trivial change, they can quickly and piece-wise upgrade. 04:12:20 foof: Yeah, if it's got postgresql I won't say lost all database support. I really don't count GDBM as a database though. 04:12:20 The FFI didn't change. What changed was macros and modules, so the macros have to be ported and the modules have to be added. 04:13:26 synx: I get the idea of versions but I'm not sold. It isn't a complete solution as I mentioned. 04:13:38 synx: And if you ask nicely, I'm sure someone will port sqlite3. 04:13:47 A complete solution? 04:13:49 hello 04:14:16 You really want to individually version each element of the export list. 04:14:50 You also need to formalize the behaviour of every part of the API, and copy that formalism in every client to make sure that it matches, by invoking a theorem prover every time you load everything up. 04:16:28 Or... You could distribute the source code to the json library, and all its dependencies, with your own source code. 04:17:11 Or... distribute a mzscheme system that has a particular version of mzscheme and a compatible set of libraries 04:17:53 Or... build a different package management client that works more like cpan does but downloads from PLaneT 04:17:59 foof: hehe, well... I'm more familiar with PLT anyway... 04:18:17 synx: Nevermind, sqlite3 is already ported. 04:18:22 ISTR that there are at least two competing sqlite eggs in progress. 04:18:23 ... 04:18:27 man am I out of the loop 04:18:39 you walk away for a week, and what happens! 04:19:11 jcowan: say nothing of the improvement I've been trying to make on the one sqlite egg that is barely functional 04:19:27 synx: You must have looked at the list of supported eggs only within the first week of Chicken 4's existence. 04:19:49 ...oh I'm sorry 04:19:54 I'm totally confused. 04:20:13 I've been trying to make improvements on a sqlite package in PLaneT. 04:21:00 -!- peter_12 [n=peter_12@S0106000f66bc42b9.vn.shawcable.net] has quit [] 04:21:29 Like how every time a call to sqlite3_step fails, sqlite3_reset must be called or it could leave the database locked and you manually have to delete the journal file. The package doesn't do that though. 04:22:10 also transactions don't nest, so that needed fixing too... 04:24:53 was trying to use sqlite3_progress to allow for scheme to run while sqlite is doing a query, but didn't work out too well... 04:34:09 http://www.pastebin.ca/1556765 04:34:17 i don't understand the problem, can anyone clarify? 04:34:39 the "2nth" is really "2^n"th power 04:35:04 and "x x + 1" is: "x |-> + x 1" 04:35:11 err, not "x + 1" 04:35:28 so, that square example 04:35:38 (define square (lambda (x) (* x x))) 04:36:03 (square (square (square (square x)))) 04:36:15 that would be the same as: f(f(f(x)))) 04:36:41 eventually, you see x^8 04:36:57 but that's as far as it goes for me 04:37:01 QinGW1 [n=wangqing@203.86.89.226] has joined #scheme 04:37:05 i don't get it after that 04:39:42 *jcowan* sees that R7RS is now replaying much of the CLtL3 discussions 04:43:05 -!- QinGW [n=wangqing@203.86.89.226] has quit [Read error: 145 (Connection timed out)] 04:43:38 QinGW [n=wangqing@203.86.89.226] has joined #scheme 04:50:32 Piratero: They want you to take (repeated square 4) and produce (square (repeated square 3)) 04:51:28 which in turn are replaying much of the R6RS discussions. 04:52:05 It's almost as if R7RS were some sort of "revision" of R6RS 04:53:18 They want you to take (revised (RNRS 6)) and produce (revised (revised (RNRS 5))). 04:55:40 -!- jonrafkind [n=jon@98.202.86.149] has quit [Read error: 60 (Operation timed out)] 04:57:27 -!- ASau [n=user@83.69.240.52] has quit ["off"] 04:59:55 big Scheme is a revision of R6RS, small Scheme is a revision of R5RS (but informed by R6RS, of course) 05:00:17 -!- QinGW1 [n=wangqing@203.86.89.226] has quit [Connection timed out] 05:03:49 Isn't revising exactly what they are supposed to do? 05:05:45 -!- rcy [n=rcy@S010600179a26267b.wk.shawcable.net] has quit [Read error: 110 (Connection timed out)] 05:06:18 Well, yes. 05:07:18 *jcowan* personally hopes the damn report will *not* be called The Revised Revised Revised Revised Revised Revised Revised Report on the Algorithmic Language Scheme. 05:07:36 reports 05:07:52 At least with R8RS they can change to Spider Scheme :p 05:08:30 Primo Levi tells of reading a kid's essay that began "Insects (insette) are so called because they have seven (sette) legs." 05:08:44 lol 05:08:55 When he pointed out that they in fact have only six, the kid retorted "The difference between six and seven is small." 05:09:45 ... and so are insects. QED 05:10:59 *jcowan* giggles. 05:11:42 The real culprit is the Latin->Italian sound change that caused Latin ct and st to collide as tt. 05:12:39 QinGW1 [n=wangqing@203.86.89.226] has joined #scheme 05:13:41 -!- eno__ is now known as eno 05:15:40 -!- ski [n=slj@c-9011e055.1149-1-64736c10.cust.bredbandsbolaget.se] has quit [Read error: 104 (Connection reset by peer)] 05:16:45 synx: so, in "((repeated square 2) 5)", would be: (square (square (square (square (repeated square 0))))) 05:16:48 what about the 5? 05:16:57 would that be whatever x is in the square procedure? 05:17:24 ski [n=slj@c-9011e055.1149-1-64736c10.cust.bredbandsbolaget.se] has joined #scheme 05:17:47 Piratero: ((repeated square 2) 5) would become (square (square 5)) 05:19:07 ah 05:19:13 so, 5*5 05:19:15 okay 05:21:17 oops, 5*5*5*5 05:22:39 lisptastic [n=user@adsl-234-145-87.bgk.bellsouth.net] has joined #scheme 05:23:07 thank you, that cleared things up a bit! 05:23:45 -!- jcowan [n=jcowan@cpe-74-68-154-139.nyc.res.rr.com] has quit ["Leaving"] 05:24:09 Are there any animals apart from starfish with an odd number of legs? 05:24:34 a dog with an amputated leg :) 05:24:36 3 05:24:40 :) 05:24:44 tessier [n=treed@mail.copilotco.com] has joined #scheme 05:25:00 okay, so if (define (square x) (* x x)) 05:25:11 is really: (define square (lambda (x) (* x x))) 05:25:30 is: (define (square x) (lambda (x) (* x x ))) really... 05:26:00 (define square (lambda (x) ((lambda (x) (* x x)))))? 05:26:11 i messed up with the parenthesis, but you guys get the idea 05:26:37 if you mess up the parenthesis, we dont! 05:27:14 choas [n=lars@91.13.228.159] has joined #scheme 05:27:39 :`( 05:28:04 thank god for parenthesis matching 05:29:10 Oh no, an odd number of parens! 05:29:26 -!- QinGW [n=wangqing@203.86.89.226] has quit [Connection timed out] 05:32:26 -!- zeroish [n=zeroish@135.207.174.50] has quit [Remote closed the connection] 05:32:59 -!- zeroish` [n=zeroish@135.207.174.50] has quit [Remote closed the connection] 05:52:48 -!- dmoerner [n=dmr@89-151.res.pomona.edu] has quit ["Leaving"] 05:57:14 -!- GeoffSK [n=chatzill@c210-49-251-180.kelvn1.qld.optusnet.com.au] has quit ["ChatZilla 0.9.85 [Firefox 3.0.13/2009080315]"] 06:10:25 -!- underspecified [n=eric@walnut.naist.jp] has quit [] 06:11:21 mmc [n=mima@cs137104.pp.htv.fi] has joined #scheme 06:11:43 underspecified [n=eric@isa7-dhcp-116-220.naist.jp] has joined #scheme 06:16:35 -!- elderK [n=zk@222-152-97-44.jetstream.xtra.co.nz] has quit [] 06:25:48 Adrinael_ [i=adrinael@rib4.kyla.fi] has joined #scheme 06:27:56 decker [n=chatzill@user-0c9h6n8.cable.mindspring.com] has joined #scheme 06:30:11 asdf 06:30:28 hi. so I'm a bit confused about conditionals at this point. One of the exercises in the scip book has you define a new if procedures like so: 06:30:31 decker pasted "new if" at http://paste.lisp.org/display/86694 06:30:40 ski_ [n=md9slj@remote1.student.chalmers.se] has joined #scheme 06:30:51 oops, procedure I meant. singular. 06:31:52 anyhow, great. try and use that new-if procedure with the square root program and you get an infinite recursion error 06:32:17 what's confusing me though, someone on the sicp wiki threw up this example: 06:33:48 -!- Adrinael [n=adrinael@barrel.rolli.org] has quit [Read error: 110 (Connection timed out)] 06:34:23 decker pasted "untitled" at http://paste.lisp.org/display/86695 06:34:50 which also has I p I came up with that I thought would maybe do the trick, but doesn't seem to... 06:35:29 ASau [n=user@host54-231-msk.microtest.ru] has joined #scheme 06:36:15 my p by itself does give an infinite recursion error, where as the (define (p) (p)) procedure doesnt. 06:37:32 anyhow I'm really confused here. I mean, it seems like conditionals go with applicative order evaluation, as the square root program doesn't work. but then it seems like normal order evaluation when you go with the example on the wiki 06:39:46 ha, does any of that make sense to anyone? 06:44:02 decker: what's the example on the wiki? 06:44:20 decker: you mean paste.lisp? 06:46:28 well yeah, I pasted the example from the wiki 06:46:50 last part of this page though if you want to take a look: http://sicp.org.ua/sicp/Exercise1-6 06:46:59 mit scheme by the way 06:48:02 decker: (define (p) (p)) is tail recursive 06:48:37 your other expression is not. Static analysis realized that, and thus the compiler/interpreter says "no". 06:48:38 -!- Adrinael_ is now known as Adrinael 06:50:18 ok, I still don't get why my p doesn't seem to break the new-if procedure though 06:51:32 decker: you know that Scheme is eager, don't you? 06:52:04 ha, not really. 06:52:08 your new-if will execute test, if-clause, and else-clause 06:52:17 ALL of them 06:53:23 -!- Quadrescence [n=quad@unaffiliated/quadrescence] has quit [Broken pipe] 06:53:28 When you call a function, all arguments are evaluated. 06:53:38 Quadrescence [n=quad@unaffiliated/quadrescence] has joined #scheme 06:53:45 so it gives up on mine? I guess one thing my p doesn't do which the sqr-iter procedure does is call new-if again. 06:54:04 Axioplase: but with the example on the wiki, he's dealing with COND; which is implemented in syntax 06:55:14 klutometis: The pase 86695 makes no sense. 06:55:55 he defines a looping 0-ary p, calls it, and then defines a 1-ary p and says he used it (where?) 06:55:59 Axioplase: yeah, it's a bit irritating decker didn't paste a self-contained example 06:56:09 decker: paste a self-contained example 06:56:16 ok 06:58:33 decker pasted "my attempt" at http://paste.lisp.org/display/86696 06:58:42 there it is 06:59:54 klutometis annotated #86696 "i figured you meant something like this" at http://paste.lisp.org/display/86696#1 07:00:38 decker: does that get at the source of your confusion, or not? 07:01:35 maybe. one sec. 07:02:27 Quadre_ [n=quad@24.118.241.200] has joined #scheme 07:04:52 yeah, sorry. I'm an idiot. I didn't notice at first that the example on the wiki isn't even using the new-if procedure. gah. sorry. 07:08:35 -!- Quadrescence [n=quad@unaffiliated/quadrescence] has quit [Nick collision from services.] 07:08:57 -!- mmc [n=mima@cs137104.pp.htv.fi] has quit [Read error: 110 (Connection timed out)] 07:09:34 -!- Quadre_ is now known as Quadrescence 07:16:05 I guess my question now though, if and cond by themselves don't go with applicative order evaluation then? They try the predicates first before evaluating any expressions? 07:16:51 I guess that would make sense. they'd have to if you had something like (if (foo) (set! ...) (set! ...)) 07:17:01 decker: apparently since they're not procedures, but syntax abstractions (i.e. macros), applicative order doesn't apply; so to speak 07:17:38 you can simulate the same thing using delay and force; but i think that comes later in sicp 07:17:41 hotblack23 [n=jh@p5B056888.dip.t-dialin.net] has joined #scheme 07:17:48 I suppose this is all explained in chapter four... 07:18:20 yeah, but sicp never really goes into macros 07:18:57 which are what if and cond are? 07:19:07 although it does bring up so-called "special forms" in the metacirc. eval. 07:19:09 yeah 07:20:34 right. moving on. 07:20:41 thanks for the help by the way 07:20:51 np 07:21:15 err, cond is a macro, but if is a special form 07:22:20 You can't (as far as I know, for as much of a sleepy guy I am today) simulate IF with anything (beside COND and such things) 07:23:28 -!- hotblack23 [n=jh@p5B056888.dip.t-dialin.net] has quit ["Leaving."] 07:25:51 decker: here's an example of simulating normal order with delay and force: 07:25:54 incubot: (letrec ((lazy-if (lambda (predicate delayed-then delayed-else) (if predicate (force delayed-then) (force delayed-else)))) (p (lambda () (p)))) (lazy-if #t (delay 0) (delay (p)))) 07:25:54 0 07:26:17 using procedures; a little ugly 07:26:49 MichaelRaskin [n=MichaelR@213.171.48.239] has joined #scheme 07:27:39