00:00:21 This and libev solve different problems. 00:00:29 arcfide: I believe that the R5RS *requires* the behavior that I think it does because I can't find anything which suggests it does *not* require it. There is no suggestion that what is on the right-hand side of a `let-syntax' form is ever expanded, and the language which specifically applies hygienic behavior to literals in the derived syntax forms does not cover `syntax-rules'. 00:00:37 faure [n=moe@CPE001217e40caa-CM0018c0c09832.cpe.net.cable.rogers.com] has joined #scheme 00:00:40 araujo [n=araujo@gentoo/developer/araujo] has joined #scheme 00:00:58 libev fails spectacularly to scale in certain dimensions. You really, really don't want to serialize every continuation of your application through a stack of event callbacks. 00:01:53 The only reason that that performs better in C programs than using threads is that C can't use a more compact general representation for its continuations than a contiguous block of memory, which is probably the main reason why a thread is so expensive in C. 00:02:06 arcfide: Thus, as far as I can determine, surrounding lexical bindings have no effect on the right-hand side of a `let-syntax' binding. 00:02:26 kiko_ [n=user@67.207.130.53] has joined #scheme 00:02:41 Scheme is not limited so; hence you don't need to waste your time serializing your continuations through a collection of event callbacks -- the thread system can do that for you. 00:03:21 (Maybe Chez is limited so, however, if it maps Scheme threads injectively to OS threads.) 00:03:49 -!- jonrafkind [n=jon@155.98.68.48] has quit [Read error: 148 (No route to host)] 00:04:01 emma [n=em@unaffiliated/emma] has joined #scheme 00:06:37 Riastradh: Well, I don't have to use native threads, but I haven't studied enough of your cml code to know whether or how this could be done, or if it should be done. 00:06:47 Conversely, while you could write an application in a completely event-driven style using CML as the main event loop, that wouldn't be a very effective use of the abstractions that CML provides, and it probably wouldn't perform very well, since it would serve as a level of interpretive overhead between your application and libev or a select loop or something. 00:08:10 If you want to read the code, start with primitive.scm, and then move to mailbox.scm or channel.scm, thinking of BASE-RENDEZVOUS as MAKE-PRV. Then read rendezvous.scm to see how negative acknowledgement fits into the picture. 00:12:32 chandler: Okay, so, it seems to me that you have two issues. One, you think that an keywords bound by a 'let-syntax' should not affect the definitions of other keywords by 'let-syntax' bindings inside the body of the first 'let-syntax'; secondly, it seems that you take issue with the rebinding of 'syntax-rules' in your grovel example. Are these the only two issues? 00:13:49 -!- Adamant [n=Adamant@unaffiliated/adamant] has quit [] 00:16:01 -!- faure [n=moe@CPE001217e40caa-CM0018c0c09832.cpe.net.cable.rogers.com] has left #scheme 00:18:24 -!- ebzzry [i=rmm@122.53.87.97] has quit [Read error: 110 (Connection timed out)] 00:20:28 chandler: Reading over R5RS, I think I can change your mind. :-) I'm surprised others have not tried to do so. 00:22:55 jcowan [n=jcowan@cpe-74-68-154-139.nyc.res.rr.com] has joined #scheme 00:23:31 *jcowan* unvanishes, whonking and ittying 00:23:44 It's a good day to be a Scheme weenie. 00:23:46 *offby1* beams generously 00:23:54 see? I'm reformed. 00:24:01 jcowan: Why? 00:24:03 *jcowan* is glad to see it. 00:24:19 Because this is the day I shrink the size of my T1 proposal 00:24:34 And there was much rejoibing. 00:24:40 Hahahaha. 00:24:45 *arcfide* shakes his head. 00:25:20 Then I have to do the work on optionality in R5RS. 00:25:25 s/optionality/implicit & 00:27:46 arcfide: And what do you mean, then, by shaking your head? 00:28:10 I just find it very entertaining to see all this discussion. 00:28:17 I have lost the race to keep up with it. 00:28:52 Although I have received some interesting private correspondence regarding strings. 00:29:03 -!- ve [n=a@94-193-95-252.zone7.bethere.co.uk] has quit ["leaving"] 00:31:37 Can you summarize it? 00:33:15 Unicode is stupid because it makes life harder than current technology warrants, and making character specific functions harder to implement doesn't make sense despite the utility of string level operations. 00:34:20 Adamant [n=Adamant@unaffiliated/adamant] has joined #scheme 00:34:28 Life is hard, I'll agree to that. But not because of Unicode. 00:34:36 anybody using scheme on a mac? 00:34:39 "Writing systems are hard, let's go shopping and just do numbers." 00:34:49 sstrickl [n=sstrickl@pool-151-199-30-68.bos.east.verizon.net] has joined #scheme 00:34:58 lighterthanair: I used to, before I drowned my Mac in a pool of unintentional Mac poison. 00:35:24 jcowan: Perhaps I should clarify. 00:35:32 -!- synx [i=synx@gateway/gpg-tor/key-0xA71B0C6A] has quit [Remote closed the connection] 00:35:41 -!- metasyntax [n=taylor@pool-71-127-125-129.aubnin.fios.verizon.net] has quit ["Nichts mehr."] 00:36:04 jcowan: Change Unicode is stupid to: why aren't we all just using 64-bit characters and external compression libraries? 00:36:27 synx [i=synx@gateway/gpg-tor/key-0xA71B0C6A] has joined #scheme 00:38:06 shame. 00:38:13 my condolences. 00:38:40 lighterthanair: It's okay, I managed to pick up a Thinkpad and put Slackware on it. Works quite well. 00:39:00 i'm so sorry to hear that. 00:40:09 lighterthanair: Did you have a question about Scheme on Mac? 00:40:33 -!- tltstc [n=tltstc@cpe-76-168-78-251.socal.res.rr.com] has quit [] 00:40:35 looking for a scheme or lisp app for a mac powerpc. 00:40:49 not the intel chip. 00:41:45 lighterthanair: Are you looking for an actual application or just a Scheme implementation? 00:42:02 what's the difference? 00:42:23 -!- pfo [n=pfo@chello084114049188.14.vie.surfer.at] has quit [Read error: 60 (Operation timed out)] 00:42:24 i want to learn scheme. 00:42:25 lighterthanair: One would be like a text editor written in Scheme, and the other would be like a Scheme compiler. 00:42:36 lighterthanair: Then you probably want a Scheme implementation. 00:42:39 okay, it sounds like i'm looking for a compiler. 00:42:53 We really don't need more than 21 bits worth of characters, until we join the Galactic Empire. In which case only the (expensive) FTL gateways will need to use EmpireCode, not the solar Internet. 00:43:16 lighterthanair: Then there are a number of options. 00:43:51 Of course, if they *conquer* us, it may be a different story. 00:44:15 Writing systems, unlike IP addresses, aren't growing even linearly, never mind exponentially. 00:44:17 lighterthanair: You can use Petite Chez Scheme (http://www.scheme.com), DrScheme/mzscheme/PLT (http://www.plt-scheme.org), MIT Scheme (http://www.gnu.org/software/mit-scheme/), and many others. 00:44:41 mit scheme doesn't seem to have a powerpc version, unless i missed something. 00:44:55 lighterthanair: I believe you can compile it for the PPC using the C Back End. 00:45:07 they said they took out that option. 00:45:10 i forget why. 00:45:10 That may require more knowledge about compiling MIT Scheme than you care about. 00:45:17 The C Back End? 00:45:19 It's there. 00:45:29 Recently added (back?), actually. 00:45:30 not according to their website. 00:45:42 There may not be a precompiled PPC version, though. 00:46:14 lighterthanair: They have their "Portable C" option. 00:46:21 how much knowledge would it require? are we talking 101 level or, like 409? 00:46:36 lighterthanair: How much about compiling unix programs do you know? 00:46:41 jack. 00:46:43 0. 00:46:46 nada. 00:46:47 zip. 00:46:56 lighterthanair: Does ./configure; make; make install mean anything to you? 00:47:20 pfo [n=pfo@chello084114049188.14.vie.surfer.at] has joined #scheme 00:47:20 i followed the instructions once for ruby. thank goodness the person wrote out each command, step by step, and wrote out what the result should be. 00:47:27 lighterthanair: If not, then you should probably consider one of the other implementations until you're a little more experienced with that. 00:47:30 i do play in the terminal, but ... 00:47:58 okay. are there any differences between chez and plt that make one better than the other? 00:48:09 lighterthanair: Basically, it consists of taking a runtime code (precompiled) and compiling the "microcode". You can then install as you see fit. 00:48:19 lighterthanair: Let's not get into that. :-) 00:48:40 okay, which version do you recommend, if any? 00:48:51 between the different options. 00:49:00 lighterthanair: If you install SWL with Petite Chez, you'll have a GUI editor and REPL, but DrScheme is well loved by some who first learned Scheme. 00:49:23 lighterthanair: I'm one of the only (the only?) guys in here who uses Chez as his daily system. Guess which one I use and recommend? 00:49:35 lighterthanair: It doesn't matter too much when you are just learning. 00:49:43 okay. 00:49:58 lighterthanair: Just pick one and stick to the basic language features until you feel more comfortable. 00:50:04 ah. 00:50:11 Then you can start learning how the different implementations work, and can pick the one you like. 00:50:17 okay. 00:50:39 Chez and PLT are both used extensively in education. 00:51:02 -!- dysinger [n=dysinger@adsl-99-181-31-201.dsl.pltn13.sbcglobal.net] has quit [Success] 00:51:26 Of course, there are about five to ten other systems that I haven't mentioned that people would recommend if they weren't asleep. 00:51:48 lighterthanair: Both PLT and Chez have complete documentation. 00:52:05 good. 00:52:12 i think i'm going to need that. 00:52:29 asleep on a friday night? i hope you mean drunk. 00:52:35 You might also play with Chicken, Gambit (C based compilers), and Scheme48 (a fan favorite around here). 00:52:41 lighterthanair, MIT Scheme worked perfectly well on Mac OS X/PowerPC when I last tried it, using the C back end. 00:53:17 lighterthanair: If Riastradh is willing to help, he could probably walk you through the MIT Scheme installation without trouble. :-) 00:53:17 thanks, riastradh, when i geek up a couple of levels, i'll look into installing it. 00:53:21 really? 00:53:34 not now, though, i need to install leopard soon. 00:53:40 lighterthanair: It's really not hard, but if you've done almost no compilation before, you may not like it. 00:53:41 As arcfide said, it is merely a matter of `./configure; make; make install'. 00:53:50 If you need it in a little more detail: 00:54:02 yeah. but if anything goes wonky, i don't know how to address it. 00:54:18 *arcfide* points at the MIT Scheme resident wizard => Riastradh. 00:54:31 Fetch the tarball, say mit-scheme-c-20090107.tar.gz. Run `gunzip -c < mit-scheme-c-20090107.tar.gz | tar xf -', and then `cd mit-scheme-20090107/src' and `./configure; make; sudo make install'. 00:54:44 *lighterthanair* copying 00:54:58 ...oops. 00:55:01 Hang on. 00:55:02 That's not right. 00:55:03 *arcfide* chuckles. 00:55:04 I forgot. 00:55:06 that's okay. i'll .. 00:55:09 delete it. 00:55:13 Riastradh: Getting rusty there. 00:55:14 Anyway, you just follow the instructions at . 00:55:34 this is for a powerpc mac 00:55:38 Yes? 00:55:55 lighterthanair: If I had to suggest my two personal favorite Schemes, it would be MIT Scheme and Chez Scheme. 00:56:07 from the site you gave me, "When built this way, the system runs slower than when it is built using the native-code compiler. For this reason, you will usually want to use native-code binaries when running on a 32-bit Intel architecture machine." 00:56:10 *arcfide* pokes chandler. 00:56:26 lighterthanair: That's a non-issue for you. 00:56:27 lighterthanair, right. But you are not on a 32-bit Intel architecture machine. Hence you want the portable C version. 00:56:43 yes, arcfide, i wanted to stick with mit because i was going to use their scip lectures. 00:56:47 or is that sicp? 00:57:15 SICP. 00:57:28 I hope you have a copy of the book? It's a great read. 00:57:34 ah, okay, Riastradh. i saw them mention intel and thought oh no. 00:57:40 the book is online. 00:57:45 and i'm on a budget. 00:58:21 brb 01:01:37 back 01:01:55 and thanks very much for the info and the site. 01:03:06 Riastradh: Do your objections to SRFI 0, expressed back in 2004 on the srfi-55@ list, still hold? 01:03:18 I agree that SRFI 0 is a hack, but sometimes people need a hack. 01:04:09 jcowan: I've been thinking on that one, and I have to say that I don't really like SRFI-0 either in most cases. I'm not sure I think it's better than using separate files for implementation specific code. 01:04:38 arcfide, to summarize my earlier discussion of CML and libev: libev knows a lot about how to deal with the operating system's notion of I/O, and would be a useful implementation for an event loop to run on each processor; CML is an abstraction for programmers to use for the processes in a program that deal with multiplexed event sources. 01:05:49 arcfide: I'm not sure what you mean by my first issue. 01:06:05 jcowan, well, that was a thoroughly fruitless discussion. Can you summarize what my views were? 01:06:18 chandler: (let-syntax ([syntax-rules ...]) (let-syntax ([blah (syntax-rules ...)]) ...)). 01:06:22 -!- sepult [n=user@xdsl-87-78-27-74.netcologne.de] has quit [Read error: 60 (Operation timed out)] 01:06:53 arcfide: OK. So I'm actually not sure what you mean by my second issue. 01:06:58 But let's deal with the first. 01:07:28 chandler: (syntax-rules () [(_ ...) (begin (define-syntax syntax-rules ...) (define-syntax blah (syntax-rules ...)))]). 01:07:44 Summary: 1) Felix is arrogant; 2) Felix doesn't listen; 3) If it doesn't work in Scheme 48, epic fail! 01:07:50 chandler: Am I correct in your assesment of the first? 01:08:30 SRFI 0 is the same as C preprocessor conditionals; they suck, but they can save you if used judiciously. 01:08:42 jcowan: Many people love to host their systems. I am personally not fond of Aubrey Jaffer's approach to things. I don't know that making such conclusions benefits us very much though. 01:09:00 sepult [n=user@xdsl-87-78-30-131.netcologne.de] has joined #scheme 01:09:16 Dark-Star [i=Darkstar@p57B57ED7.dip.t-dialin.net] has joined #scheme 01:09:17 s/to host/to advertise and argue from/ 01:09:19 arcfide: I'm not quite sure what you meant by that. I'll rephrase it: I think per the R5RS, bindings in phase 0 do not shadow bindings or affect literals in phase 1 (of which there are two examples in the R5RS, `syntax-rules' and `...') 01:09:36 jcowan, you missed `4) Riastradh was a trifle too emphatic about the discussion'. 01:09:48 arcfide: If you can point to justification in the R5RS for another behavior, I'd like to hear it. I've been unable to find it. 01:09:48 -!- sepult [n=user@xdsl-87-78-30-131.netcologne.de] has quit [Remote closed the connection] 01:10:03 arcfide: Either way, some common implementations will be in the wrong here :-) 01:10:25 chandler: Yes, I have such, but I want to address your rephrasing. Let me construct the right response once I have determined whether R5RS addresses your specific issue. 01:10:29 -!- antoszka [n=antoszka@unaffiliated/antoszka] has quit [Read error: 110 (Connection timed out)] 01:10:37 why does (foldl (lambda (x y) (and x y)) #t my-list) work (on plt-scheme) but (foldl and #t my-list) doesn't? 01:10:46 jcowan, but, in any case, most times I use cpp conditionals, I regret it later on, and seek better ways to go about doing it. 01:11:03 sepult [n=user@xdsl-87-78-30-131.netcologne.de] has joined #scheme 01:11:19 Dark-Star, AND is a name for a macro, not for a procedure. It does not have independent meaning outside the first element of a list. 01:11:21 Dark-Star: `and' is not a function. It is a macro. 01:11:33 Dark-Star, or rather, AND is a name for a macro, not for a variable (whose value is a procedure). 01:12:04 -!- ski_ [n=md9slj@remote1.student.chalmers.se] has quit ["Lost terminal"] 01:12:34 okay, makes sense. it's annoying, though ;-) 01:12:49 jcowan, moreover, Scheme ought to have a more principled system of controlling bindings than C. Cpp conditionals are mostly about testing whether this or that variable (function, macro, &c.) is defined. 01:12:54 I think it did work once, though 01:13:13 Maybe in lazy Scheme, Dark-Star. 01:14:51 -!- derdon [n=derdon@p4FC8D63C.dip.t-dialin.net] has left #scheme 01:15:46 chandler: Okay, yes, it's what I thought. 01:16:06 chandler: In my interpretation the definition of lexical scoping in R5RS *requires* the ability to shadow. 01:16:09 -!- Poeir [n=Poeir@c-98-222-133-165.hsd1.il.comcast.net] has quit [Read error: 54 (Connection reset by peer)] 01:16:22 I see a few relevant sections here. 01:16:32 Poeir [n=Poeir@c-98-222-133-165.hsd1.il.comcast.net] has joined #scheme 01:17:47 Sections 3.1 and 4.3 seem the most helpful here. 01:18:40 My first observation is that the concept of phases was introduced explicitly in R6RS, and is not a concept explicitly detailed in R5RS, or at least, I can find no occurance of the term. 01:18:44 Maybe I missed something. 01:19:05 Phases are irrelevant in the R5RS. 01:19:09 The R5RS has a concept of bindings, which can either be syntax transformers or values. 01:19:11 Riastradh: Exactly. 01:20:26 *jcowan* thinks evilly of a SRFI 7 front-end to Chicken. 01:21:21 I doubt whether there is any useful content in that discussion, jcowan, either from me or from anyone else. 01:21:36 chandler: In section 4.3, you see the following, "A syntactic keyword of a macro may shadow variable bindings," and vice versa. 01:22:14 I think ncurses sucks. 01:22:43 chandler: I read the R5RS as allowing only one binding to any given identifier at any point in the program. 01:23:01 This is true whether the code is being expanded or evaluated. 01:23:12 Right. But how does that apply to the right-hand side of let-syntax? I can't refer to any bindings there, except inside the template. 01:23:33 I'm asserting that `syntax-rules' is not a reference to a binding. 01:23:35 chandler, yes you can: you can refer to the binding of the name SYNTAX-RULES. 01:23:39 chandler: But it is. 01:24:00 chandler: Every bound identifier is bound either to a syntax transformer or to a value. 01:24:09 The R5RS doesn't say much about how that happens, as far as I know, and thereby allows both Scheme48's approach and MIT Scheme's approach. 01:24:41 In Scheme48, the right-hand side of each DEFINE-SYNTAX is evaluated one phase up, and SYNTAX-RULES is defined to be a macro that expands to an expression of a procedure; thus, the value of a SYNTAX-RULES expression is a procedure. 01:24:50 chandler: Additionally, R5RS defines lexical scoping in terms of program regions, and uses the term "program text." 01:24:52 jeapostrophe [n=jay@69.169.141.110.provo.static.broadweavenetworks.net] has joined #scheme 01:25:13 In MIT Scheme, the right-hand side of each DEFINE-SYNTAX is expanded just like any other form, and SYNTAX-RULES is a macro that expands not into an expression at all but into a transformer keyword. 01:25:24 chandler: R5RS is even explicit about the change of environment that occurs during expansion time for the body of a let-syntax form. 01:25:39 It is keywords such as ER-MACRO-TRANSFORMER and SC-MACRO-TRANSFORMER that actually do any evaluation (at one phase up, if MIT Scheme had the concept of phases). 01:25:52 arcfide: I'm asserting that since `syntax-rules' is, per the R5RS, the *only* thing that can appear on the right side of a `let-syntax' form, it is more like a literal than a bound value. But, the R5RS authors did not extend hygiene to this literal. 01:26:02 -!- Guest53038 [n=moe@p54A0C821.dip.t-dialin.net] has left #scheme 01:26:18 chandler: They identified them as syntactic keywords, not as literals. That makes a difference. 01:27:09 Right. As such, I believe it falls into a no-man's-land, where a literal `syntax-rules' is expected - even if, in phase 0, it has been shadowed. 01:27:23 *jcowan* vaguely remembers that there are procedure versions of AND and OR (without lazy evaluation, of course) in some systems. 01:27:25 chandler: Don't mistake the use of phasing here. 01:27:36 Read the semantics for 'let-syntax'. 01:27:58 Riastradh: I agree, there isn't, but I was led to read the SRFI 0 discussion as a result. 01:28:31 At any rate, I haven't yet decided whether to propose SRFI 55 for Thing One; 01:28:56 chandler, I don't think the R5RS or its authors had any particular intent about how the name SYNTAX-RULES was to be interpreted. A more important discussion is of which semantics is preferable (MIT Scheme's or Scheme48's). 01:29:09 I think that some mechanism for "load the file(s), if findable, that implement module foo" is a Good Thing. 01:29:09 "The is expanded ..."; chandler, in other words, the body of the first let-syntax is expanded in a syntactic environment containing the new bindings to any keywords that were bound in the clauses of the 'let-syntax' form. 01:29:43 Riastradh: I'm not trying to argue from intent; I'm trying to argue from what is. :-) 01:30:08 chandler: Additionally, 'let-syntax' is explicitly considered a binding form, and it binds the identifiers in its clauses. 01:30:24 jcowan, I have no more hope for the Scheme report process, but I think that any half-arsed moduloid system will get in the way of making a good one. That's similar to why I didn't want the R6RS's RAISE-CONTINUABLE. 01:31:17 According to the R5RS' definition of lexical scoping, any occurance of the identifiers bound by the 'let-syntax' form inside the body forms must refer to the new bindings, not any other bindings. 01:32:38 lighterthanair_ [n=lightert@63.118.151.12] has joined #scheme 01:32:50 -!- lighterthanair [n=lightert@63.118.151.12] has quit [Read error: 104 (Connection reset by peer)] 01:33:22 chandler: Note here that there is simply no concept of phasing taking place when speaking in terms of the R5RS semantics. 01:33:42 With low-level macros, you need a fully arsed system, but I don't see that a half-arsed one damages Thing One very much, since it only has syntax-rules. 01:34:15 I *really* don't want to see a module system standardize for Scheme that fails to control top-level environments. 01:34:29 What do you mean by "control"? 01:34:33 `standardized' 01:35:01 I don't want to see a (REQUIRE-EXTENSION FOO) that has the effect of making all the top-level bindings of foo.scm wherever the REQUIRE-EXTENSION occurred. 01:35:01 -!- lighterthanair_ [n=lightert@63.118.151.12] has quit [Client Quit] 01:35:06 arcfide: Perhaps I'm being purposefully obtuse here, but I'm still stuck on whether it *is* a reference to a binding. I can't find any references to it as such; there's no description of what it's bound to, or returns. 01:35:18 chandler: Section 3.1 details the ideas behind lexical scoping in R5RS and I believe enforces a single binding per binding per identifier model. 01:35:38 chandler: Reference to what? syntax-rules? 01:35:45 (Single binding per binding per identifier?) 01:36:14 So you object to (import foo) at the top (interactive) level? 01:37:00 REPLs may have many operations that make zero sense to standardize. 01:37:19 That was your point #6. 01:37:25 arcfide: Yes, to `syntax-rules' being bound. The use of `syntax-rules' is described as being a part of the `pattern language'. 01:37:49 davazp [n=user@79.153.148.56] has joined #scheme 01:37:52 Whether syntax-rules is bound seems to me an implementation question. 01:38:11 arcfide: clearly not every identifier mentioned in the R5RS is bound: `else', for instance, is not. 01:38:27 From a pure R5RS standpoint, it is an uninterpreted identifier which serves as part of the define-syntax, let-syntax, and letrec-syntax macros, just as else serves as part of the syntax of the cond macro. 01:38:35 What he said. 01:38:35 No, chandler, but every name has at most one meaning. You can't use COND's meaning of ELSE if you have bound ELSE to your own meaning. 01:38:47 chandler: Okay, let's assume for a moment that syntax rules is a literal for every macro that expects a transformer. 01:38:57 (at most one meaning in any particular lexical environment, that is) 01:39:43 arcfide: Aha! Now, have a look at the beginning of 4.2. 01:40:15 Riastradh: I believe that is untrue of ELSE. 01:40:31 chandler: The comparison of literals is specific in the syntax-rules definition, I believe it is a bound-identifier=?, or, if bound, they must be bound to the same "thing". When you bind syntax-rules, you are in effect "renaming" that identifier for the scope of that let-syntax form, meaning that any macro defined elsewhere to use the syntax-rules "literal" will not identify "syntax-rules" as said literal. 01:40:37 -!- blackened` [n=blackene@ip-89-102-28-224.karneval.cz] has quit [] 01:40:40 rudybot___: eval (let ((else #f)) (cond (else 'riastradh-wins) (#t 'jcowan-wins))) 01:40:41 Riastradh: your sandbox is ready 01:40:41 Riastradh: ; Value: jcowan-wins 01:40:49 ...um, I got that backwards. 01:40:53 Heh! 01:41:34 The behavior of `else' is defined by - and here I'm perhaps incorrecly asserting that it is defined *only* by - the language at the beginning of 4.2. 01:41:52 rudybot___: eval (let ((else #f)) (cond (#f 1) (else 2))) 01:41:52 jcowan: your sandbox is ready 01:42:02 ) 01:42:04 argh 01:42:08 ebzzry [i=rmm@122.53.94.172] has joined #scheme 01:42:13 -!- ebzzry [i=rmm@122.53.94.172] has quit [Read error: 104 (Connection reset by peer)] 01:42:29 rudybot___: eval (let ((else #f)) (cond (else 'riastradh-loses))) 01:42:40 lol 01:42:41 Looks like nothing ishappening. 01:42:47 See? I don't lose. 01:43:11 rudybot___: eval (list (let ((else #f)) (cond (else 'riastradh-loses)))) 01:43:11 Riastradh: ; Value: (#) 01:43:13 chandler: We can go by the macro definition given in Chapter 7. 01:43:48 arcfide, isn't concurrency more interesting than these fiddly details of the R5RS? 01:44:05 Riastradh: Yes. I just couldn't resist when chandler baited me. 01:44:09 Not for those who have chosen to undertake the burden of becoming standardizers, no. 01:44:32 arcfide: This is mostly pointless, yes. 01:44:39 I believe there is an XKCD comic on people being *WRONG* on the Internet! 01:45:23 chandler: Do you see now, though, why this should break? PLT and Scheme48 I believe change the semantics, but I don't know how they rationalize it in terms of R5RS. Of course, it is not necessary for them to do so. 01:45:46 arcfide: It does not hardly matter when `syntax-rules' and `...' are the only keywords in question, and the R6RS did define what happens here. 01:46:07 chandler: Well, it matters for cases like you created. 01:46:08 arcfide: Scheme48 works perfectly here! 01:46:26 chandler: You said that it doesn't error out in your groveler code! 01:46:30 chandler: It should error out. 01:46:42 No, it shouldn't - should it? 01:46:48 Yes, it should. 01:46:55 I didn't understand the point of that code. It doesn't introduce any new identifiers -- all of them are bound someplace. 01:47:01 Or at least, I am arguing that R5RS semantics dictates that it should. 01:47:17 jcowan: Which example? 01:47:49 chandler: When you redefine syntax-rules in your groveler macro, hygiene requires that all the "syntax-rules" identifiers in the scope of that definition be renamed. 01:47:59 The FizzBuzz example 01:48:31 chandler: lexical scoping puts that second definition in the scope of the first according to R5RS (I believe). 01:48:36 arcfide: Eh heh! But what about in (quote syntax-rules) ? 01:48:47 Where? 01:48:52 chandler: Can you show me the code again? 01:48:55 As an example. 01:49:22 It's not in the code, but I'm pointing out that not all references to the identifier are renamed. 01:49:26 The identifier is still renamed, it's just that during expansion quote doesn't care how the underlying references or wraps look. 01:49:37 Only those which could actually be a reference to the binding. 01:49:48 Semantically, it doesn't matter, because quote will discard the wraps and give you only the raw symbol. 01:50:17 And since (in Scheme48) the reference on the right-hand side *can't* refer to the binding, the wraps are discarded there as well. I think. 01:50:30 Huh? 01:50:57 Why shouldn't the first define syntax effect the syntactic environment of further macro definitions? 01:51:11 s/effect/affect/ 01:51:27 The use of `syntax-rules' on the right side of a `let-syntax' can, in Scheme48, no more refer to a surrounding phase-0 binding than a quoted use can. 01:51:46 Why do you keep talking about phase-0? 01:51:56 Because that's how Scheme48's semantics works, arcfide. 01:52:02 Because it *is* a matter of phasing. 01:52:32 So is it the case that internal define-syntax definitions are as if specified by letrec*-syntax? 01:52:35 In my understanding of it, the binding created by let-syntax *should* affect the phase-1, which is where macro expansion happens. 01:52:39 In Scheme48 *and* in PLT. Chez is the odd-man-out here for mushing all the phases together into one big happy environment. 01:53:04 jcowan: There are no internal define-syntax forms in R5RS. 01:53:05 We already knew that Chez's phase semantics is a bug, chandler. 01:53:21 jcowan: In the R6RS? Yes, definitionally I think. 01:53:51 Yes, the R6RS. I had in fact forgotten that the R5RS doesn't support them. 01:54:15 In any event, I'm too tired to keep at this, and it's really rather boring. 01:54:32 Okay. 01:54:35 On to concurrency. 01:54:36 *jcowan* urges that we move on to something more edifying. 01:54:41 Concurrency is good. 01:54:51 arcfide: What's more interesting is that even *without* the syntactic rebinding of `syntax-rules', Larceny chokes on that code. 01:54:55 jcowan, internal syntax definitions should be specified as in , and in particular either SCAN-R6RS-BODY or SCAN-R5RS-BODY (the difference between which two procedures being whether mixed definition/expression sequences are permitted). 01:55:09 http://paste.lisp.org/display/86941 for reference. 01:56:30 If anyone turns *that* in as his FizzBuzz code, *hire him.* 01:56:43 Riastradh: I'm going to try to play with your code tomorrow. 01:56:59 Riastradh: I think I see how you are approaching the IO-rendezvous stuff, and it makes sense. 01:57:08 jcowan: Or run, screaming!! 01:58:00 I'll start looking at how to implement your CML stuff tomorrow, and hopefully I'll have few questions. On the other hand, you may as well email me pointers if you have any that are not in the README. 01:58:09 *jcowan* is glad to say he has never interviewed a non-Fizzbuzzer candidate 01:59:03 jcowan: This originally started as a "what not to do on an interview programming test" example. 01:59:44 I guess it depends on where you are interviewing. 01:59:56 -!- arcfide [n=arcfide@adsl-99-50-231-131.dsl.bltnin.sbcglobal.net] has left #scheme 01:59:59 Obscure language? Check. Odd style? Check. Hairy recursive macros doing hygienic inseretion? Check. Mutation via `letrec' + `call/cc'? Check. 02:00:17 I suppose :-) 02:00:29 At Google, f'rinstance ... 02:00:52 I blew away one of my interviewers by answering his question in six words. 02:01:07 (I cheated, of course; I had read the answer somewhere and wasn't inventing it from first principles.) 02:01:58 Was it one of these answers? http://www.codeslate.com/2007/01/you-dont-bury-survivors.html 02:02:20 No. 02:02:51 "Sort this way (gesture across), then that way (gesture down)." That should be enough to reconstruct what the question was, certainly! 02:03:15 Oh. 02:03:42 *eli* questions the last occurrence of "happy" in the log 02:04:41 *chandler* vanishes for the evening 02:04:52 I think that's "happy" in the sense of "happy because unaware that there's anything to be concerned about". 02:05:10 Yeah, that sounds right. 02:05:33 There was some comedy show in israel a while ago, with the motto of "dumb but optimistic". 02:05:56 http://catb.org/jargon/html/H/happily.html 02:06:14 Sounds like the Top Gear motto: "ambitious, but rubbish". 02:16:45 So, what about them Concurrents? 02:17:34 *Fare* finds unforeseen problems when trying to cleanly build a 1500 file Lisp system. 02:18:26 *Fare* hopes that a PLT .zo requires less mucking around to load than a CL fasl. 02:19:09 Loading .zo files happens automatically, doesn't it? 02:19:20 -!- annodomini [n=lambda@wikipedia/lambda] has quit [] 02:19:26 And has exactly the effects you asked for? 02:20:36 As far as I know, anyway, it does the right thing, much better than Common Lisp does. 02:20:45 can't be worse. 02:21:02 That's a terrifying thought. 02:21:02 Common Lisp has to much open air sausage 02:21:48 *Fare* laughs at CL "fast" loading files. 02:21:52 Heheheh. 02:22:48 -!- Adamant [n=Adamant@unaffiliated/adamant] has quit [] 02:22:58 it's sure not fast when you have to do it 47284 times. Takes me about 88 minutes. 02:23:14 that's about 9 files per second average. 02:23:20 *jcowan* wonders if the Lisp (broad sense) community will be the last refuge of programmers who *understand* their programs. 02:23:40 I'd rather my program understand me. Sigh. 02:23:54 wasn't that the original promise of Lisp? 02:24:15 Sorta. 02:24:34 jcowan, anyway, the concurrent subject concurrently discussed by me and arcfide was , which I wrote yesterday. 02:25:10 This is not so much about concurrency as about communicating between concurrent processes, however. 02:25:35 are you using threads or fork? 02:25:48 or green threads, or what? 02:26:11 Communicating between concurrent shared-memory processes, to be more precise. That code is agnostic to the mechanism of concurrency, however. 02:27:58 *jcowan* tries to find the high-level explanation of CML, fails. 02:28:05 dmoerner [n=dmr@69.26.214.2] has joined #scheme 02:28:15 Here's a summary as brief as I can make it on the fly: 02:29:29 An event is something that happens once. For example, if I typed `t' at my terminal, my action if typing it would be an event. On the other end of the terminal abstraction, to the program reading the terminal, at any time, there may be the event of the next character to be typed at the terminal, for which event the program will often wait. 02:30:53 Emphatically, the objects that CML deals with are not events, although events are involved. A rendezvous is an object that at any given time is in one of two states: enabled, with some value associated with it; or blocked. Given a rendezvous, a program can wait until the event of the rendezvous' becoming enabled. In this way, a rendezvous is a sort of abstract description of a set of related events. 02:31:34 it's a stream of futures? 02:32:03 For example, the program sitting behind a terminal might have a rendezvous that is enabled when there is input pending that I have sent but that the program has not read. The program can /synchronize/ this rendezvous, which is to wait for the rendezvous to become enabled (i.e. to wait until there is pending input) and consume the pending input. 02:32:41 Rendezvous can be composed in several ways: given two rendezvous, you can make a third rendezvous that is enabled whenever either of the two is enabled, with the value of whichever of the two is enabled (or the value of an arbitrary one when they are both enabled). 02:32:45 This all sounds very Ada. 02:32:54 But more so. 02:33:17 Given a rendezvous and a procedure, you can make a rendezvous whose value is the value of the procedure applied to the value of the original rendezvous. 02:33:31 (...and which is enabled whenever the original rendezvous is enabled.) 02:33:40 -!- luz [n=davids@189.122.90.116] has quit ["Client exiting"] 02:33:59 so it's futures with user-visible synchronization state? 02:34:02 *jcowan* now understands *why* R6RS-ites want to restrict library bodies to definitions followed by expressions, at least. 02:34:16 jcowan, why? 02:34:35 Because library bodies are *bodies*, the same as non-top-level begin bodies, lambda bodies, etc. And that's the rule for them. 02:34:55 The other view holds that libraries are like top-level begin, a trivial wrapper around what remain top-level forms in essence. 02:35:00 uh? 02:35:33 The construction of rendezvous can be delayed until synchronization; for example, a rendezvous for when a server sends a reply to a particular request, will send the request when you synchronize it and then reduce to the rendezvous of waiting for a message from the server. 02:35:44 *Fare* see "toplevel forms" and dreads the conjured specter of EVAL-WHEN. 02:36:06 No, the issue is the same in a pure interpreter. 02:36:29 By rule (dating from MacLisp days), defines in a top-level begin are not scoped to the begin, but have global scope. 02:37:41 I thought they did that because they wanted to allow to compile the definitions together without having to care for the capture of continuations as forms have uncontrollable side-effects (with dynamic redefinition) in between definitions. 02:37:49 Finally, when you synchronize on a choice of such rendezvous, so that you have sent requests to two servers and are waiting for a reply from either one of them, whichever one is slower to reply may wish to know that you have selected the other one; when this happens, if you ask for it, CML can send a negative acknowledgement to the slower server. 02:38:03 Fare, no, that happens whether or not you allow expressions or definitions to be intermixed. 02:38:13 You can always turn an expression into a definition, after all, by generating a dummy name. 02:38:21 yes 02:38:25 Which is what the semantics of R6RS top-level bodies does. 02:38:39 which is why I think they're trying hard and only end pissing me 02:38:48 and making the language awkward 02:38:50 Riastradh: Thanks. 02:38:57 Quite, Fare. 02:39:24 "RnRS, the Lisp that wanted to be ML" 02:39:26 I have a guess for the real reason why the restriction exists: because it exists in Chez's MODULE form. 02:39:46 AFAIK, the reason for the top-level-begin rule is to make it possible for top-level macro calls to expand to more than a single top-level form. 02:40:36 So I was thinking of trying to make a text based game, but the parsing is just flummoxing me. 02:40:36 who'd care if they hadn't crippled non-toplevel forms. 02:40:56 it's like making a problem, then creating more problems to work around it. 02:41:05 Crippled how? 02:41:21 I'd like to have "north" be a procedure, that moves the player to whatever room is marked as northwards 02:41:36 But "go north" should also call that procedure 02:42:04 so (go north) would take a procedure as an argument, make sure it's an exit, then call it. 02:42:11 but... 02:42:17 "say hello how are you" 02:42:27 "Error: undefined symbol hello" 02:42:35 jcowan, different scoping rules. 02:42:45 I'm sorry, but you are too obscure for me. 02:42:55 quoting symbols and enclosing strings in "" is something I'm not sure players would grok all that well 02:42:58 incapacity to affect the rest of the current scope. 02:43:29 so... how to handle that? Should it be required they go {say "hello how are you"} 02:43:50 -!- Fare [n=Fare@ita4fw1.itasoftware.com] has quit ["Leaving"] 02:43:55 Write your own reader, synx. 02:44:20 but what should my reader do, is what I'm trying to figure out. 02:44:25 *jcowan* nods. 02:45:44 Without quotation marks I'm not sure there's any consistent way to do it. Otherwise you have to have an un-quotation syntax which seems tricky... 02:45:46 say hello how are you ,(name (first (send here list-people)))) 02:46:07 but then "go north" would result in "North is a symbol not an exit procedure." 02:46:12 you'd have to say go ,north 02:46:29 Figure out how to express it in S-expressions, then write a reader which accepts your preferred external syntax and generates those S-expressions. 02:46:41 -!- Nshag [i=user@Mix-Orleans-106-2-250.w193-248.abo.wanadoo.fr] has quit ["Quitte"] 02:47:05 easy to express it in S-expressions, but getting the reader to generate them I'm struggling with. 02:47:16 For an interactive user interface, synx, you may want to make the behaviour of the parser dependent on the command. 02:47:20 E.g., in Scheme48, if you type 02:47:21 ,load foo.scm 02:47:22 (say "hello how are you") and (go north) and (north) are the 3 s-exps I've considered so far 02:47:34 it gets read as (LOAD "foo.scm"). But if you type 02:47:37 ,run (+ 1 2) 02:47:41 it gets read as (RUN (+ 1 2)). 02:47:49 hmm... 02:48:28 ski_ [n=md9slj@remote1.student.chalmers.se] has joined #scheme 02:48:41 So with load and run, load would be a special syntax that stringifies its arguments? 02:49:17 The parser looks at its entry for LOAD in some table and says, `Aha! I am supposed to read one argument, a pathname.' 02:50:44 so I guess you'd specify how the arguments parse, when creating a new command? 02:51:17 Right. 02:52:03 could work I guess... trying to think how I'd write a parser that had a list of commands, with argument parsing rules... 02:53:32 antoszka [n=antoszka@unaffiliated/antoszka] has joined #scheme 02:55:19 Fare [n=Fare@ita4fw1.itasoftware.com] has joined #scheme 03:00:34 -!- jcowan [n=jcowan@cpe-74-68-154-139.nyc.res.rr.com] has quit ["Leaving"] 03:01:20 Adamant [n=Adamant@unaffiliated/adamant] has joined #scheme 03:09:01 -!- tjafk1 [n=timj@e176204227.adsl.alicedsl.de] has quit [Read error: 110 (Connection timed out)] 03:09:26 tjafk1 [n=timj@e176205170.adsl.alicedsl.de] has joined #scheme 03:10:39 -!- Fare [n=Fare@ita4fw1.itasoftware.com] has quit ["Leaving"] 03:13:08 -!- Dark-Star [i=Darkstar@p57B57ED7.dip.t-dialin.net] has quit [Read error: 110 (Connection timed out)] 03:18:41 phax [n=phax@unaffiliated/phax] has joined #scheme 03:22:37 rcy [n=rcy@S01060013104bf489.ok.shawcable.net] has joined #scheme 03:34:25 synx: MUD ? 03:34:38 syntropy_: more like MU 03:34:53 sounds neat 03:35:02 -!- eno [n=eno@nslu2-linux/eno] has quit [Read error: 54 (Connection reset by peer)] 03:36:20 *eli* rolls his eyes at jcowan 03:38:01 it is pretty spiff syntropy_. I've got a general framework for persistent objects over a sqlite thingy, just need to write a parser for it now really. 03:38:56 synx: neato, i'd be interested in testing it. I often emulate the same approach when trolling people. 03:39:16 "Where do you want to go?" 03:39:22 "Uhh, I dunno." 03:39:25