IRC log started Mon Feb 21 00:00:02 2000 [msg(TUNES)] permlog 2000.0221 -:- SignOff air: #TUNES (http://www.qzx.com/ :: sleep) -:- hcf [nef@me-portland-us640.javanet.com] has joined #tunes -:- SignOff Fare: #TUNES (Ping timeout for Fare[esmeralda.enst.fr]) -:- NetSplit: zsoldos.openprojects.net split from sterling.openprojects.net [01:18am] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [zsoldos.openprojects.net] -:- smokie [tw026024@zaalf10.twi.tudelft.nl] has joined #tunes -:- Netjoined: zsoldos.openprojects.net sterling.openprojects.net -:- water [water@tnt-9-203.tscnet.net] has joined #tunes -:- SignOff water: #TUNES (Leaving) -:- water [water@tnt-9-203.tscnet.net] has joined #tunes -:- SignOff hcf: #TUNES (Leaving) -:- SignOff smkl: #TUNES (changing servers) -:- smkl [sami@glubimox.yok.utu.fi] has joined #tunes -:- SignOff smokie: #TUNES (-> huis) anyone here? no hi foof 04:10am i'm not sure i get the essential difference between lisp and dylan's macro systems would you help me differentiate? ok let me tell you about CL macros first CL is very very different from C.. in C you have a compiler which applies a preprocessor to the code and generates .o files which are linked and then run.. ie two stages: compile and run in CL one has several stages: reading of the code, compiling of the code, loading of the code and execution of the code mhm the diff between string-replacement and recursive tree-proccessing in CL one has macros at all stages and one can change the language's syntax, semantics and mostly everything in all stages last I remember dylan just had two real stages.. load/compile and execution so you have fewer stages to apply changes in taking on the CL concept of macros you open up _everything_ at all stages for changes Dylan macros are more classic macros.. they transform code when loaded/compiled so dylan adds nothing to the macro state-of-the-art? ie a big difference... CL is a metaprogramming environment and reflective all the way dylan adds something though 04:20am that's what i wanted to know, what dylan adds it uses more classic AST-thinking when making code.. as it is not sexp it needs to names things: ie expr, classdef, etc and your output must also fit into non-sexp environment ok but what makes this better? the Dylan approach might seem good because it is more AST-style and less of a transfer from classic compiler-work the common people find that limiting compared to sexp huh? you mean cl ppl? so if you depart from sexp syntax and wish to impose restrictions on stuff allowed, dylan macros are more the right alley common people is the common lisp people.. we're quite ordinary and dull.. but we do it Right ;) yeah bigots all the way ;) Dylan is a bastard child of CL i don't see why sexp doesn't already encompass ast i mean, can't you use cl macros to get non-sexp syntax? a sexp isn't necessarily as typed as an AST yuck... typed syntax cl macros can be used to get non-sexp syntax yes.. but it takes some work and it is usually just rewriting to sexp.. ie syntactic sugar there is an infix reader-macro package hm but as I understand Slate you wish to impose a certain amount of semantics and not have sexp syntax, no? well i don't think that specific issue affects slate anyway (fixity etc) the problem right now is that my grouping operators stopped being grouping operators ( and ) now act as namespace stack manipulators you don't even always balance them balancing is good.. especially if you want editor support grr people will write unbalanced () expressions for specific reasons 04:30am and i don't give a flying fsck about ascii editor support the language is supposed to be interface-neutral -:- AlonzoTG [Alonzo@216-164-134-25.s279.tnt3.lnhva.md.dialup.rcn.com] has joined #tunes hell, my objects aren't even supposed to get clos-like repressentations the upside is uniform syntax uniform orthogonal syntax is good the downside is that clos'ers will hate it (well, i'm supposing they will) clos'ers are spoiled brats (we are :) heh too many features in your mop? :) the mop is not perfect but it is lightyears ahead of other message-passing languages for now ;) brb http://www.uio.no/~cynthiss/wwwbilder/win2001-1.jpg heh 04:40am how well behaved can garbage collectors be, if used in a gui? huh? there are threaded and even real-time guis s/guis/gc's using gcs? aha. hell, squeak works fine and there are gc's of small quanta less effective versions that preserve interactivity water: as for oo-systems, as I have expressed earlier, new models are interesting so I am looking forward to slate.. but as I have also expressed please don't make ya language with too poor macros om -:- SignOff AlonzoTG: #TUNES (Have Nice Day :)) 04:50am well, i'm just trying to figure out right now just how far i can go with macros -:- AlonzoTG [Alonzo@216-164-137-209.s463.tnt4.lnhva.md.dialup.rcn.com] has joined #tunes maybe macros to the bottom? :) :) maybe macro calculus macro calculus.. now that sounds like military intelligence well anyway since the only syntax slate has is message-passing, and since messages and objects can be swapped semantically maybe macros can be applied everywhere are macros allowed to introduce something whic is not an object or a message? um what else is there? besides my messages are objects the generated code? low-level? what do you mean exactly by generated code? you do generate code, no? sure you mean language syntactic primitives, i guess are macros allowed to take the slate code and transform it into it's own low-level code? dunno maybe you should figure that out :) umm.. I am asking if you had thought of it and it is distinct from objects and messages no that's why i've asked you about macros it shouldn't be but allowing the macros to generate it's own code is imho important for a reflective system i'm dead serious about that but it screws up semantics yes i know no it doesnt there are other ways to reflect brb.. food 05:00am do you have the generated code as part of the slate semantics? hm not well enough right now it's supposed to be why? well maybe i'm misunderstanding your question but certain stuff in the spec is marked primitive as necessary for the first implementation but is a macro allowed to change that and output non-slate code? ie x86 asm? or do you see macros as tree-transformers? explain, i think i grok but i need to be sure within slate yes? I don't grok slate.. haven't tried too hard last couple of weeks.. but think in imperative message passing.. I have a macro MY-CLASS I use it to define a class um no ok there's lots to explain then the language I use has some rules about class and stuff.. but I, the programmer, knows more about how to make this work better s/class/classes and objects/ yuck dude so MY-CLASS generates the code I want for this specific class slate works totally differently conceptually your ideas of macros probably don't apply yes.. but is a macro allowed to change that 05:10am the idea of macro i'm working with works totally differently tell me how and is still reflective i might add reflective is a gedanken thing.. everything is reflective ;) well first i'd better explain how objects relate to messages :P i mean shifting primitive code base, ok? ok anyway slate's system of objects is actually an immutable tree you only get objects by sending messages to existing objects s/immutable tree/tre of immutable values/ no worries about class definitions or such... an object consists of its own structure can I define message-classes? or message-objects? no well yes but not distinctly messages are first-class objects i mean, they're not distinct at all have you got an example of how to define .e g a MAP message? (typo-day) you don't apply arguments to messages, you send the message an accessor-message setting a slot hm a MAP message takes a list of objects (can be messages) and applies the message to all of them no fsking kidding returning a list of objects which are the results of the applied message for each object i think slate's natural thing isn't quite a map tell you what, i'll get back to you on that you can't have MAP? yes you can but i'm not sure how it would work can I use a macro to make one? the problem is the representation of objects ok a MAP is useful though i know well slate messages obviously support hof-functionality the first HOF people fall in love with yes i know where were we? 05:20am "you only get objects by sending messages to existing objects" no further down you don't apply arguments to messages, you send the message an accessor-message setting a slot the problem with using sexp macros in slate is that the ( and ) symbols don't denote structure, just namespace shifting use [ ] instead then heh what for? sexp doesn't need ().. it can use [] or <> or {} ... no kidding i'm saying why do i need structure if i have my namespaces? take a look at the bottom of the slate's semantics page structure is good url? damn topic abi slate? slate is a unifying/hybridizing of self/beta/lisp at http://www.tunes.org/~water/slate-home.html s/home/semantics/ namespace *is* structure in slate but it doesn't group things with symbols in program text you can't directly get a syntactic representation of any object ouch yeah, clos'ers will hate that my objects aren't lists, their sets... namespaces s/their/they're/ so far, the most promising semantics for slate macros have been higher-order rewrite systems ahem reflective rewrite systems 05:30am can 't you query objects for their structure either? slate will definitely be different from ordinary oo ah (make sure it is useful as well and not die like all other GREAT IDEAS ;) well, you can obtain an object representing the series of messages constructing that object you can also get a flattened representation more suited to getting the object's representation usually consisting of the minimal establishment of object structure and value so I can't get a list of messages that work on a particular object? ie that DRAW work on object SHAPE well it's via the mechanism i just explained with "add-slot" messages defining slots make sense? (add-slot is a primitive of course) ok -:- Mips [somewhere@usr49.tamcotec.com] has joined #tunes so you can query an object for it's structure ie slots yes can you also get info on who it's parent is? the parent is a slot, too a special slot? heh one with special meta-behavior, yes and a namespace contains objects with slots being an object itself, yes :) naturally am I allowed to freeze an object? objects act as slots in the namespace ie allow no more additions? that's meta-behavior yes that's why slate has/needs behavioral meta-objects is a compiler then allowed to make an optimised representation of that object? of course -:- Mips [somewhere@usr49.tamcotec.com] has left #tunes [] 05:40am can I define a macro which generates frozen objects? just like symbol tables (or run-time data-structure of choice) hm (I am now asking all the dumb questions I do with macros so you can see how they apply) -:- Downix [down@d-gnaps-184.ici.net] has joined #tunes you mean "freeze-wraps" objects? (I figured that was the best approach) morning all morning dx Fufie, reading about LISP, I have questions that none of the docs can help me with water: What's up? water: sortof.. makes the object through messages and then freezes it well i know you could even write just an ordinary message to do that downix: ask away dx: workin on slate Fufie: a) How do I do graphics with LISP b) how do I network with LISP? * Downix/#tunes thinks LISP could be used to make a great talker daemon a) depends on which lisp you use.. it's implementation-defined.. as it is in 90% of languages.. b) see a which implementation do you use? Downix: you get a nonstandard library that adds another 20 megs to your lisp image heh Fufie: clisp smkl: fine by me. 8) downix: clisp has a html-page with impelemtnation-defined stuff.. it documents sockets in clisp.. there are gui-libs in the standard clisp bundle, ie garnet and two clx versions water: did you get what I asked for? Fufie: alright. Last thing, how the hell do I compile something I've made in clisp? yes you only need a msg to do it, foof afaik dx: COMPILE, COMPILE-FILE, ... ok heh water: ok.. so through a msg you have a class/constructor? * Downix/#tunes was quite sad that ANSI Common LISP ignored that simple question dx: it does not s/class// dx: page 113-114 fuf: yeah pretty much Fufie: I didn't find it mentioned in the book (flip flip) oh, dammit, I hate it when I miss something dx: I used the index and looked at 'compile' :) you just address the bmo and have it change based on the object in question's structure bmo? abi bmo somebody said bmo was Behavioral Meta-Object (slate) Fufie: hehe. water: will a macro be allowed to do the same, possibly whipping up a bmo to do the job? yes Anyways, coding a talker daemon in LISP is not yet my project, first I have a C daemon to finish.... since a macro could do anything a msg could -:- thomas [thomas@193.217.63.152] has joined #tunes (i mean, that's the idea, isn't it? :) 05:50am a macro usually an do much more s/an// right when are macros called in slate? *when*? yes, when well when we have them ;) but seriously it shouldn't be too hard to apply them to any place in the implementation, but then that relies on having dynamic placement of primitives what I am pondering about is how slate differs between stages i haven't looked too closely at "staging" a macro during compilation usually does different things than during execution for instance, i planned from the start to keep parsing and such above the language level you're speaking pig-latin to me, man i'm not even looking at making stand-alon executables right now ie I use macros during compilation to build up structures which then are compiled and when the compiled code is loaded, all the generated structures are loaded *wham* yeah well we're not doing anything like that right now i didn't say slate was a descendent of self for nothing ok.. because my making the constructor a macro and integrating it with the compiler you would effectively have classes in slate s/my/by/ you're totally talking past me a very simple example on how macros change the norsmal system i'm not even thinking in those terms, man are you listening? I am listening what you're saying totally ignores what i'm working on and I am explaining how macros can be used and really change your slate * water/#tunes listens so if you adopt macros you must make slate changeable don't patronize or limit them just explain I don't patronise.. I am simply telling you that by allowing macros to the playing field makes everything more complex yeah but i already told you lisp is a totally different universe and you still aren't listening I know, but lispers will use macros to change slate if macros are there lispers are never happy with a shrink-wrapped thing ok when i say macros, read "slate's equivalent of macros" it;s definitely not going to look the same yes, and we set out to find out what slate macros did, no? 06:00am heh i did a long time ago so cl macros are hooks to the implementation? smkl: they are hooks to change the whole system smkl: thank you for introducing better terms fufie: less hype, please it's not hype not even buzz just explain how they work mechanically grrrrr shut up and fscking explain explain which kind of macros? regular macros? grrrr i don't know no you don't know and still you claim it to be hype very good arrogant ass CL has at least half a dozen macro types oh yuck the easiest and classic one to use is DEFMACRO heterogeny? you can take heterogeny and shove it no need to be rude yes there is you've spent more time defending your honor than giving me anything to work with DEFMACRO lives in function-space and is called where otherwise a function would be called, but the arguments are not evaluated "function space"? cl has separate places for functions and variables you know the namespace used by functions.. there are different namespaces in CL smkl: he doesn't.. he doesn't know lisp bleh more heterogeny but continue a DEFMACRO returns lisp-code ie sexp wait... when is it called? it is also evaluated in the environment where it was found and can get info about it also is macro-calling transparent? a DEFMACRO is called when it is found in code evaluated so it's a function on sexp's a DEFMACRO is a function on sexp yes i already knew this and macro-calling is transparent with function-calls except arguments aren't evaluated i know because it manipulates sexps but a macro is allowed to evaluate the sexp though "allowed"? yes.. what the macro does with the sexp it is given is up to the DEFMACRO ok then you have the SYMBOL-MACROS bleh 06:10am which works the same way only on what seems to be variables "what seems to be"? (list x)) augh.. (symbol-macrolet ((x '(foo x))) (list x)) --> ((FOO X)) the X in the call to LIST is really a symbol-macro uh sure and is transformed to (FOO X) * water/#tunes scratches his head not used very often but I have found use for it earlier but it's an x not an X lisp is not case-sensitive (in general) k then you have the compiler-macros * water/#tunes rolls his eyes which are allowed to optimise code during compilation stage i don't suppose anybody's ever considered unifying these? e.g they can turn (SQRT 4) into 2 heh ooh compiler-macros can live side-by-side as the SQRT function though s/as/with/ what does that mean? a SQRT DEFMACRO would replace any SQRT function so? SQRT is a function usually but you can through COMPILER-MACROS teach the compiler specified behaviour through compiler-macros even surpassing functions sure, but this all sounds like elaborate fixes for a non-reflective implementation not really it's elaborate fixes for people's real needs so you recompile your lisp environment a lot? :) you redefine primitive? I often recompile the code I write myself at least and when I want changes to basic CL I usually do it in my own packages you have a self-hosted lisp-to-native compiler that you regularly inspect and modify? I do inspect it often and when I am not satisfied I make my own implementation yes but I don't need to implement everything this self-hosted compiler is the actual implementation? just change parts I don't like so why do you fscking need tunes? 06:20am because it would be nice to have it integrated at a low-level.. now the system lives atop unix and unix has its problems well why aren't the macro systems unified?