IRC log started Wed Jan 27 00:00:00 1999 [msg(TUNES)] permlog 1999.0127 שםש abi [nef@bespin.ml.org] has joined #tunes !netgod:*! mccaffrey will disappear for a few min presently !wichert:*! I didn't know you were a practising magician.. שםש Closing Link: TUNES[bespin.ml.org] by asimov.openprojects.net (Ping timeout for TUNES[bespin.ml.org]) שםש Connection closed from irc.us.openprojects.net: Success שםש Connecting to port 6667 of server irc.us.openprojects.net [refnum 0] שםש Your nick [TUNES] is owned by tunes@bespin.ml.org שםש BitchX: For more information about BitchX type /about שםש Welcome to the Internet Relay Network TUNES_ (from sterling.openprojects.net) שםש Your host is sterling.openprojects.net, running version u2.10.04.resolv9.nmt.egcs4.tok.pten.tlim4.admin.upper4.whisper3.gipl.modeless7 (from sterling.openprojects.net) שםש This server was cobbled together Wed Dec 2 1998 at 02 26:39 EST(from sterling.openprojects.net) שםש sterling.openprojects.net u2.10.04.resolv9.nmt.egcs4.tok.pten.tlim4.admin.upper4.whisper3.gipl.modeless7 dioswkfcg biklmnopstv שםש [local users on irc(6)] 3% שםש [global users on irc(64)] 36% שםש [invisible users on irc(116)] 64% שםש [ircops on irc(13)] 7% שםש [total users on irc(180)] שםש [unknown connections(0)] שםש [total servers on irc(27)] (avg. 6 users per server) שםש [total channels created(54)] (avg. 3 users per channel) !sterling.openprojects.net Highest connection count: 8 (6 clients) שםש Mode change [+f] for user TUNES_ שםש Mode change [+iw] for user TUNES_ שםש TUNES_ [tunes@bespin.ml.org] has joined #tunes שםש Topic for #TUNES: metaprogramming bites שםש topic set by AlonzoTG [Tue Jan 26 21:15:37 1999] שםש [Users(#TUNES:3)] [ TUNES_ ] [ abi ] [ TUNES ] שםש Channel #TUNES was created at Wed Dec 30 09:08:43 1998 שםש BitchX: Join to #tunes was synced in 12.246 secs!! שםש SignOff TUNES: #TUNES (Ping timeout for TUNES[bespin.ml.org]) שםש Current server: irc.us.openprojects.net 6667 שםש Primary server: irc.us.openprojects.net 6667 שםש Server list: שםש [unknown] שםש 0) irc.us.openprojects.net 6667 (TUNES_) שםש [Users(#TUNES:2)] [ TUNES_ ] [ abi ] .......................LOGFILE 1999.0127 >>> You(TUNES_) are now known as TUNES שםש TMF [s720@tordivel.ii.uib.no] has joined #tunes שםש Tril [dem@bespin.ml.org] has joined #TUNES שםש Tril has changed the topic on channel #TUNES to: http://tunes.org/ - TUNES Project | http://bespin.tunes.org/free-services.html - Hosting for Free OS Projects tril, I just read that your work in tunes is your bachlors degree. How do you do that? שםש Tril has changed the topic on channel #TUNES to: http://tunes.org/ <-- TUNES Project || http://bespin.tunes.org/free-services.html <-- Hosting for Free OS Projects Fairhaven College is part of Western Washington University, and it is an experimental school in higher education formed in the early '70s. 10:20am so, you tell them what you want to do, and they consider it an education? there is no grading for Fairhaven classes, and Fairhaven has a degree called the Fairhaven Concentration where you choose an area of interdisciplinary study, and design your own degree with your own committee wow! for the concentration you have to take a good number of classes at the regular university, but you take ones that relate to your concentration. do you get assistance yeah sounds like a better deal then I'm getting! :( it's very friendly you can talk to an advisor at any time there are lots of prepatory classes to make sure you are on track and know what you are doing etc currently I have to take bogus classes just to get student _loan_! i'm very glad it was to be found near where I live! where is that? that is a system abstraction, but not sure. thanks, abi. Tril: bitte In germany? Bellingham, Washington . halfway between Seattle, and Vancouver British Columbia ok Tril, have you looked at TOOL, by the way? ahh..I need to read the last mail you wrote to me about it. my mail is in a weird state right now yeah, sure it's all tarred up from moving hard drives around. hold on while I get it accesible 10:30am found it 10:40am what was it about? Subject: The specification of TML ohh. The TML bible:D ok I understand a macro.but what about these macros inside macros? When are they expanded? what is the order of evaluation and time of evaluation etc the whole inside of a macro replaces the an occurence of a id if the inside of a macro has macro definitions the these are invoked but the scope of the inside definition is just inside the current macro? nono. the scope is only dependent on ()'s so if I write [ message [ from ] ] and then type, message, somewhere... it expands all the macros completely until none are left? so it comes out as a null string in the input after having the side effect of writing "TRil" to output it would expand to '[ from ]' which would define 'from' to be '' so every time I type 'message' it defines a from macro with global scope. Got it. only when you write 'message' and then 'from' it would expand to '' and write 'tril'. yepp ( well some current local scope, anyway ) now I need to run this thing..do I have the latest version? you sent me 2 copies of the same version, at different times latest-release.tar.gz :) I think so. Have you compiled it? there the same. I had at one time. I have to extract it from e-mail and compile it again, now. oh, current-release.tar.gz. there it is :) 10:50am hold on gotta get something to eat ok started something cooking now uh it didn't compile I don't remember if I got it compiled before or not what compiler/os are you using? undefined reference to 'sin' 'cos' 'tan' and 'pow', gcc version egcs-2.90.29 980515 (egcs-1.0.3 release) Linux tril.lnx 2.0.35 #1 Fri Sep 18 22:53:06 1998 i486 unknown try g++, if you got it hey that worked. It should be .cc if it requires g++ :) ooops :) I just wrote Hello, world in TML correctly the first time :) I don't suppose it would be too much trouble to make a lisp-like read-eval-print loop, would it? you're up to it!!!! In a way, '()' is it. well, I'd like it to have interaction, it would make it easier for me to learn it. I know, It's pretty easy though, to make it interactive. Just haven't gotten around to it yet. 11:00am reading the code.. In an interactive version, there would be an initial/global scope which was the harddrive current directory. Curretly it is possible only to read files, not write them ( except output ) hard drive? I was thinking of just a non-persistent interpretre yeah, like if you write '[ filename contents.... ]' it writes contents to a file named filename have you used lisp? seen it, never used it I want to get a prompt, type a TOOL expression, and have it run that and give me the prompt back. So I can use macros in later commands I mean TML expression yes, then you want <>'s to be interpreted as write to stdout in this initial context If I want to save some macros or objects or whatever, I should have to put them in a module and mark it as persistent (or manually save it in current systems) or send its definition to some persistent context, like a hard drive I also need to separate the read-input loop from the current state of the VM so I can run the interpreter more than once, updating the state each time basically take the macro_list outside of go_parse err, something yes, that's the place to do this thing! or just call parse multiple times with output set to stdout that would work, but how about files? currently: hold on phone 11:10am back if you write a filename, that file is invoked so you want to write files by redefining them like macros if you write a file in what syntax? [ filenam contents ] 'contents' would be the contents of the file. how so? I thought that defines a macro yes, but I could "overload" its meaning, just in the outermost scope for every macro it checks to see if there is a file by that name? in the scope where <>'s would cause writing to the outupt file. does it do this or are you just thinking of doing this it does. create a new file in the directory you are in! put some TML code in it have another TML source "call" the file, and it works as if it was a macro! do you have full file system support (chdir mkdir rm ls ...)? no, just this file transparency if you are conflating files with macros you should have a "list macros" = ls and "chdir" = change scope for reading the semantics should be the same if you dont want all macros to be files then there should be a special form for the file access if you look at TOOL, you will see that it's objects are on the form [ objectname identity structure state ] It's not that I do explisit file access. It's just that it's convienient to say that the directory you're in, is part of the inititial scope what does an object have to do with a file? look at how the object would be written to a file! I haven't even gotten to the () operator yet, much less TOOL! 11:20am it's just that object mab well to files, too. So well infact that in TOOL you would never need explisit files. much less mkdir and ls! but, read about ()'s. It will make it all clearer. () redirects the output of an expression to the evaluator? it reevaluates the output of what's inside the ()'s it's a trick:D But if I define a macro inside (), it doesnt affect the outside unless it has output? but if it does have output it has to be evaluated.. that seems to me to be combining two features into one. 1. nesting scope 2. evaluating an expression Yes like perl, you create syntax which is convenient for you, but is not designed for the long term general-purpose use. what if I wanted to do one, but not the other? (< [ this is an example ] >) this -> is an example this -> is a rule :D look at ()'s as a way to interpret something then the way you interpret something is dependent on the context. ()'s can be seen as a meta context, which you specialize by defining/redefining terminology used to interpret things in the context. 11:30am but the specifics of one context shouldn't implisitly effect the global context ... confusing well, I'm aware of this combining of two things in one, and I have considered splitting it into two constructs, but during the design of TOOL, I've never seen the need in practice I want to see more examples of using TML without TOOL, some common programs that do simple things like loops, conditional execution, arithmetics ? yeah, I have a couple of those in a message you sent, but I'd like to see more. Perhaps some programs that combine 2 or 3 of the above anything longer than a snippet try writing a loop, by recursive macro definition which terminate did you write a turing machine in it when you proved it was turing complete? I'll look at that.. :) yeah, well, never tested it on anything. you can have the source, and look at it. 11:40am I'm looking at it now. It's not a good example. mostly concerned with maintaining uniqe integers for tape positions. I just wrote a read-eval-print loop using shell script. Only problem is it erases the macro definitions every time. yes, you want to store them somewhere - not easy. haha.. I just crashed TML 11:50am how did you do that? [ loop loop ] ( < loop > ) Out of memory ok, I don't detect recursive definitions. how much memory does linux give programs anyway? don't know, but [ loop loop ] will use it in a hurry ( I'll be back in five minutes ) Looks like about 8 megs before stopping must be encoded in the binary, because netscape uses more than that :) why does [ ] () not print bar 12:00pm and why does [ ] (<>) print foo??? [ <...> ...] isn't defined hey, stop finding all those bugs, man! (<>) should print nothing, right? it should print foo <> prints that is ok because (<>) evaluates to the (<>) works correctly, it's just that [ ...] does nothing ack! I got it to read a file, accidentally! () has an error because the file "foo" has errors that could be it anyways..what about this (<>) what exactly does it do in what order nevermind I think i understand it I guess the best example is TOOL itself, then? yes, TOOL's a good example. how the heck did you debug this? how did you get the right punctuation..did you define a TML mode for Emacs? (I'm looking at the tool source) I implemented TOOL in TOOL code and hand translated it to TML It's a bitch to debug right! so... now you just need to implement reflection for TOOL on top of TML 12:10pm TOOL _is_ reflective so you have the ability to manipulate TOOL's definition (the version written in TOOL code), and it will take effect in the TML version? tool can generate its own TML code now, right? TML and TOOL are not different versions yes is there a way to get the definition for a macro in TML? TML is the complete meta layer, while TOOL is the consistent object layer no, not currently, but once TML interpreter is written in TOOL, this is possible. TOOL is a generic compiler, which currently compiles to TML code, but it may well be made to write C code, for example what OO language is TOOL most like? Smalltalk: that's what I thought.. the tool file can be seen as an image In comparison, though, smalltalk is somewhat of a hack :D so you use a message passing metaphor? objects contain state and methods? yes in a way 12:20pm methods are objects too btw, where is the source for TOOL written in TOOL you write it that way, so it must be easier to read wrote I haven't got it here and it's not complete ok. go on about methods and state >>> Tril [dem@bespin.ml.org] requested PING 917468676 676095 from #TUNES state is composed of the state of all the dynamic attributes that an object has static attributes are not part of state are there classes and instances? the difference betweem classes and instances is only that instatnces have no static attributes. there is no practicle need for this distinction, so in the new version of tool, it will no be there ( yeah, I'm working on a better TOOL, now that I have the experience of making it work ) do you know the lambda calculus? the lambda calculus is abtract, and doesn't seem very useful at first sight. still you can make useful things with it NOT YOU DOOFUS I know what it is, but I have never used it. how is it different from traditional OO? 12:30pm lambda calculus isnt OO it's a model of computing. I think 'tis a lot like TML ok, in what way? I dont know TML well enough to explain. but it is based on expansion of macros I'm searching for l.c. now. do you know any quick reference :| ? Fare likes this one: http://www.cs.rice.edu/~matthias/ I have more if you need them I can't find anything explisitly on l.c. 12:40pm you're right..he has some papers about LC in the Papers section but they don't look introductory..here's some other links: http://www.math.macalester.edu/~vsharma/papers/cs47_lambda.html http://www.cs.monash.edu.au/~lloyd/tildeProgLang/Paradigms/Lambda/Ch/01.Calc.html http://www.eg.bucknell.edu/~cs208/subpages/Problems/LambdaCans/LambdaCans.html ok, I got one introduction that I'll print and read. you said a method is an object, I think that is good, but how does it work? Can you put any object "inside" any other object, and what does it mean to do so? having a method object inside another object just means that the method is some kind of attribute a method in itself, contains 'action', a token, and two classes : the input type and the output type does that mean 4 things or 3 (is the action a token or are they two different things)? 12:50pm the action is a token. a token is a string that can be evaluated. I might well call it 'string' or 'text', since token is not a good word eighter. 'message' mabe. whatever it is called, it should be an object for consistency it's an object all right! even the punctuactions are objects! to alter the meaning of '.' in the 'object' context, write ( object ( '.' ( meaning ( = (:<...new meaning here...>:) !) .) .) .) not just the character cells, but the language structures denoted by the whole string should be objects too. And those objects should be usable without the string, since the string just describes them. ? I don't like strings being the primary representation for source code. When the string is parsed, the parser forms abstract structures which represent the structure of the program These structures should be directly editable by people and programs! in fact in my system there won't be strings to start with, just structures (made out of objects) yes, but these structures _are_ the source imagine that you are writing an abstract syntax tree directly exactly. instead of something that will be translated into one so, what I was saying, is ideally a method doesn't have a string but instead an abstract syntax tree (object). yes, I see what you mean 01:00pm and it might be worth looking into! I think it makes everything simpler how would you store the ast in the method? Mainly, in your case, I am concerned about not typing all that annoying punctuation! well the ast is just a tree of objects. How do you store the reference to the input/output types? I imagine you'd do it the same way. Punctuation, operator precedence, etc, is all the syntactic details of a language. These details are what make languages incompatible, even when their ideas are so much the same! yes, but the syntax needs a certain overkill, in comparison to syntax for treating numbers, in order to be able to manipulate general object sorry if I am talking on two subjects at the same time. that's three. But it's ok:) I'm trying very hard not to bring up another! :) hit me! well first what do about the storing of the AST, how DO you store references to input and output types? a reference is ... a STRING ! :DDDD well, I don't like that, for reasons I have postedf about before. But ok. but seariously, if you look at the way TML works, the "strings" I'm talking about is more than that. they _are_ the code at one level, viewed from somewhere else I need to get a better grasp at the way TML works. I'm interested to heear you compare it to LC, but first maybe we can just teach it to me a little better I would more than very much like to teach it to you! LC? what is your offset to GMT (i.e. when are you leaving) 01:10pm lambda calc, sorry ok, l.c.! שםש Yery [ost@benetnash.ffke-campus.mipt.ru] has joined #tunes it's 10:10, soon midnight are you at work now? hi Oleg, we are talking about TMF's TML I'm at home I have class in 3 hours but I can talk until then OK, so you are +0100 hi. what is TMF's TML? I'll be kicked out by security in 1 hour 50 minutes. I'm at a computer lab at the university in Bergen Yery, it's my baby:D :) baby I guess you didn't post much about it on the list. anyway, I understand <> and [] alone... no, frankly, I'm a bit protective. ( read paranoid :) does <> do anything except print what's in between, to the current output, exactly? no, just copy the characters one by one any macro expansion, or treatment of other operators inside it (like other <>, or [] or ()) ? I think it does all of the above no, it just copies them, without thinking hmm, maybe not! why? I just tried it :) what happend? > < ( < [ foo hi ] foo > ) > ( < [ foo hi ] foo > ) > (the first > is my custom prompt to TML) 01:20pm so is the second one err, the last one it just copies, right? yep how about macro definition. [ m def ] just copies def exactly as it stands, into the definition for m? if I use any other [] <> () inside def it gets stored inside the macro definition, right? actually, it just remembers 'm' and the position right after. On macro invoke, it jumps. well, that's the implementation (doesn't matter, and it might change) yes anyway, it doesn't modify anything or try to build some meta data okay, next is macro expansion.. if I define m, then use m by itself it gets expanded. the definition of m is evaluated instead of m, exactly as if it had been there in its place, right? yes well, that's easy. the only other thing is the (), right? TML _is_ easy, and fun too!) yes, but you might notice something that is impossible to write what could that be? שםש SignOff Yery: #TUNES (Ping timeout for Yery[benetnash.ffke-campus.mipt.ru]) a macro cannot expand to its literal name, instea it gets evaluated endlessly well, no, you can [ m ] I dont know anything you can't write, if you mean characters, because anything can be put in a <> literally. what about '<'? שםש Yery [ost@benetnash.ffke-campus.mipt.ru] has joined #tunes tml starts eating up ram.. well, that's not what you said then. You said anything between <> is taken exactly as it is. You mean, anything except for < and > ??? (btw, it is eating up ram a lot SLOWER than when I did [loop loop ]) 01:30pm yes, duplicating one character at a time < < > -> out of memory !!! should I say, "< < >" = out of memory you said I can't print a <, so I tried it, and you're right. Why is that? well, you said "what about <", implying I can't print it < < > > prints "< >" just fine. yes, < and > is the characters that you cannot print it's because <>'s are nested, like all the parenteses in TML that seems like it is bad. is it good? it's a nessessity why, just because it is easier to program for some reason? no, because I don't have enough information to stop. you should just see a <, then find the right most >, and print everything in between. you do know where the END of the input is, since you require it to be a file! yes, but in most cases, that would mean the whole source code so, '<' would mean print until you shoke on it! it should stop at EOF, anyway. That's a bug. yes, another one. TML.c cares little for incorrect code :) but it's not a problem to fix you aren't going to fix? why do you have a list of bugs? someone might fix them for you, if you post the source with a list of bugs :) שםש hcf [nef@escher.sdi.agate.net] has joined #tunes I assume it is because you are going to implement TML in TOOL, and TML.c will be unneeded then? yes, basically, TML.c is just a bootstrap but, the <<> problem has nothing to do with the bug. 01:40pm the interresting thing is: ok, so let me get this right: inside a <> it prints everything exactly, BUT any block structure punctuation must be in matching pairs, and correctly nested, including <> () and [], am I right? only <>'s <[ hello > is legal why? don't you parse all the blocks simultaneously? obviously you don't. <<> isn't really illegal, it just doesn't mean what you at first think it would, and then you discover that something is infact impossible to write interresting: <<> and <>> are not nessacary when implementing a turing machine but: when implementing TOOL, you'd need tables beyond the frontier of the known universe to get it to work TMF: I think TML is overkill, because lambda calculus is turing complete with ONLY macro expansion. You could implement all the other operators using just macro expansion. (and you should, too! :) does this mean that TML with the option to write <<> and <>> is more complete than a turingmachine. Could a turing machine simulate TOOL* It's not possible to implement an assignment with just macro expansion ok, you're right, macro expansion AND macro definition. for example x = x + 1 the notion of macros is not adequate. just using macros, x would here be recursively defined! l.c. has something called alpha-conversion, which means the macro names are renamed to not conflict. even if I did [ a x ] [ x a + 1 ] it would still be rec.def. besides, you are mixing up assignment with macro definition. They are not the same thing. 01:50pm But I need to "emulate" assignment, because it is not a primitive operation. ( < [ x > ++ x <]> ) does the job. how? wel, that is beyond my expertise, because I have not actually built a turing machine in l.c., but that's what Church did or someone, maybe not Church [ a <[x> ] [ b <]> ] [c a ++ x b ] c , that is exactly what I did in my reflective C program..it's a hack. exactly! () makes it easier, though. speaking of that I didnt get to it yet I have: "( exp )" evaluates exp in a temporary context, then evaluates the output in the current context." is that ok? yes, the temp context has one level deeper scope than the current context wait, what I wrote must be wrong somehow, otherwise you wouldnt need the () operator.