freenode/lisp - IRC Chatlog

8:04:52 theos morning beach

8:06:03 beach Hey theos.

8:11:59 beach The more I think about it, the more it seems like a really stupid thing to do to use string comparison for checking the type of an object. Why don't they introduce something similar to the class precedence list. Hell, since everything is static, it can be computed at compile time, or at least at link time. Much easier than what CLOS has to do.

8:13:00 beach Maybe this is additional evidence for my standard phrase: "it is impossible to write a C++ program that is both fast and maintainable".

8:21:27 beach Now, the main question remains, of course: Why do so many people in the software industry still believe it is possible?

8:22:25 beach Sorry about the rant. I'll stop now.

8:23:58 Shinmera It seems like people can just become infatuated with C++ for some reason. It happened to a friend of mine and I don't know why.

8:25:50 beach That this can happen to people with no formal training and no budget responsibility may have an explanation, but I can't understand why it happens in a professional setting.

8:26:15 ASau` ** NICK ASau

8:26:40 Shinmera My point is more that personal preferences are highly influencing and people will often do whatever they can to get that preference satisfied.

8:26:45 Shinmera So it spills over into the workplace.

8:28:00 beach I guess so.

8:54:55 drmeister Slime works with cclasp now. I can exit from sldb and it doesn't crash

8:58:34 beach Great!

8:58:50 beach Is that the problem that stassats talked about?

9:04:04 drmeister Yes.

9:04:31 beach What next? Sleep I guess. Then what?

9:04:59 drmeister I've got to get back to the chemistry code and get ready for the Google talk.

9:05:40 beach Oh. When is the talk again?

9:06:47 drmeister Cleavir+Clasp is only 4x slower than C code with inlining. With type inference and a that one T/NIL optimization it will be even faster. That's the proof-of-concept I was looking for to be able to say that cclasp will be a performant compiler.

9:06:51 drmeister Aug 4.

9:06:53 drmeister Not long.

9:07:14 drmeister I've lost a week to this exception handling bug.

9:08:33 beach Right.

9:09:03 beach The 4x is for fixnum arithmetic?

9:09:17 beach Or is it more general?

9:11:38 drmeister The 4x is for fixnum arithmetic.

9:12:04 drmeister But with more inlining it will speed up other things.

9:12:27 beach Can bclasp do inlining?

9:12:38 drmeister llvm inlining.

9:12:50 beach Ah, but not at the Common Lisp level?

9:12:54 drmeister Right

9:14:08 drmeister I'm going to get some sleep.

9:14:17 drmeister I'll talk to you later.

9:14:23 beach OK.

9:40:14 jackdaniel hello all (:

9:41:19 beach Hello jackdaniel.

9:45:14 jackdaniel ccl init file is .cclrc?

9:45:31 Shinmera .ccl-init.lisp

9:45:36 jackdaniel thanks

9:51:48 jackdaniel hm, asdf doesn't seem to be much of backward compatible

9:52:12 sbe hi how do I hide the top window in emacs

9:52:20 jackdaniel top window?

9:52:21 sbe file, edit and those pesky icons

9:52:58 beach (menu-bar-mode 0) (tool-bar-mode 0)

9:52:59 jackdaniel oh, M-x menu-bar-mode

9:53:07 jackdaniel ↑

9:53:09 beach ... in .emacs

9:54:45 sbe tyvm

9:55:09 sbe im on eeepc which is a bit limited with screen space :P

9:55:58 jackdaniel consider changing default font size if you use X client

9:56:00 beach sbe: While you are at it, do M-x define-global-abbrev<RET>tyvm<RET>thank you very much<RET>

9:57:28 dmitigr beach: it's possible to write maintainable code of C++ which is fast (e.g. Node.js runtime)

9:57:37 jackdaniel not everybody use erc :p

9:57:38 sbe Thank you very much beach

9:57:51 beach Heh! You are welcome, sbe.

9:58:05 sbe whats your abbrev for you're welcome?

9:58:12 dmitigr jackdaniel: hi

9:58:17 jackdaniel dmitigr: hello

9:58:18 beach Don't have one actually.

9:59:17 jackdaniel well, you can write C in C++, but what's the point of using C++ then? ;) C++, as something distinct, is a mess imo

9:59:40 dmitigr C++ is a horrible language because of it's ugly semantic.

10:00:00 beach dmitigr: Since we can't measure maintainability, we could argue that point forever. I'll decline.

10:00:04 dmitigr so, it is impossible (or very very difficult) to create useful IDE for it

10:00:31 jackdaniel problem with C++ is that when language creators had to choose between two viable solutions, they implemented both

10:00:44 dmitigr parsing C++ code is painful because it too context-denepdend

10:00:53 dmitigr dependend

10:01:04 Shinmera *dependant

10:01:09 dmitigr right

10:01:18 Shinmera Get a spell checker.

10:01:26 dmitigr Shinmera: :)

10:01:39 beach M-x flyspell-mode

10:01:42 beach Easy.

10:01:53 jackdaniel unless not using erc ^_^

10:02:04 dmitigr I use erc

10:02:42 dmitigr oops, rcirc

10:02:51 beach Sometimes I hit RETURN before having time to look at what the spell checker said.

10:05:35 dmitigr another problem of C++ is ugly template language -- the C++ language unavailable in templates. CL macros much better because the whole host language are available on the compile stage

10:06:03 sebbee ugly template language?

10:06:28 dmitigr CL macros are primitive and that makes them so powerful :)

10:06:49 jackdaniel I'd say they are far from primitive

10:07:10 beach He means built-in, maybe?

10:07:25 dmitigr jackdaniel: look at the Racket macros and see the difference :)

10:08:16 jackdaniel don't ceed into me "looking for a difference", comparison isn't method of measuring sophistication

10:08:22 sebbee isn't racket a scheme and not comparable with clisp

10:08:33 sebbee atleast thats what I've been told last time I mentioned racket

10:08:46 jackdaniel sebbee: you're right, but clisp is implementation, cl is language ;)

10:08:50 beach sebbee: CLISP is an implementation of Common Lisp.

10:09:14 beach sebbee: So you can't compare a language (Scheme) with an implementation (CLISP).

10:09:41 sebbee oh

10:09:50 sebbee i was talking about common lisp

10:09:55 sebbee guess clisp is not common lisp

10:09:57 dmitigr sebbee: C++ templates allow to do some kind of metaprogramming, but the C++ unavailable on the compilation stage. only ugly "template" language are available. you can read The Modern C++ Design by Alexandresku to see what I mean :)

10:10:16 sebbee i have to work with c++ at university at the moment

10:10:27 sebbee i know the pain but where comes lisp in

10:10:30 beach sebbee: CLISP is an implementation of the Common Lisp language.

10:14:12 dmitigr jackdaniel: by primitiveness of CL macros I meant that the programmer can just use the host language i.e. the CL and without any hygiene (e.g. as in Racket)

10:16:01 jackdaniel hm, what hygiene prevents given programmer uses gensyms ?

10:16:21 jackdaniel programmer not using hygienic macros°

10:18:19 jackdaniel ok, I've found, someone can shadow built-in operators

10:21:19 dmitigr jackdaniel: nothing prevents. but hygienic macros are less primitive than macros w/o one

10:21:39 dmitigr jackdaniel: the CL macros are strict subset of Racket macros

10:21:51 dmitigr have to go! see you latter

10:21:53 jackdaniel so you say, that a men in a hat is less primitive then men with a hat?

10:21:58 jackdaniel s/is/are/

10:22:06 jackdaniel s/with/without/

10:23:11 dmitigr jackdaniel: 13:08 <jackdaniel> don't ceed into me "looking for a difference", comparison isn't method of measuring sophistication :)

10:23:30 dmitigr jackdaniel: see you latter bro :)

10:24:02 jackdaniel it is irrevelant as far as I can process - you claim that hygiene is matter of sophistication, I say it is constraint ruling out many things

10:35:03 whiteline_ ** NICK whiteline

10:44:49 scymtym_ Shinmera: re notifying users of upcoming breaking changes: the next SBCL release (or whenever sf recovers) will have a DEPRECATED declaration applicable to functions, variables and types that causes compile-time warnings when deprecated things are used. i think Xach mentioned he would be willing to collect such warnings during quicklisp builds and notify downstream projects. maybe this combination could be a first step towards a

10:44:49 scymtym_ solution.

10:45:15 Shinmera Hmmm. That sounds great.

10:47:02 Shinmera Reminds me to get going on writing a github API lib (unfortunately doing so requires me getting a github subscription) so that things like issue ticket submission on QL build failures can be automated.

10:58:50 pjb` beach: I think that given the VCS and bug tracker histories we have now available, it should be possible to establish a statistical model of the maintainability, in function of the changes requested (and implemented). A tool could be written to analyse project history and determine a "maintainability" coefficient.

11:00:24 pjb` jackdaniel: indeed, there are more than half a dozen irc clients on emacs… http://www.emacswiki.org/emacs/InternetRelayChat

11:24:08 loz1 greetings #lisp

11:24:25 loz1 I want to write a language and have some ideas in mind

11:25:44 pjb` Have you read the dragon book?

11:26:36 loz1 I'm in the process

11:27:11 ggole Do you want to design a language, or construct a compiler for one?

11:27:24 ggole (I wouldn't do both at the same time.)

11:27:32 loz1 the thing is I'm thinking about lisp with types checked in read time

11:28:39 loz1 and there are not many attempts in this field afaik

11:29:02 ggole There are various stabs at a typed lisp around

11:29:59 pjb` Yes, but not in read time.

11:30:18 sebbee What is the 'Dragon' book

11:30:29 pjb` Even amazon knows!

11:30:38 ggole It's an (outdated, imo) compiler text

11:30:42 loz1 even wikipedia knows)

11:31:00 ggole Mind you, I haven't read the latest edition.

11:31:10 pjb` http://www.amazon.com/Compilers-Principles-Techniques-Tools-2nd/dp/0321486811/ref=sr_1_1?ie=UTF8&qid=1437305464&sr=8-1&keywords=dragon+book+compiler

11:31:16 loz1 ggole: shen, typed racket, what else?

11:31:32 sebbee oh pjb` , i got your memo but was unable to reply

11:31:38 ggole Chris Done's thing

11:31:39 pjb` loz1: Common Lisp.

11:31:41 ggole Gimme a sec

11:32:03 loz1 pjb`: it checks types at runtime, isn't it?

11:32:15 ggole https://gist.github.com/chrisdone/516489f4f27846712225#file-anintro-md

11:32:16 pjb` tell that to sbcl..

11:32:22 ggole Bit of a toy

11:33:31 loz1 ggole: is it deriving types?

11:33:43 schjetne loz1: if you just need to change the read-time behaviour there's no need to build a whole compiler or whole new language. Try writing a type-checking reader for Common Lisp.

11:34:01 ggole Yeah, pretty much standard prenex polymorphism AFAICT

11:34:50 loz1 schjetne: sounds like what I can afford in few evenings, ye

11:35:09 schjetne Should be enough to test your hypothesis

11:36:25 schjetne What's your main rationale for a statically typed Lisp? Performance? Correctness?

11:36:51 loz1 correctness

11:37:28 pjb` This cannot be done at read time DUH!

11:37:50 loz1 I was thinking about dynamic language for scripting which I can actually put in use

11:37:57 pjb` If you had sayd for crazyness, I wouldn't mind, but you won't achieve correctness at read time.

11:38:30 loz1 pjb`: what exactly do you mean?

11:39:04 schjetne Besides, static types isn't a panacea, it prevents type errors, and in my experience type errors is only a small fraction of bugs, at least in the software I write.

11:39:14 pjb` When you read: (lambda (x) you cannot know the type of x, because you haven't read the body yet!

11:39:48 loz1 language can force you to specify types in such places I guess

11:39:58 pjb` loz1: Please read: http://www.nhplace.com/kent/PS/Ambitious.html

11:40:13 loz1 I will, thx

11:41:27 lisper29` ** NICK lisper29

11:48:08 rritoch loz1: Clojure already uses non-strict typing by applying metadata. You may want to look at their syntax before you re-invent the wheel.

11:50:27 pjb` rritoch: Common Lisp already has a perfectly good syntax to declare the types!

11:50:32 pjb` There's no wheel to invent here.

11:51:07 pjb` and in anycase, you DO NOT WANT to declare types. You want the compiler to INFER them.

11:55:23 rritoch pjb`: So how do you do that in a case of defun where your trying to limit, or hint to the compiler, the argument type?

11:55:48 theos CL>Clojure

11:56:31 pjb` rritoch: you don't want to limit types. Write generic functions, it's better.

11:57:05 rritoch pjb`: Doesn't that invalidate your previous statement that Common Lisp already has a perfectly good syntax to declare the types?

11:57:20 pjb` rritoch: no. Just: do not use it!

11:59:17 rritoch pjb`: I really can't find an example of it, and I've never seen it done before, which is why I ask. Though I have seen it in Clojure.

12:04:16 rritoch pjb`: I'm also looking for the methods of reading function metadata in general, such as the documentation sring.

12:14:13 loz1 rritoch: is clojure metadata similar to cl's symbol's properties?

12:14:21 pjb` rritoch: (with-input-from-string (source "(defun f (x y) (declare (integer x)) \"does something\" (declare (character y)) (+ x (char-code y)))") (multiple-value-bind (docstrings declarations body) (com.informatimago.common-lisp.lisp-sexp.source-form:parse-body :lambda (cdddr (read source))) (declare (ignore declarations body)) (first docstrings)))

12:16:36 sebbee read a file character by character in lisp?

12:16:50 pjb` Why would you do that?

12:16:55 therik hello

12:17:15 pjb` Use com.informatimago.common-lisp.cesarum.file:text-file-contents to read wholesale!

12:18:08 loz1 pjb`: i have read Kent's article

12:18:29 pjb` Do you understand the difference between read-time and compilation-time now?

12:19:02 loz1 well, yes, and reasons not to mix them

12:19:28 lispguy1 I want to fork from SBCL

12:19:36 pjb` loz1: type inference is done by the compiler at compilation time.

12:20:22 rritoch pjb`: You can only get the docstrings when they're being read?

12:20:30 loz1 maybe its just about terminology

12:20:35 pjb` loz1: you can write a static analyser, that will basically be a compiler that doesn't generate code, but that does all the early phases the compiler does: code walker, environment tracking, type inference. Plus any checking you'd like to do, like global analysis.

12:21:04 pjb` rritoch: here's cl:documentation, but no guarantee.

12:21:08 pjb` +t

12:21:39 loz1 pjb`: that sounds like what I'm thinking of

12:22:53 lispguy1 SBCL's maintainers did nothing to correct the situations with sourceforge unavaibility. I want to fork to start my own Lisp implementation.

12:23:14 loz1 lispguy1: you have my blessing

12:23:18 christoph_debian lol. because auf 5 days downtime of their website?

12:23:22 pjb` loz1: notice that to do that, you will need to read the source like the compiler does it, including macroexpanding toplevel forms, including evaluation of any toplevel form in (eval-when (:compile-toplevel) …), in particular, in-package and setf *readtable* forms, and #. etc.

12:23:22 rritoch loz1: Yes, clojure metadata is mostly identical to symbol properties

12:23:35 pjb` loz1: basically, you're writing a compiler.

12:23:58 pjb` loz1: perhaps you could work with/on sicl.

12:24:51 loz1 pjb`: for now I want something really smaller than cl

12:24:59 pjb` loz1: when I say global analysis, I mean, analysis beyond the compilation-unit boundaries.

12:25:15 pjb` loz1: compile-file must work on a compilation-unit, and perform separated compilation.

12:26:03 loz1 looks like a separation of compiling process to me

12:26:13 pjb` loz1: but what one would want, is to save a lisp image, and therefore, perform an analysis on all the code that has been compiled and loaded in the lisp image, to validate the types globally, and possibly eliminate dead code (three shaking).

12:26:41 loz1 you compiled one file, second, and don't get anything you can run until you compiled whole thing

12:27:08 rritoch pjb`: Yeah, cl:documentation is what I was looking for, at least it is a step forward

12:27:23 pjb` loz1: So your compiler would have to be able to "compile" compilation units, noting all the info it requires, and then have a global analysis phase once all the compilation units have been loaded, before saving the executable image.

12:28:09 loz1 you are making things complex, aren't you?)

12:28:22 pjb` They are this way.

12:28:29 pjb` Alternatively, it could do both, like ecl is able to process a asd file to generate a library or an executable, but it may be less easy for the user.

12:29:05 pjb` loz1: you have semantic problems because of the Turing complete evaluations possible at read time, at compilation time and macroexpansion time.

12:29:20 pjb` loz1: This is why you have to reproduce the minimal compilation algorithm.

12:30:23 loz1 this language is not intended to be metacircular

12:31:11 loz1 at read-time I want to use only what is available on underlying platform

12:31:29 pjb` If you're not discussin Common Lisp, the change channel.

12:32:02 loz1 there is no channel for my language yet

12:32:16 lispguy1 loz1: so create one

12:32:49 loz1 pjb`: maybe you can suggest any channel to discuss such things?

12:33:06 pjb` or ##lisp

12:33:22 lispguy1 loz1: why do you need you own language?

12:33:40 lispguy1 loz1: lets start the own implementation by forking from SBCL!

12:33:43 pjb` when all you want is a global static analyser of CL programs…

12:34:07 loz1 lispguy1: I don't want CL this is the thing

12:34:08 H4ns you don't make anything into being "your own" by forking it

12:34:21 H4ns besides, https://github.com/sbcl/sbcl

12:34:53 pjb` it might be easier to work with sicl.

12:35:14 loz1 pjb`: I don't quite understand what sicl is

12:35:19 pjb` a compiler.

12:35:29 Shinmera loz1: If you don't want CL then you also don't want this channel.

12:35:33 loz1 is it implementation or some kind of description of practicies?

12:35:42 pjb` an implementation.

12:35:54 loz1 on cl itself?

12:36:05 pjb` like sbcl.

12:36:16 pjb` https://github.com/robert-strandh/SICL

12:37:30 pjb` There's only one letter difference.

12:37:43 lispguy1 H4ns: William Harold Newman forked CMUCL to create SBCL. I want to fork in that fashion to create LispGuyCl

12:37:45 loz1 Shinmera: I want this channel cause I can discuss language construction here)

12:37:55 lispguy1 H4ns: what wrong with it?

12:38:07 Shinmera This channel is for CL. If you don't want to write a CL or write it in CL you are in the wrong place.

12:38:07 loz1 Shinmera: if you have a better one in mind, tell please

12:39:00 loz1 Shinmera: I will write first implementation in CL, ok?)

12:39:04 H4ns lispguy1: what is your contribution that warrants putting your own name on something that many smart people contributed to?

12:39:16 Shinmera loz1: Good. Then now ask CL questions related to that.

12:41:00 lisper29 ** NICK vibs

12:41:08 vibs ** NICK vibs29

12:41:14 vibs29 ** NICK lisper29

12:41:46 jackdaniel why fork? I'm sure SBCL will gladly take sensible contributions

12:42:35 rritoch loz1: In miy experience your not going to be able to find much help with compiler design, there are VERY few people who do it. What I can say is that you seem to be headed directly for a dead-end if your going to write your first language from scratch.

12:42:59 pjb` jackdaniel: because that's the normal way with github: you fork, you modify your repo, then you send a merge request.

12:43:08 rritoch loz1: Your better off using a compiler-compiler like yacc, sablecc, etc.

12:43:12 lispguy1 H4ns: I'm not satisfied how SBCL is maintained hence I want to fork. I haven't contributed anything, but what can prevent me to fork?

12:43:33 loz1 rritoch: aren't they just generate a reader/parser for you?

12:43:33 lispguy1 H4ns: the license does not prevents me

12:43:51 jackdaniel pjb`: I believe it's not what lispguy1 wants to do - my question was context-dependant

12:44:10 loz1 lispguy1: you could do it like 50 times instead of talking here)

12:44:15 rritoch loz1: They build the AST for you, potentially saving you a year+ of development time.

12:45:13 loz1 rritoch: I don't quite understand, what is the format of this AST, how can I use it?

12:45:43 rritoch loz1: You need to hit the books. If you don't know what an AST is, your not ready to create your own language.

12:46:00 ggole There's #compilers, but it's very quiet

12:46:22 rritoch loz1: AST = Abstract Syntax Tree

12:46:46 loz1 rritoch: I know that, just how do I put semantics on top of it?

12:47:51 loz1 rritoch: or do these tools generate me the whole thing?

12:48:53 rritoch loz1: It really depends on what compiler-compiler you use, but either you'll edit the generated AST or you'll extend the generated AST, in most cases.

12:49:39 ggole ASTs aren't all that much work

12:49:43 rritoch loz1: I like sablecc since it produces java which is easy to work with, you can literally get your first language working within a week.

12:50:46 rritoch loz1: Just read the documentation of whatever compiler-compiler you use, they explain how to create an implementation from the generated code.

12:50:52 ggole And to get started you can skip lexing and parsing and write ASTs of some test programs by hand

12:51:30 ggole Going back to the start once you know more or less what you are doing.

12:53:07 loz1 rritoch: so the output is some kind of program on my host language which I can use and extend?

12:53:48 rritoch loz1: Yes, compiler-compiler's generate source code for you to edit and/or extend.

12:54:43 rritoch loz1: You write out the syntax of your language, typically in a BNF grammar file, and it generates half of the code for you.

12:57:24 loz1 >once you know more or less what you are doing

12:57:24 loz1 I remember one guy on local meetup said 'when you are starting to write program you often don't even know what do you want'

12:59:30 ggole loz1: oh, irken was the other thing I had in mind

12:59:33 ggole Just remembered it

12:59:55 ggole It's more like ML in Lisp clothing than a statically typed Lisp though

13:00:02 rritoch To get back on topic, are there any compiler-compilers which generate lisp?

13:01:37 rritoch There must be since there seems to be this mistaken belief that lisp is the language to write compilers in, even though I've never seen a single compiler written in lisp, other than lisp compilers themselves.

13:02:21 loz1 rritoch: there is cl-peg

13:04:37 loz1 rritoch: hm, btw, if I write grammar for particular compiler-compiler, does that mean it can be easily ported to different platforms?

13:04:57 rritoch loz1: Cool, well that would proabbly be a good starting point for you, assuming you can find the documentation.

13:06:50 pjb` rritoch: you must be quite blind.

13:07:26 loz1 and one more, are there any more or less popular languages created using compiler-compiler?

13:09:17 rritoch loz1: Your confusing languages with implementations. Most, if not all, modern languages can be implemented using a compiler-compiler.

13:10:40 rritoch loz1: C, Java, and Ruby are languages that I've seen BNF grammars for. There is probably one for lisp also though I immagine it is boring.

13:11:05 loz1 rritoch: ok, I see

13:11:44 pjb` rritoch: no, there's no bnf for lisp.

13:11:49 rritoch loz1: I'm not sure how accurate this is, but here... http://cui.unige.ch/isi/bnf/LISP/BNFlisp.html

13:11:51 pjb` rritoch: there's a lisp reader algorithm.

13:12:00 pjb` rritoch: it is totally wrong.

13:12:01 rritoch pjb`: ^^^

13:12:32 beach Good afternoon everyone!

13:12:39 pjb` rritoch: try in your REPL: (set-syntax-from-char #\( #\a)

13:13:09 rritoch pjb`: I wouldn't say that, it is over-simplified and not-CL compliant, but it is still LISP

13:13:40 rritoch pjb`: Even if you include reader-macros in the BNF it wouldn't be that complex.

13:13:44 beach rritoch: We don't to LISP anymore. We haven't for decades. Now we do Lisp and that grammar is totally wrong.

13:13:49 pjb` LOL!

13:14:45 beach rritoch: If you are going to state things with such certainty, I recommend you check your facts first. There are quite a lot of very knowledgeable people here in #lisp, and they would very much appreciate that.

13:15:10 beach rritoch: And pjb` is definitely one of those people.

13:15:54 rritoch beach: myob, I wasn't disrespecting him in any way. But while he may know Lisp, I know compiler design, which is what we are talking about.

13:16:18 beach rritoch: Then you are off topic. This channel is about Common Lisp.

13:17:20 rritoch beach: And that has already been brought up previously, but not helping someone just because it's off topic is elitest garbage, especially in a dying language like lisp that could use as many supporters as possible.

13:17:58 beach rritoch: But I will remember your domain of competence the day I need to ask questions about the best algorithm to use for PRE that can handle conditional branches in MIR as well.

13:20:03 rritoch beach: No, can't help you there. I don't deal with compiler optimization, I just care about getting things to work, I don't care much about efficiency due to moore's law.

13:20:27 beach I am beginning to understand.

13:20:51 rritoch beach: Most of my work is in web development, efficiency is rarely an issue there since the network itself provides most latency

13:21:08 beach Like I said, I am beginning to understand.

13:21:10 ggole beach: are you starting to seriously write optimisation passes?

13:21:13 Guthur` rritoch: it's apparently ending so you might have to start worrying, hehe

13:21:32 Guthur` mainly because it was never actually a law, just an observation

13:21:42 beach ggole: I already implemented some stuff, but I think it has bitrotted at this point.

13:22:15 beach ggole: I will need liveness analysis soon, and that one is still working, modulo a few API changes.

13:22:39 beach ggole: Not a very serious optimization pass, though.

13:22:51 rritoch Guthur: It isn't ending, and again far off topic. Just google Xion PHI, NVidia Tesla, and AMD firepro

13:22:56 ggole Right. It's not a bad idea to leave it until later anyway.

13:23:10 jackdaniel Guthur`: aren't laws in physical world just models based on observation? ;)

13:23:28 beach ggole: But the liveness stuff is needed in the type inferencer, and that one I am currently working on.

13:23:55 loz1 and they are saying compilers are off the topic!

13:23:56 schjetne Until I can download a faster processor I very much care about efficiency

13:23:58 Guthur` jackdaniel, true, they don't tend to shift as often as Moore's though

13:24:18 pjb` loz1: compilers of random languages. But not compilers of Common Lisp :-)

13:24:22 beach loz1: Common Lisp compilers are on topic.

13:24:39 jackdaniel arguably, but I see your point

13:25:01 loz1 and only processors which can run CL?)

13:25:13 ggole beach: so reaching definitions? I'm actually not very familiar with how to infer types in languages like CL.

13:25:50 beach ggole: It can be done fairly easily, using forward flow analysis.

13:26:06 rritoch loz1: I'm confidient Xion PHI can run CL since it's intel architecture. The rest I'm not sure about since it would require a micro version of lisp that can fit in a GPU shader.

13:26:08 jackdaniel rritoch: claiming sloppy design and not carying about optimization is justified by moors law is bs I hear to often

13:26:18 jackdaniel s/to/too/

13:26:49 beach ggole: Define yourself a finite lattice representing some simplified type system and just do a `join' when your flow converges.

13:27:11 jackdaniel and in a matter of fact experience, when surfing through overdesigned web apps with more js then actual content

13:27:23 rritoch jackdaniel: Code golf and optimization have it's place, but not wasting your time over-optimizing isn't equivalent to sloppy code.

13:27:46 ggole beach: do you do anything like path sensitive refinement (eg, (if (consp x) <x is evidently a cons here> ...)?

13:27:54 beach Guthur`: Strictly speaking, Moore's law is still fairly valid. It is a common misunderstanding that it has to do with performance. It has to do with transistor count.

13:27:57 ggole That seems like the tricky part to me.

13:28:07 beach ggole: Yes, definitely.

13:28:55 beach ggole: In Cleavir HIR, I define an "instruction" called TYPEQ that branches according to a statically known type. That instruction is used by the type inferencer.

13:29:10 beach ggole: Functions such as CONSP are defined in terms of that instruction.

13:29:21 beach ggole: ... and then inlined.

13:29:24 ggole Ah, I see.

13:29:37 beach Pretty simple, really.

13:30:12 Guthur` beach: yep, was aware of the misconception

13:30:26 beach So you just saw your chance, huh. :)

13:30:39 beach I might have taken it too. :)

13:31:36 Guthur` it's been in the news a bit recently with IBM doing some nice stuff in the lab, and Intel having trouble at the fab

13:33:23 ggole I was thinking about introducing explicit refinement instructions and rewriting uses to point to them, eg (if (consp x) (foo x)) becomes (the IR equivalent of) (if (consp x) (let ((_x (presume-type 'cons x))) (foo _x)))

13:34:25 ggole I think your approach is better. This way makes it very easy for an analysis to know about the refinement, but it's a pain to have to "look through" the additional binding to get to the original definition.

13:36:44 beach ggole: I see.

13:36:59 pjb` ggole: (if (consp x) (foo (the cons x)))

13:37:07 pjb` from (if (consp x) (foo x))

13:37:30 beach It will end up being the same after transformation into SSA.

13:37:55 beach ggole: Your approach does a bit of SSA as well.

13:38:09 beach ... by separating x and _x

13:38:11 ggole Yeah.

13:38:35 ggole With similar benefits and drawbacks, I would guess.

13:38:59 beach I would think so, yes.

13:39:16 ggole eg, IR size increase, having to do so-called phi-translation for some algorithms, etc

13:40:53 beach Which reminds me, I still have to go through the literature and check which algorithms that require SSA actually take advantage of the fact that each assignment occurs only once. I know one, but by doing that, the algorithm is less effective than it could be.

13:43:30 beach That work could become an interesting survey paper.

13:47:12 beach ggole: Do you write technical papers for publication?

13:48:07 ggole No, every time I "invent" something good it's been done better twenty years ago -_-

13:48:17 beach Ouch! :)

13:48:36 ggole So far that's happened for PRE, TRMC, and "tagless" GC

13:48:38 ggole Sigh

13:49:17 beach As long as it doesn't happen ALL THE TIME it is fine.

13:49:39 beach Inventing something like that is quite good actually. It means you have good ideas.

13:49:41 rritoch Is there any free source for the metaobject protocol specification, or at least a list of the functions provied by implementations that support it?

13:50:02 rritoch provied=provided

13:50:11 beach http://metamodular.com/CLOS-MOP/table-of-contents.html

13:50:28 rritoch beach: ty

13:50:38 beach ywlcm

13:51:18 beach This is actually the start page: http://metamodular.com/CLOS-MOP/

13:51:27 beach It has some "about" information on it too.

13:52:18 lispguy1 jackdaniel: why to fork? because concurrency is always good. and because I don't want waste my time to ask mainteiners to commit my desire

13:54:07 jackdaniel depends on wording, what you call concurrency I call splitting effort, or even reiterate same work in parallel

13:54:08 loz1 guys, guys

13:54:20 jackdaniel s/reiterate/reiterating/

13:55:16 beach lispguy1: Go right ahead. What is keeping you?

13:55:25 loz1 I just imagined ORM which looks at your database, checks tables and fields and creates classes to access data and everything

13:55:50 loz1 and I'm sure there were attempts to make something like this

13:55:58 loz1 do you know any?

13:56:13 rritoch loz1: Lol, your trying to get a lawsuit arn't you?

13:56:45 rritoch loz1: Here's some free legal advice, AVOID publishing ANY ORM implementation.

13:56:51 loz1 no, I want to check if this idea bad or good and why

13:57:21 rritoch loz1: It's a bad idea beacuse it is already patented, in the most illegitamite piece of garbage you'll ever read, if you take the time to read it.

13:57:22 loz1 rritoch: I dont have any implementation, just got the idea

13:57:50 rritoch loz1: The patent virtually covers any possible ORM implementation

13:57:52 jackdaniel isn't it software, which isn't patentable right now?

13:58:10 rritoch loz1: There is technically "prior art" supposedly found by sun, but google paid their way out of their lawsuit.

13:58:12 loz1 rritoch: are you talking about something particular?

13:58:51 rritoch jackdaniel: Possibly, there's no case-law yet to establish it yet as a fact until someone takes it to court and actually uses the supreme court decision as their defense, and that defense is accepted.

13:59:16 rritoch jackdaniel: It will take someone with deep pockets to go down that road

13:59:26 H4ns rritoch: please take that discussion somewhere else?

13:59:56 loz1 I don't really care about laws and such things

14:00:09 loz1 what about the *idea*?

14:00:43 H4ns loz1: have you looked at the many orm implementations that exist for common lisp yet?

14:00:44 loz1 every ORM I have used made db from your code

14:01:18 loz1 H4ns: CLSQL only

14:02:18 H4ns loz1: actually, now that i think of it, there are not too many that i could recommend to look at.

14:03:01 H4ns loz1: while the idea is rather obvious, making an orm that actually works is a lot of hard work, and the work becomes much less fun once you recognize that orm's basically suck :)

14:03:44 loz1 H4ns: and which way for dealing with db doesn't suck?

14:04:05 H4ns loz1: that said, i think dwim.hu had something that was complete enough to work at some point.

14:04:08 lispguy1 loz1: Postmodern is nice

14:04:19 lispguy1 loz1: and ORM are suck IMO

14:04:45 H4ns loz1: i find postmodern rather nice, too. it just requires giving up on the idea that a language object is a domain object is a database object.

14:05:58 loz1 H4ns: and how exactly does it maps objects on each other?

14:06:31 lispguy1 loz1: why do you want to do mapping? It very painful. Just use SQL

14:06:48 H4ns loz1: i cannot tell you. i wrote an orm in c++ back in the day and recognized that it sucked.

14:07:57 lispguy1 loz1: all that "nice" ORM propaganda is created to force people do overcomplicated job

14:08:19 loz1 lispguy1: well, sql is painful

14:08:25 lispguy1 loz1: haha

14:08:28 rritoch loz1: I'm with lispguy1 on this one. ORM's are extremly inefficient. Since your already familiar with AST's you can actually create an AST for SQL which makes developing complex queries much easier.

14:08:41 H4ns loz1: it is really much less painful with postmodern

14:09:02 loz1 ok, looks like I just have to try it

14:09:02 lispguy1 loz1: SQL is not painful, it's a very simple and powerful DSL

14:09:28 H4ns it cannot be painful, it is a dsl!

14:09:55 lispguy1 loz1: if you like Lisp or functional programming SQL should feels nice

14:10:27 lispguy1 loz1: the best "ORM" for Lisp is AllegroCache, probably

14:10:45 H4ns lispguy1: AllegroCache is not an orm at all

14:10:58 lispguy1 H4ns: note quotes

14:11:34 H4ns lispguy1: the quotes don't make your statement any more valid. you are spreading misinformation. and besides, allegrocache is pretty sucky in my book.

14:13:48 lispguy1 H4ns: it does in sence that with AllegroCache SQL does not needed at all

14:14:04 lispguy1 H4ns: but of course it is not an ORM

14:16:17 lispguy1 H4ns: what book you are writting?

14:16:58 H4ns lispguy1: it is titled "wrong things that people said in #lisp1". today, i have added three chapters already

14:17:29 lispguy1 H4ns: haha

14:17:34 lispguy1 H4ns: nice

14:19:03 lispguy1 so, there is no usable DBMS written in Common Lisp

14:19:07 jdz H4ns: haha, that's a good one. i'd definitely buy that book.

14:19:25 jdz lispguy1: does PostgreSQL count?

14:19:48 H4ns lispguy1: because allegrocache sucks, there is no usable dbms written in common lisp now?

14:20:00 lispguy1 jdz: PostgreSQL written in C

14:20:09 lispguy1 H4ns: tell me please that RDMBS

14:20:23 lispguy1 H4ns: I want to use it

14:21:02 pjb` lispguy1: postgresql was written in lisp originally.

14:21:17 lispguy1 pjb: lol

14:21:21 lispguy1 pjb: I know

14:27:33 lispguy1 AFAIK, Edi Weitz wanted to write a DBMS, but he have'nt time/money for it

14:27:59 sebbee AST? DBMS?

14:28:08 lispguy1 I read his interview

14:28:16 lispguy1 sebbee: wut?

14:29:05 lispguy1 sebbee: AST? DBMS?

14:29:16 sebbee ah, nvm

14:29:19 sebbee you are talking databases

14:55:02 sandybatman Hi, does anybody else have a problem with slimv? I connect the first time around and it does nothing apart from displaying all those messages. I can get to work with it only *after* I connect the second time (again)

14:55:27 sandybatman by connect - I mean the SWANK server. I use sbcl on 64 bit linux

15:37:01 theos everything that starts with Allegro- is sucky in my books

15:41:35 jackdaniel they made possible many things

15:42:51 lispguy1 jackdaniel: but too expensive :(

15:43:33 jackdaniel lispguy1: a see you are first-class troll, therefore I'm adding you to ignore list of my client. Sorry about that

15:43:37 jackdaniel s/a/I/

15:45:28 vibs29 ** NICK Guest70392

15:47:19 lispguy1 jackdaniel: np

15:50:12 Xach theos: get better books.

15:51:39 clique i would like to know which kinds of books

15:51:41 p_l I once got free enterprise licenses for ACL and AGraph for a project

15:51:48 p_l guess how? I asked :D

15:52:22 lispguy1 theos: or get better compilers. For example, GCC :) It's available to download right now, not as sbcl.

15:53:30 lispguy1 @p_l: you are happy!

15:53:43 H4ns was lisp discussed on reddit or something?

15:54:12 lispguy1 p_l: even LispWorks does'nt provide discounts

15:54:21 p_l lispguy1: you'd be surprised

15:54:50 lispguy1 p_l: not, I'm not

15:55:32 p_l GCC is kinda a susprise it turned out as good as it did

15:56:26 p_l anyway, GCC is usually not the topic, so EOT on that

15:56:40 therik Hello, I'm playing with a game I'm making and I want to implement large world map. For this map that contains various entities (moving and stationary), I need a function find-entities-in-rectangle that'd give me collection of entities in rectangle, marked by x1, y1, x2, y2 coordinates

15:57:03 lispguy1 p_l: I just want to add, that new JIT feature of GCC as awesome! :)

15:57:14 therik I'm looking at R-trees and R*-trees, BSP methods and so on, any suggestion about which way to go?

15:57:16 lispguy1 s/as/is

15:57:22 p_l therik: implement somethink along the line of R-tree?

15:57:31 p_l ah, I see you already looked at them

15:57:37 axion therik: quadtrees?

15:58:21 p_l therik: I'd suggest experimenting with few approaches from the area, and see how well they mesh with the rest of your engine

15:59:09 beach What on earth are all the arguments used for in the MOP function FIND-METHOD-COMBINATION?

15:59:10 beach mop find-method-combination

15:59:11 specbot http://metamodular.com/CLOS-MOP/find-method-combination.html

15:59:47 therik ok, so you confirmed that I'm looking in right direction.. I'll definitely experiment with it. Just one more doubt, if you'd be kind to dismiss: Can r-trees or quadtrees handle lot of moving entities on the map?

15:59:57 beach Compare to this one:

15:59:57 beach mop g-f-m-c

15:59:58 specbot Matches: generic-function-method-class, generic-function-method-combination.

16:00:10 beach mop generic-function-method-combination

16:00:10 specbot http://metamodular.com/CLOS-MOP/generic-function-method-combination.html

16:00:34 p_l therik: IMHO moving or not should not be too much of a problem, though it would depend on how well specific data structure handles updating. It would be one of the things I'd be testing for

16:00:38 lispguy1 ** TOPIC Common Lisp, the #1=(programmable . #1#) programming language <http://cliki.net/> <http://paste.lisp.org/new> logs:<http://ccl.clozure.com/irc-logs/lisp/>|contact op if muted| SBCL 1.2

16:00:53 lispguy1 ** TOPIC hahaha

16:00:55 p_l I think updating BSP can be a bit of pain

16:01:31 H4ns therik: maybe https://github.com/hanshuebner/bos/blob/ab5944cc46f4a5ff5a08fd8aa4d228c0f9cfc771/web/quad-tree.lisp can serve as inspiration or basis for experiments

16:01:32 Shinmera * sighs

16:01:41 p_l ** MODE +b *!*@*109.165.53.246

16:01:43 beach lispguy1: Can you please not do that.

16:01:59 p_l anyone got a log of previous name?

16:02:16 H4ns ** TOPIC Common Lisp, the #1=(programmable . #1#) programming language <http://cliki.net/> <http://paste.lisp.org/new> logs:<http://ccl.clozure.com/irc-logs/lisp/>|contact op if muted| SBCL 1.2.13, cl-launch 4.1.3, flexi-streams 1.0.15, Hunchentoot 1.2.34, Drakma 2.0.1

16:03:22 beach p_l: newtolispguy

16:03:38 beach p_l: lispnewb

16:06:18 p_l ** MODE -b *!*@*109.165.53.246

16:06:21 therik thanks a lot guys!

16:06:32 p_l beach: I removed the ban

16:07:09 p_l but I think that noticing that you can change the /topic shouldn't result in immediately changing it to "hahahahaha"

16:08:14 theos * has changed the topic to: dont ban me!

16:09:02 H4ns ** TOPIC Common Lisp, the #1=(programmable . #1#) programming language <http://cliki.net/> <http://paste.lisp.org/new> logs:<http://ccl.clozure.com/irc-logs/lisp/>|contact op if muted| SBCL 1.2.13, cl-launch 4.1.3, flexi-streams 1.0.15, Hunchentoot 1.2.34, Drakma 2.0.1

16:09:43 theos hah! looks like H4ns uses an old irc client

16:09:47 H4ns what's it with all these bottom feeders in this channel lately? reddit? something in the drinking water? paul graham republished one of his 2002 essays?

16:09:49 cohakr ** TOPIC Common Lisp is obsolete. Period.

16:10:00 H4ns p_l: please

16:10:11 cohakr ** TOPIC Common Lisp is obsolete.

16:10:13 theos cohakr is same lispguy

16:10:17 jackdaniel how childish

16:10:35 cohakr hi all

16:10:37 theos H4ns to make things clear, i didnt change the topic. it was a /me

16:10:39 H4ns jackdaniel: you might be so close to the truth.

16:11:02 H4ns theos: thank you for explaining it again! i appreciate it!

16:11:12 cohakr hi all

16:11:14 cohakr hi all

16:11:16 jackdaniel why topic is accessible by ordinary users anyway?

16:11:27 H4ns jackdaniel: because most people are good people

16:11:29 Xach jackdaniel: It has not been much of an issue for 15 years.

16:11:50 jackdaniel ok, both arguments are perfeclty valid

16:12:01 theos ** TOPIC Common Lisp, the #1=(programmable . #1#) programming language <http://cliki.net/> <http://paste.lisp.org/new> logs:<http://ccl.clozure.com/irc-logs/lisp/>|contact op if muted| SBCL 1.2.13, cl-launch 4.1.3, flexi-streams 1.0.15, Hunchentoot 1.2.34, Drakma 2.0.1

16:12:06 p_l ** MODE +b *!*@www.nisys.be

16:12:47 p_l jackdaniel: it's due to using topic for announcements

16:13:05 sebbee is sourceforge down for anyone else?

16:14:03 p_l sebbee: sourceforge as a whole is dead - their storage arrays uptime gone like tears in the rain

16:14:05 jackdaniel sebbee: it's down since yesterday, or even day more

16:14:37 sebbee well that's unfortunate for my sbcl installation efforts

16:14:48 p_l yeah

16:16:09 p_l also, at this rate, we might need a ban-tracking not

16:16:49 lambda-smith sebbee: Sourceforge is accessible to me

16:17:00 marekk lambda-smith: it's in static mode afaik

16:17:19 lambda-smith marekk: really? so the effort to kill it has paid off?

16:17:30 marekk who tried to kill it?

16:17:34 lambda-smith but what about sbcl?

16:18:07 p_l they pulled a killswitch on themselves when they laid off old staff

16:18:15 lambda-smith marekk: there was a lot of uproar when people were aware that sf bundle adware with the installer and hijack project websites

16:18:38 schjetne Didn't they start injecting malware years ago? I've always wondered why it took so long.

16:19:04 therik whoah, I see quad trees are nice simple option, but it seems like if I have full map of items, and I have items that span more than one tile and I can have multiple items per tile... The whole tree is gonna be big

16:19:24 lambda-smith this seems like a good summary of the (latest) incident http://www.infoworld.com/article/2929732/open-source-software/sourceforge-commits-reputational-suicide.html

16:21:53 p_l joel135: you need to quote it like '(1 2 3) or (quote (1 2 3))

16:22:42 sebbee thanks for the link lambda-smith

16:24:00 p_l minion: tell joel135 about pcl

16:24:00 minion joel135: please look at pcl: pcl-book: "Practical Common Lisp", an introduction to Common Lisp by Peter Seibel, available at http://www.gigamonkeys.com/book/ and in dead-tree form from Apress (as of 11 April 2005).

16:24:12 beach p_l: Was that remark as a result of a PM?

16:24:25 p_l good resource if you program already in another language

16:24:46 beach p_l: I don't see any uttering by joel135.

16:25:11 lambda-smith Pardon my ignorant, but who is lemonodor?

16:25:16 p_l beach: it was lost in talk about sourceforge

16:25:25 lambda-smith Saw his name in the SLIME's start quote

16:25:28 p_l lambda-smith: a famed lisp blogger

16:25:55 lambda-smith p_l: I see, it seems that he is no longer active, is he?

16:26:10 lambda-smith Also, did he worked on any projects?

16:27:20 p_l lambda-smith: the quote is related to how it used to be that the major way of getting publicity about your project was to be mentioned on lemonodor

16:27:50 lambda-smith p_l: Oh, I see

16:28:02 lambda-smith Thanks for the context, p_l!

16:28:43 lambda-smith Too bad he seems to be no longer active

16:29:40 sebbee but i am active :)

16:29:45 lambda-smith :D

16:35:10 sebbee i have yet to make progress on lisp

16:37:15 p_l as for where bottom feeders come in - there was a project that got linked from HN

16:39:04 p_l btw, about sourceforge issues: https://www.reddit.com/r/sysadmin/comments/3do9k0/sourceforge_is_down_due_to_storage_problems_no_eta/ct77o49

16:45:00 lambda-smith p_l: Oh my God, that's a nightmare

17:04:15 akkad * ponders why any project would still use sf

17:05:53 lambda-smith akkad: SBCL is still on sf, though

17:05:59 lambda-smith as well as Maxima too, I think

17:06:17 lambda-smith Oh wow: http://www.sbcl.org/

17:06:27 lambda-smith You can't even reach SBCL right now

17:07:16 lambda-smith Maxima as well,

17:08:47 jackdaniel akkad: changing hosting involves a lot of burden

17:09:09 k-stz What is the meaning of the exclamation mark as a naming convention: (foo! 1 2) etc. ?

17:09:37 jackdaniel it is schemism, it hints program reader, that it is destructive function

17:10:08 jackdaniel and (foo? 1 2) indicates, that it's predicate

17:10:28 k-stz thanks

17:10:30 jackdaniel CL uses (foo-p 1 2) for latter, while former would be (nfoo 1 2)

17:10:45 jackdaniel but it's convention, I personally find ? and ! more appealing and readable

17:11:01 k-stz nfoo looks bad, I think nobody follows this convention outside the standard function?

17:11:13 jackdaniel you're probably right

17:11:20 Brucio-92 ** NICK beslyrus

17:11:35 Bike it's also not terribly consistent, hence no nappend

17:11:39 k-stz also the "n" hasn't even a meaningful mnemonic iir pcl correctly

17:12:13 jackdaniel if you start project from scratch, you won't infridge consistency if you'll use "?" and "!"

17:15:12 akkad sourceforget.net is the place to stay then :P

17:18:54 ghost_runner On a scale from one to ten, rank SICP. Just curious as to what you Lispers think about the book.

17:19:09 Bike nine

17:19:49 ghost_runner Hmmm... What prevents it from being a ten for you?

17:20:23 Bike habitual distrust of perfect ratings

17:20:32 jackdaniel :)

17:20:41 k-stz don't find the time to read it; still at chapter 2

17:20:54 k-stz the video lectures are nice

17:26:35 ghost_runner Bike: That being your only reason? :P

17:28:37 k-stz from just the first chapter it taught me stuff im grateful to know. I think the problem with rating a book like sicp is to weigh the didactic representation of knowledge with the knowledge present in the book.

17:30:41 k-stz "What knowledge does the book teach" "how well does it teach it"

17:31:55 jasom * has never read SICP

17:33:54 jebes read it.

17:34:05 jebes now.

17:35:21 ghost_runner jebes: What do you think? :P

17:35:32 Bike don't need to be pushy

17:36:24 ghost_runner wat

17:36:31 wat yes?

17:36:39 wat :)

17:37:49 ghost_runner lol

17:42:40 clique critical reading

17:44:55 DrWat wat

18:43:01 phf is david lichteblau still maintaining closure-html? is there official fork that quicklisp is pulling from? i have a lot of patches, that i produced as a result of running closure-html against a lot of html from the wild, and i'm not quite sure who to send them to

18:46:14 PuercoPop phf: you could always check in https://github.com/quicklisp/quicklisp-projects/ to see where it is being pulled from

18:47:30 PuercoPop it appears david lichteblau is still maintaining it

18:51:28 phf ok

18:52:09 phf come to think of it, i haven't actually tried simply emailing him

18:53:23 ziocroc2 ** NICK ziocroc

18:54:21 sebbee u should try that

19:13:01 therik any way to put q-trees into an array?

19:13:21 therik uh, now I'm thinking, it's dumb, cause it's not balanced tree

19:16:01 therik any other way to efficiently store q-tree nodes, that wouldn't involve keeping all the pointers around, together with info about sizes/levels/positions?

19:18:46 therik I mean, there's this trick with keeping balanced binary tree in an array, where parent node is (floor (/ i 2) and children are on (1+ (* 2 i)) and (+ 2 (* 2 i)), I wonder if there are some tricks for q-trees

19:20:52 ggole q-trees being quad trees?

19:24:10 clique adopting bthta notation an o-tree is an hilbert q-tree

19:24:55 clique that, uff

19:25:32 therik ggole, yes

19:27:46 ggole I think you can use the skip trick to represent n-ary trees without pointers (at the expense of making insertions difficult)

19:28:58 ggole The idea is that you represent a tree node as whatever info it contains, plus the offset to the next sibling

19:29:12 ggole Children are simply the next thing in memory.

19:34:16 therik yaix, that doesn't sound very convenient

19:34:38 ggole Traversals are easy, insertions...

19:34:41 ggole Don't ask about insertions.

19:35:36 therik to get to children and siblings would be ok, to get to parent I'd have to move back until I hit something that has pointer to sibling that's after the element I search the parent for

19:36:09 ggole Or always maintain a stack of positions

19:36:16 ggole Which may or may not be possible.

19:37:56 therik honestly, I'm not quite sure what I need yet, but I'm aiming for a self-expanding map that can potentially get huge, and the first task I decided to tackle was to find set of entities on the map in certain radius from a given point

19:39:35 ggole That suggests kd-tree

19:40:23 ggole I suppose quadtrees would also work

19:45:13 pjb` Bike: it should have been twelve.

19:47:57 pjb` therik: so you will add objects in your map at the bacterial level, then at the atomic level, then at the string level?

19:48:55 pjb` therik: there's also #lispgame where you can discuss game implementation strategies.

19:57:33 therik pjb, seems like that.. well, at the beginning, all the map is going to be empty, once bit of it is generated, that portion should get indexed. I'll have small tiles, probably 10x10 pixels. Right now, tiles are in leaves of the tree, but I'm thinking of splitting the info from tiles into tree nodes

19:58:12 pjb` Sorry, #lispgames

19:59:43 pjb` therik: I mean, you should decide from the start the granularity of your game. If you have a grid of 1 meter², then your quadtree will have cells corresponding to 1 meter², and you won't be able to put billions of objects in that square.

20:00:00 pjb` therik: now, if you discover another continent, you just add a new quad tree.

20:00:03 pjb` or another planet.

20:00:32 pjb` Remember, planets means moving body, and continents move relative one to another (plaque tectonics).

20:04:09 therik I'm not really implementing moving continents

20:04:27 pjb` In anycase, you wouldn't grow your tree from the leaves, but from the root. So you could have two levels, one a small quadtree of quadtrees where you use normal nodes with pointers, and big quadtrees of objects of fixed size that you store in arrays without pointers.