libera/#lisp - IRC Chatlog

12:53:59 aindilis_ hey folks, I am always doing Knowledge Representation and Reasoning stuff, and I use Emacs as my primary interface. I was wanting to integrate Emacs and UTexas' KM (Knowledge Machine).

12:53:59 aindilis_ Can anyone recommend the best way to integrate these? I talked with Claude, and iterated towards some kind of bridge between ELisp and CL.

12:54:02 aindilis_ However, the only one I know of is not maintained presently. Can anyone recommend anything relevant?

13:24:32 varjag aindilis_: does it run in a lisp-based REPL?

13:24:41 varjag then SLIME or SLY could be a start

13:25:27 varjag from the user manual it does look a lot like lisp repl

14:04:26 aindilis_ well, I have it mostly working using sly

14:05:32 aindilis_ varjag: oh, just saw your message. ty! yeah, I was told to use sly, and then Claude helped me get it. KM itself has compilation issues, pretty sure I have a more modern version lying around somewhere, and I'll release and post here and #emacs if/when it works.

14:05:59 aindilis_ mostly integrated at this point

14:06:10 aindilis_ will be cool to have KBS capabilities "inside" Emacs

14:10:19 bjorkintosh aindilis_: what are they good for now that LLMs are all over the place?

14:16:32 aindilis_ bjorkintosh: I would also like a Text version of a KBS too, now that you mention it, which has all clauses in English and does inference inside the LLM.

14:17:26 aindilis_ but, you know, I am using it for just simple KRR, not anything too advanced. The basic relationship are still valid despite there being other, perhaps better subsymbolic approaches that treat things as gradients as opposed to binaries.

14:17:54 bjorkintosh aindilis_: any good references on the topic?

14:18:06 aindilis_ in fact, I think I owe causative an apology, because causative always argued in favor of NNs and I didn't think at the time that they were going to get as insanely capable as they have

14:18:21 aindilis_ bjorkintosh: on what topic?

14:18:31 bjorkintosh what you're doing.

14:18:37 bjorkintosh KRR

14:18:44 bjorkintosh (as opposed to KKR)

14:18:51 aindilis_ what is KKR?

14:18:59 aindilis_ oh, well dmiles work is the best in that regard as regards Prolog-based KRR imho

14:19:12 aindilis_ I'm just trying to deploy it for these use cases: one sec:

14:19:22 bjorkintosh aindilis_: the defintion of an evil vampire squid corporation.

14:19:37 aindilis_ https://github.com/aindilis/free-life-planner/blob/master/README.md

14:19:39 aindilis_ https://www.academia.edu/116194435/The_Free_Life_Planner

14:19:42 aindilis_ https://frdcsa.org

14:19:56 aindilis_ my specific use case that drove me to wanting an Emacs KB was:

14:20:43 aindilis_ delineating which keybindings were mnemonics: e.g.: \C-clmfao vs were transliterations, e.g. \C-comf for "open-main-file" or something like that

14:21:01 aindilis_ I mean, drove me this morning. I've wanted Emacs KB+ATP as long as I can remember, for quick deductive work

14:21:16 aindilis_ I plan on integrating EmacsKB w/ Cyc

14:21:29 aindilis_ using whatever preexisting KM Cyc integration exists, which is manifold

14:21:44 aindilis_ googling KKR

14:22:25 bjorkintosh aindilis_: don't. they're the epitome of evil vampire squid capitalism. they destroyed Toys R US for the fun of it.

14:23:05 bjorkintosh aindilis_: you've no doubt played around with interlisp and Cyc, yes?

14:23:28 aindilis_ have they no decency! I loved Toys R Us! There is no curse in Elvish, Entish, or the tongues of Men for this treachery.

14:23:52 bjorkintosh exactly.

14:24:10 aindilis_ yeah I hate the firms that buy hospitals and part them out, leaving communities without healthcare too

14:24:22 aindilis_ That's why I'm writing that health planner thingie mentioned above

14:24:43 aindilis_ well I set up interlisp using someone else's stuff for AM/Eurisko, and yeah I'm a long time Cyc user

14:25:20 aindilis_ but I don't really know interlisp. I value that it worked, and I understand that like Prolog and FOL don't change, they're still just as useful as ever, and I assume the same with interlisp

14:25:36 aindilis_ *the basic rules of Prolog & FOL

14:27:49 aindilis_ I'm sorry, I should properly credit the work of setting up AM/Eurisko here. I just have a bad memory these days due to meds. It was seveno4.

14:29:29 aindilis_ White_Flame and I came across Eurisko in the SAILDART repos. I also wrote cyc-mode-2, an Emacs major mode for Cyc using the Java API.

14:30:02 aindilis_ how's it going btw, long time since we last chatted bjorkintosh

14:30:18 bjorkintosh good, thank you. you?

14:30:26 aindilis_ better than it was thankfully

14:31:00 aindilis_ things were hard for a long time. Now if the same thing that happened to Toys R Us doesn't happen to the United States, then I will probably be fine for a while unless I have some health issue crop up.

14:32:10 aindilis_ do you still do much Prolog and/or chat with dmiles? I haven't heard directly from him in ages

14:32:29 aindilis_ he's working on MeTTa, I think I might try to help out

14:33:26 aindilis_ oh btw, re LLMs, I'm heavily invested in them at this point

14:33:27 bjorkintosh nah. but prolog is on my back burner for now.

14:33:41 aindilis_ cool

14:33:42 bjorkintosh good. I'm not yet. need to find time.

14:33:50 aindilis_ oh what is taking up your time?

14:34:02 aindilis_ sorry to pry

14:34:05 bjorkintosh w$rk. databases.

14:34:13 aindilis_ ah, nice

14:34:15 bjorkintosh life too.

14:34:22 aindilis_ yeah

14:34:36 bjorkintosh the usual AFK stuff.

14:34:36 aindilis_ I hope you get a respite and time to do programming

14:34:39 aindilis_ yeah

14:34:51 aindilis_ I always say "things are looking up now" and then bam

14:35:02 aindilis_ vaguely similar to drowning

14:35:27 aindilis_ so yeah, things are looking up now

14:35:29 aindilis_ :)

14:35:34 aindilis_ happy hacking!

14:45:20 efti o/ my name is efti

14:45:51 Lumie o/

14:45:59 efti if I have an object which is say a 2D rect with h = 5 and w = 3

14:47:29 efti and then I have a method to transform this into an actual representation, so in this case it will be 5*3 = 15 char long list ( ... ) based on the two parameter values and how it should be represented in the base case which is length = 1 ( . )

14:47:53 efti what do you call that method/translation? I called it 'draw' but they didn't like it

14:50:58 efti question 2, I have a method with formal parameters (arguments) dst for destination and src for src, it writes src onto dst, if and what was there before that is lost. so if dst is 'xxx' and src is 'yy', the result on dst will be 'yyx' - what do you call that? I called it 'write-over' but they again didn't like it

14:51:05 efti o/ Lumie, do you know?

14:52:50 paulapatience Who is they?

14:53:24 efti those who reviewed the software

14:53:31 paulapatience and I'm not sure I understand your first situation

14:53:59 paulapatience In CL the second operation is just REPLACE

14:54:17 efti ah, not bad!

14:55:44 efti first case is some change in representation, like gold coins: 5 -> { ***** } where * is one coin, ha

14:56:30 efti a less abstract way of representing the same thing?

14:56:55 efti I have never heard of that before I wrote it, and they are not happy with the name :)

14:57:48 efti rewrite, encode, substantiate?

15:03:14 efti initialize, concretize, solidify

15:03:27 efti make-solid, I take that. case closed

15:07:12 efti paulapatience, yeah, CL is good in this case as the code is eieio.el which is Elisp's implementation of CLOS

15:40:20 x8dcc If (car '(abc xyz)) returns the symbol "abc", why does (car '(nil nil nil)) return the empty list instead of the symbol "nil"?

15:41:30 x8dcc I have spent a lot of time trying to figure out why (null? (car '(nil nil))) is false in my Lisp, and it's because it was checking the symbol "nil". It kind of makes sense.

15:44:00 x8dcc Oh, wait... in Emacs Lisp (type-of (car '(nil nil))) is symbol. Then `null' accepts the "nil" symbol?

15:45:35 dutchie nil and 'nil are the same object

15:46:10 x8dcc I don't understand. 'nil is a symbol, not the empty list, right?

15:46:27 dutchie it's both

15:46:40 dutchie a list is just a bunch of cons cells with nil at the end

15:46:43 dutchie or nil

15:47:35 x8dcc But if you quote a symbol it's not evaluated.

15:48:38 paulapatience nil and t are self-evaluating symbols

15:48:39 x8dcc If I define the symbol "sym" to a list, (listp 'sym) is false

15:48:59 paulapatience Basically: (defconstant nil 'nil)

15:49:02 x8dcc paulapatience: But even if they evaluate to themselves, why are evaluated in the first place? If I quote them

15:49:31 paulapatience Also, nil is '() which is ()

15:49:45 x8dcc ... I don't think it's clear what I mean

15:50:13 x8dcc https://bpa.st/ELFO4

15:50:44 paulapatience Your paste makes sense

15:51:02 x8dcc Yes. But now look: https://bpa.st/KAOGO

15:51:15 x8dcc That's what doesn't make sense to me. Why is 'nil a list and not a symbol.

15:51:25 paulapatience Yes, that makes sense too because like I said: nil is 'nil is () is '()

15:51:41 dutchie well, the symbol nil is by definition a list

15:51:50 paulapatience nil is a list and a symbol

15:52:02 dutchie it is the only object that is both a list and a symbol

15:52:16 x8dcc Very weird

15:54:10 x8dcc Now I am not sure if I should add this exception to my Lisp as well. Doesn't make much sense

15:54:14 paulapatience efti: A possibly more conventional term would be materialize

15:54:45 dutchie there are two exceptional things about nil: 1) it's both a list and a symbol 2) it's self-evaluating

15:55:12 paulapatience x8dcc: Scheme does it differently

15:55:26 dutchie (there are probably others as well, but those are the ones relevant to this conversation)

15:55:33 x8dcc The symbol "t" is also self-evaluation, but that's logical. Having 'nil evaluate to a list when explicitly quoting doesn't make much sense to me

15:55:40 paulapatience Nil is just a symbol, () is the empty list, and #f is false

15:55:50 paulapatience You might prefer that instead

15:56:10 x8dcc Yeah but "nil" doesn't have any meaning in scheme

15:56:21 paulapatience That's what I said

15:56:34 paulapatience Do you want nil to have some meaning?

15:57:38 x8dcc Right now I was treating "nil" like a normal variable, defined to the empty list. And empty lists evaluate to themselves in my lisp.

15:58:31 dutchie so the main difference there is that you distinguish the symbol nil and the empty list, whereas in CL

15:58:33 x8dcc I might do it "half-way". Keep "nil" as a normal symbol, but use both the "nil" symbol and the emtpy list as false.

15:58:39 dutchie ...and elisp they are the same thing

15:59:00 x8dcc dutchie: I haven't checked in CL, give me a sec

15:59:27 paulapatience I would advise you to choose either the CL or the Scheme convention.

15:59:42 paulapatience An in between seems like the worst of both worlds.

15:59:59 x8dcc (type-of (car '(nil nil))) => NULL

16:00:01 x8dcc Ah, yes.

16:00:24 levitating_ ** NICK levitating

16:00:30 paulapatience x8dcc: You might be interested in reading this: https://www.tfeb.org/fragments/2022/12/16/the-empty-list/

16:03:19 dutchie that's a good explanation

16:03:39 x8dcc paulapatience: Very interesting, thanks

16:03:55 x8dcc My approach seems to be closer to CL

16:04:57 x8dcc It's just that this whole "nil is both a list and a symbol" thing doesn't sound right to me

16:05:05 Bubblegumdrop oh yeah did we talk about how sometimes nil isn't '() ?

16:05:09 Bubblegumdrop is that scheme?

16:05:16 x8dcc Scheme doesn't have nil

16:05:19 Bubblegumdrop yeee

16:05:23 Bubblegumdrop You were asking about your cons pairs last night

16:05:50 x8dcc Yeah, this is yet another different adventure filled with suffering and parentheses

16:05:59 Bubblegumdrop are we consing yet?

16:06:36 Bubblegumdrop actually that's kind of a funny punny. list traversal subroutines going "are we there yet? Nope"

16:06:52 x8dcc Not at all. Yesterday I tried adding something to the standard library and I thought I broke something because my symbol "nil" is not false. But that was actually expected. I guess.

16:07:08 Bubblegumdrop I think in CL "NIL" is a special symbol...

16:07:17 x8dcc God almighty not again

16:07:20 Bubblegumdrop like you had your ExprType enum? It would be EXPR_NIL

16:07:29 Bubblegumdrop I'm not an expert on this though.

16:07:30 x8dcc I tried adding that and someone bullied me somewhere

16:07:39 x8dcc I don't think it's a good idea anwway

16:07:44 x8dcc anyway*

16:07:57 Bubblegumdrop You're farther along than I am!

16:08:08 x8dcc I am not so sure about that :)

16:10:40 x8dcc This just feels so wrong: https://bpa.st/QSEXU

16:24:16 Bubblegumdrop that's quite a bit

16:25:33 Bubblegumdrop I think in other implementations they use something called "tagged pointers" for this kind of thing...

16:25:50 Bubblegumdrop https://en.wikipedia.org/wiki/Tagged_pointer

16:26:27 Bubblegumdrop like storing your data in a pointer then setting some bit of the pointer to indicate the type or nil

16:26:37 Bubblegumdrop then you "undo" the tagging to get the value

16:27:48 x8dcc I don't think I understand the relation, sorry

16:28:05 x8dcc What does this have to do with nil?

16:33:57 Bubblegumdrop https://en.wikipedia.org/wiki/Tagged_pointer#Null_versus_aligned_pointer

16:34:43 Bubblegumdrop Some implementations just use a pointer value of 0 to represent nil

16:34:55 Bubblegumdrop so your code would be come return e == NULL; or something similar

16:35:37 x8dcc Yeah, that might make sense when implementing lists as cons pairs. In my case nil is an empty list, so an EXPR_PARENT with NULL children.

16:35:46 Bubblegumdrop it reads clearly enough to me

16:36:04 Bubblegumdrop it is quite a mouthful though

16:36:19 Bubblegumdrop up to personal preference I suppose being limited by actual hardware

16:36:36 x8dcc In any case, the "symbol exception" is not related to this. I had a hard time understanding why the *symbol* "nil" was equal to the empty list in Emacs Lisp, even before evaluating the symbol

16:39:55 Bubblegumdrop oh hm I don't know much about elisp

16:40:07 Bubblegumdrop https://www.gnu.org/software/emacs/manual/html_node/elisp/nil-and-t.html

16:40:11 x8dcc I think it's similar in CL

16:40:28 Bubblegumdrop looks like an implementation detail

16:40:46 Bubblegumdrop Yes I know for certain in CL ,(eq '() nil)

16:41:05 Bubblegumdrop oh is the bot not in here? Maybe I did it wrong.

16:41:26 x8dcc I knew that, I didn't understand why (eq '() 'nil)

16:41:41 Bubblegumdrop Are you aware of the differences between eq eql equal and equalp?

16:42:10 x8dcc Yes, but there is only one equal? in my Lisp, since I don't have garbage collection by now and no pointers are re-used

16:42:22 Bubblegumdrop indeed

16:42:23 Bubblegumdrop https://www.cs.cmu.edu/Groups/AI/html/cltl/clm/node74.html

16:42:31 efti paulapatience, yeah

16:42:38 Bubblegumdrop in the order I listed, they become more specific

16:42:57 Bubblegumdrop or maybe it's the other way around?

16:43:03 Bubblegumdrop equalp is more general than eq...

16:43:09 Bubblegumdrop @_@

16:43:41 Bubblegumdrop Yes I was mistaken the first time. in the order I listed they become more general with eq being the most specific.

16:46:36 Bubblegumdrop I'm not sure if you can see at the top of that page I linked, " Common Lisp the Language, 2nd Edition "

16:46:36 Bubblegumdrop That's a great book to have for writing a lisp implementation.

16:46:37 x8dcc Thank you for the link

16:46:42 Bubblegumdrop I believe it's from CLtL2

16:47:20 Bubblegumdrop Is that Guy Steele?

16:47:31 Bubblegumdrop Yep

16:47:52 Bubblegumdrop It's available on archive.org

16:53:20 x8dcc I think this is accurate: https://i.imgur.com/ZWoqgwO.png

16:58:32 Bubblegumdrop Oh man that's a doozy, (eq 'nil nil)

16:58:36 Bubblegumdrop @_@

16:59:04 Bubblegumdrop Oh That last bit too! oof

17:15:43 efti anyone knows eieio.el or CLOS assuming it is the same how do I arrange so that a method is called for an object immediately and always after instantiation? now I call them manually every time which is ugly and error-prone compared to automation, of course

17:17:35 efti like if one could to :after the constructor or something? how is that setup?

17:23:47 Bubblegumdrop well for clos I think you'd do something like (defmethod initialize-instance :after ((obj your-class) &rest initargs &key &allow-other-keys) ...)

17:24:23 Bubblegumdrop efti have you read "The Art of the Metaobject Protocol" ? it's pretty good

17:24:43 Bubblegumdrop There are actually different ways to combine methods too

17:25:00 Bubblegumdrop most specific first, least specific first, nconc (in order of creation I believe)

17:38:38 efti yes I know, I read the eieio.el documentation

17:39:43 efti yeah, I think you are right! but run out of steam, I try it in one hour :)

17:40:03 efti and this isn't in the manual, I don't know why or maybe I didn't see it

17:41:26 x8dcc Normally, when tracing Lisp functions, the interpreter traces the actual function, right? Not the symbol

17:42:48 x8dcc So I can define `my-func' to be `list', and if I trace `my-func' then all calls to `list' are traced as well. Right?

20:02:28 levitating_ ** NICK levitating

20:14:49 edgar-rft_ x8dcc: not with SBCL -> https://plaster.tymoon.eu/view/4610#4610

20:35:52 x8dcc edgar-rft_: Hmm... So you can only trace named functions, then?

20:49:19 keeliro Which Lisp community is the welcomest?

20:50:08 keeliro I think that vertical groups suck.

20:50:43 keeliro The only thing vertical which does not suck is software integration.

20:57:40 owen what is a vertical group?

21:59:17 Bubblegumdrop keeliro all of them

21:59:27 Bubblegumdrop x8dcc I think you can pass &rest to (trace ...)

21:59:45 Bubblegumdrop Invoking trace with one or more function-names causes the denoted functions to be ``traced.''

21:59:50 Bubblegumdrop This is from hyperspec

22:01:00 x8dcc Hmm... I changed it to use symbols but I think I rather use functions

22:02:17 Bubblegumdrop in my implementation (sbcl) trace is a macro so you don't pass symbols to it

22:02:23 Bubblegumdrop er, unquoted symbols

22:02:29 Bubblegumdrop (trace func) does what you expect

22:03:37 x8dcc And what do I expect? Tracing the symbols or the function itself?

22:05:57 Bubblegumdrop https://www.lispworks.com/documentation/HyperSpec/Body/m_tracec.htm

22:06:40 Bubblegumdrop Neat, I just learned about STEP https://www.lispworks.com/documentation/HyperSpec/Body/m_step.htm#step

22:06:52 Bubblegumdrop Did not know I could step through forms

22:07:06 Bubblegumdrop Very handy.

22:15:12 keeliro #emacs community is welcoming...its just that the ops are too bossy

22:15:38 keeliro and I hate harassment

22:16:06 keeliro I was very enthusiastic at the start

22:16:29 Bubblegumdrop don't stress about it, irc is life for some people

22:16:41 Bubblegumdrop bossy ops just doing what they think is best

22:16:47 x8dcc Never felt that way about #emacs

22:16:56 x8dcc But what is a vertical group?

22:16:58 keeliro All I know is that strict rules kill creativity

22:17:09 Bubblegumdrop not true :)

22:17:11 x8dcc That is very arguable

22:17:36 keeliro creativity flourishes with freedom...rules quench freedom

22:17:41 keeliro rules are impositions

22:17:48 x8dcc I would say strict rules can foment creativity to overcome those rules

22:18:04 keeliro what ?

22:18:13 x8dcc But I still don't get the point. I don't think IRC rules have anything to do with creativity

22:18:32 x8dcc More importantly: What is a vertical group?

22:18:38 keeliro strick rules make people be creative to circumvent them, is what you mean?

22:18:49 x8dcc Yes, sometimes

22:19:06 keeliro If you have a nazi regime, that is totally uncreative

22:19:12 keeliro only obedience counts

22:19:33 keeliro but there are different amounts of rules

22:19:59 keeliro and the more strict the rules are, people are forced to do instead of motivated

22:20:22 keeliro a vertical social structure is one with hierarchies

22:20:41 keeliro a horizontal one is one with equality

22:20:46 keeliro a democratic one

22:20:50 x8dcc No offense, but what do you expect? No rules, no ops, and no hierarchies?

22:21:05 keeliro of creative spaces, yes

22:21:10 keeliro for armies, no

22:21:48 x8dcc I don't know what a "creative space" is, but I am pretty sure IRC is not one.

22:22:14 keeliro it is, in some places

22:22:49 x8dcc So you don't want rules in IRC

22:22:49 keeliro a creative space is one where people are motivated to be creative

22:23:05 keeliro only rulemakers like rules

22:23:11 x8dcc Right...

22:23:15 keeliro the rest are obedient

22:23:23 keeliro obedient people are not creative

22:24:02 keeliro motivation does not require rules

22:24:05 tumdum ACTION takes out popcorn

22:24:30 Alfr tumdum, oh, thanks!

22:24:54 x8dcc Thank God for motivation, if you know what I mean... (I don't)

22:24:59 keeliro you will never see an artist be forced by rules to produce his art

22:25:13 ieure I see that my immediate reaction of "by 'channels that suck,' they probably mean 'ones where they get sick of my bullshit'" is turning out to be prescient.

22:25:17 keeliro motivation moves the world

22:25:30 x8dcc Amen to that, brother

22:25:49 keeliro of course...nobody promotes something where he is unwelcome

22:26:03 x8dcc I am a bit lost. Promoting what?

22:27:01 keeliro never mind...I just made a comment... I do not want to explain what is just rejected

22:27:31 x8dcc Right, sorry.

22:27:43 keeliro if yo do not share my view, I do not need to convince you

22:28:04 keeliro you*

22:29:26 x8dcc I don't think I understand your view, that's all

22:29:40 bjorkintosh keeliro: well that's no way to gather followers, you have to convince them! unless you don't want followers.

22:30:14 x8dcc I just don't see the relationship of all this with #emacs :/

22:30:21 Bubblegumdrop keeliro is this what you were talking about in #emacs? It seems a bit off topic to me. I can understand why the ops might feel "bossy" towards you.

22:30:32 Bubblegumdrop don't sweat it...