freenode/lisp - IRC Chatlog

11:47:06 MichaelRaskin Technically I did not say it is better than Tue 18:00, just the same — but indeed it covers well the people overseas for whom 18:00 CEST was bad.

12:05:52 froggey fe[nl]ix: I'm here

12:23:47 _death hmmm.. planet.lisp.org cert needs renewal

12:25:59 phoe oh goodness

12:41:30 jackdaniel because it is so unsafe without a valid certificate!

12:41:46 jackdaniel at least that's what my web browser said

12:46:18 phoe I know, right

13:16:04 dlowe it's just to prevent there from being time to crack the private key

13:16:27 dlowe it doesn't become unsafe at any particular point

13:16:41 dlowe (well, any known particular point)

13:30:24 MajCon f

13:32:12 MajCon phoe:Hi! Are you the person responsible for the Online Lisp Meeting series?

13:32:33 beach He is.

13:32:56 phoe MajCon: oh gods they found me

13:33:01 phoe ACTION runs for cover

13:33:16 phoe MajCon: sure, it's me! what's up?

13:33:50 jackdaniel hand your wallet and report none of it to the police - we know your address

13:34:14 phoe ACTION hands over the wallet with 384 MB worth of conses

13:34:35 jackdaniel clearly you use sbcl and you did not type anything in the repl yet!

13:34:59 MajCon Haha :-). It's a bit early (like 1 month early) but I've been using Racket+Redex for doing my master's thesis on using formal methods for WebAssembly. Is that the kind of stuff you think people would be interested in hearing about?

13:36:08 phoe MajCon: yes please

13:36:25 phoe this is a Common Lisp place, so we are just slightly off topic; mind if we take it to #lispcafe?

13:36:53 MajCon Just need to register this name, be there in a bit

13:57:50 flip214 when using cl-jupyter, only the current result is shown. How can I configure it to have a history, only changing previous cells by explicit intent?

14:07:56 flip214 Oh, I need to click "+" (insert cell) every time...

14:30:42 yottabyte hi all, I'm trying to make a simple executable to test, but when I run "(sb-ext:save-lisp-and-die "hello.exe" :toplevel #'main :executable t)" in slime I get "Cannot save core with multiple threads running."

14:30:46 yottabyte I was just following this: https://stackoverflow.com/questions/14171849/compiling-common-lisp-to-an-executable

14:31:05 beach Do it outside of SLIME.

14:31:06 dlowe yottabyte: don't do it in slime

14:31:11 yottabyte where do I do it

14:31:18 dlowe start sbcl on a terminal

14:31:35 yottabyte in the directory where my hello.lisp is?

14:31:39 yottabyte or do I load the file?

14:31:56 dlowe start sbcl in a terminal, load your code, then save your image

14:33:59 yottabyte neat! a hello-world is 45 megabytes. is that okay?

14:34:14 dlowe it doesn't go up by much when you write a big program

14:34:24 dlowe your hello world also includes an IDE :)

14:34:47 dlowe well, minus the editor :p

14:35:09 pve yottabyte: you can use ":compression t" if you want

14:37:21 yottabyte so I can compile a program that accepts a file as an argument and loads and compiles it so I don't have to start a sbcl session to make an executable

14:38:00 dlowe sure, just sbcl -eval '(load "hello.lisp")

14:38:14 dlowe -eval '(sb-ext:save-lisp-and-die ...)'

14:38:16 phoe yottabyte: also turn on core compression, you'll get 11 MB rather than 45 MB

14:38:35 phoe oh, pve alredy said it

14:47:57 jmercouris I'm trying to make something like $EDITOR support for Next

14:48:04 jmercouris the first thing that I think about is UIOP:RUN-PROGRAM

14:48:19 jmercouris however, probably I should launch that in a new thread to avoid blocking, right?

14:49:54 phoe clhs ed

14:49:55 specbot http://www.lispworks.com/reference/HyperSpec/Body/f_ed.htm

14:50:17 patlv Hi all, I've been using the Sxql package from Eitaro Fukamachi with great success, now I'm trying to call a stored procedure in mysql with it, it's not documented, anybody knows if it is possible?

14:51:20 jmercouris THATS UNBELIEVABLE!

14:51:23 jmercouris they thought of everything!

14:51:25 jmercouris phoe: thanks

14:51:54 Bike i don't think many implementations put effort into ed any more, and it might block

14:51:59 Bike might be worth a shot, i guess

14:52:31 jmercouris even if it blocks, it is very explicit

14:52:45 jmercouris I can still pop it into a new thread

14:53:13 jmercouris hmm, doesn't work on my machine

14:53:42 jmercouris "Don't know how to ED /Users/jmercouris/Downloads/file.txt"

14:54:00 jmercouris *ED-FUNCTIONS* is NIL

14:54:05 jmercouris hm, that is problematic

14:55:01 jmercouris this looks good: https://github.com/sanel/magic-ed

14:59:59 jackdaniel sbcl defines a hook mechanism for ed

15:00:05 jackdaniel so you may put there anything, i.e emacs

15:00:17 jackdaniel ecl afair uses the EDITOR env variable value

15:01:02 jackdaniel having hooks is clearly better, because you may feed it with something what's already part of the image (i.e climacs or some sed-like repl)

15:02:09 jackdaniel http://www.sbcl.org/manual/#Customization-Hooks-for-Users

15:03:24 phoe ACTION clicks magic-ed

15:03:55 jmercouris the README is lies

15:04:02 jmercouris it is available on standard quicklisp dist

15:04:15 phoe ...this is a portability library for cl:ed

15:04:30 jmercouris that's exactly what it is, but it also respects $EDITOR

15:04:32 jmercouris supposedely

15:04:40 jackdaniel does the readmy state, that it is not available on ql?

15:04:44 jackdaniel readme°

15:05:07 jmercouris says something about quicklisp

15:05:08 Bike it says you can get it from quicklisp but recommends ultralisp.

15:05:09 jmercouris ultralisp

15:05:22 jmercouris I looked at the source

15:05:24 phoe "Install via Quicklisp, which updates distro about once a month."

15:05:26 jmercouris it does respect $EDITOR

15:05:28 jmercouris if you have CCL or SBCL

15:05:29 jackdaniel "says something about quicklisp" is not the same as "does not work on quicklisp", is it?

15:05:30 phoe "For the daily updates track, install via Ultralisp."

15:05:52 jmercouris LIES!

15:05:56 jmercouris yes, I was overzealous

15:05:58 phoe jmercouris: yes, it is available on standard quicklisp dist, and the readme of that project explicitly states that

15:06:55 jmercouris :'-(

15:06:57 jmercouris Yes

15:07:35 jmercouris I don't know if this is worth a dependency

15:07:36 jackdaniel also: does it *not* respect EDITOR variable on other implementations?

15:07:44 jmercouris it doesn't explicitly add support for them

15:08:22 jmercouris effectively yes

15:17:22 phoe no idea

15:17:59 phoe there's only SBCL and CCL getenv in there

15:18:22 phoe probably could be made better via uiop's getenv

15:20:34 jmercouris Yes

15:20:38 jmercouris That was my thinking

15:20:51 jmercouris also I was thinking that the scope is wrong, why eval text gotten from ed?

15:20:58 jmercouris why not just load it into string?

15:21:13 bfig_ ** NICK bfig

15:21:18 jmercouris also, why do we need to write it to a file?

15:22:07 jmercouris I guess we must use a temporary file, this is Unix after all...

15:23:54 borei hi all

15:24:13 jmercouris hello borei

15:24:34 borei finally getting back on my lisp learning curve, last months were just killing me

15:29:38 borei well, after getting some result (doing 3D learning project, linear algebra etc) i started to understand that my approach is not in lisp way. Im using CLOS - it works very well, but starting to read "Let Over Lambda" - pointed me that there is alternative, and that alternative is even more powerful then CLOS, but not easy to understand.

15:30:22 beach And what is that alternative?

15:30:32 beach Don't tell me it's "closures".

15:31:01 borei hi beach: :-)

15:31:11 beach Generic functions and standard classes provide way more flexibility than closures.

15:31:13 borei yeah it closures

15:31:42 beach You are the second person in the past few days to have fallen into that trap.

15:31:47 borei i can be wrong, im not experienced yet, mayb i just need to understand it deeper

15:32:32 beach Closures can be used as a poor man's object system, but it is definitely not recommended when you have CLOS.

15:32:34 jmercouris OO is certainly one of the most powerful paradigms, even if itsn't as cool

15:32:43 pjb Well, it depends on the scope of the project. closures are anonymous objects with one (or very few "methods").

15:32:53 pjb Often you don't need more.

15:33:00 beach OK, here we go again.

15:33:10 pjb And a bunch of functions in CL prove it (reduce, mapcar, etc)

15:33:20 borei just understanding how they work in connection with lisp macro system will bring me to the higher level

15:33:33 jmercouris no, practice will bring you to a higher level

15:33:36 beach I think #lisp has picked up a habit of contesting everything I say, so I will just shut up.

15:33:41 jmercouris solving artificial problems doesn't make you better

15:33:59 jmercouris beach: seid nicht so sauer! neimand hat dass gesagt

15:33:59 borei so now it's very academic and learning approach

15:34:00 pjb beach: it's not contesting everything. It's that everything is gray, and depends on the context.

15:34:00 beach borei: Good luck with your closures.

15:34:08 borei :-)

15:34:11 borei im trying

15:34:26 pjb OO is good, closures are good. one needs to understand both, and to use them in the right circumstances.

15:34:46 borei yep, that is my final goal

15:35:06 pjb If you make the wrong choice, you get overly complex code.

15:35:19 Bike closures aren't really an alternative to clos. you could implement clos using closures as objects, probably.

15:35:21 borei so question is very basic, but i can't jump over that problem, i even don't know if my approach is correct

15:35:47 jmercouris Bike: that sounds like an exercise in frustration

15:35:59 Bike i don't think it would be a great idea, no

15:36:03 borei (defmacro defnode (name)

15:36:03 borei (let ((x 0))

15:36:03 borei `(defun ,name()

15:36:03 borei (format t "Hello, from Macro~%")

15:36:03 borei (incf x)

15:36:03 borei (format t "New X: ~a~%" x))))

15:36:12 jmercouris borei: please, use a pasteservice

15:36:14 Bike please use a pastebin service if you're going to put in more than one line.

15:36:32 pjb try http://termbin.com

15:36:33 borei yep yep, really forgot about paste service

15:36:45 Bike the binding of X takes place at macroexpansion time, so in the expansion X is unbound.

15:37:03 jmercouris while Bike is correct, I don't think borei will get it

15:37:12 borei yep

15:37:42 borei function is defined in lexical environment - should it has an access to varaibles ?

15:37:52 pjb borei: no it is not.

15:37:59 borei i can't find answer

15:38:16 pjb borei: the function is defined in the lexical environment where the defnode macro call is expanded from.

15:38:28 pjb borei: Read what Bike said!

15:38:42 Bike The function defined by DEFUN doesn't have X in its local environment.

15:38:51 pjb borei: since your code is erroneous, it's difficult to guess what you want to do…

15:39:07 Bike The code that DEFNODE expands into will need to bind X itself.

15:39:14 pjb borei: so you should first write down in English what it is you're trying to achieve.

15:39:57 borei need some time, wife is just behind me, breakfast, and she has very heavy argument :-) gimme 10-15

15:40:48 pjb if it is to generate a function that will return a different value each time it's called, then I would advise against it. Such procedures are not mathematical functions (their results are not computed in function of their arguments), therefore they are difficult to prove, to debug and to test.

15:41:03 beach pjb: So even the fact that I am contested was contested.

15:41:11 beach I think I get the idea.

15:41:14 pjb beach: sorry about that. :-)

15:43:46 jackdaniel let over lambda is a very bad book for a beginner in common lisp

15:44:00 Josh_2 ^

15:46:19 jmercouris anything wrong with having UIOP:GET-ENV in a slot initform?

15:46:21 jackdaniel it is as if you were learning a language with tongue twisters: neither they are representative nor complete from the language perspective

15:46:43 jmercouris as much as i love to disagree with jackdaniel on any topic of any nature, he is wholly correct about this

15:47:46 jackdaniel jmercouris: you could skip your snarky remarks, I have enough of being personally addressed by you in a way which is impolite

15:47:50 borei ok, main goal (i know it can be achieved with CLOS, and i can do it pretty easy) - im building pipeline, which is directed acyclic graph, nodes of that pipeline supposed to do very simple operation, for example to do high pass filtering.

15:47:54 jmercouris jackdaniel: ebenso

15:48:05 X-Scale` ** NICK X-Scale

15:48:20 Bike jmercouris: initforms can be whatever and it should be fine. they're not executed if you're loading an object from a fasl for example.

15:48:20 borei so i decided to try to use closure as a foundation for nodes definition

15:48:48 jmercouris Bike: aha, that.. could be a problem, how do I know what is executed/not-executed when loading from image? that has brought me pain several times

15:49:04 jackdaniel "likewise" would be appropriate if I did that at least once, which I did not.

15:49:06 Bike clhs make-load-form

15:49:07 specbot http://www.lispworks.com/reference/HyperSpec/Body/f_mk_ld_.htm

15:49:10 Bike jmercouris:

15:49:14 borei but i still want to modify variables defined lexical environments of the corresponding closures

15:49:46 jmercouris Bike: I'll read, thanks

15:49:48 Bike yes, you'll need to have other closures that modify the variables, or add messages to your existing closures to let you tell them to modify a lexical variable.

15:49:57 Bike with CLOS you can just use accessor functions. it's easy.

15:50:14 Bike jmercouris: basically, you can customize it pretty much entirely.

15:51:06 jmercouris that's pretty cool

15:55:56 jmercouris OK to use UIOP's with-temporary-file?

15:56:13 jmercouris that is what I plan to pass to ED to edit, and then grab the results from that

15:56:28 borei Bike: "yes, you'll need to have other closures that modify the variables, or add messages to your existing closures to let you tell them to modify a lexical variable." - can you give me simple example if possible ?

15:56:50 jmercouris borei: what is your goal in all of this? I can tell you that you won't be learning lisp

15:56:51 Bike No, because I told you about this complication to encourage you to use CLOS instead.

15:57:07 jmercouris nobody would ever write such a program in the manner described

15:57:13 jmercouris unless it was strictly an exercise in masochism

15:57:20 Bike let over lambda should cover that kind of thing. if it doesn't, you are not prepared

15:59:36 borei that is true, i'm not tat that level ... yet

16:04:32 jackdaniel another analogy with language twisters is that even when you know them all, it won't make you know the language. also speaking in language twisters on daily basis it annoying to others (at best)

16:05:32 MichaelRaskin jmercouris: I thinh my debugger helper to fight compilers optimising away local variables did more or less this… but yeah, if was not normal code, and if such things are not obvious yet, one should not do them

16:07:41 jmercouris MichaelRaskin: yes :-)

16:07:56 Josh_2 borei: you see how confusing this gets after a while? https://imgur.com/AHROXd6.png and this example is only trivial. You really should just use CLOS where applicable

16:10:15 _death borei: On Lisp has a basic chapter about network represetation by closures

16:10:44 kinope Hey! I was that second person who fell into constructing my classes from closures, gave me a chuckle. The conversation has got me wondering how many people wonder in here bathed in the radiant light of Let Over Lambda

16:11:30 Josh_2 From what I have seen in here Let Over Λ has some controversial takes

16:11:46 Josh_2 I've read 4 chapters

16:12:08 kinope I've seen some of the controversy in the past too

16:12:15 kinope around the web

16:12:19 _death such examples are also given in SICP.. I think it's good to learn, even if the CLOS style often makes sense in practice

16:12:55 TMA it is neat that it is possible, yet it is not exactly practical (and more suited for scheme, where a decent object system is not standardized in the older revisions of RnRS)

16:13:36 borei ok, let put in this way - im not looking to build my own CLOS, im looking to build foundation for statefull functions

16:14:22 beach TMA: Be careful what you say. I said something similar the other day, resulting in an avalanche of utterances contesting what I said.

16:14:58 kinope Yeah, I think it's worth exploring all of the language constructs even if there are better ways to tackle the common use cases.

16:15:30 beach Closures are essential, but mostly for other things.

16:15:37 Josh_2 ^

16:15:47 TMA borei: did you look into funcallable instances? http://mop.lisp.se/www.alu.org/mop/concepts.html#funcallable-instances

16:16:36 phoe I guess that would work, the funcallable instance is a CLOS instance with slots *and* a function at the same time

16:17:03 borei nope, i didn't know about them till now

16:17:04 phoe you can use that to have a function that stores its own state "inside itself".

16:17:43 beach phoe: The term "CLOS instance" is meaningless. Every Common Lisp object is an instance of some class, so every Common Lisp object would then be a "CLOS instance".

16:17:51 _death if someone does not know recursion, it makes sense to give exercises in recursion that in practice are solved by other means such as loops.. the same goes for closures

16:18:19 phoe beach: fine, s/CLOS instance/standard object/

16:18:26 beach Thank you.

16:18:35 kinope Rewrote my simple queue closure today as a class and honestly it's much easier to grok

16:18:50 phoe thanks as well

16:19:21 phoe borei: an advantage of a funcallable instance over closures is that it's much easier to inspect the state in case of debugging.

16:20:01 phoe like, you literally check the slot values of the funcallable instance, as opposed to using implementation-dependent means to get the values and what else from the places that are closed over

16:20:42 borei https://pastebin.com/qeTicWt3 <- this form do what im looking for, it points me that i don't understand how does macros works

16:21:21 kinope I had a look at cl-speedy-queue afterwards and whoa! the performance blew mine out of the water.

16:21:53 TMA beach: I am trying to be careful. I too have been rebuffed in a similar way on too more ocasions than I care to count.

16:21:59 phoe that's a closure

16:24:28 Bike borei: compare (macroexpand-1 '(defnode ...)) to that form, and observe whta is different

16:28:16 kinope Apparently cl-speedy-queue is non-consing. I'm wondering how that works. My queue class operates on lists and I think I have a constant complexity algorithm but it's still very slow. How does one create a data structure that operates faster than consing lists, I thought that the list was as deep as you could go in lisp. Am I making sense?

16:28:41 borei after macroexpand-1 let form just gone, meaning that there is no lexical environment - is it correct statement ?

16:28:42 jackdaniel kinope: you can define an array

16:28:49 jackdaniel with a fill pointer

16:29:00 jackdaniel then you simply fill the array

16:29:26 jackdaniel that may require some modulo arithmetic if you want to make it rotary, but either way that should not cons

16:29:31 jackdaniel (I don't know how the library is built)

16:30:35 jackdaniel or you can of course cons a large list and make it circular

16:33:04 beach Or you could use Flexichain which is designed for this kind of situation.

16:33:08 kinope jackdaniel: Okay thanks, I'll look into that

16:34:01 borei why let gone ???

16:35:38 kinope Flexichain huh, alright I'll give that a squiz too.

16:35:45 kinope cheers

16:38:57 borei and this form works as expected - https://pastebin.com/EhDxNwHr

16:48:04 bfig -part

16:54:51 kinope Reading the design document for Flexichain now, it seems to tackle a problem that is a bit larger than I need right now. Interesting none the less. If I ever build an application with my framework flexichain seems like a good candidate for text fields and such. I'm going to study the design of cl-speedy-queue for my own edification, but I'm probably going to use it for the message queue for now.

17:02:08 kinope One has to specify the size of the queue when using speedy-queue, but when I think about it I don't think the message queue needs resizing, or even needs to be so big. Since a stable actor network shouldn't have a queue of unprocessed messages growing however fast or slowly towards infinity anyway. Should be possible to determine the size of queue needed for a given application.