freenode/#lisp - IRC Chatlog

3:10:31 pierpa` ** NICK pierpa

3:25:10 loke Hello Lisp!

3:45:24 neoncont_ ** NICK neoncontrails

4:06:00 beach Good morning everyone!

4:12:36 loke Good morning Beach!

4:17:06 pillton G'day beach.

4:18:28 beach * needs another apprentice.

4:18:54 aeth Is there a non-destructive version of (progn (setf (nth 2 foo) 42) foo)?

4:19:37 Bike usually just copy the list

4:20:22 Bike you could also do (list* (first foo) (second foo) 42 (nthcdr 3 foo)), but i mean, ugh, man

4:20:55 aeth I don't need anything particularly fancy, I just am temporarily using lists for some data structure I haven't discovered yet

4:21:31 aeth And it looks like I want to make two or more copies of a list, with only one or two differences

4:28:25 slyrus RichardPaulBck[m: around?

4:28:33 slyrus morning beach

4:28:45 beach Hey slyrus. Long time no see.

4:28:46 slyrus what happened to your last apprentice?

4:28:59 slyrus indeed! hope you and your projects are well!

4:29:02 beach He is still around, but he is working on unrelated stuff.

4:29:10 slyrus ah, too bad

4:29:47 beach It's his choice. This is not a formal agreement. He hangs out here as `phoe'.

4:30:33 beach slyrus: Will you come to ELS2017?

4:31:35 slyrus hmm... thanks for the pointer. I'll see what I can do.

4:32:50 beach That would be great!

4:32:51 loke * really wants to go to ELS.

4:35:27 beach loke: You definitely should.

4:35:49 Fare * wonders if his current ASDF work is worth presenting at ELS... just how many times can I present work on ASDF???

4:35:54 loke What date was it again?

4:35:59 Fare where is it?

4:36:22 beach Brussels.

4:37:20 beach http://www.european-lisp-symposium.org/editions/2017/

4:41:47 beach Fare: I don't think the program chair (whoever that will turn out to be) will have the ambition of turning ELS into a highly filtering event. Therefore I think there is room for a paper on ASDF progress/ideas.

4:42:16 loke * wishes he had something to present. :-)

4:42:27 loke But I don't write papers.

4:42:46 PuercoPop beach: btw nice work on SICL's loop. I was reading the code today and is the most understandable loop implementation I've found

4:43:17 beach PuercoPop: Oh, thank you! Nice of you to say so.

4:44:24 PuercoPop and it doesn't appear to use CLOS for much so I'm thinking it could be adapated for JSCL replacing the defmethod with typecase/cond. I'll attempt it one of this days

4:44:47 beach ?

4:44:54 beach It uses CLOS a lot for the parsing.

4:45:49 beach Perhaps this is a different implementation of SICL. :)

4:46:20 PuercoPop For the parsing you use combinatory parsing and use defclass as record. For code generation you use a generic functions

4:46:31 beach Right.

4:46:40 PuercoPop (and defclass can be replaced with defstructs :D)

4:46:46 beach Yeah, sure.

4:47:19 beach I am currently working on a version of the sequence functions that I am hoping will be fast, maintainable, and portable. With a bit of luck, I'll present it at ELS.

4:47:30 PuercoPop and I don't see any around methods so, the generic functions could be adapted to normal funs with a typecase for each version (not ideal)

4:47:52 beach PuercoPop: You might be right. I don't remember the details.

4:49:23 PuercoPop beach: well I haven't done it yet. I'll let you know if I succeed

4:49:39 beach Yes, please keep me up to date!

5:00:46 beach My version of the sequence functions requires a "sufficiently smart compiler". As it turns out, the SBCL compiler is sufficiently smart, and now the pressure is on to make the Cleavir compiler sufficiently smart as well.

5:04:36 Bike ah? what is needed? i've only looked at those million functions a little bit.

5:05:08 beach I am ditching the old version with explicit functions for each case.

5:07:35 beach Very preliminary: http://metamodular.com/sequence-functions.pdf

5:08:33 beach The basic technique is illustrated in section 3.

5:09:37 Bike seems like all that has to be done is extend type inference to array types, and maybe have an inline for typep -> typeq.

5:10:45 beach You may be right.

5:11:51 Bike of course, doing that for every upgraded type is kind of eh...

5:12:13 beach It is messy, yes.

5:12:22 beach But I think it is necessary.

5:14:51 Bike i guess for variations in test and so on you just need a compiler macro to inline the call.

5:16:30 beach Yes, there is a macro WITH-TEST-AND-TEST-NOT.

5:16:51 beach It duplicates the body for specific combinations of the two, in particular EQ and EQL.

5:17:22 beach Similarly, there is a special case for certain KEY functions such as IDENTITY, CAR, and CDR.

5:17:53 Bike does it need to be specific to eq and eql? i feel like FIND is basic enough that you could always inline it

5:19:48 beach Maybe. But it gets messy with things like :FROM-END. A compiler macro could make such a decision.

5:20:33 Bike right... and you had your own thing for from-end, too.

5:22:09 beach Yes. Messy, but all hidden in a macro: FOR-EACH-RELEVANT-CONS.

5:28:30 Bike of course, this just reminds me that cleavir can't inline functions with &key lambda lists right now

5:37:28 beach A compiler macro would replace such a call with one with no &key anyway I would think.

5:38:39 Bike yeah probably.

6:20:23 pillton This kind of reminds me of why I wrote specialization-store.

6:21:24 Bike yeah, it's sort of like static gfs.

6:22:49 pillton Specialising with respect to each array element type isn't so bad.

6:23:42 Bike not so bad, but getting a canonical list of upgraded array element types is tricky

6:24:45 pillton Yes, there is that.

6:45:44 ack006 i literally cried my eyes out when i read this excercise (highschool outreach project, 1999!) and the example solution: https://www.cs.umd.edu/Outreach/hsContest99/questions/node8.html

6:46:05 ack006 this would have been done sooo much nicer in lisp...

6:47:30 ack006 iirc this is taken almost literally from some mit ai class, except the fact that lisp has left the building.

6:49:00 jasom ack006: were you in HS or at UMD then?

6:53:01 ack006 jasom: no :) stumbled on it while looking for tree comparison algorithms.

6:53:24 ack006 original paper: http://www.grantjenks.com/wiki/_media/ideas:tree-to-tree_editing_problem.pdf

6:53:41 ack006 no lisp here either, strange given that it is from 1977.

6:57:39 ack006 :-) http://mercury.lcs.mit.edu/~jnc//humour/lisp.tree

7:50:49 dwts morning

7:59:25 beaky hellow dw

8:05:15 beaky so mccarthy in his history of lisp outlines what it would take to obsolete lisp. have there been any serious efforts to conceive of such a superlisp

8:20:50 JuanDaugherty http://www-formal.stanford.edu/jmc/history/lisp/node6.html i take it; with the practical domination proviso, prolly not

8:51:03 razzy89___ beaky: Artificial General inteligence?

8:53:27 razzy89___ Tell her what to do, and it will do it :]

9:55:08 Jucaa Hellow

10:57:19 beach Jucaa: Are you new here?

11:00:28 Jucaa An acquaintance pointed me actually not quite know what the subject of that channel, but he pointed out to me as being about artificial intelligence etc, would it?

11:01:18 jackdaniel Jucaa: this channel is about Common Lisp (general purpose programming language)

11:01:37 edgar-rft we prefer human intelligence :-)

11:01:41 jackdaniel it may be used for artificial intelligence of course

11:01:57 beach The language, not the channel. :)

11:02:06 jackdaniel right ^_^

11:03:56 Jucaa It is all artificial, until we are kkk

11:08:24 |3b| Jucaa: "kkk" might not mean what you intended in english

11:10:16 hydraz ** NICK SIGINT

11:10:28 Jucaa forgot, do not speak English very well, I'm from Brazil, meant laughs with "kkkk"

11:11:43 beach |3b|: How did you catch that?

11:12:52 SIGINT ** NICK amused

11:14:08 |3b| beach: Jucaa said something in the past using same phrase that google translate thought was Portuguese, so thought it might be regional slang, and a searching for portuguese internet slang found a list including the term

11:14:35 beach I am impressed.

11:17:44 Jucaa Ever wanted to learn English, I opted for the Spanish, now regret.

11:25:20 |3b| Jucaa: you could also try #lisp-br #lisp-pt or #lisp-es not sure any of those are about artificial intelligence in general though

11:26:52 Jucaa thank you

12:32:11 |3b| is there any way to find the source for a function object in slime/sbcl?

12:33:43 |3b| ah, i guess this function doesn't have any source anyway:/

12:48:30 Xach_ the immaculate function?

12:52:52 ASau`` ** NICK ASau

13:06:08 msmith hello all. I assumed that when you start another process in sbcl via sb-ext::run-program from slime, control would be returned like if you spawned a new thread. Was this an incorrect assumption?

13:06:39 Xach_ msmith: it depends on how you start it

13:06:58 Xach_ msmith: there is a :WAIT argument that controls what happens.

13:07:10 msmith hey Xach_

13:07:24 Xach_ msmith: i don't recall the default, but if :wait is nil, the call returns an object you can check for the process status.

13:09:09 msmith I start it without the wait argument but the program that is run has an infinite loop.

13:10:27 msmith Xach_: thanks

13:10:28 Xach_ msmith: omitting the wait argument is the same as providing the default, practically

13:10:36 Xach_ msmith: and it looks like the default is T

13:11:26 msmith Xach_: I ran it with wait as nil and it performed as you described which is the desired behavior. I think :-).

13:12:30 |3b| Xach_: nah, just generated then compiled with (compile nil ...)

13:12:51 Xach_ msmith: very good!

13:12:52 msmith Xach_: The end goal is to be able to send data to the process via a stream, preferrably reading and to file descriptors.

13:13:37 |3b| you can ask for lisp streams, not sure about FDs

13:14:01 |3b| probably could extract FDs from the streams if really needed though

13:15:03 msmith 3b: thanks I'll look into that.

13:16:01 |3b| :wait nil :output :stream :input :stream

13:16:51 |3b| then (sb-ext:process-output foo) where foo is return from run-program, similar for sb-ext:process-input

13:17:25 Xach_ sockets are a form of IPC, too

13:22:13 |3b| and don't forget finish-output/flush-output, might also want read-char-no-hang

13:39:46 pipping msmith: there's also uiop:run-program/uiop:launch-program, which is meant to be a portable wrapper around e.g. (sb-ext:run-program ... :wait t) and (sb-ext:run-program ... :wait nil)

13:41:08 |3b| * found uiop's run-program to be lacking in that area when i tried it :(

13:42:36 |3b| (might have been a windows specific problem though, don't remember... not that i got things to work right anyway since the program i wanted to run didn't like being run that way on windows)

13:43:12 Fare or you could simply use uiop:run-program

13:43:55 |3b| can it give you in/out streams for a long-running child process?

13:44:23 Fare only in the latest soon-to-be-released 3.1.7.x series

13:44:39 |3b| ah, cool

13:45:12 Fare but portably to abcl allegro clozure cmucl ecl (and lispworks os-unix) mkcl sbcl scl

13:45:17 |3b| not that it would fix windows emacs :/

13:45:35 Fare what's broken on windows emacs?

13:45:41 |3b| but i guess i could still use it to ssh to linux emacs when i got to that point

13:45:59 |3b| it uses windows console for text mode

13:46:23 Fare there's an emacs that uses the window graphic mode, you know

13:46:25 |3b| so need ugly hacks at best to get access to its in/out, and even then you only get 1 text-mode window

13:46:38 |3b| Fare: doesn't help if i want to display the text myself :)

13:47:01 Fare oh. what are you trying to do?

13:47:04 |3b| (for example when wearing an HMD so i can't see the graphics mode window)

13:47:18 Fare time to resurrect hemlock?

13:48:05 |3b| yeah, one of the CL editors is an option, do any of them have equivalent of slime?

13:48:34 |3b| * crashes lisp too much to run my editor in same image, unless it has very good state crash recovery

13:49:00 |3b| (equivalent of the remote connection part of slime i mean)

13:50:45 |3b| then there's all the random emacs bits i'd miss or need to spend time reimplementing, and i'm already a few stacks of yaks away from even being at the point of displaying the result anyway without replacing my editor:p

13:51:10 dlowe and that's why we all still use emacs

13:51:15 raydeejay "Use ED!"?

13:52:41 |3b| dlowe: yep :(

13:53:44 dlowe Seems to me like some person with more time than me could mash Qt/GTK and guile elisp together to make an emacs that's more sane under the hood.

13:55:09 foom GNU emacs already mashed in gtk. :)

13:55:12 |3b| * also doesn't have the time for the other option of hacking a shared-mem or whatever backend for gnu emacs

13:55:24 dlowe foom: that did not make it more sane.

13:55:41 foom dlowe: no....no, it did not.

13:56:13 dlowe though after using emacs in daemon mode exclusively, I think a system with editor server/clients is the way to go.

13:57:03 |3b| * also considers trying to control it over a socket and just dump screen contents back as fairly raw text, but i don't know enough elisp/emacs to tell how hard that would be

13:57:38 |3b| yeah, as long as it doesn't implement the display in the server :)

13:57:59 dlowe exactly.

13:58:51 foom it does implement the display in the server, doesn't it?

13:59:03 foom I thought emacsclient just sent the TTY to the server, and the server ran the whole thing

13:59:13 |3b| sshing from windows to emacs running on linux connecting to lisp on windows is ugly, but would probably work fairly easily

13:59:31 foom |3b|: you *could* just use unix emacs on windows

13:59:32 |3b| foom: right, which is why it only has 1 console on windows

13:59:44 |3b| foom: on windows 8?

13:59:48 |3b| .1

14:00:12 foom well, cygwin has an emacs

14:00:21 |3b| and even the linux-on-win10 thing sounded like it would not be much better than remote emacs

14:01:00 |3b| * tried cygwin, its pty emulation didn't work with normal sbcl host, anbd was flaky even when i tried it from another cygwin emacs

14:01:18 |3b| (and emacs requires a pty for text mode, and can't just talk to stdin/out)

14:01:36 foom I thought you just wanted to run emacs in the pty, not sbcl

14:01:44 |3b| right, from sbcl

14:02:06 foom oh, that's pretty complicated. :)

14:02:09 dlowe that's what pseudo-ttys are for

14:02:14 |3b| and windows doesn't have PTY, so non-cygwin process can't talk to cygwin fake pty

14:02:17 dlowe right.

14:02:28 joga http://www.theregister.co.uk/2016/10/31/google_man_drags_emacs_into_the_1990s :)

14:02:58 dlowe I'm surprised that escaped into mainstream media

14:03:03 |3b| heh

14:03:32 pipping |3b|: https://gitlab.common-lisp.net/asdf/asdf/blob/master/test/test-run-program-unix.script#L194-621 should give you an impression what run-program/launch-program can currently do

14:03:44 foom funny, i've never noticed that problem

14:03:55 amused ** NICK hydraz

14:04:25 foom Maybe because my display has a compositor, so everything's buffered anyways? (I dunno how this crap works, but that sounds plausible)

14:04:42 pipping Fare: do you have any thoughts on https://bugs.launchpad.net/asdf/+bug/1636903 ?

14:05:06 pipping * will only be able to read replies in a few hours, unfortunately, gotta run now

14:05:28 dlowe I haven't noticed, but I bet emacsclient is hiding it somehow

14:05:29 |3b| pipping: yeah, from what Fare said, the feature i was missing will be out soon :)