freenode/#lisp - IRC Chatlog

8:10:08 theothornhill You could extend cl-font-lock-built-in-mode?

8:23:23 splittist I would like to use a Docker container to build an executable from a lisp project. There are many fine lisp images, but what's the best way to give the container access to the local projects (i.e. the components not in quicklisp) that are needed to create that executable?

8:29:01 lukego splittist: Maybe docker's `-v` to bind-mount ~/quicklisp/local-projects from your master repos on the host? https://docs.docker.com/storage/bind-mounts/

8:30:04 splittist Thtat

8:30:37 splittist (new keyboard). lukego: that seemed the obvious way, so I assumed it was wrong (:

8:31:21 lukego new keyboard! I remember seeing your old one, it looked like the best one available on the market ;-)

8:32:36 lukego Maybe `-v` isn't ideal really since that's more of a runtime rather than build-time thing. If you want `docker build` to do the whole job then maybe that doesn't work. I guess you could have your Dockerfile COPY stuff into the same location instead?

8:33:13 lukego disclaimer: I've used Docker for various things but usually just cutting through the shortest path without necessarily knowing what is correct..

8:36:44 splittist I sort of have this feeling I should be copying the local projects and letting quicklisp pull things into the container. Particularly since I have a locally-patched version of some things to make them work on my Windows dev machine and the container will be linux, of course.

8:39:18 lukego That makes sense to me but take that with a grain of salt :)

12:24:30 Xach splittist: i would be inclined to pin the quicklisp version for that kind of use

12:39:00 jmercouris CLPM

12:46:51 splittist Xach: that seems like a good idea.

12:47:12 Xach unless your goal is to test ongoing compatibility, that is

12:48:01 splittist I think I'd call that 'homework mode'

12:50:29 splittist My problem is that the executable is to use in a non-lisp project (<gasp!>), and I was hoping to use a multi-stage dockerfile to build the executable then COPY -from=lisp-building-container-thingy etc. But dockerfiles can't pick up context outside their root.

12:52:41 splittist (I'm mangling the terminology, sorry.) So either I (a) copy the relevant local-projects into the non-lisp project; (b) develop them under this non-lisp project (but some of the lisp projects are general libraries, not limited to this use case); or (c) create my own quicklisp distribution ???

12:53:26 Xach you don't need a distribution if you just want to get some code. it could be a tarball that you fetch and unpack.

12:55:19 splittist OK. (It's just annoying that the files are sitting there in a sibling directory and I have to do something outside the dockerfile to get at them. I understand the reasons, but... I think I'm thinking about this the wrong way.)

13:10:40 Xach every time you fetch quicklisp stuff in a CI pipeline it costs me a picodollar

13:11:35 jackdaniel (loop (vroom :clean t))

13:18:21 flip214 Xach: during the next real-world ELS I can offer you a nanoeuro

13:19:47 jackdaniel or a nanocrown in some countries, and the crown is trendy lately

13:47:57 Xach μbles?

13:54:41 shka_ Xach: good one

15:05:02 edgar-rft with a nonocrown you're nanoking! welcome to your nanokingdom

15:22:41 dbotton Working now on "thick" bindings with clog now - Working api for desktop environment in and now adding dialogs etc :)

15:25:36 mfiano Congratulations on all the positive feedback on CLOG and your hard work learning CL

15:42:55 splittist So my solution is to create the lisp-builder with executable image separately. Only my dump-exe.lisp script, which runs fine from outside the container, and fine from inside the container, does not run from a RUN command in the Dockerfile.

15:59:18 dbotton mfiano thanks!

16:00:39 ldbeth good midnight

16:00:50 ldbeth ACTION lol

16:15:41 flip214 Is there an equivalent of CHECK-TYPE that evaluates the type? So that I can pass in expected array sizes?

16:16:31 jackdaniel flip214: (assert (typep)) ?

16:16:32 flip214 Like (check-type input `(array fixnum (,size ,size)))

16:16:45 jackdaniel you can't do it at runtime, that's the point of check-type

16:17:06 jmercouris you can do type-of

16:17:23 jackdaniel jmercouris: and what of it?

16:17:27 jmercouris you could do it

16:17:36 flip214 jackdaniel: well, I want to tell the compiler that the used indizes _must_ be in the array, so no bounds checks are necessary in a loop

16:17:40 jmercouris type-of object => typespec

16:17:46 jmercouris compare if it matches your expectations... at run time

16:18:04 jackdaniel flip214: ah

16:18:23 jmercouris also you are a tomato

16:18:29 jmercouris sorry, wrong channel

16:20:21 jackdaniel flip214: all I can think of is typecase with known sizes put in there and a generic otherwise clause

16:22:07 scymtym flip214: do you want the code to still be safe? if not, (locally (declare (optimize (sb-c::insert-array-bounds-checks 0))) …) would be one (non-portable) option

16:23:04 jackdaniel (declare (safety -1))

16:24:00 jmercouris no idea what is going on here: http://dpaste.com/DUJWF7SF9

16:24:09 jmercouris I don't understand this error message

16:24:12 flip214 scymtym: I'd like to still be safe... OTOH, if I check the input types beforehand, the arithmetic can't go wrong - so SAFETY 0 might be acceptable

16:24:43 jmercouris (INTERN (SYMBOL-NAME INTERFACE) (PACKAGE-NAME (SYMBOL-PACKAGE INTERFACE))) is not a string designator.

16:24:47 jackdaniel jmercouris: the first argument to define-user-class is not evaluated

16:24:50 jmercouris do I have to evaluate it or something

16:24:55 jmercouris ah damnit I was just about to say something like that

16:25:03 jmercouris OK, so how should I rewrite this?

16:25:08 flip214 well, I've got an (ASSERT (TYPEP input `(array fixnum (,size ,size)))) and (DOTIMES (i (1- size)) (DOTIMES (j (1- size)) ...))

16:25:27 flip214 and still get HAIRY-DATA-VECTOR-REF

16:25:34 jackdaniel (eval `(define-user-class ,(intern …))

16:25:49 jmercouris or otherwise make a macro myself?

16:26:06 jmercouris I guess I can eval it for now

16:26:47 jmercouris thanks jackdaniel

16:27:23 scymtym flip214: or compile specialized versions dynamically if jackdaniel's suggestion is impractical: (lambda (a) (let ((table (load-time-value (make-hash-table :test #'equal))) (dims (array-dimensions a))) (funcall (alexandria:ensure-gethash dims table (compile nil `(lambda (a) (declare (type (array t ,dims) a)) a))) a))) (not thread-safe)

16:27:30 jackdaniel sure

16:28:17 flip214 scymtym: uh, ouch

16:28:59 scymtym i for one am glad that we /can/ do something like that if needed

16:32:02 contrapunctus Anyone know why SLIME autodoc might work correctly for a local image, but stop working over SSH?

16:32:38 prxq contrapunctus: paths

16:32:40 jackdaniel maybe it tries to find locally the file?

16:33:30 prxq contrapunctus: what exactly do you mean 'over SSH'? You log in via SSH remotely? Or are u using swank?

16:33:52 flip214 scymtym: well, I'd like to tell the compiler that the array is a specific size; and DOTIMES won't go over the limit (that should be known already)

16:34:46 scymtym flip214: i can comment on that later

16:39:48 contrapunctus prxq, jackdaniel: oh, I was following the steps here - https://common-lisp.net/project/slime/doc/html/Connecting-to-a-remote-lisp.html but now that I think about it, I can't remember if I did the path translation thing. I'll take a look at that.

16:44:46 flip214 even (DOTIMES (i (array-dimension input 0)) ...) doesn't help.... what am I doing wrong?

16:45:08 jackdaniel you are assuming the sufficiently smart compiler

16:45:11 jackdaniel compilers are dumb :)

16:49:45 flip214 https://paste.debian.net/hidden/146f51f1/

16:50:03 flip214 jackdaniel: I got some similar stuff working with sbcl already

16:50:13 flip214 just right now I'm too dumb, I believe

16:50:42 Bike yeah i'm not sure if sbcl can actually track that kind of dependent type

16:51:27 Bike also, you have it as an array rather than a simple-array

16:51:34 Bike that might screw up doing any optimization whatsoever

16:53:22 Nilby yes, my experience is that you ususally need a simple-array to get the fast path

16:56:53 Nilby although I guess, like the case in the paste, sometimes it's obvious it's not displaces, fill-pointered, or adjusted

16:57:20 Bike i don't see anything in the paste that would force M to be a simple array

17:11:42 Nilby ACTION fell down an asm diffing rabbit hole

17:12:28 contrapunctus prxq, jackdaniel: thanks, it was indeed missing path translation config! ^^

17:40:33 X-Scale` ** NICK X-Scale

17:52:10 scymtym flip214: as bike said, to generate optimized array accesses, SBCL must infer the simpleness of the array, the array element type (even if t) and the array dimensions. i'm certain that it can't do that based on a dynamically constructed TYPEP assertion. however, if you write something like (dotimes (i (array-dimension a 0)) …), SBCL can infer a constraint between I and A so that accesses in the body will not require a bounds

17:52:10 scymtym check

17:54:49 scymtym here is a small example that shows the constraint: https://techfak.de/~jmoringe/sbcl-infer-array-in-bounds.png

17:55:55 scymtym not sure how this would work for ranks other than 1, though

18:42:06 flip214 scymtym: Bike: https://paste.debian.net/hidden/0601d838/ has simple-array in the assert, and uses array-dimension - still, there's a HAIR in the soup...

18:43:44 flip214 ah, with (DECLARE (TYPE (SIMPLE-ARRAY... ))) it looks better

18:44:09 scymtym flip214: 1) what is MATRIX? 2) as i said, i don't think the (assert (typep …)) with a dynamically constructed type specifier allows inferring anything

18:46:25 scymtym also, speed 3 with safety 0 will elide the bounds irregardless of type inference and constraints

18:47:45 scymtym (compile nil (lambda () …)) (as opposed to (compile nil '(lambda () …))) does not usually compile

19:18:00 nij Any more interesting examples that make use of monads, as in https://common-lisp.net/project/cl-monad-macros/monad-macros.htm ?

19:19:10 _death https://8c6794b6.github.io/posts/Delimited-continuations-with-monadic-functions-in-Common-Lisp.html

19:22:16 nij Oh great! I missed this while searching :-)