IRC channel logs

<lilyp>autotools would typically pull those from the system (or your guix shell, whichever matters), not sure how hall does it

<apteryx>hm, guile 3.0.9 has reproducibility issues when building

<apteryx>packages such as mumi or cuirass don't build reproducibly in Guix

<jackhill>lispmacs[work]: here's something I did will hall and guix four years ago. Not saying it's pretty, but it works. Guile script: https://gitlab.oit.duke.edu/jackhill/guix-installed Guix package definition: https://gitlab.oit.duke.edu/jackhill/guix-at-duke/-/blob/master/guix-at-duke/packages/endpoint-management.scm?ref_type=heads#L20-92

<endersending>That's crazy there is a whole guile chat room

<endersending>I am trying to get guile working but I have no idea where to begin, how to use it. I have so many questions

<endersending>I am using debian, and I did apt install guix, and then guix install guile-ncurses

<endersending>when I run a test program I get this error: no code for module (ncurses curses)

<endersending>is guile only a C/C++ thing? Can I run python-curses-guile?

<rlb>sneek: later tell civodul it's actually hanging in the scm_spawn_thread for the signal_delivery_thread. Vaguely wondering if there might be some race when under load.

<sneek>rlb, you have 1 message!

<sneek>rlb, civodul says: we’ll need to investigate; currently we don’t see that on x86_64-linux-gnu FWIW: https://ci.guix.gnu.org/jobset/guile

<sneek>Got it.

<rlb>sneek: later tell civodul right wrt investigate -- note that this hang only occurs right now with the parallel test change, so it might or might not be a "real" bug.

<sneek>Got it.

<lilyp>endersending: pretty much all modern-day FFI goes through C first. So you can bridge Python<->Guile if you want to, but it'll most likely not be a pleasant experience juggling two APIs claiming objects :)

<lilyp>for ncurses, try `guix shell guile guile-ncurses -- guile`

<lilyp>Guix requires you to specify the interpreter along with all other packages to get load paths rolling

<lilyp>e.g. `guix shell python-numpy -- python3` might even give you a command not found

<lilyp>whereas `guix shell python python-numpy` does what you'd expect

<lilyp>(all the stuff that I wrote for guix shell also apply to guix install, with the difference that shells are transient)

<natmeo>can i do anything to suppress the display of very long arguments (vectors) in error output?

<lilyp>natmeo: ice-9 pretty-print has a function to print arguments in a truncated fashion

<lilyp>(IIRC it's truncated-print)

<natmeo>lilyp: is there a printer for exceptions that I can hook it up to?

<natmeo>probably there is

<lilyp>natmeo: set-exception-printer!

<natmeo>naturally

<natmeo>i have to say, now that i grok scheme a little i kinda enjoy it

<natmeo>with some luck i will never have to write any php again

<cow_2001>HARK! i am adapting the https://standards.scheme.org/corrected-r7rs/r7rs.html output to texinfo: https://codeberg.org/kakafarm/guile/src/branch/work-in-progress-texinfo-r7rs-small/doc/r7rs-small/r7rs-small.texinfo

<endersending>thank you lilyp

<endersending>So i Ran this: guix shell guile guile-ncurses -- guile

<endersending>and it installed huile again and I think ncurses.. but I still get the same error running ./hello.scm

<endersending>guile*

<mwette>works for me: https://paste.debian.net/1305456/

<endersending>mwette, It works that way for me I guess

<endersending>why can't I run ./hello.scm?

<mwette>not sure what's in hello.scm. You might try adding `-L .` to the end of the `guix shell' command

<endersending>this hell world example is what I am using: https://www.gnu.org/software/guile-ncurses/manual/html_node/Using-the-Guile-Curses-module.html#Using-the-Guile-Curses-module

<mwette>or `guix shell ... -- guile -L . ./hello.scm'

<endersending>mwette, yes that worked

<mwette>Ah.

<endersending>thanks for the help. have to go for now. I will try some more later

<mwette>and that script maybe should have first line #!/usr/bin/env guile

<mwette>If you do that, and then type `guix shell guile guile-ncurses` you get a good environment; then you can do `user@host [env]$ ./hello.scm`

<mwette>snook: later tell endersending, use `guix shell guile guile-ncurses` (no `-- guile') and change top line of `hello.scm' to `#!/usr/bin/env guile' and it should work

<mwette>sneek: later tell endersending, use `guix shell guile guile-ncurses`

<mwette> (no `-- guile') and change top line of `hello.scm' to `#!/usr/bin/env

<mwette> guile' and it should work

<sneek>Will do.

<mwette>yikes

<ttz>Hi, all! I am wondering how to write a macro to cons `n` times a given form. I first tried using a procedure, but then the form gets evaluated all list elements point to the same object, which is not what I want.

<ttz>When I try replacing `define` with `define-syntax-rule`, it seems that I create an infinite loop:

<ttz>(define-syntax-rule (repeat-macro n expr)

<ttz> (if (< n 1)

<ttz> '()

<ttz> (cons expr (repeat-macro (- n 1) expr))))

<mwette>try syntax-rules : https://www.scheme.com/tspl4/further.html#./further:h1

<mwette>That won't work.

<mwette>You might need syntax-case. So you want (repeat-macro 20 ex) => (cons ex (cons ex ...)))

<ttz>Thanks, I will have a look at syntax-case, then.

<rlb>ttz: is the goal to write a macro, or rather, what's the top-level problem?

<rlb>Was mostly just wondering whether say the scheme equivalent of clojure's (repeatedly n fn) would suffice.

<ttz>No the goal is not to write a macro. But my first implementation unsing a procedure led all repetions to point to the same location. After calling (repeat 2 '(1 2)) => $1 = ((1 2) (1 2)), setting the car of the first element impacts them all: (set-car! (car $1) 0) then $1 = ((0 2) (0 2))

<rlb>Ahh.

<rlb>(repeat 2 (lambda () (list 1 2)))?

<rlb>or similar...

<lilyp>pro tip: avoid procedures with a !

<ttz>Yes sure

<mwette>Do you want solution? I have something working.

<lilyp>there's typically a non-destructive alternative

<rlb>i.e. build a new one for each repition.

<rlb>It's actually not iirc defined wrt set! on constant lists?

<rlb>iirc with compilation guile may put those in a read only elf section, or am I completely misrembering?

<ttz>what does iirc stand for?

<ttz>mwette: sure if you have one

<rlb>sorry -- "If I recall correctly"

<ttz>rlb: about wrapping the argument in a lambda: I wanted a "cleaner" interface (one that does not rely on the caller to know what it's doing)

<mwette> https://paste.debian.net/1305461/

<ttz>thanks for the code mwette

<ttz>It works when passed (list 1 2) for example. Bur if I pass '(1 2) it still shares the same value accross repetitions...

<ttz>Anyway, I still have a lot to learn about macros. Thanks for answering :)

<ttz>Maybe a last newbie question: is there a standardized or usual way to represent dense matrices in scheme? (I initially tried to write the repeat procedure to initialize a matrix of arbitrary dimensions)

<bjoli>Ahoy hoy friends!

<mwette>that's a differnt problem: you can't mutate literals

<mwette>i.e., (set-car! '(1 2) 3) is invalid code

<ttz>Yes, but one could want to allocate twice the same literal in a list, such that the resulting list elements do not point to the same location. When I create a list with (repeat 2 '(1 2)), I would have wanted to get a list ((1 2) (1 2)) such that the two sub-lists do not share data.

<mwette>for kicks you could try starting guile, (use-modules (system base compile)) (default-optimization-level 0) then load and use repeat-macro ; maybe the CSE optmization will make separate instances

<rlb>ttz: the key thing is that that modifying a constant isn't allowed by the language, so you can't assume anything in particular, i.e. guile could crash if you do that, ignore the modification, re-use the same storage, etc.

<rlb>And whatever guile 3.0 does, 4.0 might do something different.

<mwette>I could see that he would want non-atomic(?) literals to be not-`eq?' ((sxml xpath) depends on that, for example) If you know these are lists then one can use deep copy from (ice-9 copy-tree)

IRC channel logs

2024-01-27.log