IRC channel logs

<dthompson>civodul: the reason for needing a record is that we need something to encapsulate the bytevector and offset. whether or not to build a type hierarchy for strong typing could be up for debate. personally I see value in it and it's what ftypes do

<dthompson>civodul: but hey I'm glad you were able to try it so quickly and that you like it! can't think of a much better endorsement :)

<dthompson>you can also directly get/set into bytevectors without an intermediate record btw! bytestruct-pack and bytestruct-unpack do that

<civodul>oh, that’s excellent

<dthompson>yeah, it's a must-have for me for packing gpu buffers with vertex data

<dthompson>but the wrapper structs are nice when you want encapsulation

<dthompson>and to give someone an API that looks like a regular Scheme record

<dthompson>(bytestruct-unpack <vec2> ((y)) (f32vector 1.0 2.0) 0) would return 2.0

<civodul>very nice

<civodul>the best of both worlds!

<freakingpenguin>Hi all! Is the reason this snippet https://paste.debian.net/1309708/ fails because the define-syntax macro is hygienic?

<freakingpenguin>I have to define a public function with some shared logic in several different modules due to some other code and figured macros were the least code-dupe way to handle it.

<freakingpenguin>I should be able to cludge something together with functions but a (define-foo) macro feels like it would be cleaner.

<rlb>freakingpenguin: yes, I think more or less -- try (macroexpand '(test 'foo))

<rlb>But you can likely use syntax case, and break hygiene if you like, or you could (less safely, depending on the rest) use define-macro.

<rlb>freakingpenguin: perhaps related - https://www.gnu.org/software/guile//manual/html_node/Hygiene-and-the-Top_002dLevel.html

<rlb>And here with a syntax-case variant of your macro:

<rlb>scheme@(guile-user)> (macroexpand '(test 'foo))

<freakingpenguin>rlb: Thanks a bunch! From what I've read of syntax-case in the manual it looks like datum->syntax is how I'd intentionally break hygiene, is that right?

<chrislck>sneek: botsnack

<sneek>:)

<rlb>freakingpenguin: yep

<rlb>Though note that you may not *need* it depending on what you're doing, i.e. here lokke creates the top-level binding for a clj protocol without it: https://codeberg.org/lokke/lokke/src/branch/main/mod/lokke/datatype.scm#L129-L132

<rlb>s/i.e./e.g./

<rlb>ACTION double-checks that name wasn't introduced some other way, now that he said that...

<rlb>Also, if your end-goal is a module, I'd probably work immediately with that, so you have the semantics (top-level and otherwise) you want from the start.

<rlb>Those define-generics just below the link are also "top level" defs, fwiw.

<freakingpenguin>Yeah, this is roughly what I'm going for if you're familiar with Guix and/or Cuirass: https://paste.debian.net/1309712/.

<rlb>I'm not, yet.

<freakingpenguin>Thanks for the example! Is... that project an implementation of Clojure in Guile?

<rlb>fsvo implementation -- it's intended to be heading in that direction :)

<freakingpenguin>"For a more Scheme oriented experience in Scheme" I love this community. Everywhere I look there's some cool new project.

<rlb>Yeah, up to a point I've tried to keep some bits that might also be useful to guile implemented first as mostly scheme-friendly, and then I've build the clj side on top.

<rlb>some of that's in here fwiw: https://codeberg.org/lokke/lokke/src/branch/main/mod/lokke/scm

<mwette>Due to the day job, I missed the entire struct conversation. bytestructures provides macros to generate offsets of struct-in-struct-...-in-structs.

<mwette>I have been using that in ffi-helper, but a little while ago I started looking at new form. My original interest was in supporting cross architecture access. (e.g., accessing data structures snarfed from microcontrollers).

<mwette>Also, I plan to parse C structures down to primitive types: u8, i16, f32, etc. so that equality predicates are doable.

<flatwhatson>bytestructs sound great, just need something like nyacc to autogen them from C headers

<old>mwette: I've had some syntax that does that

<mwette>does what?

<old>defining complex C-struct around a bytevector

<old>with accessors

<old>by complex I mean compound one, not just flat structure. With union and enum

<old>it is just very slow unfortunatelly

<old>I'm not sure if that's what you are refering to by bytestructures, I have not seen the conversation

<mwette>OK. nyacc will generate bytestructure descriptors for all structs inside include files. The bytestructures package provides accessors, but no in the flavor of scheme records.

<mwette>this is bytestructures: https://github.com/TaylanUB/scheme-bytestructures

<old>hm it looks similar to what I have

<old> https://git.sr.ht/~old/guile-ev/tree/master/item/ev/ffi.scm.in

<old>these are the bindings for libev I made, manually written

<old>using the following syntax mess: https://git.sr.ht/~old/guile-ev/tree/master/item/ev/foreign.scm

<old>but it is nice to see we can have something portable across Scheme with bytestructures

<mwette>nyacc autogenerates the scheme code. For glib gobject.h the file is ~13,000 lines of code; for gtk2.h it's 98k lines of code.

<mwette>example of gtk2 program using all scheme ffi code: https://git.savannah.nongnu.org/cgit/nyacc.git/tree/examples/nyacc/lang/ffi-help/gdk2-03.scm

<mwette>gtkgl example: https://git.savannah.nongnu.org/cgit/nyacc.git/tree/examples/nyacc/lang/ffi-help/gtkgl-01.scm

<mwette>mixing C and Scheme has its idiosyncrasies

<jmes>How do I make (file-exists? "$XDG_RUNTIME_DIR/frumble") understand the environment variable?

<jmes>Generally, how do "expand" strings with environment vars?

<jmes>how does one*

<mwette>(file-exists? (string-append (getenv "XDG_RUNTIME_DIR") "/frumble"))

<mwette>you could also use (format ...)

<jmes>mwette: Thanks a bunch. And the universe seems to be poking fun at me because as soon as I bother you about it I finally find the relevant docs online...

<mwette>no problem; glad to help

<ArneBab>Does anyone have a high performance implementation of skip lists for Guile?

<sneek>Yey! chrislck is back :)

<dsmith>sneek, botsnack

<sneek>:)

<dsmith>!uptime

<sneek>I've been serving for 10 days

<sneek>This system has been up 48 weeks, 4 days, 21 hours, 14 minutes

<freakingpenguin>Hi all! I'm trying to add a fender to a syntax-case macro to restrict input to only "list of X", but I can't seem to get it working. https://paste.debian.net/1309766/

<freakingpenguin>Pretty sure the problem is from wrapping the list in a syntax but I can't get rid of it.

<freakingpenguin>In this case I guess I could turn it into a datum but in my actual use case that list would contain variables, not constants, so I don't think that would work.

<mwette>maybe expand to (define foo (if (every ...) bar (error "")))

<old>freakingpenguin: fender is not what you want if the elements are not constant I think

<freakingpenguin>Alrighty, thanks! Do you know of any good resources on this topic? The manual is a bit hard to parse sometimes.

<old>On syntax fence?

<freakingpenguin>Huh, it does actually work with syntax->datum and an eval in the fender, it's just really ugly. https://paste.debian.net/1309776/

<old>freakingpenguin: right because the expander can deduce that hello is a constant number

<freakingpenguin>I did also try it cross-module with some records, seemed fine.

<old>try it inside a function that received some parameter

<old>on the parameter

<old>anyway, I would not mix eval and fence, you might encouter issue in the future

<freakingpenguin>You're right, it can't handle that. Putting in a conditional error clause is definitely cleaner, just wish I could understand all the finer points of macros. Guess that just comes with practice.

<dsmith>sneek, macros?

<sneek>I could be wrong, but macros is http://hipster.home.xs4all.nl/lib/scheme/gauche/define-syntax-primer.txt

<dsmith>freakingpenguin, ^^ Though that's only syntax-rules, not syntax-case

IRC channel logs

2024-03-07.log