IRC channel logs


back to list of logs

<KarlJoad>Does anyone have documentation or examples on how use the guile-sqlite3 library? I am finding it difficult to work with just the exported procedures.
<old>And here's a query example:
<old>Note that I would not recommend to do like I did. I found it very cumbersome to do it that way, but I was in a rush
<old>Any way to determine the distribution in a portable way with guile?
<old>There's uname but it does not say for example ubuntu/archlinux
<KarlJoad>old: I believe the general-ish solution for determining distro is to read /etc/os-release.
<KarlJoad>Thanks for the links. These will be super helpful!
<old>oh awesome thanks for /etc/os-release! is that part of FHS?
<old>Happy to help :-) Like I said, my usage of it is terrible. I'm sure you will find a better way of doing things :-p
<KarlJoad>As far as I can tell, it is not part of FHS, but something systemd introduced that many other systems have picked up.
<KarlJoad>Even Guix provides it now, as it is a simple thing to have present which makes many things a lot easier to handle.
<dadinn>i can see that syntax-locally-bound-identifiers is defined in (ice-9 psyntax) module, but some reason when I try to use-module it says that no code exists for it
<dadinn>it seems the psyntax module was last edited in 2012, so I would assume my Guile 3.0.5 should haves it!
<cow_2001>how do i have a sort of (runtime), as in SICP exercise 1.22?
<KarlJoad>cow_2001: Not exactly equivalent to (runtime), but similar in this exercise is (current-time). Rather than measuring the uptime of the system, it takes the time since UNIX epoch. If you need finer granularity, I believe you can use (times). See (guile) Time.
<cow_2001>oh oops
<cow_2001>to be honest i don't quite understand what they are asking me to do
<KarlJoad>Can doc-comments be added to macros like they can for procedures, e.g. (define-syntax macro-name "documentation" ...)?
<lilyp>KarlJoad: at least in syntax-case they can
<KarlJoad>lilyp: I'm using (define-syntax macro-name (syntax-rules ...)), if that changes things.
<lilyp>that'd change my answer to "I don't know"
<jpoiret>syntax-rules is just a macro that extends to syntax-case, so theoretically yes
<chrislck>sneek: botsnack
<sneek>dsmith-work: wb
<dsmith-work>sneek, botsnack
<dsmith-work>Howdy Guileers
<dadinn>hi all
<gnucode>heyo dadinn
<dadinn>I am looking at the docs:
<dadinn>it says there should be a syntax-locally-bound-identifiers procedure
<dadinn>I can't find it in Guile 3.0.5
<dadinn>any suggestions where to find it?
<dadinn>I am thinking I would need it, because I am writing a syntax-case macro, and want to ensure that an identifier is defined in the lexical context of the expansion, else use some default values instead it.
<dadinn>I can see in the sources that it should be defined in (ice-9 psyntax) module, but it is missing for me for some reason:
<dadinn>hmmm... thought there is no define-module of any (ice-9 psyntax) anywhere in this file, which might explain why I cannot load it! :/
<dthompson>yeah (ice-9 psyntax) is special
<sneek>dthompson, you have 3 messages!
<sneek>dthompson, old says: here's with getters/setters inside srfi-9 record:
<sneek>dthompson, old says: there's couple of things to get fix. First module resolving of type with (system foreign) means no composition of structure
<sneek>dthompson, old says: then there's the referencing with bytevector with I don't like.
<dthompson>the manual says that syntax-locally-bound-identifiers is available in (system syntax)
<dadinn>dthompson: ah, thanks... didn't notice that piece of code! :'D
<dthompson>I've never used anything in that module but I was poking around that section of the manual recently. good luck!
<old>oh hey dthompson
<old>what do you think of my lastest version:
<old>(example at bottom)
<dthompson>old: whoa that looks pretty cool! does it borrow from guix's code? I saw someone point to that in the log
<dthompson>I'd have to try it out to really understand but the point example looks good!
<dthompson>something like this should really make its way into guile core
<old>Nope. daviid did point to me the guix code, but I was unaware of it
<dthompson>ah okay, they look a bit similar. probably a good thing!
<mwette>are those structs or classes (add methods)?
<old>Guix version is way better (I'm bad at syntax-case)
<old>But I support nest type. I don't see Guix supporting that
<old>mwette: What do you mean?
<dthompson>nested types are important!
<dthompson>mwette: they are structs
<mwette>i see define-c-struct ... copy-point. Is copy-point data, C function, or schemem procedure?
<dthompson>that's a scheme procedure
<old>yes. And you get a reference. For example (rectangle-point rect) will give you a srfi-9 record with a slice of the underlying bytevector of the rectangle
<old>mwette: All Scheme
<old>Nothing C
<dthompson>old: that is awesome! bytevector slices are great.
<dthompson>great new feature.
<old>So great
<old>Now I want to support pointer to structure and not just composition
<old>Also right now you have to define <point> before using it in <rectangle> which make sens. But if I were able to make a two passes compilation that could change!
<dthompson>I maaaay see some sources of unnecessary allocation in here but I'm not sure
<old>But I think syntaxes have the same probleme in Scheme. Order is important
<dthompson>fwiw I think it's fine that you have to define <point> before <rect>
<mwette>ever read about the julia type system?
<dthompson>old: the record type only has a bv field, so every time it needs converting to a pointer via bytevector->pointer a new pointer object is allocated.
<dthompson>if you're making lots of calls, the allocations add up.
<old>hmm true I did not think of that
<dthompson>I have some performance sensitive code where I cache the pointer in a record field
<old>I could probably cache the pointer in a member instead
<dthompson>yes, I recommend that!
<old>will do
<dthompson>no need to allocate the pointer until someone asks for it, either.
<old>it depends
<old>Having set in the CTOR ahead of time as its advtange
<old>Doing if-statement in a hotpath can be costly
<dthompson>yeah it's a trade-off
<dthompson>maybe in this case it would be better to allocate it up front.
<old>I think so. Memory is not that much sacred anymore. We can spare the 8 bytes :-)
<dthompson>I have some code that only sometimes needs pointers. plenty of objects are created that will never need a pointer.
<old>hmm true
<dthompson>so it slows things down to create the pointer for every object
<old>I was thinking of adding some configuration to the macro
<old>(defin-c-struct <type> (#:packed #:aligned cache-line)) stuff like that
<old>maybe a (define-c-struct <type> (#:lazy)) could be it
<dthompson>personally I don't have bottlenecks around lazily allocating pointers
<dthompson>it might be better to just not cache if there isn't one way to settle it
<old>it's good to give the possibility though. This way it's just a matter of flipping a f to a t to see the difference
<mwette> when I am banging on a lot of C I often do: (define &x (pointer-to x)):
<mfiano>Hmm, what is the signum function called in guile?
<mfiano>or scheme for that matter