IRC channel logs

<daviid>ArneBab: using this repo git://github.com/ijp/guildhall.git, which seems to have your patches, guild update raises an excepton using 2.1.7.10-f261eaf, is there a working repo somewhere or is that what we have for now?

<amz3>there is port of industria to guile in that repo

<daviid>knowing i have 'discrete' float values, such as 1.0, 2.0 ... is there a faster way to convert to exact then our inexact->exact ?

<spk121>daviid: all the numbers have zero after the decimal place?

<daviid>spk121: yes

<spk121>daviid: I don't have a better idea. In the general case (inexact->exact (round x)) is often faster than (inexact->exact x)

<spk121>daviid: but probably not for your

<daviid>spk121: trying to boost some image processing relayed stuff, in this case, im-scrap, where 40% is spent in inexact->exact

<daviid>s/relayed/related

<daviid>and I know they are 'discrete' floats, I need their 'integer' value to access a vector (which is cache to boost performance precisly...)

<daviid>I was hoping there is a math 'shortcut' in case ...

<amz3>maybe (inexact->exact (round x)) triggers the optimisation, dunno

<amz3>apparently it's possible to solve the package dependency algorithm using minisat https://en.wikipedia.org/wiki/ZYpp

<xkapastel>amz3: reminded me of https://research.swtch.com/version-sat

<xkapastel>particularly this quote: One reaction is to embrace the complexity and be thankful that SAT solvers are as good as they are. Another reaction is to ask whether this can possibly be a good idea. Maybe we should not be building tools that require solving this problem. Maybe something has gone wrong in the way we develop software.

<amz3>tx!

<amz3>They are arguing that sat solver is overkill for the subset of the problem of package management basically

<amz3>that package management can be made a simpler problem, so that it isn't NP-complete

<xkapastel>yeah

<amz3>yes, but I need to learn minisat or 3-sat solvers, because that's what grammar link use for deciding the correct grammartical parse of a natural language sentence

<xkapastel>it does seem like allowing two different versions of a package is a simple solution

<amz3>yes

<xkapastel>well, sat solvers are good for all sorts of things

<xkapastel>so don't let this turn you off them :P

<amz3>prolly! I need to find a 3-sat solver with which I can work with now

<spk121>daviid: I decided to goof off with your prob for a second. exact->inexact is very complicated

<spk121>daviid: a func like this http://paste.lisp.org/+7APG. is 100x faster

<spk121>daviid: i made a tiny binary module to test it out

<spk121>daviid: so it is a solvable problem. But, it requires some C hacking in upstream.

<daviid>sneek: later tell spk121 thanks! i was hoping to reduce the time still using scheme only, but thanks for the pasted code, let's talk (maybe) about this .. thanks again

<sneek>Okay.

<marcus>HELP

<ryanwatkins>Hi guys, I am trying to configure a guile library from source on GuixSD, any idea how to solve the error 'No Guile development packages were found'?

<ryanwatkins>It is the chickadee game engine library :D

<jmd>ryanwatkins: Guile is probably not in you profile

<Amynka>hey guys :) https://bugs.gentoo.org/show_bug.cgi?id=610052 any idea?

<slyfox>Amynka: looks like a ~known gcc-6 bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71020

<Amynka>slyfox: that would make sence if the guy does not have gcc 4.9.4

<slyfox>-Os might also trigger an old gcc bug

<Amynka>slyfox: yea i mostly wonder what upstream thinks

<Amynka>slyfox: i will try though good tip :)

<Amynka>thanks

<amz3>let's look at this sat resolver for computing package dependencies

<amz3>I stumbled upon a scheme version of sat solver, but without licence file, I am wondering if I had the right to read that code...

<amz3>Someone added gnu guix to http://guildhall.hypermove.net/

<amz3>:)

<amz3>It seems like I need to write the sat solver myself :/

<amz3>I was reading again my code, and it's there is minor differences betweeen webui.scm @ master and webui.scm @ guildhall.hypermove.net

<amz3>fun! In git lab you can create merge requests between branches

<amz3> https://framagit.org/a-guile-mind/culturia.one/merge_requests/1.diff

<amz3>basically in master I use the graph database with guix-web run-server handler

<amz3>in guildhall branch it's minikanren, some validation of input and another run-server handler that dispatch on http method too

<amz3> http://culturia.one/

<amz3>So I think i have sat solver ready

<amz3> http://paste.lisp.org/display/340542

<amz3>now i need to test to test it to resolve dependencies

<wingo>good evening

<paroneayea>hello *

<wingo>heya

<daviid>hello

<paroneayea>ACTION boots up haunt again

<paroneayea>time to update https://activitypub.rocks/ now that we have an implementation reports template

<paroneayea>probably should update the 8sync site too pointing to the videos from chicagolug / fosdem too

<wingo>ACTION trying to bang together my spam classifier with tekuti

<paroneayea>wingo: ooh I am interested in that btw for different reasons

<paroneayea>wingo: I'd like to play with your code once you got it going... I want to experiment with anti-spam, and eventually anti-abuse, systems for federated networks

<amz3>expressing a dependency seems so simple in predicate logic, that it looks unreal: (or (not haunt) guile)

<amz3>paroneayea: how many implementation a activitypub are on going?

<paroneayea>amz3: not 100% sure, but Pubstrate (Guile server and minimal scheme'y client) and socie-el (emacs client) are the main testbeds I'm hacking right now, and I know Benjamin Goering has one, and this summer MediaGoblin will be implementing it

<paroneayea>additionally, I *believe* Pump.io will be implementing it, as well as Mastadon, probably uh, the OwnCloud replacement

<paroneayea>what's it called

<paroneayea>NextCloud

<paroneayea>that's the one

<paroneayea>and Amy Guy has an implementation

<amz3>oh

<amz3>lot of people seems to be interested

<paroneayea>so a few, but we could use more, and we could use help with submitting implementation reports :)

<paroneayea>if you want to take a crack at it:

<paroneayea> https://github.com/w3c/activitypub/blob/gh-pages/implementation-reports/TEMPLATE.md

<amz3>right now the only "thought" as I have is that one could create a P2P network on top of activitypub but not sure how

<amz3>paroneayea: I need to finish my current mini project, I'd like to improve guildhall package dependency solver using sat algorithm

<amz3>it seems like apt has use also a sat solver if you request it

<amz3>maybe it can be useful for guix

<wingo>paroneayea: i don't know much about the field. bayesian classification works ok when there are just two categories, attacker not too too active, and you can tokenize well; but it seems that other strategies are the state of the art these days

<wingo>would be nice if we had a text classification community in guile; we have a nice dev environment, so it's possible one day i guess ^^

<paroneayea>:)

<wingo>paroneayea: so are you free of w3c things now, activitypub-wise?

<paroneayea>wingo: I wish

<paroneayea>wingo: Activity*Streams* made it to Proposed Recommendation, which means it's left the hands of the group and is on to w3c advisory committee to approve

<paroneayea>which is great, because that's the vocabulary backbone of ActivityPub

<paroneayea>however, ActivityPub still needs implementation reports and a test suite

<paroneayea>but!

<paroneayea>now at least I have some nice async options to build the test suite on in Guile :)

<wingo>paroneayea: have you been able to try fibers yet?

<wingo>i know there were some gouix snafus before

<wingo>*guix

<paroneayea>wingo: not yet. I want to soon!

<paroneayea>wingo: now that it's in guix, it's a finding a spare moment thing :)

<paroneayea>wingo: I have read the manual several times though :)

<wingo>:)

<paroneayea>wingo: I have been thinking a lot about how to put 8sync on top of fibers

<paroneayea>I think it'll be doable but the "juggling actors" bit might be a bit trickier, or would have to be redone... I'll see :)

<paroneayea>explaining that probably isn't worth it without trying it

<paroneayea>so soon! I'll know better

<daviid>folks, yesterday spk121 posted http://paste.lisp.org/+7APG, to help me boosting (inexact->exact (round val)) where val is always a 'discrete' float (such as 1.0, 2.0 ...) but (a) I don't know how to use this from scheme :), and (b), it uses SCM_I_MAKINUM, so that is part of our VM, so the quizz is is there a way to access this SCM_I_MAKINUM from scheme 'immediately'?

<daviid>in the above I used round because spk121 said it somehow makes inexact->exact faster

<daviid>though I know all floats are already rounded

<daviid>and the above does not seems true, I removed the call to round and win 0.03 seconds

<spk121>oh hi. The round trick only works when you have non-integers (eg 1.234) that your trying to convert to integers. It doesn't speed up 1.0, etc

<sneek>spk121, you have 1 message.

<sneek>spk121, daviid says: thanks! i was hoping to reduce the time still using scheme only, but thanks for the pasted code, let's talk (maybe) about this .. thanks again

<daviid>ah spk121 hello!

<daviid>spk121: how do i use what you posted yesterday from scheme?

<spk121>daviid: The whole story? lemme find a place to post it...

<daviid>spk121: what i dont get is SCM_I_MAKINUM is 'in' guile, so why can't i 'just bind it or call it?

<amz3>meh... my sat solver is buggy

<daviid>so i would use the guile ffi to bind a libguile function, how is that? :)

<spk121>daviid: maybe? SCM_I_MAKINUM is a gnarly macro, so it can't be bound w/o making it a function first

<daviid>spk121: I see, too bad

<daviid>but at least it exists

<spk121>daviid: but you might be able to bind scm_to_double and scm_from_int32

<spk121>daviid: and join them that way

<daviid>spk121: why scm_to_double? and i guess you mean scm_to_int32? I'msorry i'm very unknowledgeable of that side of guile

<daviid>my floats are single float (f32)

<spk121>daviid: because the C procedure scm_to_int32 will throw an exception when passed a non-integer SCM

<daviid>spk121: I read i the manual that scm_from_int32 expects an integer, s I don't see how to convert ...

<spk121>daviid: so SCM -> scm_to_double -> double -> cast to int -> scm_from_int32 -> SCM

IRC channel logs

2017-03-04.log