IRC channel logs

2016-09-03.log

back to list of logs

<paroneayea>annnnnnd
<paroneayea>rebooted, I still can't import guile-gnome
<paroneayea>so, that makes no sense
<paroneayea>the only things I can think of between it working, killing the process, and it not working, is the first time it compiled everything in the cache, but I wiped that
<daviid>paroneayea: what is the error?
<paroneayea>daviid: pretty straightforward, ERROR: no code for module (gnome gobject)
<paroneayea>I mean, I know what that error means in theory
<paroneayea>but why dit it stop working between REPL sessions?
<daviid>paroneayea: you have to ,use (gnome-2) first
<daviid>(use-modules (gnome-2))
<paroneayea>oh
<paroneayea>huh
<paroneayea>yeah that was it
<paroneayea>daviid: what's ,use (gnome-2) ?
<paroneayea>daviid: well I'm glad my system wasn't mutated in some insanely strange way!
<daviid>paroneayea: it sets %load-path so it finds guile-gnome modules
<paroneayea>ACTION doesn't remember doing it the first time
<paroneayea>but what do I know, I guess
<daviid>paroneayea: then you had a script and used '... exec guile-gnome-2 ... which also sets things 'up' to find modules...
<paroneayea>daviid: huh, maybe :)
<paroneayea>daviid: well thanks for the help!
<daviid>paroneayea: anyway, I recommend you #:use-module (gnome-2) in any of your guile-gnome modules ...
<paroneayea>guess I didn't need to restart :)
<daviid>welcome!
<daviid>sapientech: have to go afk for a bit. you should update the configure.ac as we've talked about... then I suggest you look at http://git.savannah.gnu.org/cgit/g-golf.git/tree/, you may copy the 'am' content [guile.mk] (some of us call this subdir 'mk', then look at Makefile.am and pre-inst-env.in, the meta subdir as well ... feel free to copy anything ...
<daviid>sapientech: note that in g-golf, I 'improved' guile.m4 a bit, because I wanted GUILE_GLOBAL_SITE_DIR and GUILE_SITE_CCACHE_DIR, you may copy that version as well if your are interested
<daviid>bbl
<sapientech>daviid: you pulled dev branch?
<amz3`>bad news again, another of my program is killed by oom killer, but this time there is no streams at all
<amz3`>I'll try using guile 2.0
<amz3`>and then fill a bug
<amz3`>in the meantime here is list of 500k urls extracted from HN: http://hyperdev.fr/data/hn/hn.urls.txt.xz (8 MB)
<amz3`>I should propably take this as an opportunity to install guix but I'm too lazy
<janneke>amz3`: what's the opportunity?
<amz3`>I need to install guile 2.0.12 to test a program
<janneke>ah
<amz3`>If I had installed guix, it would be easy to switch from 2.1.3 to 2.0.12
<amz3`>is that correct?
<amz3`>janneke: ^
<janneke>yes, it helps
<janneke>you don't have to sudo dpkg -i g2/*deb -- g2.2/*deb all the time
<janneke>if you're on debian, i have a guix package that can be easily built and installed an you run guix (the package manager) on top of debian first
<amz3`>that's the case, I run debian jessie (stable)
<amz3`>janneke: where is the guix package?
<amz3`>please
<janneke> https://gitlab.com/janneke/debian-guix
<janneke>amz3`: you still need to build the .deb; see the recipe in README
<amz3`>janneke: does it work on ubuntu?
<janneke>amz3`: yes
<amz3`>cool I'll do the same on my personal computer then
<amz3`>janneke: that's super cool, thx for your time
<janneke>Oh, it's courtesy Diane Trout -- I just cloned it and made some small fixes and updates that are not in Diane's repo -- guess they lost interest
<amz3`>janneke: sorry, but there is an issue when I run `uscan' I get the following message: uupdate: new version 0.10.0-1 <= current version 0.10.0-1; aborting!
<amz3`>I think there is an issue
<janneke>amz3`: hmm... you should get 0.11.0 because that's current
<amz3`>I must switch to develop in guix repo?
<janneke>i built it on ubuntu for 0.11.0, let me have a look
<janneke>amz3`: no, it builds from the release tarball
<amz3`>let me recap, I git clone the guix repo, then I git clone guix-debian as debian directory in guix repo and run uscan in guix repo, is that correct?
<janneke>amz3: no, just mkdir deb-guix; cd deb-guix; git clone debian-guix.git debian; (no CD-ing); uscan ... dpkg-build etc
<janneke>amz3`: i pushed an update for 0.11.0; please re-clone or hard reset to master, sorry ;-)
<janneke>you can clone guix later and have the guix binary from the deb package work with the latest guix repo
<amz3`>ok I'm trying that
<amz3`>ok
<amz3`>on debian stable, I got an error during dpkg-buildpackage regarding unmet guile-json dependency, guile-json is not in apt
<amz3`>I'm trying the same on ubuntu
<amz3`>on ubuntu 16.04.1 it's building...
<janneke>hmm, guile-json now how did I get around that
<amz3`>probably you installed it manually?
<amz3`>is it ok the send a 8MB to the bug tracker?
<amz3`>is it ok the send a 8MB *file* to the bug tracker?
<amz3`>test of guile 2.0.12 in progress
<rekado>Is there an SVG image creation library for Guile? I’m looking for abstractions to compose simple SVGs from within Guile.
<rekado>I know there are bindings to librsvg, but I think they are not for creating SVGs.
<rekado>now I’m just using ad-hoc functions to generate some sxml expressions
<amz3`>IIRC cairo as svg output
<amz3`>has
<rekado>amz3`: I’m looking for something where I can write (translate #:x 10 (circle x y r)) and it produces the corresponding SVG for a translated circle.
<amz3`>I updated the documentation regarding guile-wiredtiger with pointers to the official doc, it's available at https://amirouche.github.io/Culturia/doc/wiredtiger.html
<amz3`>I should make some benchmarks...
<amz3`>ACTION 's running gag
<wingo>hehe
<wingo>it would be great to rewrite the reader in scheme and compile to c, that way we could possibly create syntax objects instead of s-expressions just by modifying the compiler!
<amz3`>I pushed a new documentation regarding wiredtigerz, the high level API for interacting with wiredtiger
<amz3`> https://amirouche.github.io/Culturia/doc/wiredtigerz.html
<amz3`>The main addition to wiredtigerz is that you use a wiredtiger database from multiple thread seemlessly via the environment and call-with-cursor
<amz3`>also it has a mini-dsl to define a table with indices
<amz3`>does length procedure goes through all of the lst to compute the length?
<janneke>amz3`: guile/libguile/list.c:scm_ilength ()
<janneke>it's worse: it uses a tortoise and hare to make sure not to get stuck in circular lists
<otjura>why isn't guile -c (+ 1 2) working?
<amz3`>thx
<otjura>I want to eval sexps in bash and man says "Stop argument processing, and evaluate EXPR as a Scheme expression."
<otjura>with -c option
<jmd>otjura: Try guile -c '(display (+ 1 2))'
<otjura>aha thanks
<amz3`>preliminary benchmark, show that there is 1k assoc inserted per second, and 1k5 assoc read per second
<amz3`>where an assoc, is an alist with 10 symbol keys of length 30 associated with a string value of length 128
<amz3`>it seems honest to me
<amz3`>later I'll publish a graph over 10^7 assoc inserted
<amz3`>later
<amz3`>here is the graph http://i.imgur.com/Oa3TZiC.png
<amz3`>this is before the end of the benchmark
<amz3`>there is already 3.5G of assoc in the database
<amz3`>I expect better performance when wiredtiger is used like a RDBMS
<amz3`>now I should compare this at least to sqlite but I'm too lazy
<amz3`>also the insert time include the generation of the assocs
<amz3`>I mean the insert RPS value include the time required to randomly generate an assoc
<daviid>amz3`: when comparing to sqlite, you must put the write i a transaction because otherwise unfair :) (unless in a transaction, sqlite open and closes the db at each write ... )
<amz3`>ah yes, I did not use txn, I think I should try
<amz3`>I'll do that later
<amz3`>thx for the pointer daviid
<daviid>wc!
<djcb>is there function in the c-api to convert a sexp to a string?
<djcb>ie. the equivalent of (format #f "~S" sexp)
<wleslie>like scm_display sort of?
<random-nick>shouldn't it be scm_write or something?
***holomorp1 is now known as holomorph
<amz3`>so now I need to convert my boolean-keyword search engine to use a graphdb
<amz3`>to make it more *flexible*
<amz3`>my search engine will *flex*, it *flex* around doing fancy stuff
<amz3`>I have all the stuff around but I'm lazy, I'm looking for something else
<amz3`>I'm wondering whether I should restart git bindings or implement flexbox layout
<djcb>amz3`: would xapian fit the bill?
<amz3`>djcb: no!!!!!
<amz3`>:)
<amz3`>I am doing a research project, I want maximum flex, flex, flexibility
<djcb>you sound like a duran-duran song :-)
<amz3`>ahah, listenning to duran-duran right now
<amz3`>how to explain what I am doing?
<amz3`>It seems to me that doing search primary goal is given a way to search for things, you build to most performant data layout
<amz3`>given an algorithm, you build the best index
<amz3`>that's what xapian does for full text search and boolean keyword search
<amz3`>but I'd like to do more than boolean/keyword/fulltext search
<amz3`>like I'd like to resolve synonyms
<djcb>aha
<djcb>xapian does have support for that
<amz3`>really
<djcb>(though perhaps not as flex flex flexible as you want)
<amz3`>more generaly I want to create a playground for NLP software
<amz3`>I am experimenting with that goal using a search engine
<amz3`>also I want to do a lot of guile ;)
<djcb>ah, i'm combining a bit of guile and a bit of xapian
<djcb>xapian "does the job" and it does it pretty well
<djcb>but perhaps not give you all the flexibility you need
<amz3`>ah
<amz3`>djcb: what is it for?
<amz3`>djcb: does it support i18n?
<djcb>i'm using it for e-mail indexing (mu/mu4e/mu-guile)
<amz3`>ah
<djcb>the existing bindings are a bit... pedestrian... so i'm trying to design some new ones
<djcb>xapian uses std::string which I feed with utf8; so that part of i18n is there
<amz3`>I wonder whether the fulltext search engine customize stemming based on language
<amz3`>in principle it should take a language argument as input with the indexed text, because you can't reliably guess the language just with the text
<amz3`>it's doable tho
<amz3`>*with the text to index*
<djcb>thought about that, doing a little ngram/bayesian thing. xapian has some support influencing stemming
<djcb>but i don't think it is "document"-specific
<djcb>so perhaps you'd need a database per language (it's possible to combine databases for queries though)
<amz3`>I don't support i18n anyway
<paroneayea>djcb: you don't happen to know of any guile bindings for xapian, do you? :)
<paroneayea>though I guess iirc mu is itself C or C++, and connecting to guile
<paroneayea>(iirc xapian is C++ so the FFI couldn't be used)
<djcb>paroneayea: indeed, i don't happen to know...
<djcb>i do a little bit of c++ to wrap the parts of xapian i need; most of the rest of the code is in c, and guile talks to the c code using manual bindings
<djcb>would prefer to use e.g. guile-gir, but that project seems on hiatus
<paroneayea>djcb: xapian has GIR bindings?
<paroneayea>whoo boy wow it does!
<djcb>i think the endless folks did something?
<paroneayea>djcb: yeah you're right
<paroneayea> https://github.com/endlessm/xapian-glib
<djcb>now all we need is to have the guile-gir bindings finished
<djcb>that would open up quite a treasure chest of software
<amz3`>+1
<webshinra>yep, time is comming to get ride of C, or at least reserve it to really specific tasks
<stis>hey guilers!
<amz3`>héllo stis