Go Language Resources Go, golang, go... NOTE: This page ceased updating in October, 2012

--- Log opened Sat Nov 20 00:00:17 2010
00:05 -!- ShadowIce [~pyoro@unaffiliated/shadowice-x841044] has quit [Quit:
00:09 -!- kanru [~kanru@118-160-169-118.dynamic.hinet.net] has joined #go-nuts
00:11 -!- cco3 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has quit [Quit:
00:11 -!- cco31 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has joined #go-nuts
00:14 -!- belkiss [~belkiss@drn13-1-78-235-168-105.fbx.proxad.net] has quit [Quit:
KVIrc Insomnia 4.0.2, revision: 4740, sources date: 20100627, built on: 2010-10-19
12:51:39 UTC http://www.kvirc.net/]
00:15 -!- krutcha [~krutcha@remote.icron.com] has quit [Quit: Leaving]
00:16 -!- cco31 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has quit [Ping
timeout: 276 seconds]
00:25 -!- zozoR [~zozoR@5634798d.rev.stofanet.dk] has quit [Quit: Morten.  Desu~]
00:28 -!- iant [~iant@nat/google/x-npcilkzwsiuwszdo] has quit [Quit: Leaving.]
00:34 -!- drc_ [~drc@dh165049.res.uvic.ca] has joined #go-nuts
00:41 < drc_> Does anyone know of a gcov like tool for statement coverage of
a go program?
00:41 -!- iant [~iant@nat/google/x-vpuoaruixyrdmstl] has joined #go-nuts
00:42 -!- mode/#go-nuts [+v iant] by ChanServ
00:43 -!- artefon [~thiago@] has quit [Quit: bye]
00:47 -!- iant [~iant@nat/google/x-vpuoaruixyrdmstl] has quit [Ping timeout: 245
00:59 -!- drc_ [~drc@dh165049.res.uvic.ca] has left #go-nuts []
01:05 -!- Tv [~tv@gige.bur.digisynd.com] has quit [Ping timeout: 245 seconds]
01:09 -!- sauerbraten_ [~sauerbrat@p508CEEF3.dip.t-dialin.net] has quit [Remote
host closed the connection]
01:28 -!- adu [~ajr@pool-173-66-11-168.washdc.fios.verizon.net] has joined
01:30 -!- kanru [~kanru@118-160-169-118.dynamic.hinet.net] has quit [Read error:
Operation timed out]
01:31 -!- iant [~iant@] has joined #go-nuts
01:31 -!- mode/#go-nuts [+v iant] by ChanServ
01:39 -!- devrim [~Adium@] has quit [Quit: Leaving.]
01:46 -!- Tv [~tv@cpe-76-168-227-45.socal.res.rr.com] has joined #go-nuts
02:08 -!- niemeyer [~niemeyer@189-10-175-46.pltce701.dsl.brasiltelecom.net.br] has
quit [Remote host closed the connection]
02:15 -!- gummibaerchen [~timm@archlinux/trusteduser/gummibaerchen] has quit [Ping
timeout: 240 seconds]
02:24 -!- bjarneh [~bjarneh@232.80-203-20.nextgentel.com] has joined #go-nuts
02:25 -!- ExtraSpice [~XtraSpice@] has quit [Ping timeout: 240
02:28 -!- devrim [~Adium@cpe-72-225-239-227.nyc.res.rr.com] has joined #go-nuts
02:34 < |Craig|> whats the preferred method for converting from numeric
types to bytes?  In my specific case I need to do uint32s.  I can compute each
byte manually, but that seems inefficient as what I want already exists in memory.
02:40 -!- SoniaKeys [~soniakeys@c-76-118-178-209.hsd1.ma.comcast.net] has joined
02:49 -!- ymasory [~ymasory@c-76-99-55-224.hsd1.pa.comcast.net] has joined
02:54 -!- tensorpudding [~user@] has quit [Remote host closed the
03:07 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has joined
03:09 -!- Ccod [~peter@c-76-18-5-93.hsd1.fl.comcast.net] has joined #go-nuts
03:10 < skelterjohn> |Craig|: you can use unsafe
03:10 -!- matti___ [~mumboww@c-98-207-108-218.hsd1.ca.comcast.net] has quit [Ping
timeout: 250 seconds]
03:11 < skelterjohn> i think it goes like
03:11 < skelterjohn> err
03:12 < skelterjohn> [8]byte
03:12 < skelterjohn> not 64
03:12 < |Craig|> skelterjohn: I figured it might require that.  Most
languages I use have buffers with write methods for different types, I guess I
could make one
03:13 < Ccod> Im on the last leg of installation, having a problem getting
my $PATH to work.  (need that compiler command)
03:13 < Ccod> hello by the way
03:13 < skelterjohn> hi Ccod
03:13 -!- iant [~iant@] has quit [Quit: Leaving.]
03:14 -!- Eridius [~kevin@unaffiliated/eridius] has quit [Ping timeout: 245
03:14 < Ccod> is it cool to ask this kind of thing here??
03:14 < skelterjohn> you haven't asked anything :)
03:14 -!- tensorpudding [~user@] has joined #go-nuts
03:14 < Ccod> oh, well What is my PATH suppose to look like
03:15 < Ccod> its in .bash_profile right?
03:15 < skelterjohn> what happens when you >ls $GOBIN
03:15 < skelterjohn> i put paths in .bashrc, but maybe i'm a weirdo
03:15 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has joined #go-nuts
03:16 < Ccod> nothing useful I think
03:16 < Ccod> just shows me the same ls list
03:17 < skelterjohn> so, GOBIN is not defined
03:17 < skelterjohn> is GOROOT defined?
03:17 -!- tensorpudding [~user@] has quit [Remote host closed the
03:17 -!- matti___ [~mumboww@c-98-207-108-218.hsd1.ca.comcast.net] has joined
03:18 < Ccod> ">" is suppose to hold a bit of significance right?  unless
i take it away I get nothing
03:18 < Ccod> in both cases
03:18 < skelterjohn> i just used that to indicate you should type something
on the command line
03:18 < Ccod> but GOROOT is not likely defined
03:18 < Ccod> right
03:18 < skelterjohn> to get a go installation working you need to define a
few environmental variables
03:19 < skelterjohn> GOROOT and GOBIN
03:19 < skelterjohn> GOROOT should be wherever you downloaded the go
directory to
03:19 < skelterjohn> GOBIN can be $GOROOT/bin, for now
03:20 < |Craig|> Ccod: All I have for go is the go/bin on my path
03:20 < Ccod> hmm
03:20 < Ccod> one sec
03:20 < |Craig|> I think the binaries can figure out the rest unles you want
to move files from the default locations
03:21 < |Craig|> my $GOROOT is not set
03:21 < skelterjohn> weird
03:21 < Ccod> hmm
03:21 < Ccod> well then
03:21 < skelterjohn> didn't know you could get along without setting goroot
03:21 < skelterjohn> i guess the go binaries will look at ..  if nothing is
03:21 < Ccod> how does your path statement look...  cause I probably have it
wrong then
03:22 < skelterjohn> export PATH=$PATH:$GOROOT/bin
03:22 < skelterjohn> that statement is in my .bashrc
03:22 < |Craig|> PATH=$PATH:/Developer/go/bin in my /bash_progile
03:22 < Ccod> in $HOME directory right?
03:22 < skelterjohn> what do you mean "in $HOME directory"
03:23 < Ccod> where you created the .bashrc file
03:23 < skelterjohn> oh yes
03:23 < skelterjohn> that is in my home directory
03:23 < Ccod> right
03:23 < skelterjohn> that only makes sense if you are running bash
03:23 < Ccod> just making sure
03:23 < skelterjohn> > echo $SHELL
03:23 < skelterjohn> to find out
03:23 < |Craig|> Ccod: What os are you on btw?  (Mac here)
03:23 < Ccod> same
03:28 < Ccod> k..  just to be sure.  To compile you type 6g thing.go in the
command line right?  its not in the file or some other method.
03:28 < skelterjohn> yes
03:28 < Ccod> well..  im going to go make a bashrc and see if that does
03:29 < skelterjohn> did you echo $SHELL ?
03:29 < |Craig|> then 6l thing.6 ; ./6.out
03:29 < Ccod> yeah, says bin/bash
03:29 < |Craig|> you mean /bin/bash ?
03:30 < skelterjohn> then making a .bashrc will do what you need
03:30 < Ccod> yes that
03:30 < Ccod> k
03:30 < skelterjohn> though i'm certain it already exists
03:30 < Ccod> probably
03:30 < |Craig|> skelterjohn: at least on mac .bash_profile works
03:30 < |Craig|> what these 2 files do on mac is slightly different than
linux I believe
03:30 < |Craig|> (I think both would work)
03:31 < skelterjohn> oh - on my mac i have no .bashrc it's true
03:31 < skelterjohn> i use .profile
03:31 < |Craig|> I've never used /bashrc
03:31 < skelterjohn> on linux it works
03:32 < Ccod> well, made the file and put PATH=$PATH/me/go/bin and no dice
03:33 < skelterjohn> you have to open a new terminal now
03:33 < skelterjohn> loading those files only happens when the term opens
03:33 < skelterjohn> or if you > source .bashrc
03:33 < Ccod> oh
03:33 < skelterjohn> but like |Craig| implied, .bashrc might be wrong for
03:33 < skelterjohn> if that still doesn't work, try .profile
03:33 < skelterjohn> that definitely works for me on my mac
03:35 < Ccod> just checking to rules on adding to paths that are already
there.  Just going to a new line is fine here right?
03:35 < |Craig|> CcoD did you export PATH?
03:35 < skelterjohn> you mean for editing your .profile?
03:35 < skelterjohn> just add a line at the bottom
03:35 < Ccod> yeah, opened it and see something there
03:35 < skelterjohn> and yeah, you need to export it
03:35 < Ccod> k
03:35 < skelterjohn> export PATH=$PATH:whereYourGoStuffIs
03:36 < Ccod> k
03:39 -!- tensorpudding [~user@] has joined #go-nuts
03:41 < Ccod> what to do now...
03:41 < skelterjohn> what happens when you > 6g
03:41 < Ccod> bash 6g: command not found
03:41 < skelterjohn> > echo $PATH
03:42 < Ccod> its there in the end...  among a sea of other stuff
03:43 < Ccod> the*
03:43 < skelterjohn> what's in the directory?
03:43 < Ccod> :peter/go/bin
03:44 < skelterjohn> that isn't the directory
03:44 < skelterjohn> it should be /Users/peter/go/bin
03:44 < skelterjohn> or something like taht
03:44 < skelterjohn> it's a global path
03:44 < Ccod> ah
03:44 < Ccod> I'll try that
03:46 < skelterjohn> if you cd into the bin directory and do > pwd
03:46 < skelterjohn> it will tell you exactly the right directory for $PATH
03:46 < skelterjohn> pwd = print working directory
03:46 -!- aho [~nya@fuld-4d00d1f9.pool.mediaWays.net] has quit [Quit:
03:46 < Ccod> k
03:49 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has quit
[Quit: Bye!]
03:50 < Ccod> yeah, still not happening even after checking with pwd
03:51 < Ccod> that 6g thing is not workin
03:53 < skelterjohn> what did it print when you said pwd
03:54 < |Craig|> Ccod: for a test, you can just use the full path to 6g
03:55 < |Craig|> Ccod: you did remember to relaunch your terminal after
editing your .bash_profile right?
03:55 < Ccod> yea, I echoed $PATH just to make sure it was there
03:56 < Ccod> so in the directory of the .go file stick the path and 6g in
03:57 < Ccod> is that what you mean by the test?
03:57 < skelterjohn> Ccod, tell me what it prints when you go to the go bin
directory and do > cwd
03:58 < skelterjohn> we don't have to do any half-measures here
03:59 -!- jochang [~identd@] has joined #go-nuts
03:59 < |Craig|> Ccod: you can run 6g using the full path to it (and thus
not relying on $PATH) try it
03:59 < Ccod> cwd: command not found.  is that a shorthand?
03:59 < skelterjohn> !
03:59 < |Craig|> pwd
03:59 < skelterjohn> oops
03:59 < skelterjohn> pwd
03:59 < skelterjohn> sorry
03:59 < Ccod> lol
03:59 < Ccod> yeah
03:59 < Ccod> I did that one already
03:59 < |Craig|> its cwd on windows isn't it?
04:00 < skelterjohn> and what did it print
04:00 < Ccod> Users/peter/go/bin
04:00 < skelterjohn> beginning with a slash?
04:00 < Ccod> yes
04:00 < |Craig|> the slash is important
04:00 < skelterjohn> and /Users/peter/go/bin is on the end of > echo
04:01 < skelterjohn> hmm.  is there anything in this directory?
04:01 < skelterjohn> :)
04:01 < skelterjohn> have you built go?
04:01 < Ccod> yes it is all there
04:01 < Ccod> one sec
04:01 < Ccod> gonna try something
04:01 < |Craig|> /Users/peter/go/bin/6g should run it
04:02 < Ccod> got it
04:02 < Ccod> !
04:02 < skelterjohn> what did you try?
04:02 < Ccod> it was that / you pointed out
04:03 < skelterjohn> ah
04:03 < skelterjohn> yeah, it matters
04:03 < Ccod> in front of Users
04:03 < Ccod> yeah...  now it says non-declaration statement outside
function body :P...  which means I know it at least tried
04:04 -!- krutcha [~krutcha@S010600045a27676a.vs.shawcable.net] has joined
04:04 < skelterjohn> now it's just a compile error
04:04 < skelterjohn> that kind of thing is much easier to deal with
04:04 < Ccod> yep
04:04 < Ccod> well i certainly learned a few lessons over that last couple
hours or so
04:04 < Ccod> thanks alot
04:05 < Ccod> I think I will be better prepared next time I need to se up an
enviorment :P
04:05 < skelterjohn> it's a learned skill
04:05 < Ccod> set*
04:05 < skelterjohn> and no problem at all
04:06 < Ccod> now I can finally start playing with this thing :P...  so what
have you guys been doing with it?
04:07 < skelterjohn> i write experimentation software using go
04:07 < skelterjohn> i'm a phd student doing machine learning stuff
04:07 < |Craig|> I've been writing a MMOG server
04:08 < Ccod> 0.0 awesome
04:08 < Ccod> didnt think there was an emoticon for that one
04:08 < skelterjohn> the lack of a nice debugger and go-oriented IDE is a
bit annoying
04:08 < skelterjohn> but overall i still write code faster because it's go
code that i'm writing
04:09 < skelterjohn> it's very concise - you get to the interesting bits
right away
04:09 < skelterjohn> and don't spend hours designing inheritance trees
04:09 < skelterjohn> to pick an example completely at random
04:09 < |Craig|> I can't say I develop go code faster than python, writing
go code is faster than writing cython code that will run as fast to use with
04:09 < Ccod> I can't wait
04:10 < skelterjohn> oh - python is also very fast to write
04:10 < skelterjohn> but i find that python loses readability over time
04:10 < skelterjohn> the language that takes longer to write for me is java
or C++
04:11 < |Craig|> when I have better tools for go, I think ill pass my python
coding speed for it
04:11 < skelterjohn> both involve large amounts of boilerplating
04:12 < Ccod> what do you use for writing?  TextMate?
04:12 < |Craig|> I use TextWrangler
04:12 < |Craig|> I tried Goclipse, but its clearly not ready for use
04:12 < Ccod> I have been forcing myself to use vim lately since I got the
syntax highlighter to work there.
04:12 < skelterjohn> i use xcode on my mac
04:13 < skelterjohn> and emacs on linux
04:13 < Ccod> ah
04:13 < |Craig|> aparently there is a pretty sweet vim plugin for
autocompleation with go, but I'm not a vim guy
04:13 < Ccod> oh
04:13 < Ccod> I need to go find it then
04:13 < skelterjohn> if you're interested in that, nsf (in this chatroom) is
the guy who wrote it
04:13 < skelterjohn> and he might be happy to help you set it up
04:13 < skelterjohn> but he is a notorious curmudgeon, so it varies
04:13 < |Craig|> video: http://nsf.110mb.com/gocode-demo.swf
04:14 < Ccod> not familiar with curmudgeon
04:14 < Ccod> ??
04:14 < skelterjohn> google ;)
04:14 < |Craig|> Ccod: put you mouse over the word and hit command control d
04:15 < skelterjohn> uh - what is that suppose to do, craig?
04:15 < skelterjohn> perhaps you have a different irc client
04:15 < |Craig|> pops up a defination
04:15 < Ccod> haha nice
04:15 < skelterjohn> doesn't do anything for me
04:15 < |Craig|> its a mac thing, not in the client
04:15 < skelterjohn> oh - i thought it was just ctrl-d
04:15 < skelterjohn> neat!
04:15 < |Craig|> Ccod: another mac feature, in text edit, the esc key is
auto compelate
04:16 < Ccod> heh
04:16 < |Craig|> I was amazed when I discovered that
04:16 < |Craig|> no good for coding though
04:16 < skelterjohn> hah that'd be neat
04:17 < skelterjohn> if it brought up the declaration of whatever you
04:17 < Ccod> hmm, the autocomplete thing is for compiled variable names
04:17 < skelterjohn> mac's intrinsic autocompletion picks out other things
you've typed recently that begin the same
04:17 < Ccod> ah
04:17 < skelterjohn> nothing to do with the structure of your code
04:17 < Ccod> hmm...  sounds like it could get annoying
04:18 < skelterjohn> it is annoying, for code
04:18 < Ccod> getting the wrong word
04:18 < |Craig|> Ive never used it
04:18 < skelterjohn> one of the things i don't like about xcode
04:18 < Ccod> I see
04:18 < skelterjohn> it pops up wrong autocomplete bubbles all the time
04:18 < Ccod> the vim thing is the same thing?
04:18 < skelterjohn> and god forbid i try to do a newline then
04:18 < skelterjohn> instead of giving me a new one, it inserts this wrong
04:18 < skelterjohn> the vim thing analyzes the go source
04:18 < Ccod> ah
04:18 < |Craig|> that vim thing is based on gocode and knows all about the
types and such
04:18 < skelterjohn> and also knows about the libraries
04:19 < Ccod> sounds far more useful
04:19 < Ccod> :P
04:19 < |Craig|> yes
04:20 -!- trophi [~phi@pool-71-178-14-3.washdc.fios.verizon.net] has quit [Ping
timeout: 245 seconds]
04:27 < Ccod> how long have you been using go?
04:27 < Ccod> since your making an MMOG server I guess you are quite
04:29 < skelterjohn> i started taking a look at it when i saw the slashdot
article last november
04:32 -!- rejb [~rejb@unaffiliated/rejb] has quit [Ping timeout: 276 seconds]
04:33 < skelterjohn> gnight
04:33 -!- skelterjohn [~jasmuth@lawn-gw.rutgers.edu] has quit [Quit: skelterjohn]
04:45 -!- jochang [~identd@] has quit [Ping timeout: 264 seconds]
04:45 -!- Stiletto [7f000001@] has quit [Ping timeout: 276 seconds]
04:49 < |Craig|> Ccod: This is my second project in go.  I made a simple
fractal generator to learn go
04:49 < |Craig|> I have a bit of a history in writing partially finished mmo
04:49 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has joined
04:51 < |Craig|> I do a lot of stuff with Panda3D and python.  Its a great
game engine, and has been used for commercial mmogs, but networking in it is
poorly documented, and is not the high performance stuff that Disney wrote for
their mmo server.  I decided to make a decent server for it in go because it
looked fun
04:51 < Ccod> well thats better than me though...  I hardly have any
experience with that kind of stuff
04:53 < |Craig|> go just looked so fun, and a server looked like the best
thing to write with it, and the only servers I know anything about are game
servers, so thats what I'm writing
04:53 < Ccod> I can't decide what kind of project to start on :P
04:54 < Ccod> well I have no idea how someone goes about doing something
like that...  definitely interested though
04:55 < Ccod> wonder where to look more into that sort of thing, my club has
been doing XNA stuff lately
04:57 < |Craig|> well, you can look at my source:
https://github.com/Craig-Macomber/Grains--Vegetarian-Zombie-Rising (I started it
this week, its not like I've gotten very far)
05:04 < Ccod> nice
05:04 < Ccod> will do
05:10 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has quit [Ping
timeout: 240 seconds]
05:12 -!- BlaSux [7f000001@] has joined #go-nuts
05:21 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has quit [Remote host
closed the connection]
05:22 < |Craig|> ha, I forgot to commit my most significant go file, fixed
05:25 -!- Adys [~Adys@unaffiliated/adys] has quit [Quit: I ♥ Unicode]
05:25 -!- Adys [~Adys@unaffiliated/adys] has joined #go-nuts
05:27 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has joined #go-nuts
05:28 < Ccod> got it
05:28 < Ccod> :P
05:28 -!- kanru [~kanru@61-228-153-104.dynamic.hinet.net] has joined #go-nuts
05:29 < |Craig|> I just DOS attacked my server.  10000 connections was a tad
overkill, though the server was fine (as expected), but there was no way my python
client running on the same machine could deal with it
05:30 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has joined
05:30 < Ccod> sounds like we have a winner :P
05:31 < |Craig|> it even fails with 20, though with the server spamming at
max throughput, I guess thats expected
05:31 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has quit [Remote host
closed the connection]
05:31 < |Craig|> go can write to tcp faster than poorly optimized python can
read it, not to mention that the server compresses it once and the clients each do
it seperatly
05:32 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has joined #go-nuts
05:32 < Ccod> I'm even more impressed with go from that 00
05:33 < |Craig|> I think the client died befor it connected more than 100 or
05:33 -!- devrim [~Adium@cpe-72-225-239-227.nyc.res.rr.com] has quit [Ping
timeout: 255 seconds]
05:34 < |Craig|> I was spamming over 1000 compressed messages a second.  at
100 connections, that would have been running 100000 decompressions in python per
second, and it just failed to keep up
05:34 < Ccod> so your going to pull your python code for go?
05:35 < |Craig|> nah, in real use case terms, that would have been 10000
user's computers, they would have been able to keep up
05:36 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has quit
[Quit: skelterjohn]
05:36 < |Craig|> rather than just my laptop with the server already eating
all the cpu
05:36 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has joined
05:36 < Ccod> ah
05:36 < Ccod> I see
05:37 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has quit
[Client Quit]
05:37 < |Craig|> Im impressed that with things slowed down a bit python
(using twisted) can track 200 pretty high traffic connections ok
05:38 < |Craig|> and through all this abuse, my little go server that I
wrote today didn't crash or lock up once
05:39 < Ccod> very impressive
05:39 -!- matti___ [~mumboww@c-98-207-108-218.hsd1.ca.comcast.net] has quit [Ping
timeout: 276 seconds]
05:39 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has joined
05:40 < Ccod> well, sounds it at least...  I don't have much experience in
that field
05:41 < |Craig|> 100 connection requests piled on it all at once, all got
through without dropping.  I guess the TCP server thingy in the net package is
pretty nice
05:42 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has joined
05:43 < Ccod> have you seen the google cast thing on go?
05:43 < |Craig|> I saw the inital launch video
05:44 < |Craig|> thats about it though
05:45 < Ccod> yeah, I think thats it.  There was this thing when they typed
./eval1 ...  which i thought was like a go console
05:45 < Ccod> or something
05:45 < Ccod> ...  is there something like that somewhere?
05:45 < |Craig|> 247 is the magic number.  No matter how much I slow things
down, I can't get more connections than that.
05:45 -!- matti___ [~mumboww@c-98-207-108-218.hsd1.ca.comcast.net] has joined
05:46 < |Craig|> Ccod: there is the go playground on the golang site
05:49 -!- fabled [~fabled@] has joined #go-nuts
05:49 -!- iant [~iant@adsl-71-133-8-30.dsl.pltn13.pacbell.net] has joined #go-nuts
05:49 -!- mode/#go-nuts [+v iant] by ChanServ
05:50 < Ccod> hmm, not really what I was looking for.  I wanted a kind of
trace on the execution of the code...
05:50 -!- devrim [~Adium@cpe-72-225-239-227.nyc.res.rr.com] has joined #go-nuts
05:52 < krutcha> trying to wrap my head around pkg/xml, anybody played with
it much?
05:52 < Ccod> not yet
05:53 < |Craig|> I finally made my server die, some "too many open files"
05:53 < |Craig|> well, it locked up until I took the load off.  It didn't
05:57 < Ccod> heh
05:58 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has quit
[Quit: skelterjohn]
05:58 < Ccod> later
05:58 -!- Ccod [~peter@c-76-18-5-93.hsd1.fl.comcast.net] has left #go-nuts []
05:59 < |Craig|> oh, its the mac default open files limit for a process,
makes sense.  I should be able to get around that
06:22 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has quit [Ping
timeout: 252 seconds]
06:46 < krutcha> I don't suppose mac/bsd have ulimit?
06:46 < krutcha> on linux ulimit -n shows that number
06:49 -!- mssm [~mssm@ip-95-221-86-196.bb.netbynet.ru] has joined #go-nuts
06:52 < |Craig|> krutcha: I figured it out
06:52 < |Craig|> ulimit just says unlimited, but there is a per process max
default I avoided by adding a config file
06:58 < krutcha> nice
06:58 < krutcha> I'm still lost in XML rfc's and vomiting at the size of
some of the streams I'm seeing
06:59 -!- ExtraSpice [~XtraSpice@] has joined #go-nuts
07:01 -!- mssm [~mssm@ip-95-221-86-196.bb.netbynet.ru] has quit [Ping timeout: 245
07:05 -!- noktoborus [~noxless@gateway/tor-sasl/noktoborus] has quit [Quit:
07:06 -!- Chopinn [~Chopin@ti0018a380-dhcp1590.bb.online.no] has quit [Quit:
07:13 -!- pingveno [~pingveno@c-98-246-133-8.hsd1.or.comcast.net] has quit [Quit:
07:14 -!- mssm [~mssm@ip-95-221-86-196.bb.netbynet.ru] has joined #go-nuts
07:14 -!- pingveno [~pingveno@c-98-246-133-8.hsd1.or.comcast.net] has joined
07:16 < krutcha> is it possible to directly update a struct's data in a map
or must one copy out/update/clobber?
07:17 < SoniaKeys> i would expect it to be possible
07:17 < SoniaKeys> but it's easy enough to try a little experiment
07:24 < SoniaKeys> well, it's no problem if you store a pointer to the
struct in the map.
07:26 < krutcha> yeah, I'm not going to go there in this case.  it's easy
enough to copy out/in for my needs.
07:26 -!- nsf [~nsf@jiss.convex.ru] has joined #go-nuts
07:31 -!- ExtraSpice [~XtraSpice@] has quit [Quit: Leaving]
07:33 < krutcha> hm when removing a struct from a map, do you have to pass
in a valid struct value?
07:33 < krutcha> nil and _ don't seem to work
07:34 < |Craig|> krutcha: nil only works for pointers
07:34 -!- ShadowIce [~pyoro@unaffiliated/shadowice-x841044] has joined #go-nuts
07:34 < krutcha> effective go shows the example: timeZone["PDT"] = 0, false
07:35 < krutcha> but my values are structs, it seems paintful to have to go
{"","","",""}, false
07:35 < |Craig|> krutcha: you can use the struct's 0 value
07:36 < |Craig|> new(typeName)
07:36 -!- ucasano [~ucasano@host153-182-static.227-95-b.business.telecomitalia.it]
has joined #go-nuts
07:36 < |Craig|> also, if you specify a name for one of the values in the
composite literal, you can leave out the others
07:37 < krutcha> but new would allocate an empty struct on the heap just to
delete it and the one in the map
07:37 < krutcha> wouldn't it?
07:38 < |Craig|> no matter what you do, you are going to allocate something
right?  (Not the best design in my opinion)
07:39 < krutcha> in this rare case I'd prefer a _ on the right side.  Or
maybe an interface function like func (*this map) delete (key)
07:43 < SoniaKeys> the compiler might optimize away the pointless
allocations.  still, i agree it seems silly to have to supply a value that will be
07:44 < krutcha> yeah it's a bit wierd, I found a thread on it here:
07:44 < krutcha> I don't see any examples in there where the value is a big
honkin struct tho
07:44 < Archwyrm> timeZone["PDT"] = _ // would seem rather unambiguous, IMHO
07:45 < krutcha> it makes the m[k] = x,false quickly look a lot less clean
07:46 < |Craig|> krutcha: if someone had a big struct, they would use a
07:46 < krutcha> I guess in go I could use pointers there, as the GC would
save me
07:46 < Archwyrm> Hence I have mostly been using 'nil, false' which isn't
too bad.
07:46 < |Craig|> its not like maps already screw the cache, and the pointer
dereferance won't kill you
07:47 < |Craig|> *I mean it is like maps already screw the cache
07:47 < krutcha> alright references it is
07:47 < krutcha> easy change in go :)
07:47 < krutcha> no .'s to change to ->'s
07:48 < Archwyrm> ;)
07:48 < SoniaKeys> i've appreciated that a few times
07:49 < krutcha> though it does sort of hose my initial insertion loop
07:49 < Archwyrm> The syntactical change in C/C++ was to remind you to clean
up after yourself.  ;)
07:49 < krutcha> I had map[k] = tempBigStruct in a loop of updates
07:49 < krutcha> I surely don't want to pass in 1000 references to the same
every changing temp struct
07:50 < krutcha> time to bust out new(
07:50 < |Craig|> just use &{blah}
07:50 < |Craig|> or new if you want 0 values
07:50 -!- adu [~ajr@pool-173-66-11-168.washdc.fios.verizon.net] has quit [Quit:
07:50 -!- noktoborus_ [~noxless@gateway/tor-sasl/noktoborus] has joined #go-nuts
07:51 < krutcha> can you call functions in that paramater list?
07:51 < |Craig|> I don't see why not
07:51 < krutcha> to populate from returned data {x(),y(),z}?
07:51 < krutcha> good
07:58 -!- cco3 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has joined #go-nuts
08:06 -!- noktoborus_ [~noxless@gateway/tor-sasl/noktoborus] has quit [Remote host
closed the connection]
08:08 -!- SoniaKeys [~soniakeys@c-76-118-178-209.hsd1.ma.comcast.net] has quit []
08:11 -!- |Craig| [~|Craig|@panda3d/entropy] has quit [Quit: |Craig|]
08:12 -!- Tv [~tv@cpe-76-168-227-45.socal.res.rr.com] has quit [Ping timeout: 276
08:12 -!- noktoborus_ [~noxless@gateway/tor-sasl/noktoborus] has joined #go-nuts
08:14 -!- cco3 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has quit [Ping
timeout: 245 seconds]
08:26 -!- bjarneh [~bjarneh@232.80-203-20.nextgentel.com] has quit [Quit: leaving]
08:43 -!- Project_2501 [~Marvin@] has joined #go-nuts
08:49 -!- pingveno [~pingveno@c-98-246-133-8.hsd1.or.comcast.net] has quit [Quit:
Lost terminal]
09:09 -!- photron_ [~photron@port-92-201-176-25.dynamic.qsc.de] has joined
09:22 -!- rlab [~Miranda@] has joined #go-nuts
09:25 -!- awidegreen [~quassel@p5DF1D7D4.dip.t-dialin.net] has joined #go-nuts
09:31 -!- krutcha [~krutcha@S010600045a27676a.vs.shawcable.net] has quit [Quit:
09:32 -!- skejoe [~skejoe@] has joined #go-nuts
09:40 -!- zozoR [~zozoR@5634798d.rev.stofanet.dk] has joined #go-nuts
09:44 -!- fabled [~fabled@] has quit [Read error: Operation timed
09:46 -!- tsykoduk [~tsykoduk@] has quit [Ping timeout: 276 seconds]
09:48 -!- tsykoduk [~tsykoduk@] has joined #go-nuts
09:53 -!- jo__ [3a72cb23@gateway/web/freenode/ip.] has joined
09:54 -!- belkiss [~belkiss@drn13-1-78-235-168-105.fbx.proxad.net] has joined
09:59 -!- fabled [~fabled@] has joined #go-nuts
10:01 -!- terrex [~terrex@] has joined #go-nuts
10:03 -!- araujo [~araujo@gentoo/developer/araujo] has quit [Quit: Leaving]
10:07 -!- virtualsue [~chatzilla@nat/cisco/x-lloixnpwewaklifc] has quit [Ping
timeout: 240 seconds]
10:10 -!- belkiss [~belkiss@drn13-1-78-235-168-105.fbx.proxad.net] has quit [Quit:
KVIrc Insomnia 4.0.2, revision: 4740, sources date: 20100627, built on: 2010-10-19
12:51:39 UTC http://www.kvirc.net/]
10:21 -!- femtoo [~femto@95-89-197-196-dynip.superkabel.de] has joined #go-nuts
10:28 -!- Project_2501 [~Marvin@] has quit [Quit: E se abbasso questa
leva che succ...]
10:36 -!- tensorpudding [~user@] has quit [Read error: Connection
reset by peer]
10:36 -!- devrim [~Adium@cpe-72-225-239-227.nyc.res.rr.com] has quit [Ping
timeout: 245 seconds]
11:08 -!- Fish [~Fish@bus77-2-82-244-150-190.fbx.proxad.net] has joined #go-nuts
11:10 -!- terrex [~terrex@] has quit [Quit:
11:12 -!- zozoR2 [~zozoR@563461e3.rev.stofanet.dk] has joined #go-nuts
11:15 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has joined #go-nuts
11:21 -!- virtualsue [~chatzilla@host81-148-100-129.in-addr.btopenworld.com] has
joined #go-nuts
11:23 -!- wrtp [~rog@] has joined #go-nuts
11:26 -!- virtualsue [~chatzilla@host81-148-100-129.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
11:28 < nsf> 8 files changed, 666 insertions(+), 465 deletions(-)
11:28 < nsf> omg
11:28 < nsf> what a patch
11:29 -!- skejoe [~skejoe@] has quit [Quit: Lost terminal]
11:31 < nsf> I've just added arrow keys scrolling (per item) to alt go docs
11:34 -!- virtualsue [~chatzilla@host81-139-165-122.in-addr.btopenworld.com] has
joined #go-nuts
11:35 -!- wrtp [~rog@] has quit [Quit: wrtp]
11:35 -!- zozoR2 [~zozoR@563461e3.rev.stofanet.dk] has quit [Quit: Morten.  Desu~]
11:53 -!- virtualsue [~chatzilla@host81-139-165-122.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
11:54 -!- devrim [~Adium@cpe-72-225-239-227.nyc.res.rr.com] has joined #go-nuts
11:57 -!- ucasano [~ucasano@host153-182-static.227-95-b.business.telecomitalia.it]
has quit [Quit: ucasano]
11:59 -!- virtualsue [~chatzilla@host81-139-66-80.in-addr.btopenworld.com] has
joined #go-nuts
12:16 -!- wrtp [~rog@] has joined #go-nuts
12:24 -!- artefon [~thiago@] has joined #go-nuts
12:25 -!- lmoura_ [~lauromour@] has quit [Ping timeout: 245
12:27 -!- virtualsue_ [~chatzilla@nat/cisco/x-hstxevtxcouglhhg] has joined
12:47 -!- fabled [~fabled@] has quit [Quit: Ex-Chat]
12:49 -!- araujo [~araujo@] has joined #go-nuts
12:49 -!- araujo [~araujo@] has quit [Changing host]
12:49 -!- araujo [~araujo@gentoo/developer/araujo] has joined #go-nuts
12:52 -!- rejb [~rejb@unaffiliated/rejb] has joined #go-nuts
13:03 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has joined
13:12 < hokapoka> Is there any means to load structs from others packages
13:17 < hokapoka> For example, if I had a package that would say serialize
structs to a file.  Would it be possible to build another, seperate, package that
exported some structs and some interface that I could interact with for the
serialization package, without having to import & re-compile it?
13:17 < hokapoka> s/for/from
13:17 < taruti> like json?
13:19 < hokapoka> The serialization was just an example, I'd like to beable
to use the reflect package that would be imported to & use the structs.
13:19 < hokapoka> s/&//
13:21 -!- virtualsue [~chatzilla@host81-139-66-80.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
13:21 < hokapoka> A bit like plugins, so that I could have a standard
package that would load structs and call exported functions within the other
package, assuming the other package Exposed some standard interfaces.
13:21 < taruti> there is no dynamic loading.
13:21 -!- virtualsue_ [~chatzilla@nat/cisco/x-hstxevtxcouglhhg] has quit [Ping
timeout: 255 seconds]
13:22 < hokapoka> okay, so if I wanted todo this I would need to writen
another package that would generate the imports and re-complie the main package.
13:22 -!- sauerbraten [~sauerbrat@p508CC67A.dip.t-dialin.net] has joined #go-nuts
13:24 < nsf> skelterjohn: I've added your feature request to my go docs
13:24 < nsf> skelterjohn: I hope your notebook has arrow keys :D
13:24 < hokapoka> I thought that was the case, and as go builds so fast it's
no a big deal.  In theory it would be quite simple to achive via some standard
approach.  Just wanted to be sure that I couldn't load other packages dynamically.
13:27 -!- virtualsue [~chatzilla@host81-148-115-114.in-addr.btopenworld.com] has
joined #go-nuts
13:29 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.pa.comcast.net] has quit
[Quit: skelterjohn]
13:32 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has joined
13:35 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has quit
[Client Quit]
13:39 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has joined
13:43 -!- skelterjohn [~jasmuth@c-76-98-219-171.hsd1.nj.comcast.net] has quit
[Client Quit]
13:48 -!- wrtp [~rog@] has quit [Ping timeout: 245 seconds]
13:54 -!- wrtp [~rog@] has joined #go-nuts
14:02 -!- fenicks [~christian@log77-3-82-243-254-112.fbx.proxad.net] has left
#go-nuts []
14:03 -!- virtualsue [~chatzilla@host81-148-115-114.in-addr.btopenworld.com] has
quit [Ping timeout: 252 seconds]
14:09 -!- virtualsue [~chatzilla@host81-139-108-139.in-addr.btopenworld.com] has
joined #go-nuts
14:15 -!- rinzai [~rinzai@host81-154-14-173.range81-154.btcentralplus.com] has
joined #go-nuts
14:19 -!- meanburrito920 [~john@] has joined #go-nuts
14:19 -!- meanburrito920 [~john@] has quit [Changing host]
14:19 -!- meanburrito920 [~john@unaffiliated/meanburrito920] has joined #go-nuts
14:22 -!- speps [d529952a@gateway/web/freenode/ip.] has joined
14:29 -!- Guest92113 [d529952a@gateway/web/freenode/ip.] has quit
[Quit: Page closed]
14:29 -!- rinzai [~rinzai@host81-154-14-173.range81-154.btcentralplus.com] has
quit [Quit: Leaving]
14:36 -!- virtualsue_ [~chatzilla@nat/cisco/x-ewgclewgzmepusib] has joined
14:55 -!- meanburrito920 [~john@unaffiliated/meanburrito920] has quit [Ping
timeout: 265 seconds]
15:00 < nsf>
15:00 < nsf> warning: trolling insdie
15:00 < nsf> inside*
15:00 < nsf> "Go is the Paris Hilton of programming languages."
15:00 < nsf> it's always good to hear alternative opinion :)
15:03 < bartbes> but is it annoying and omnipresent?
15:04 < nsf> I don't know, I'm just boring, when I'm boring I read flamewars
15:04 < nsf> Go vs.  D is kind of fun to me
15:05 -!- femtoo [~femto@95-89-197-196-dynip.superkabel.de] has quit [Quit:
15:05 < nsf> bartbes: so..  I guess I'm noticing alt opinions only when I'm
15:09 -!- skelterjohn [~jasmuth@lawn-gw.rutgers.edu] has joined #go-nuts
15:11 < nsf> lol, Alexandrescu says that Go has issues with semicolons
15:11 < nsf> I have written like 10k lines of Go code and had zero problems
with semicolons
15:11 < nsf> :\
15:13 < bartbes> the thing is, Go can't help him with his not-quite colon
15:13 < bartbes> sorry, those jokes were horrible
15:13 * bartbes leaves
15:20 < nsf> the funnies part it all things they complain about is lack of
generics (of course, if language doesn't have generics you can't write tons of
books about it) and semicolons :)
15:20 < nsf> funniest*
15:20 < nsf> yet, still no one was able to provide real example for generics
use case in real application
15:21 < nsf> sure, there are some, but are they frequent enough
15:22 < nsf> In C the most thing I use preprocessor for is a generic vector
15:22 < nsf> and that's it
15:22 < nsf> but Go has slices
15:22 < nsf> and even maps
15:22 < nsf> I just don't understand D authors
15:23 < nsf> I guess they are jealous that Go takes much more attention than
D, when it was born
15:23 < nsf> D is a total failure
15:23 < nsf> :P
15:25 < skelterjohn> what do you mean, no one can provide a real example for
generics use
15:25 < skelterjohn> i certainly can
15:25 < nsf> can you?
15:25 < nsf> do it
15:25 < nsf> :)
15:25 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has quit [Read error:
Connection reset by peer]
15:25 < nsf> it should be from a real, existing application
15:25 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has joined #go-nuts
15:25 < skelterjohn> I have an algorithm that I work with, developed in my
15:25 < skelterjohn> it's called FSSS
15:26 < skelterjohn> it is a search tree algorithm that works on somethign
we call an oracle
15:26 < skelterjohn> and the oracle has an underlying type
15:26 < skelterjohn> often this can be an integer
15:26 < skelterjohn> but sometimes it would be nice to have an array
15:26 < skelterjohn> or a string, doesn't matter, as far as FSSS is
15:27 < skelterjohn> i could use interface{}, but that would have literally
millions of type assertions that don't need to happen
15:27 < skelterjohn> and speed is important
15:27 < skelterjohn> and i don't feel like making identical copies of FSSS
for the different types i want
15:27 < nsf> make a generator
15:28 < skelterjohn> add generics to the language
15:28 < nsf> but why?
15:28 < skelterjohn> so people don't have to make generators
15:28 < nsf> generic algorithms is a rare
15:28 < nsf> why should we have something like that in the language, that
isn't essential
15:28 < skelterjohn> because it makes the programmer's life easier
15:29 < skelterjohn> also, the C++ standard template library is a great
example of appropriate use of generics
15:29 < skelterjohn> container classes that don't need type assertions
15:29 < nsf> and compiler programmer's life harder and tools programmer's
life harder
15:29 < skelterjohn> and can be space efficient
15:29 < skelterjohn> yes - there are many more "regular" programmers than
compiler programmers
15:30 < skelterjohn> we could have the compiler programmer's life really
easy and just use assembly
15:30 < skelterjohn> save them lots of work
15:30 < nsf> yes, they don't like to spend two hours a day to write a
15:30 < nsf> yet*
15:30 < skelterjohn> i don't follow that point
15:31 < skelterjohn> also, literally one of the first things i learned about
how to write code (from my dad when I was 13) was "if you're copying code around,
you're doing something wrong"
15:31 < nsf> generics in a language touch every feature
15:31 < skelterjohn> now, that may or may not be true
15:31 < nsf> basically adding generics to a language
15:31 < nsf> means making a different language
15:31 < skelterjohn> but it's sort of a deep belief of mine now, since i
learned it when i was so young
15:32 < nsf> as a matter of fact it's a lie
15:32 < skelterjohn> the fact that it's hard to add generics to the language
is orthogonal to the fact that generics would make the language nicer for some
people, like me
15:32 < nsf> if you're copying code around then you have something to change
if you can't, then you need to write a generator
15:32 -!- Fish [~Fish@bus77-2-82-244-150-190.fbx.proxad.net] has quit [Ping
timeout: 255 seconds]
15:33 < nsf> but also generics will brought tons of crappy unreadable code
15:33 < nsf> like it is now with C++
15:33 < skelterjohn> btw i know you don't mean offense by calling my dad a
liar, but a better way to say that might be "as a matter of fact it's not true" :)
15:33 < nsf> and with D
15:33 -!- Fish [~Fish@bus77-2-82-244-150-190.fbx.proxad.net] has joined #go-nuts
15:34 < nsf> your dad isn't a liar because he belives that it's true
15:34 < nsf> he's just mistaken
15:34 < skelterjohn> i know that's what you meant, that's why i didn't take
15:34 < nsf> there is literally no way to avoid that kind of "copy" action
15:34 < nsf> it's just a nice signal to you
15:34 < nsf> "time to do some refactoring"
15:35 < nsf> the worst thing here is to ignore it
15:35 < skelterjohn> right - if i feel the need to copy code somewhere, i
refactor the code instead
15:35 < nsf> but by mistake people think they should avoid copying
15:36 < nsf> but still there is something profound wrong with the
understanding of the problem
15:36 < skelterjohn> but having code generators is just as bad - you have
multiple pieces of code that have exactly the same behavior, and a bug in one is a
bug in all of them, but to fix this bug you have to fix it in all those places
15:36 < nsf> if you think that copying is bad, you're trying to make
everything generic
15:37 < nsf> otherwise you just write code, and when there are copies you
transform them to generic code
15:37 < nsf> code generators aren't bad
15:37 < nsf> Go compiler is written using a code generator
15:37 < nsf> called yacc
15:37 < skelterjohn> the programming language for the compiler is yacc, not
15:37 < skelterjohn> even if it creates C code as an intermediate step
15:38 < nsf> it's not a programming language, it's sort of a description of
rules and it generates C code :)
15:38 < nsf> type is just a description in templates
15:38 < skelterjohn> it's a language that creates a program
15:38 < skelterjohn> aka a programming language
15:38 < nsf> templates are a built-in code generator
15:38 < skelterjohn> i don't think go should have templates
15:38 < skelterjohn> it doesn't make sense with go
15:39 < nsf> but then it goes to the point, what Go should have?  no one
knows yet
15:39 < skelterjohn> that is a different question than "would some sort of
generics be useful?"
15:39 < nsf> nope :)
15:39 < nsf> how one can solve generic algorithms problem in Go
15:40 < nsf> I vote for code generators
15:40 < skelterjohn> they are related, but it's a different question.
15:40 < skelterjohn> here's something that makes code generators not work
15:40 < skelterjohn> someone made GoLLRB - a nice binary search tree package
15:40 < skelterjohn> uses interface{} as a type
15:40 < skelterjohn> this guy should have to distribute his source
15:41 < skelterjohn> not interface, sorry
15:41 < skelterjohn> but an interface
15:41 < skelterjohn> actually that's a bad example
15:41 < skelterjohn> here's a better one
15:41 < skelterjohn> i made a package called gohash
15:41 < skelterjohn> it has a hashmap type that has interface{} as its value
15:41 < skelterjohn> i shouldn't have to distribute source to let someone
use it
15:42 < nsf> why?
15:42 < skelterjohn> which means they can't use a code generator to make a
version for their own custom type
15:42 < skelterjohn> because i have a proprietary super algorithm inside
that i want to sell
15:42 < nsf> well, you know, selling algorithms is stupid, sorry for my
15:42 < skelterjohn> and THAT is certainly a different issue entirely
15:42 < exch> Rent doesn't pay itself :p
15:43 < nsf> exactly, it just means Go isn't a language you want
15:43 < skelterjohn> but go allows you to distribute only binaries and
should continue to do so
15:43 < skelterjohn> well it is, because i don't ahve to distribute source
15:43 < skelterjohn> anyway - in that scenario, code gen doesn't work
15:43 < skelterjohn> but some sort of generics could let the person using
the code be type safe and efficient at the same time
15:44 < nsf> haha, if you think that having a binary will stop people from
knowing what your algorithm is
15:44 < nsf> you're wrong
15:44 < skelterjohn> irrelevant
15:45 < nsf> it is relevant, because you're complaining about binary vs.
source form
15:45 < skelterjohn> it is completely irrelevant - source and binary are not
15:45 < nsf> nothing can stop other people from using your algorithm, except
law (at some level)
15:45 < skelterjohn> otherwise microsoft would include a source DVD with
windows 7
15:45 < exch> The legal difference is what it's all about
15:46 < exch> But, you could license out the source itself as well
15:46 < nsf> and as a matter of fact it's not even about the language
15:46 < nsf> it's about implementation
15:46 < skelterjohn> you're moving the goal-posts.  the scenario i've
explained is why i think generics are useful.
15:47 < nsf> also C++'s and D's templates work only in a source form
15:47 < skelterjohn> so?
15:47 < exch> templates != generics
15:48 < nsf> ok ok
15:48 < nsf> nevermind then
15:49 < skelterjohn> you know what i think would be useful?  an
interface[]{} type
15:49 < skelterjohn> not an array of interfaces
15:49 < skelterjohn> but an interface on an array
15:50 < skelterjohn> so if i have type FooInter interface{ Foo() }
15:50 < skelterjohn> hmm..  i don't know what the syntax would be
15:50 < skelterjohn> maybe FooInter[]
15:50 < skelterjohn> type FooInter[] could wrap a slice of things that all
implement FooInter
15:51 < skelterjohn> it would have to keep size info for the things in the
slice, so when i have "x FooInter[] = ...  ; x[2].Foo()" it could do it properly
15:51 < nsf> I don't think it's a good idea
15:52 < skelterjohn> it would allow things that take arbitrary types to not
have the issue of rebuilding lists
15:52 < skelterjohn> what about it do you think is a bad idea?
15:52 < exch> Isn't that the same as just defining a slice of an interface
15:53 < skelterjohn> no - because I can't say var X []FooInter =
15:53 < skelterjohn> because that slice is the actual structs, not the
interfaces wrapping the structs
15:53 < exch> yes, but thats irrelevant.  If they all implement the given
interface, you can treat em as such
15:54 < exch> regardless of what they really are underneath
15:54 < skelterjohn> but to do so i'd have to rebuild the slice
15:54 < skelterjohn> for i := range myOtherSlice { myFooSlice[i] =
myOtherSlice[i] }
15:54 < nsf> skelterjohn: see how sort.Interface does that
15:54 < exch> oh right, I think I get what you mean
15:54 < nsf> it wraps the whole array to an interface
15:54 < nsf> instead of asking for each element
15:54 < skelterjohn> does it use reflect on the inside
15:54 < skelterjohn> ?
15:54 < nsf> no
15:55 < skelterjohn> oh - but that's because the itnerface defines the
necessary array operations
15:56 < nsf> but I don't like sort.Interface, because it involves virtual
function calls on each operation
15:56 < skelterjohn> right
15:56 < nsf> it will be as slow as ruby or php
15:56 < nsf> :)
15:56 < nsf> well, actually faster
15:56 < nsf> but still slow
15:56 < nsf> :)
15:56 < skelterjohn> so if it took as a type interface[]{ Less(other
interface{}) bool }
15:56 < nsf> even compared to C's qsort
15:56 < skelterjohn> modulo readable syntax
15:56 < skelterjohn> it could be faster
15:57 < nsf> no
15:57 < nsf> any interface is a virtual function calls
15:57 < skelterjohn> but it has fewer of them
15:57 < skelterjohn> only Less would require it
15:57 < skelterjohn> not len or swap
15:58 < skelterjohn> that's half as many virtual function calls
15:58 < nsf> uhm..  you can't implement swap for interface[]{}
15:58 < skelterjohn> you don't have to
15:58 < nsf> without doing a virtual dispatch
15:58 < skelterjohn> you can just say mySlice[i], mySlice[j] = mySlice[j],
15:58 -!- niemeyer [~niemeyer@189-10-175-46.pltce701.dsl.brasiltelecom.net.br] has
joined #go-nuts
15:58 < nsf> yes, but what code it will generate
15:58 < skelterjohn> the type knows the size of the things in it
15:59 < skelterjohn> so it knows exactly which bytes are "mySlice[i]"
15:59 < nsf> ah, yes
15:59 < skelterjohn> and it can copy them
15:59 < nsf> ok, swap solved
15:59 < nsf> other issues are still here
16:00 < nsf> the only reason why I need templates (or code generator) is
16:00 < skelterjohn> what are the other issues with interface[]{}?
16:01 < nsf> it doesn't solve other generic algorithms problems
16:01 < nsf> trees you've mentioned
16:01 < skelterjohn> it's not supposed to :)
16:01 < skelterjohn> just another thing that i'd find useful
16:01 < nsf> you need more than that to add a feature to the language
16:01 < nsf> imho
16:02 < skelterjohn> if we add interface[]{}, then the only thing that I
would need from generics is type safety
16:02 < skelterjohn> because we can just use interface{} and interface[]{}
underneath the hood
16:02 < skelterjohn> though interface[]{} is clearly a horrible way of doing
the syntax
16:03 < nsf> maybe simply: interface[] :)
16:03 < skelterjohn> the interface needs some methods
16:03 < skelterjohn> interface[]{Foo()}, eg
16:03 < nsf> interface[ Less(other interface{}) ]
16:03 < nsf> :D
16:03 < skelterjohn> haha
16:03 < skelterjohn> it also needs to be stackable
16:03 < skelterjohn> interface[][]{}
16:04 < nsf> iSlice, lol
16:05 < skelterjohn> i'm gonna bother the ML about it
16:06 < skelterjohn> and see if they make fun of me
16:06 < wrtp> i will :-)
16:06 < skelterjohn> haha
16:07 < wrtp> if arrays why not maps too?
16:07 -!- Tv [~tv@cpe-76-168-227-45.socal.res.rr.com] has joined #go-nuts
16:07 < nsf> iMap
16:08 < nsf> iChannel
16:08 < nsf> why not
16:08 < nsf> :)
16:08 < skelterjohn> well, i don't see any reason why not
16:08 < wrtp> anyway you can have arrays that correspond to arbitrary
16:08 < skelterjohn> it certainly makes sense to want to use it like that
16:09 < skelterjohn> i don't know where you're going with that - can you
16:10 < wrtp> when i want to make a list.List typesafe, i just do: type
MyTypesafeList list.List
16:10 < wrtp> and define methods to do the requisite type checking
16:10 < wrtp> i mean, methods taking and returning the required type
16:11 < skelterjohn> right - that's a way to deal with type safety, but not
a way to deal with asserts that can be known to succeed at compile-time
16:11 < skelterjohn> both are issues
16:12 < wrtp> i don't think you can deal with the latter without generics
16:12 < skelterjohn> i agree
16:13 < wrtp> because interfaces are "there exists" types not "for all"
16:13 -!- danslo [~daniel@s5593965d.adsl.wanadoo.nl] has quit [Quit: Leaving.]
16:14 < wrtp> gotta go
16:14 < skelterjohn> peace
16:16 -!- artefon [~thiago@] has quit [Read error: Connection reset
by peer]
16:19 -!- rlab [~Miranda@] has quit [Ping timeout: 240 seconds]
16:27 -!- rlab [~Miranda@] has joined #go-nuts
16:28 -!- ExtraSpice [~XtraSpice@] has joined #go-nuts
16:28 -!- napsy [~luka@] has quit [Ping timeout: 255 seconds]
16:30 -!- rlab_ [~Miranda@] has joined #go-nuts
16:31 -!- artefon [~thiago@] has joined #go-nuts
16:31 -!- rlab [~Miranda@] has quit [Ping timeout: 240 seconds]
16:41 -!- aho [~nya@fuld-4d00d53f.pool.mediaWays.net] has joined #go-nuts
16:45 -!- napsy [~luka@] has joined #go-nuts
17:09 -!- ymasory_ [~ymasory@c-76-99-55-224.hsd1.pa.comcast.net] has joined
17:09 -!- ymasory_ [~ymasory@c-76-99-55-224.hsd1.pa.comcast.net] has quit [Read
error: Connection reset by peer]
17:19 -!- nsf [~nsf@jiss.convex.ru] has quit [Quit: WeeChat 0.3.3]
17:28 < wrtp> skelterjohn: so in your example that you posted, if you tried
to assign an object of type MyInterface to y that didn't happen to be a Fooer,
what would happen?
17:29 -!- Tv [~tv@cpe-76-168-227-45.socal.res.rr.com] has quit [Ping timeout: 245
17:32 -!- mssm [~mssm@ip-95-221-86-196.bb.netbynet.ru] has quit [Ping timeout: 245
17:32 -!- awidegreen_ [~quassel@p5DF1E44E.dip.t-dialin.net] has joined #go-nuts
17:33 -!- awidegreen [~quassel@p5DF1D7D4.dip.t-dialin.net] has quit [Ping timeout:
240 seconds]
17:34 -!- napsy [~luka@] has quit [Ping timeout: 264 seconds]
17:36 -!- napsy [~luka@] has joined #go-nuts
17:58 < skelterjohn> wrtp: can you be more specific?
17:59 < skelterjohn> oh - i see
17:59 < skelterjohn> interesting point
18:00 < skelterjohn> ok - the type MyInterface[] knows its type at runtime
18:00 -!- cco3 [~conley@c-69-181-138-209.hsd1.ca.comcast.net] has joined #go-nuts
18:00 < wrtp> so the type check is dynamic?
18:00 < wrtp> that seems wrong to me, if so.
18:00 < skelterjohn> i'm thinking
18:01 < skelterjohn> well, when dealing with interfaces, runtime checks
happen all the time
18:01 < skelterjohn> if generics were done as a layer on top of that the
check could be skipped
18:01 < skelterjohn> so i don't think that a dynamic check for assignment to
a slice-interface is necessarily a bad thing
18:02 < skelterjohn> it'd be equivalent to y.([]NotAFooer)[2] =
18:02 < wrtp> dealing with interfaces, runtime checks that can result in an
error only occur when you specifically ask to do them
18:02 < skelterjohn> that's true
18:03 < wrtp> i think you mean y[2] = notAFooerInstance.(Fooer)
18:03 < skelterjohn> no - Fooer is the non interface type that happens to
have a Foo() method
18:04 < skelterjohn> and you could never convert notAFooerInstance to a
Fooer instance
18:04 < wrtp> exactly
18:04 < skelterjohn> but it could still satisfy MyInterface
18:04 < wrtp> but y only accepts Fooer instances
18:04 < skelterjohn> you only know that at runtime
18:04 < wrtp> you might as well use an array of Fooer
18:05 < skelterjohn> you don't know at runtime that Fooer is the solid type
you want
18:05 < skelterjohn> you just know it's an array of things that can go Foo()
18:05 < skelterjohn> i may have chosen the names poorly in my example
18:07 < skelterjohn> but that's a good point about type checking only being
allowed to fail when you ask for it
18:07 < skelterjohn> anything else would not be typesafe, clearly
18:08 < wrtp> yup
18:08 < wrtp> gotta go again!
18:08 < yiyus> and how do you know, at runtime, if a value can be asigned to
an interface[] value?
18:09 < skelterjohn> the interface{}[] type keeps track of the internal type
in the same way that interface{} does
18:09 < skelterjohn> it's just an interface for a slice, and needs to know
how big its elements are
18:10 < yiyus> but with interface{} you just have to check if you implement
interface{}, which makes sense
18:10 < yiyus> but with an interface{}[] you have to check that the
underlying types are the same, which i find much less evident
18:10 < skelterjohn> to assign something of type []T to type MyInterface[],
you just check T against MyInterface
18:11 < skelterjohn> i don't follow
18:11 < yiyus> but you cannot asign whatever that satisfays MyInterface
18:11 < yiyus> because Type1 and Type2 can both implement MyInterface
18:12 < yiyus> but you cannot mix them in arrays
18:12 < skelterjohn> that's right
18:12 < skelterjohn> this is the issue that wrtp brought up
18:12 < skelterjohn> and i haven't come up with a satisfactory answer yet
18:12 < skelterjohn> but i am thinking
18:12 < yiyus> indeed, it's the same problem
18:13 < skelterjohn> a necessary (but not sufficient) condition for this to
be useful is to be able to say y[i] = y[j]
18:13 < skelterjohn> in this case, you know that the types match
18:13 -!- virtualsue [~chatzilla@host81-139-108-139.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
18:13 < skelterjohn> another thing that would be nice to be able to do is
say z := y[j]; y[i] = z
18:14 < skelterjohn> but if that is allowed, it is allowed for any z
18:14 < skelterjohn> and i can't figure out a way to get around needing a
type assertion there
18:14 < skelterjohn> that could potentially fail
18:14 < skelterjohn> you also can't say y.([]TheRightType)[i] = z, because
you don't know TheRightType
18:15 -!- virtualsue_ [~chatzilla@nat/cisco/x-ewgclewgzmepusib] has quit [Ping
timeout: 240 seconds]
18:15 -!- |Craig| [~|Craig|@panda3d/entropy] has joined #go-nuts
18:15 < skelterjohn> y[i] = z could do a silent type assertion, but that
will never happen
18:16 < skelterjohn> if this is all wrapped up in generics, though, the type
assertion succeeds automatically
18:17 < skelterjohn> so there is that
18:21 < skelterjohn> Well, I posted that.  But the argument put a bit of a
dent in my idea, though
18:23 -!- virtualsue [~chatzilla@host81-148-89-136.in-addr.btopenworld.com] has
joined #go-nuts
18:24 -!- jo__ [3a72cb23@gateway/web/freenode/ip.] has quit [Quit:
Page closed]
18:33 -!- virtualsue [~chatzilla@host81-148-89-136.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
18:38 -!- napsy [~luka@] has quit [Quit: Lost terminal]
18:38 -!- napsy [~luka@] has joined #go-nuts
18:39 < wrtp> skelterjohn: if you've got generics, you don't need this, i
18:39 < skelterjohn> generics need an implementation
18:39 < skelterjohn> you can pretend to do generics without arrays by using
18:39 < skelterjohn> all you lack is type safety
18:39 < skelterjohn> if you ahve this, you can do it with arrays
18:43 -!- virtualsue [~chatzilla@host81-148-59-74.in-addr.btopenworld.com] has
joined #go-nuts
18:44 < wrtp> generics are about much, much more than type safe arrays
18:45 < skelterjohn> it's not the solution, just a part of one possible
18:45 < wrtp> think about pointers to functions of data structures defined
in terms of the parameterised type, for example
18:45 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has quit [Remote host closed the connection]
18:46 < wrtp> interface types cannot be compatible with generic parameter
18:46 < skelterjohn> i don't think you can have a data structure defined in
terms of a parameterized type without source analysis....  not and have it be a
continuous block of data
18:46 < wrtp> of course you can - all the major functional programming
languages allow it
18:47 -!- femtoo [~femto@95-89-197-196-dynip.superkabel.de] has joined #go-nuts
18:47 < skelterjohn> in java it's not a continuous block of data
18:47 < skelterjohn> the parameterized types are reference types
18:47 < wrtp> sure, that's one approach
18:47 < skelterjohn> in C++ it uses source analysis
18:47 < wrtp> C# does it right
18:47 < skelterjohn> i'm not aware of the other ways this is done
18:48 < wrtp> most functional programming languages, i believe, use a single
size for all data types (pointer sized, i'd guess)
18:48 < wrtp> so that makes it easy
18:48 < skelterjohn> doesn't work if your parameterized type is a larger
18:49 < wrtp> it does if you keep all structs on the heap
18:49 < skelterjohn> but functional programming languages are safe if the
values are const
18:49 < wrtp> C# relies on runtime code generation, i think
18:49 < skelterjohn> if the values are const, then copying a structure that
contains a pointer is ok, because you know what it points to can't change and have
18:50 < wrtp> sure, but i'm not just talking about haskell here
18:50 < skelterjohn> i don't know how that could be safe otherwise
18:51 -!- virtualsue [~chatzilla@host81-148-59-74.in-addr.btopenworld.com] has
quit [Ping timeout: 245 seconds]
18:51 < wrtp> you don't copy the structure
18:51 < wrtp> you have code that can deal with all possible sizes of type
18:51 < skelterjohn> this is all starting to sound inefficient
18:51 < wrtp> either several versions of the code, or one version that does
memcpy all the time
18:51 < wrtp> yup, and that's the dilemma
18:53 < |Craig|> I'd be happy with a way to import files that does a find
and replace on the source to swap a type.
18:53 < wrtp> there's already something that does that
18:53 < |Craig|> wrtp: link?
18:53 < wrtp> can't remember
18:53 < skelterjohn> search the mailing list for roundy gotit
18:54 < |Craig|> I;ll look around, I think I might have saw it once too
18:54 < skelterjohn> roundy is the guy's name
18:54 < skelterjohn> gotit is the (poorly named) package
18:54 < wrtp> it doesn't work so well, as it doesn't work at more than one
18:54 < |Craig|> I already manually duplicated my container's source for one
type so some automation would be nice
18:55 < skelterjohn> this would be appropriate for you
18:55 < |Craig|> excelent
18:56 < wrtp> thing is, most types are actually a very small range of sizes
18:57 < wrtp> byte, word, double word, triple word.  not much code passes
larger structs by value
18:57 < wrtp> so a memcpy hit probably isn't too bad
18:57 < |Craig|> I'm very aware of how horribly messy generics can be
(java).  I mostly see them as useful for containers
18:57 < wrtp> (in the latter case)
18:57 -!- virtualsue [~chatzilla@host81-148-75-88.in-addr.btopenworld.com] has
joined #go-nuts
18:57 < wrtp> a channel is a form of container...
18:58 < skelterjohn> it'd be nice to be able to use something other than
primitive types as keys for a map
18:59 < wrtp> yup - with generics that should be possible (add an argument
to map make)
18:59 < wrtp> that means you wouldn't have to privilege some arbitrary
method name
18:59 < |Craig|> wrtp: I have a data structure that has good reason to use
larger structs by value (up to several words).  Its sole purpose is to store
objects side by side in memory in hopes of good cache performance and perhaps
reduced GC load and fragmentation
19:00 < wrtp> yeah, i've used them too
19:00 -!- tensorpudding [~user@] has joined #go-nuts
19:01 < |Craig|> Does one copy to remove an object, and one to add an
object, indexed as pointers.
19:01 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has joined #go-nuts
19:01 < |Craig|> Gotit seems to be missing: https://github.com/droundy/gotit
19:03 < wrtp> hitting the sweet spot for generics is hard.  that's why they
haven't been done yet!
19:03 < |Craig|> ah, renamed https://github.com/droundy/gotgo/ (better name)
19:04 < |Craig|> I think part of the issues is that there are reasons to use
multiple different of generics, and no one wants to clutter the language with
19:05 < |Craig|> go already has built in generic stuff (like arrays, maps,
channels), and some odd functions like close closed and cap.
19:05 < wrtp> generality *and* efficiency...
19:06 < |Craig|> how len closed and cap work seems rather odd to me, arn't
they exceptions to how things generally work in go?
19:06 < skelterjohn> the built in functions cheat
19:07 < skelterjohn> overloaded, generic, etc
19:07 < |Craig|> if builtins have to cheat in several places, it indicates
something is missing from the language, and when fixed, we will have a real
solution + builtin hacks that work different (well maybe gofmt will save us from
that and ending up like java)
19:08 -!- SoniaKeys [~soniakeys@c-76-118-178-209.hsd1.ma.comcast.net] has joined
19:09 -!- virtualsue [~chatzilla@host81-148-75-88.in-addr.btopenworld.com] has
quit [Remote host closed the connection]
19:09 < |Craig|> the more hacks like that that get added (such as append)
the bigger the language gets and the less intensive there is for real generics to
rationalize it
19:10 < skelterjohn> i think that's the idea
19:10 < |Craig|> add more mess so we can avoid a feature?
19:10 < skelterjohn> seems that way to me
19:12 < |Craig|> length and cap are fields of an slice, and could be non
exported and accessed through methods, but they arn't because generics are
missing, so we use some global name space polluting special functions.  That kind
of thing is the only real issue that bugs me about go
19:13 < Namegduf> They aren't because the design decision was made that
built-in types do not have methods.
19:13 < Namegduf> Has nothing to do with generics.
19:13 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has quit [Remote host closed the connection]
19:13 < |Craig|> Namegduf: ok, that explanation works for me
19:13 < |Craig|> interesting choice
19:14 < skelterjohn> it does indirectly - because they pollute the namespace
where if we had generics they could be "regular" functions defined in packages
19:14 < Namegduf> They could have a magic package as much as magic
19:14 < Namegduf> Hell, we do; unsafe.
19:15 < |Craig|> Thats really different from python where the difference
between builtin types and others is almost non existent.  Do you know why they
decided not to give the builtin types methods?
19:15 -!- Project_2501 [~Marvin@] has joined #go-nuts
19:15 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has joined
19:15 < skelterjohn> well, in python the built in functions call methods
19:15 < skelterjohn> len(x) calls x.__len__(), etc
19:16 < Namegduf> It's like the worst of both worlds!
19:16 < Namegduf> :P
19:16 < |Craig|> skelterjohn: yes, and you can make something from scratch
that is identical to an int
19:16 < skelterjohn> __plus__ as well?
19:17 < |Craig|> skelterjohn: all operators call methods
19:17 < |Craig|> in python
19:17 < skelterjohn> hey - it beat's C++'s operator overloading
19:17 < skelterjohn> beats
19:18 < |Craig|> in python, you can make immutable, and give them math
operators, and make a true int, string or what ever (aside from some performance)
without leaving pure python.  In go, you can't make any of the built in types.  I
won't say thats bad, but its really different and it took a while to sink in what
it meant as far as implementing containers.
19:19 < |Craig|> My containers can never be like the built in ones
19:19 < Namegduf> Actually
19:19 < Namegduf> *can* you?
19:19 < Namegduf> Python does special optimisations for the built-in int.
19:19 < Namegduf> Can you match its performance, or get close to?
19:20 < skelterjohn> "(aside from some performance)"
19:20 < |Craig|> Namegduf: in python over load __new__ for immutable and
interning.  Anyway, performance is implimentation specific, not a language thing
for python
19:20 < Namegduf> I agree that it's not a language thing for Python
19:21 < |Craig|> you can auctually hack it to claim its class is an int too
19:21 < Namegduf> But it's kind of important, especially for the example of
19:21 < Namegduf> Not that I see the limitation in not being able to
reimplement built-in types.
19:22 < |Craig|> Having built in containers and user made ones work
differently in the very least adds complexity to learning the language, and lack
of interchangeability
19:23 < Namegduf> But does it add complexity compared to alternative
proposals that would make them work the same?
19:23 < Namegduf> You'd have to make other language changes.
19:23 < Namegduf> You can't compare "Problem vs Not Having Problem", you
have to compare "Problem vs Alternative"
19:23 < |Craig|> It would require generics
19:24 < |Craig|> which yes, would add complexity
19:24 < Namegduf> Also operator overloading
19:24 < Namegduf> x[foo] syntax
19:24 < |Craig|> I'm not saying I have a solution, its just that I don't
have any other problems with go to discuss :)
19:24 < skelterjohn> i don't mind not having operator overloading
19:24 < skelterjohn> it doesn't bother me to say
19:25 < Namegduf> I agree that overloading is the enemy of simplicity by way
of expanding the number of rules you need to remember.
19:25 < Namegduf> Also less helpful than you'd think with matrixes unless
you can add new operators for the additional operations that'd be nice to have.
19:26 < |Craig|> ha, you could allow operators as method names (kinda like
scheme/lisp) myMatrix.+(anotherMatrix) (and I will not advocate allowing that)
19:26 < skelterjohn> it has been suggested
19:26 < skelterjohn> i think 1) it is ugly
19:26 < skelterjohn> that's about as far as i went
19:26 < Namegduf> I think there's a problem with associativity with that.
19:27 < Namegduf> And m1 + m2 + m3 is still ugly.
19:27 < Namegduf> Not a problem for Lisp, obviously.
19:27 < |Craig|> I was really glad there was no operator overloading in go
19:28 < |Craig|> its a whole mess I don't have to deal with
19:28 < Namegduf> I wonder what a solution to make Go handle large amounts
of matrix multiplication would be.
19:28 < Namegduf> Or whether such a program would be better done in
something else.
19:29 < Namegduf> I think it might just be best to take the hit and use
methods everywhere.
19:29 < skelterjohn> if you want to write linear algebra, use matlab
19:29 < skelterjohn> if you want to do matrix stuff in go, use gomatrix ;)
19:29 < skelterjohn> certainly matlab will be faster than gomatrix, because
the guys who made matlab are way smarter than i am
19:29 < |Craig|> Namegduf: skelterjohn's matric package works, but having it
be generic for matrix sizes (so I could type something as a 3*3 and still have it
have methods) would be nice.
19:30 -!- virtualsue [~chatzilla@nat/cisco/x-funmgovthqtqktvk] has joined #go-nuts
19:30 < Namegduf> What would be particularly interesting would be to enforce
matrix sizes using the type system.
19:30 < skelterjohn> now it's getting messy
19:30 < Namegduf> I wonder if there's a generics proposal that would allow
19:30 < skelterjohn> i believe you can do that in C++
19:30 < |Craig|> I have a package that only works with n*n and n*1 matrixes.
I think you could do that generically in C++
19:31 < |Craig|> I just have to have size checks everywhere, of hope its
19:31 < |Craig|> *or
19:31 < skelterjohn> it will throw up on you if it's wrong
19:32 < |Craig|> yes, it will, and being able to do that costs time at
19:32 < skelterjohn> not compared to the actual matrix operation
19:32 < |Craig|> Im dealing with 2*2 and 2*1 only here, I think it might be
19:32 < Namegduf> Depends on matrix size, yeah.
19:33 < skelterjohn> it's exactly one comparison
19:33 < Namegduf> Right now I'm playing with 4x4 in C.
19:33 < |Craig|> makes the code messy too as the sizes are only in the
19:33 < skelterjohn> m1.cols == m2.rows
19:35 < |Craig|> skelterjohn: initing a 2*2 takes 4 calls, all with size
checks for row and column, not a set(a,b,c,d) (but you could add that with the ...
thing maybe?  I need to learn about that).  It also many operations would not need
to return errors if the sizes were fixed
19:36 < skelterjohn> you can initialize it with a []float64{a,b,c,d}
19:36 < |Craig|> type matrix2x2 [4]float <-- I guess I could do that
19:36 -!- virtualsue [~chatzilla@nat/cisco/x-funmgovthqtqktvk] has quit [Ping
timeout: 250 seconds]
19:36 < |Craig|> skelterjohn: I didn't realize that
19:36 < skelterjohn> it's in dense_basic.go, i think
19:37 < |Craig|> ok
19:37 < skelterjohn> func MakeDenseMatrix(elements []float64, rows, cols
int) *DenseMatrix
19:37 < skelterjohn> in dense.go
19:38 < |Craig|> skelterjohn: you made your matrixes structs right?
19:38 < |Craig|> I wonder how making them slices or arrays would come out
19:39 < skelterjohn> same length array for different width/height
19:39 < skelterjohn> since it's just their product
19:39 < skelterjohn> could do it with [][]float64, but i do some clever
things with a flat array that make that inefficient
19:40 < |Craig|> I had a struct that just contained an int, and things ran
faster when I made the type an int (type someName int) (I'm tempted to call it a
subclass, but I know thats wrong).
19:41 < |Craig|> optimization in go is fussy.
19:41 < skelterjohn> call it a subtype
19:41 < |Craig|> I added a name to a return value (though didn't use it in
the code) and my runtime went down 2%.
19:41 < skelterjohn> neat
19:41 < skelterjohn> that's just because the compiler is young, though
19:42 < |Craig|> ya
19:42 < skelterjohn> also, 2% is usually within the margin of experimental
19:42 < |Craig|> skelterjohn: I ran several tests on over 10000 loops, I can
cosistanct to 3 digets
19:42 < |Craig|> *had consistency
19:43 < |Craig|> I think it was just a cache vs time and place of allocation
19:44 < SoniaKeys> i flattened some four dimensional arrays a program of
mine.  it made code slightly faster, made the data take a little less space, but
the bigger surprise for me was that the code was just as easy to understand and in
some cases much simpler.
19:46 -!- skelterjohn [~jasmuth@lawn-gw.rutgers.edu] has quit [Ping timeout: 255
19:47 -!- skelterjohn [~jasmuth@lawn-gw.rutgers.edu] has joined #go-nuts
19:47 < skelterjohn> got booted, sorry if it seemed like i was being
19:48 < |Craig|> I had some horrible C code (my first C project) that used 3
dimensional arrays and pointer arithmetic.  I replaced this line with using the
arrays instead of painter arithmetic, and it ran faster (because GCC is smarter
than me): *(pIn+*ptrans++)+=*pIn2++; It makes me doubt the value of pointer
19:48 < skelterjohn> sometimes when you do stuff with arrays the compiler
can vectorize it
19:49 < skelterjohn> where if you do the same thing with pointer arithmetic
it cannot
19:50 < SoniaKeys> but in go we don't have multidimensional arrays.  i'm
sure arrays of arrays are trickier to optimize, slices of slices more so.
19:51 < skelterjohn> i think |Craig| meant an array of an array of an array
when he said 3 dimensional array, since he is talking about C
19:52 < |Craig|> multidimisnional arrays in go: this is valid:
19:52 < |Craig|> so SoniaKeys we do have them
19:52 < skelterjohn> she means like they have them in, for example, C#
19:52 < SoniaKeys> or fortran
19:52 < skelterjohn> int[10,10]
19:52 < SoniaKeys> right
19:52 < skelterjohn> used like theArray[x,y] = v
19:53 < |Craig|> is it really any different as far as how its structured in
memory?  Its just a bunch of arrays end to end
19:54 < skelterjohn> it's more like, you're telling the compiler how you
want to use the arrays
19:54 < skelterjohn> so it can sometimes take advantage of that information
19:54 < skelterjohn> i'm getting a weird compile error: (node AS2DOTTYPE)
used as value
19:54 < skelterjohn> when i'm trying to type assert an interface to another
19:54 < skelterjohn> i can't figure out what is causing it
19:55 < skelterjohn> oh duh, nevermind
19:56 < SoniaKeys> i think numpy is wonderful.  it'd be wonderful to see
something like that for go some day.
19:56 < skelterjohn> i did "if a, ok := b.(A) {...  }"
19:56 < skelterjohn> what about numpy would you like to see in go?
19:57 < SoniaKeys> well, speaking memory layout, numpy separates the working
shape from the memory layout
19:57 < skelterjohn> i don't know what you mean
19:58 < skelterjohn> i'm not really a nuts-and-bolts kind of guy though
19:59 < SoniaKeys> it stores everything flat internally, but keeps the
shape--dimensionallity information in a struct.  so you can change the shape
dynamically without moving stuff around in memory or reallocating.
19:59 < skelterjohn> you can do that with gomatrix :)
20:00 -!- pingveno [~pingveno@c-98-246-133-8.hsd1.or.comcast.net] has joined
20:00 < SoniaKeys> ok, i confess, i haven't looked at it.  i really
20:00 < skelterjohn> gomatrix stores the elements in a flat slice, keeps the
width and height, but also "step"
20:01 < skelterjohn> and this step parameter allows you to have a sub-matrix
use the same data correctly
20:01 < skelterjohn> assuming that's what you meant by changing the shape
20:01 < SoniaKeys> a stride, in numpy speak, right?
20:01 < skelterjohn> quite possibly, though i'm not familiar with numpy's
20:01 -!- virtualsue_ [~chatzilla@nat/cisco/x-onyljcakngevmxpr] has joined
20:01 < SoniaKeys> yes
20:02 < skelterjohn> it's equivalent to the number of columns in the
original matrix
20:02 -!- jochang [~identd@58-114-203-35.cable.dynamic.giga.net.tw] has quit [Ping
timeout: 265 seconds]
20:02 < skelterjohn> so to go down one row, you add step to the index
20:02 -!- wrtp [~rog@] has quit [Ping timeout: 240 seconds]
20:03 < SoniaKeys> (installed.  i'm looking at it now)
20:04 < skelterjohn> i'd be glad to hear any comments.  it was the first
major chunk of go that i wrote, and some things are silly
20:05 < skelterjohn> |Craig| pointed out weirdness in the error returns, for
20:06 < SoniaKeys> well i think it's cool that it's one of the more
installed packages, according to the dashboard
20:06 -!- Tv [~tv@cpe-76-168-227-45.socal.res.rr.com] has joined #go-nuts
20:07 < skelterjohn> um - probably because when i test it i use goinstall
20:07 < SoniaKeys> i wrote something similar in C++ years ago.  there are
plenty of design decisions, i know.
20:07 < skelterjohn> that number has stayed pretty steady since i got it
20:07 < skelterjohn> i'd like to see the dashboard count unique IP installs
20:07 < SoniaKeys> haha.  i don't think so.  i think it counts unique ip
addresses or something, to avoid that
20:08 < skelterjohn> hey now
20:09 -!- wrtp [~rog@] has joined #go-nuts
20:13 < SoniaKeys> do you have the cross product in there somewhere?
20:13 < skelterjohn> no
20:15 < SoniaKeys> that was one i needed recently.  it's just one line of
code though for the 3x3 matricies i had, so it was no big deal.
20:15 < skelterjohn> might be interesting to have a function that takes n-1
n-vectors and gives one that is perpendicular to all of them
20:16 < skelterjohn> dot is in there, anyway :)
20:16 < SoniaKeys> i saw that one.
20:17 < skelterjohn> well, it's just a.Transpose.Times(b).Trace()
20:17 < skelterjohn> do i actually have it in there as dot somewhere?
20:17 < SoniaKeys> mm, i just ran godoc.  i think i saw something like it...
20:18 < skelterjohn> it appears in a comment
20:19 < SoniaKeys> ah
20:19 -!- wrtp [~rog@] has quit [Ping timeout: 245 seconds]
20:24 -!- wrtp [~rog@] has joined #go-nuts
20:49 < SoniaKeys> nice variable name in svd
20:50 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has quit [Read error:
Connection reset by peer]
20:50 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has joined #go-nuts
20:52 < skelterjohn> using greek letters in code was new to me, so i went
all out
20:52 < skelterjohn> kind of annoying now :)
20:52 < skelterjohn> especially since it's easy for me to type them on my
mac, but not on my linux box
20:55 < SoniaKeys> i think there are ways to make it easier.  i haven't used
them enough to bother yet.  ^v u memorized unicode...
20:55 < Tv> nice way to reduce other people's contributions to copy-paste
20:56 < skelterjohn> on my mac i set it up to switch to the greek keyboard
by hitting cmd-opt-space
21:02 -!- photron_ [~photron@port-92-201-176-25.dynamic.qsc.de] has quit [Ping
timeout: 255 seconds]
21:03 -!- SirPsychoS [~sp@c-24-13-132-255.hsd1.il.comcast.net] has joined #go-nuts
21:04 < SirPsychoS> Is it not possible to have a const array in Go? "const a
[1]byte = [1]byte{0}" gives "const initializer must be constant"
21:04 -!- napsy [~luka@] has quit [Quit: Lost terminal]
21:05 < skelterjohn> consts have no type
21:05 < SirPsychoS> "const a int = 5" compiles fine
21:05 < SirPsychoS> consts _can_ have no type
21:06 < skelterjohn> you can say, for example "const a = 2"
21:06 < skelterjohn> ah
21:06 < skelterjohn> then i dunno
21:06 -!- artefon [~thiago@] has quit [Read error: Operation timed
21:08 < SirPsychoS> looks like const only applies to scalar values, and not
to lists of values like slices, maps, and arrays
21:08 -!- nsf [~nsf@jiss.convex.ru] has joined #go-nuts
21:09 -!- napsy [~luka@] has joined #go-nuts
21:09 < SirPsychoS> huh, even structs can't be consts
21:12 -!- htoothrot [~mux@66-169-185-121.dhcp.ftwo.tx.charter.com] has quit [Read
error: Connection reset by peer]
21:19 -!- niemeyer [~niemeyer@189-10-175-46.pltce701.dsl.brasiltelecom.net.br] has
quit [Ping timeout: 245 seconds]
21:35 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has joined #go-nuts
21:35 -!- zozoR [~zozoR@5634798d.rev.stofanet.dk] has quit [Quit: Morten.  Desu~]
21:50 -!- femtoo [~femto@95-89-197-196-dynip.superkabel.de] has quit [Quit:
22:07 -!- watr_ [~watr@] has joined #go-nuts
22:08 -!- sauerbraten [~sauerbrat@p508CC67A.dip.t-dialin.net] has quit [Remote
host closed the connection]
22:11 -!- tensorpudding [~user@] has quit [Ping timeout: 255
22:11 -!- SoniaKeys [~soniakeys@c-76-118-178-209.hsd1.ma.comcast.net] has quit []
22:12 < |Craig|> I can't seem to make any of gotgo's example make files to
do anything, but the binary sure works.  I guess I'll stick with my shell script
for building now, and learn about makfiles later
22:14 -!- Fish [~Fish@bus77-2-82-244-150-190.fbx.proxad.net] has quit [Quit: So
Long, and Thanks for All the Fish]
22:15 < skelterjohn> i never tried the makefiles
22:15 < skelterjohn> i don't know if they were around when i fooled with
22:15 < skelterjohn> if you post in the ML i bet dave would be happy to help
22:15 < |Craig|> I don't know how to write make files, and I'm building with
a shell script anyway
22:16 < skelterjohn> ah
22:16 < skelterjohn> he might not feel like teaching you make :)
22:16 < skelterjohn> but it is a good tool to have in your belt
22:16 < |Craig|> I just wanted to run the tests after I built gotgo, and I
22:16 < skelterjohn> i recommend trying to learn it, if you have the time
22:17 < |Craig|> I keep intending to learn make.  I wrote a single package
make file, but when I found an example multi package one it was 10 times as long
and super confusing, and I used a 1 line shell script instead
22:17 < |Craig|> I guess that ended up a 3 line shell script auctually
22:18 < |Craig|> but still trivial
22:18 < skelterjohn> what i do right now is have the template makefile for
each package, and a script that runs all of them :s
22:18 < skelterjohn> not very make-pro-ish
22:19 < |Craig|> my shell script is now 5 lines, for 2 packages including
using gotgo
22:19 < skelterjohn> i am juggling 20 different packages and 10 different
cmd binaries
22:19 < skelterjohn> when i run my build script i feel silly
22:20 < skelterjohn> whoah, i exaggerate.  8 pkgs and 7 cmds
22:20 -!- tensorpudding [~user@] has joined #go-nuts
22:24 < |Craig|> maybe I'll get a go project to rival my python project with
52 python modules and 5 cython modules
22:24 < |Craig|> I've been keeping my package count low in go due to build
22:25 < |Craig|> thats one thing that goclipse has down.  It auto generates
22:25 < skelterjohn> the thing that troubled me was having them depend on
each other
22:25 < skelterjohn> so i just have my shell script build them in a
topological order
22:26 < skelterjohn> i didn't feel like typing in all the dependencies
22:27 < |Craig|> just write packages that only have dependices to ones with
names earlier in the allphabet
22:28 -!- cenuij [~cenuij@] has joined #go-nuts
22:28 -!- cenuij [~cenuij@] has quit [Changing host]
22:28 -!- cenuij [~cenuij@base/student/cenuij] has joined #go-nuts
22:28 < uriel> |Craig|: really, if you are having trouble dealing with
building your go projects, you are doing somewhing wrong
22:28 < uriel> the makefiles are *trivial*
22:28 < skelterjohn> ordering them isn't the issue - it's having them do the
minimal build necessary that is irritating
22:30 < |Craig|> I looked at this (number 4 there)
22:30 < skelterjohn> eep
22:30 < skelterjohn> use the templates
22:30 < |Craig|> lookes far worse than something like 6g tcp/*.go
22:31 < skelterjohn>
http://code.google.com/p/gomatrix/source/browse/matrix/Makefile eg
22:31 < |Craig|> skelterjohn: that works for 1 package
22:31 < |Craig|> what if I like lots of tiny packages
22:31 < skelterjohn> oh didn't see the "multi" part
22:33 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has quit [Remote host closed the connection]
22:34 < uriel> what are 'subpackages' anyway?
22:34 < skelterjohn> ideally a build tool would peek at the source to
discover dependencies
22:34 < vsmatck> When I originally started using Go I thought packages were
equivalent to classes.  So I broke up everything such that no piece of code has
access to variables it didn't need.  It didn't turn out well although I'm not sure
it was a bad idea.
22:34 < uriel> there are packages, period
22:35 < uriel> vsmatck: thinking of packages as 'classes' is certainly not a
good idea,
22:35 -!- rlab_ [~Miranda@] has quit [Quit: Miranda IM! Smaller,
Faster, Easier.  http://miranda-im.org]
22:35 < vsmatck> yes, but I'm not sure the rest was a bad idea.
22:35 < uriel> (thinking of *anything* as 'classes' is a bad idea)
22:35 -!- awidegreen_ [~quassel@p5DF1E44E.dip.t-dialin.net] has quit [Remote host
closed the connection]
22:35 < vsmatck> Although what I was doing created complex dependencies that
couldn't be easily managed manually.
22:35 < uriel> i have no clue what you mean by ' such that no piece of code
has access to variables it didn't need.', isn't that just the way one build
22:36 < uriel> and i have no clue what that has to do with packages
22:36 < vsmatck> Go doesn't encourage that IMO.
22:36 < uriel> functions/methods and structs provide plenty of encapsulation
22:36 < uriel> doesn't encourage what?
22:37 < |Craig|> uriel: can you point me to a trivial multi package make
22:37 < vsmatck> I'm thinking like private data members of a class.  On that
22:37 < uriel> |Craig|: no, because for the most part you should have one
Make file per package
22:37 < uriel> (at least that is what I do, and I don't see what the problem
with that is)
22:37 < vsmatck> Encapsulation in my go programs is not as complete as in my
C++ programs.  But it doesn't really feel like much of a problem.  *shrug*
22:38 < uriel> vsmatck: structs?
22:38 < skelterjohn> because you can have packages that depend on other
22:38 < skelterjohn> and you need them to be rebuilt when the other package
22:38 < uriel> vsmatck: buy you were talking about *variables*
22:38 < |Craig|> uriel: thats tolerable, so how to I make my whole project
then?  I make make files referance eachother?
22:38 < uriel> vsmatck: and unless you are using plenty of globals, I don't
see what the problem is (and you obviously should not be using plenty of globals)
22:39 < vsmatck> I didn't say there was a problem.  :)
22:39 < vsmatck> I just can't program C++ in Go dammit.  heh
22:40 < uriel> |Craig|: see src/pkg/Makefile
22:40 < uriel> vsmatck: I would say that is a very good thing ;P
22:41 < vsmatck> *shrug* It's relative to what you're doing.
22:41 < skelterjohn> uriel: that makefile depends on running a dependencies
22:41 < nsf> makefiles in Go tree are crappy
22:42 < vsmatck> I have the idea that dividing a go program in to super
granular packages is not a good idea.  I'm not sure if this is because of lack of
fancy build tools though.  I think that's a good what to put what I was thinking.
22:43 < uriel> skelterjohn: unless you got hundreds of packages, managing
that by hand is trivial
22:43 < nsf> also using shell script is fine
22:44 < nsf> unless the speed of build doesn't bother you
22:44 < vsmatck> I always thought it was really hard to organize my C++
programs in to a tree.  Not sure that's necessary tho.
22:44 < skelterjohn> i don't want to manage it by hand.  i hate writing
build tools
22:44 < skelterjohn> the information is all there - it should figure it out
22:44 < nsf> skelterjohn: only for simple packages and executables
22:44 < nsf> take a look at runtime package in Go's std lib
22:45 < skelterjohn> it does ok with $GOROOT/src/pkg
22:45 < skelterjohn> anything with cgo gets complicated
22:45 < skelterjohn> that's why i never do anything with cgo
22:45 < uriel> is listing each package and its dependencies really so hard?
if you don't know what your packages depend on, you are probably making a big mess
22:45 < nsf> well, I have really good makefiles right now, but they are not
ready for mass usage yet
22:45 < vsmatck> It can get hard.
22:45 < nsf> and you still need to understand make very well
22:46 < nsf> and if you haven't written your makefiles then probably you
22:46 < nsf> :)
22:46 < uriel> nsf: I don't understand make very well, and have had no
22:46 < uriel> I just copy what the existing Go packages do
22:46 < vsmatck> That's one data point.
22:47 < nsf> uriel: do you have at least one app that uses multiple local
22:48 < nsf> because the way makefiles are now, it's hard to have that kind
of thing
22:48 < nsf> templates are supposed to be used in a recursive fashion
22:48 < uriel> I will admit that having multiple local packages can be a bit
akward, but goinstall mostly takes care of all that
22:48 < skelterjohn> how does goinstall do anything with local packages?
22:49 < nsf> well, I have system that solves it, but as I've said..  it
doesn't really work yet :)
22:49 < nsf> lately I've been too lazy to finish it
22:49 < uriel> hmmm...  this makes me think, why not make goinstall work
with local packages?
22:49 < uriel> although that would encourage non-portable code, probably a
stupid idea, nevermind
22:49 < nsf> ugh, I don't even use goinstall
22:50 < nsf> the worst tool I've ever seen
22:50 < nsf> :D
22:50 < skelterjohn> i think it's great
22:50 < nsf> it just downloads one Go package (can't work with cgo) and
installs it
22:50 < nsf> and that's it
22:50 < nsf> what is so special about it
22:50 < skelterjohn> it also downloads other goinstallable packages that it
depends on
22:50 < nsf> it doesn't even track anything
22:51 < skelterjohn> it's nice because for someone to use something i made,
i don't have to make a distribution
22:51 < skelterjohn> i just put it on googlecode and tell them what the path
22:51 < nsf> but it's easy to do, the problem is it doesn't work for a lot
of packages (which are cgo based)
22:51 < skelterjohn> sure
22:51 < vsmatck> I like how goinstall understands certain websites (like
github).  So instead of having to download a tar file I can just point goinstall
at the website.  It's simple but useful IMO.
22:52 < vsmatck> Makes it easy for me to keep packages up to date.
22:52 < nsf> right, you need to run 1 command instead of 3
22:52 < nsf> what a feature
22:52 < skelterjohn> yep
22:52 < vsmatck> 1 > 3
22:52 < skelterjohn> haha
22:52 < nsf> :)
22:53 < vsmatck> :)
22:53 < skelterjohn> also you can have it track versions by convention
22:53 < skelterjohn> gomatrix.googlecode.com/hg/1.0/matrix will never change
22:53 < skelterjohn> and w/out the 1.0, it's the latest commit
22:56 < nsf> still, I don't like goinstall and I won't fix it, because
lately I've been writing too much Go dev tools
22:56 < nsf> not mentioning that I don't know how to fix it :)
22:57 < nsf> it's sad that opengl is broken in go
22:57 < nsf> I'd like to do some graphics stuff :(
22:57 < uriel> sdl should work fine
22:57 < uriel> and exp/draw
22:57 < nsf> http://code.google.com/p/go/issues/detail?id=1262
22:57 < nsf> sdl works
22:58 < nsf> opengl doesn't
22:58 < skelterjohn> i feel like the whole plan9 thing makes it a lot harder
for go to work with existing stuff
22:58 < nsf> no, it's not plan9
22:58 < nsf> more like unusual execution model
22:58 < skelterjohn> that's what i mean, i just don't know the right way to
talk about it
22:58 < nsf> goroutines, not-compatible stack, calling conventions
22:58 < skelterjohn> the stack and calling conventions
23:02 < uriel> no, it is the Bell Labs/Unix(the real Unix) culture, that
clashes with the abomination it has turned into outside the Labs over the last 30
23:03 < uriel> the non-compatible stack thing is pretty much necessary if
you take as a core language feature that you don't have to worry about stack size
23:04 < uriel> the calling conventions is simply a result of the compilers
used, and I guess is more of a somewhat arbitrary divergence, other that it
originates from the compilers used
23:04 < nsf> I don't know about culture, but frankly I don't like current Go
implementation..  sadly it's the best implementation we have today, gccgo is not
an option because its compilation takes hours
23:04 < skelterjohn> well, i don't know enough about architecture to comment
23:04 < nsf> well, it's less than an hour, but still
23:04 < nsf> I'd like to see LLVM based Go implementation
23:05 < nsf> even though LLVM is C++, it's designed as a library so you
don't need to recompile it each time
23:05 < uriel> nsf: wtf is wrong with the current implementation?  works
pretty damned well for me
23:05 < uriel> llvm is awful
23:05 < uriel> it only looks good by comparison with gcc
23:05 < nsf> uriel: lack of dynamic linking support, undocumented hackerish
C code
23:05 < skelterjohn> haters gotta hate
23:06 < nsf> I don't like that
23:06 < uriel> nsf: what is 'hackerish' and 'undocumented'?
23:06 < uriel> Go's C code is some of the cleanest around
23:06 < nsf> I don't think so
23:07 < uriel> anyway, would be nice if somebody plugged the Go frontend for
gcc into llvm, but I certainly will have no use for it, it would be cool because
diversity of implementations is good
23:07 < nsf> it's all purely subjective anyway
23:07 < uriel> nsf: so, point me to a cleaner C codebase of comparable size
23:07 < skelterjohn> the "hackerish" quality may be intrinsic to C, rather
than the fault of the people writing it
23:08 < uriel> skelterjohn: exactly
23:08 < nsf> but what's the point of pointing you..  there is only one Go
compiler written in C that works :)
23:08 < nsf> and I will use it until there will be a better implementation
23:08 < nsf> but subjectively, I don't like it
23:08 < uriel> skelterjohn: and lets remember, the people writting it pretty
much invented C, ken at the very least was the first C programmer and heavily
influenced C in many ways
23:08 < uriel> skelterjohn: and he wrote most of the C in the Go
23:08 < skelterjohn> that doesn't mean it's easy to read
23:09 < skelterjohn> or not hackerish :)
23:09 < uriel> sure, but I do find it easier to read than most other C code
I have ever seen
23:09 < skelterjohn> i try not to read C code
23:09 < nsf> I had experience debugging Go's runtime
23:10 < uriel> as for 'hackerish', hell, the Lion's comentary on 6th ed Unix
shows plenty of 'hackerish' tricks, I find that is part of the beauty of it
23:10 < nsf> don't want to do it again
23:10 < nsf> although now it supports DWARF
23:10 < nsf> at least
23:10 < uriel> nsf: I'm not sure how the DWARF support really helps
debugging the runtime
23:11 < nsf> uriel: print statements work too
23:11 < uriel> and I have had to deal with bugs in gcc, and since then I try
to avoid touching gcc like the plague
23:11 < nsf> but in the case I had (race condition) the debugger is
23:11 < nsf> well, again..  by me
23:11 < nsf> I don't know about others :)
23:11 < uriel> yes, but what I mean is that when debugging the runtime, as
opposed to debugging Go code, the DRARF stuff shouldn't make any difference afaik
23:12 < uriel> as for race conditions, I'm not sure debuggers help much,
unless they are concurrency-aware debuggers
23:12 < nsf> but before that you simply wasn't able to debug Go's runtime
using gdb
23:12 < uriel> often running a program that is blowing up because of a race,
the race will go away if you run it under a debugger
23:12 < nsf> debugger is good at catchin memory corruptions
23:13 < uriel> I certainly hope something acid-like with
concurrency-oriented features for debugging will come up som day
23:13 < nsf> that was my case
23:13 < nsf> I had spent two days juggling print statements around
23:13 < nsf> :D
23:13 < nsf> the fix was one liner :D
23:13 < uriel> I might be ignorant, but I still can't quite see how a
debugger would hvae helped
23:14 < uriel> (or how gdb would have helped anyway)
23:14 < nsf> well, I don't know
23:14 < nsf> I haven't tried it
23:14 < nsf> but for some reason I thought it should be easier
23:15 < nsf> and I still think that way
23:15 < uriel> I have had to deal with such bugs in C code, and didn't find
gdb of much help, but then I barely know how to do more than printing a stack
trace or setting a breakpoint in gdb
23:15 < uriel> print based debugging is also much less disruptive of
timing-related bugs
23:15 < nsf> in gdb you can set up a breakpoint for write/read access to the
specific memory location
23:16 < nsf> it helps sometimes
23:16 < nsf> for example, one functions locks and does something with data
23:16 < nsf> and you need to make sure that it's the only thread that
accesses this data
23:17 < uriel> so you only access it from one goroutine ;P
23:17 < uriel> just (half) kidding
23:17 < nsf> in runtime it was the actual threads
23:17 < nsf> it was a scheduler bug
23:17 < nsf> were*
23:18 < uriel> well, hopefully those are not too common ^_^
23:18 < nsf> that one with OpenGL
23:18 < nsf> is one of them
23:18 < nsf> somewhere
23:20 < |Craig|> so I tried to make my 3 go file 2 package project build
with Makefiles, and I failed, it look like its not building my other package
before main and thus failing before building anything:
23:21 < |Craig|> any suggestions?
23:21 < |Craig|> I'm pretty clueless when it comes to make
23:21 < nsf> you totally don't know how to use make
23:21 < nsf> read manual :)
23:21 < |Craig|> nsf: exactly
23:21 < nsf> it is useful knowledge in a perspective
23:22 < nsf> anyway
23:22 < |Craig|> people said it was triviall and yo use the examples, so I
did, and I failed
23:22 < nsf> :)
23:22 < nsf> you've defined a var DIRS
23:22 < nsf> defining a VAR doesn't do anything else magically
23:23 < nsf> make is about defining rules and their dependency relationships
23:23 < |Craig|> the example trivial make file for go I was prointed to
included autogenerated files and was over 200 lines...  I got confused
23:24 < nsf> it defines the DIRS variable as well
23:24 < |Craig|> I need something like main: tcp right?
23:24 < nsf> but then it substitutes it to a rule dependency
23:24 < nsf> ugh
23:24 < nsf> really
23:24 < nsf> read the make manual
23:24 < nsf> http://www.gnu.org/software/make/manual/make.html
23:24 < nsf> it is long
23:24 < nsf> and boring
23:25 < |Craig|> thanks :)
23:26 < |Craig|> I'm just a bit confused why some people say make files for
go are trivial and do what pkg.Makefile does when that file is 200+ lines any my
shell script i'm using is 5 lines
23:26 < uriel> and it is gmake, which is evil
23:27 < nsf> all makefiles are trivial, because make is not the C++
23:27 < nsf> but if you don't understand how make works
23:27 < nsf> everything is non-trivial
23:27 < uriel> |Craig|: as far as I can see, you are not specifying any
23:27 < nsf> of course
23:28 < |Craig|> uriel: I'm aware of that
23:28 < uriel> I don't know how make works, (certainly not how gmake works)
23:28 < uriel> but copy-pasting from other projects is rather trivial
23:28 -!- tsykoduk [~tsykoduk@] has quit [Ping timeout: 245 seconds]
23:29 < nsf> copy-pasting without understanding
23:29 < nsf> is a waste of time
23:29 < |Craig|> make looks more confusing than go
23:30 < nsf> as uriel have noticed, gmake is indeed :)
23:30 < nsf> but still it is quite simple
23:30 < nsf> preprocessor-like language for defining rules, actions and
23:31 < uriel> |Craig|: make is certainly more confusing than Go
23:32 < |Craig|> at least make has a manual for me to read.  I'll be back
when I have learned it's basics, nsf thanks for the link and not being too mad at
my foolish attemts to do before understanding
23:32 < nsf> oh, it's not that easy to make me mad
23:32 < nsf> :D
23:33 < |Craig|> asking for help without reading the manual was pretty rude
of me though
23:33 < uriel> ok, here is a ridiculously trivial example of how to build a
multi-package Go program:
23:33 < nsf> and saying thanks to a guy who sent you to rtfm, well..  it's
very nice of you
23:33 < uriel> https://github.com/jackpal/Taipei-Torrent
23:33 < nsf> :)
23:33 < uriel> |Craig|: I don't expect anyone to read the gmake manual, I
actually hope nobody reads it
23:36 < nsf> I have different opinion
23:36 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has quit [Read error:
Connection reset by peer]
23:36 -!- noam [noam@IGLD-84-229-71-149.inter.net.il] has joined #go-nuts
23:36 < nsf> it's nice to know how the tool works
23:36 < nsf> especially if everyone uses it
23:37 < skelterjohn> always fun - looking up the deadline for a conference
you're working towards so you can make a schedule and seeing that the abstract was
due yesterday
23:37 < skelterjohn> i just submitted a "fantastic" abstract to ICAPS 2011
23:38 -!- skelterjohn [~jasmuth@lawn-gw.rutgers.edu] has quit [Quit: skelterjohn]
23:38 -!- XenoPhoenix [~Xeno@cpc5-aztw24-2-0-cust39.aztw.cable.virginmedia.com]
has joined #go-nuts
23:42 < |Craig|> uriel: that project just uses a bash script to run all the
make files.  Is that considered good design?
23:42 < nsf> no (imho)
23:43 < |Craig|> I could use a bash script to do everything, and it would be
as long as one that runs the make files
23:43 < nsf> I've said that to you earlier
23:43 < nsf> nsf | also using shell script is fine
23:44 < nsf> the problem with shell script
23:44 < nsf> it doesn't scale
23:44 < nsf> and it knows nothing about dependencies
23:44 < nsf> and it knows nothing about rules
23:44 < nsf> etc.
23:44 < nsf> make is mostly like shell, but adds an ability to describe
23:45 < |Craig|> I think gotest needed make files to be used easilly
23:45 < nsf> it uses makefiles
23:45 < nsf> I don't use gotest :)
23:45 < |Craig|> I see
23:47 < uriel> |Craig|: how many hundreds of packages you have or what?
23:48 < |Craig|> at the moment I have 2 trivial packages.  I started this
project this week
23:48 < uriel> so what the hell is the problem?
23:48 < uriel> just put the thing in a shell script if you like, and be done
with it
23:48 < nsf> |Craig|: I use shell scripts even for my C++ projects
23:49 < nsf> at least for few first weeks
23:49 < uriel> one Makefile per package, and a trivial shell script to run
them all
23:49 < uriel> it is ridiculously simple, the Makefiles are like five lines,
and the shell script like four
23:49 < nsf> then it grows and I add cmake or something
23:49 -!- watr_ [~watr@] has quit [Ping timeout: 260 seconds]
23:50 < |Craig|> I liked gotest, so it looked like I should have makefiles,
and I was lead to believe they were rather easy to learn and setup, but apparently
not so.  I think i'll stick with my shell script for now
23:50 -!- littlebobby [~bob@unaffiliated/littlebobby] has joined #go-nuts
23:50 < uriel> |Craig|: you should certainly have a Make file per package
23:50 < uriel> but again, that is trivial as you can see in any individual
package and their make file
23:51 < uriel> the only question is whatever having a project-wide make file
or shell script, but any project with <10 packages, a simple four line shell
script should be more than enough
23:58 < plexdev> http://is.gd/hvfM9 by [Ken Thompson] in go/src/cmd/gc/ --
more on dynamic hash in compound literals.
--- Log closed Sun Nov 21 00:00:17 2010