03:35:21 <dimday> hi, happstack newb here. I'm trying to run happstack-plugins to get auto-recompile working. I've seen the blog post from 2010 and tweaked it to get test to compile. But it's not working for me. Is it correct that I'm using runhaskell to do this?
03:35:48 <stepcut> hmm
03:36:02 <stepcut> I don't remember if runhaskell works or not
03:36:18 <dimday> stepcut: what would the normal way to do this?
03:36:25 <dimday> ghc --make ?
03:36:29 <stepcut> yeah
03:36:37 <stepcut> though, maybe plugins is broken again
03:37:00 <dimday> ah ok. I'll keep that in mind. it's basically a yak-shaving trip. :)
03:37:32 <stepcut> I have not tried using plugins in a while.. though i do want it to work :)
03:37:52 <stepcut> I've actually been working on a different part of the plugins stack the past few days
03:41:27 <dimday> stepcut: thanks for the heads up. it sounds great that it's still being worked on. I'm a refugee from ruby and just loved shotgun just removed the need for ctrl-c + up. :)
03:42:15 <stepcut> yeah, it's definitely on the radar, but could be higher :)
03:42:42 <stepcut> though, I believe the maintainer for plugins-auto is integrating the new cross platform file watching code, so that could be good :)
03:42:54 <stepcut> also, what do you mean by 'not working for you' ?
03:43:48 <dimday> well I've basically done the "hello, world" example and then changing hello to goodbye and then refreshing the browser. it didn't change.
03:43:59 <dimday> I've made sure I removed the Plugins.Static import just to be sure
03:43:59 <stepcut> can you paste the code on hpaste?
03:44:02 <dimday> ok
03:49:23 <dimday> ack. I found my error as I pasted it onto hpaste. apparently I didn't remove the #ifdefs from the correct file so it was still using the static library. anyway here it is. http://hpaste.org/76350
03:49:29 <stepcut> heh
03:49:31 <dimday> it works okay now
03:49:34 <dimday> using runhaskell
03:49:44 <dimday> thanks!
03:52:11 <dimday> will now continue wrapping my head around getting this and hamlet and webroutes all together working. onwards and upwards!
03:53:23 <stepcut> sweet!
03:53:27 <stepcut> let me know if you have questions!
06:43:43 <dimday> FYI. Okay I've just got hamlet, webroutes and happstack-plugins compiling and running. however the autoreload of the hamlets don't seem to be working. I'm guessing it has to do with the QQ and TH intricacies.
06:45:10 <dimday> if hamlet and hapstack-plugins dynamic is supposed to be able to work, I must've done something wrong. I'd be happy for any tips.
08:46:08 <Palmik> Hmm, when comparing the performance of data-store with Map and IntMap, I now compare it against Map k Int (or IntMap Int), I should probably compare it against Map k (Set Int) (or IntMap (Set Int)), since those can mimic the functionality of 1D data-store, right?
12:07:39 <donri> stepcut: http://www.yesodweb.com/blog/2012/10/measuring-warp acme-http benchmarks!
12:17:20 <HugoDaniel> !
12:23:22 <HugoDaniel> interesting
12:23:48 <HugoDaniel> -Nx makes scaling very bad
12:27:23 <donri> IIRC it's something like, -threaded uses select and only non -threaded is epoll, but there's work on fixing this
14:36:45 <alpounet> i love the 'recover' trick from the pipes-2.4 post
14:46:44 <Palmik> Hmm, did anyone here have problems installing syb-with-class (needed for ixset for example) (fresh, with cabal-dev).
15:29:35 <dcoutts_> donri: other way around
15:29:43 <dcoutts_> donri: -threaded uses epoll
15:29:50 <dcoutts_> non-threaded uses select
15:30:45 <dcoutts_> the -threaded one scales to lots of clients, but it's true that it doesn't scale well across multiple cores at the moment -Nx
15:34:04 <alpounet> dcoutts_, so johan and bos' IO manager is only used when -threaded ?
15:34:20 <dcoutts_> yep
15:34:39 <dcoutts_> the IO manager design relies on -threaded
15:35:00 <dcoutts_> there is no IO manager as such in the single-threaded rts
15:37:00 <donri> aha
15:39:45 <dcoutts_> I'm looking forward to the next article about making the IO manager scale better on multiple cores
15:40:18 <dcoutts_> they've done some work on stripeing the IO manager lock(s)
15:40:42 <alpounet> oh? i wasn't aware the IO manager was still being worked on
15:41:15 <dcoutts_> well, not as actively as when tibbe and bos were hacking it
15:41:31 <dcoutts_> we all know it doesn't scale, but exactly why is not clear
15:41:38 <dcoutts_> though there's lots of reasonable guesses
15:41:47 <dcoutts_> we lack the profiling tools to really be sure
15:42:42 <alpounet> like, knowing when one core is waiting on a lock or something?
15:43:25 <dcoutts_> though we're about to release some tools that can let us see linux perf events in the context of the ghc events system
15:43:43 <dcoutts_> alpounet: right, or even just MVars
15:43:55 <dcoutts_> but yes, OS level too
15:44:05 <dcoutts_> the OS level is what our new tools will hopefully address
15:45:22 <alpounet> ok, cool
15:45:25 <dcoutts_> profiling MVars is tricky, since they have no identity really
15:45:44 <dcoutts_> so hard to track, even by doing event logging from within the rts
15:47:03 <alpounet> dcoutts_, the traceMarker (or something like that) mentionned on #ghc a few days ago could help with that
15:47:08 <alpounet> i think?
15:47:33 <dcoutts_> you could do that already with traceEvent
15:47:38 <dcoutts_> no need for traceMarker
15:48:17 <dcoutts_> alpounet: yeah, maybe
15:48:53 <alpounet> dcoutts_, yeah but traceMarker puts that sweet pink line on the threadscope graph :P
15:49:07 <dcoutts_> traceEvent events show up too
15:49:10 <alpounet> i'm not sure of what traceEvent can do
15:49:15 <dcoutts_> but more suitable for high frequency
15:53:53 <alpounet> ok
18:24:39 <donri> stepcut: alright i have generic PathInfo working except... without using the names of constructors. i managed to use the names for toPathSegments, but haven't figured out how to parse them in fromPathSegments
18:24:50 <stepcut> ooo
18:25:22 <donri> so, Article 0 <-> /1/0, where the 1 means "second constructor"
18:26:03 <donri> Home <-> /0, where the 0 means "first constructor"
18:26:29 <donri> i might consult the generics ML about it
18:27:06 <donri> dreixel claims generic Read is possible, so
18:27:25 <donri> but maybe he meant depending on a generic Show... as in my current solution
18:28:58 <donri> http://hub.darcs.net/dag/web-routes/patch/20121016182552-6eb02 not sure if worth pulling just yet, but if you wanna look
18:33:20 <stepcut> did you look at web-routes-regular at all?
18:34:04 <donri> nope
18:34:45 <donri> oh hey that looks very similar to what i have
18:39:12 <donri> ah, undefined :: C ...
18:39:21 <donri> that might be the trick that escaped me
18:44:42 <stepcut> :)
18:45:04 <stepcut> someday I should revive web-routes-quasi
18:45:15 <stepcut> shouldn't be hard
19:04:13 <donri> stepcut: http://hub.darcs.net/dag/web-routes/patch/20121016190642-6eb02
19:04:34 <stepcut> nice
19:09:30 <donri> do you say "GHC 7.2 or {newer,better,later}"?
19:11:11 <stepcut> later
20:54:05 <donri> stepcut: http://hub.darcs.net/stepcut/web-routes/patches kinda nice hub feature :)
20:54:08 <donri> added docs
20:54:24 <stepcut> nice
20:54:33 <stepcut> i wonder what will happen if I press 'merge selected'
20:54:40 <stepcut> I'm going to do it!
20:54:55 <stepcut> Patches merged!
20:55:09 <stepcut> \o/, http://hub.darcs.net/stepcut/web-routes/changes
20:55:18 <stepcut> :-/
20:55:21 <donri> there is such a button? not for me ...
20:55:52 <stepcut> there is.. I pressed it and it said 'patches merged' but then nothing seems to have changed
20:56:01 <donri> in deed
20:56:02 <donri> sm!
21:05:49 <stepcut> I guess I could do it the old fashion way
21:05:54 <stepcut> unless you got more coming
21:06:55 <donri> don't think so!
21:13:31 <sm> hey what what
21:15:21 <donri> sm: apparently there's a "merge selected" button, and apparently it doesn't work?
21:15:28 <donri> stepcut: btw did you actually, you know, select anything?
21:15:58 <stepcut> no :)
21:16:06 <sm> heh
21:16:20 <sm> it should not give a success alert in that case, known issue
21:17:46 <sm> so, check a few boxes and it'll do it
22:12:13 <donri> stepcut: web-routes-quasi isn't dead, just not based on web-routes anymore. web-routes-qq maybe?
22:17:20 <donri> hm actually it's kinda dead, not updated since january and has outdated upper bounds
22:32:43 <stepkut> yeah