00:08:25 so anyway 00:08:35 does anyone know why gitit always uses 1% of cpu idling? 00:08:41 seems likely a happstack 00:08:43 thing 00:10:32 mmm 00:10:40 huh, indeed. 00:10:44 it's irritating - clutters up my top 00:10:51 that's a bit odd. 00:11:10 can you write a report, say that i confirmed it, so someone hunts down what is going on. 00:11:23 i'm seeing 2% cpu. so something is waking up every second or such like. 00:11:26 of course I confirmed it, I would hardly be complaining otherwise 00:11:35 right, but it is good to have other users. 00:11:45 you use any plugins? 00:11:47 nope 00:11:56 vanilla gitit 00:12:16 ACTION is a little surprised, but eliminates one possible source of variance 00:14:43 wonder if we could check with another app 00:16:53 @hoogle listenON 00:16:53 Network listenOn :: PortID -> IO Socket 00:17:09 suspicious looking. wonder if it just loops 00:24:30 well, presumably at the base gitit/happstack sits in an accept loop 00:24:47 sleeping the main process until the kernel says data is ready on a watched socket 00:26:52 listen (MkSocket s _family _stype _protocol socketStatus) backlog = do modifyMVar_ socketStatus $ λ status → do if status ≠ Bound then ioError (userError ("listen: can't peform listen on socket in status " ++ s 00:26:57 how status)) else throwSocketErrorIfMinus1Retry "listen" (c_listen s (fromIntegral backlog)) >> return Listening 00:27:15 wonder what modifyMVar_ does, efficiency wise 00:27:23 this is all the way down in Network.Socket 00:36:15 dons: but if the process is really sleeping in modifyMVar, why would we see it in top? 00:43:54 dons: alright, I'm just emailing the gitit newgroup 00:43:58 if you have any input 00:44:27 it won't show up. 00:44:33 ? 00:44:34 i bet it is other threads in the system doing something 00:44:47 checking modify times on files or some such 00:44:51 rather than watching sockets 00:48:09 well, it could only be 'wiki' or 'dir' 00:49:12 wisince the final line of main is 00:49:13 simpleHTTP serverConf $ msum [ wiki conf' 00:49:16 , dir "_reloadTemplates" reloadTemplates 00:49:20 ] 00:50:18 but no, simpleHTTP' doesn't run the second argument until a response hits 00:50:34 gwern: hey, maybe try threadscope on gitit? 00:50:42 how does threadscope work? 00:50:47 you'll learn a new tool, and prob. get a good idea of what is running when gitit is supposed to be idle. 00:51:06 it attaches to the event log , you get the event log from the running app by launching it with a flag. 00:51:12 oh. might need 6.12... 00:51:18 profiling perhaps then. 00:53:43 doesn't go too hot with plugins 00:54:07 let's see how threadscope goes 00:55:40 you'll need to compile it all with ghc 6.12. is that possible with happstack? 00:56:00 why with 6.12? 00:56:08 threadscope seems to compile 00:56:50 to get event logging from the runtime. 00:57:02 oh 00:57:02 you need gitit running on the 6.12 runtime 00:57:16 that's even worse than recompiling gitit and editing my setup to run profiled :) 00:57:20 yes 00:57:24 (don't want to upgrade just yet) 00:57:26 but you get better info about threads. 00:58:33 and spend hours dealing with errors and issues in various places 01:03:04 right. 03:34:36 anybody know how to make HSX generate sane XML indentation? 03:43:26 quuxman: you just need to realize that it is already sane 03:44:31 somebody: well, actually I didn't mean XML. I'm using html4 03:44:32 whitespace in the cdata of XML is (often) significant (aka, meaningful), so you can't just stick random whitespace between the tags.. 03:44:43 somebody: yeah I know that's why it looks like that 03:44:50 somebody: but there's a flag to turn that on and off 03:44:55 in the XML spec 03:45:01 or at least the XHTML spec 03:45:36 but even so, it doesn't indent properly. It indents siblings 03:45:47 well, for the

 and  tags I don't think you can turn it off
03:46:06 <somebody> but if you want to change the formating you need to make an alternative version of renderAsHtml
03:46:18 <somebody> which is found in HSP.HTML I believe (in the hsp package)
03:47:13 <stepcut> there is definitely no way to turn it off right now. Another option would be to add a filter which runs the output through html tidy
03:48:46 <stepcut> why do you need it in a more human readable format?
03:49:55 <stepcut> the easiest workaround right now might be to use something like firebug or google chrome that has a view to show you the DOM...
10:31:13 <paulvisschers> I've been trying to install happstack-data (both 0.4.1 from hackage and 0.4.3 from darcs)
10:31:26 <paulvisschers> ghc: dist/build/Happstack/Data/Default.o: unknown symbol `sybzmwithzmclasszm0zi6zi1_DataziGenericsziSYBziWithClassziInstances_constrZMad6dZN_closure'
10:31:41 <paulvisschers> is the error I get, how do I install it?
10:44:07 <Muad_Dibber> I wouldn't know paulvisschers, as I'm new to happstack myself, and when I last dove into haskell, Cabal/hackage were almost non-existant and darcs was just coming up;
10:44:15 <Muad_Dibber> but I noticed there were more people with the same problem:
10:44:17 <Muad_Dibber> http://groups.google.com/group/happs/browse_thread/thread/c66c74294d8eabf
10:46:25 <paulvisschers> Muad_Dibber: I'll check it out, thanks
10:51:55 <paulvisschers> cabal install -O0 happstack-data (this does work, with thanks to Erik Hesselink)
19:29:27 <stepcut> I think the constant gitit load is due to idle time garbage collection
22:06:17 <mrsolo> so i am running typical issue with RoR and is in active process of switching framework
22:06:37 <mrsolo> is there some unbiased assessment of happstack that i can read up somewhere?
22:06:57 <mrsolo> all i have found so far is the performance comparison between it and ror, erlyweb etc..
22:07:33 <mightybyte> mrsolo: What criteria would you like assessed?
22:07:47 <mrsolo> reliability and performance
22:08:05 <mrsolo> and framework's near term future
22:08:23 <aavogt> the latter is quite intangible, no?
22:08:59 <mrsolo> yes but people whom have worked with happstack will have better vibes about it
22:09:17 <aavogt> they will also be biased :)
22:09:37 <mrsolo> well right but i will take that too over none :-)
22:09:52 <mightybyte> Recently there has been a decent bit of activity.
22:11:45 <mightybyte> I think that will probably continue at a similar pace, but the mass of users isn't large enough to insulate against a core developer's efforts going in a different direction.
22:12:00 <mrsolo> no
22:12:05 <mrsolo> but it will make it rather difficult
22:12:11 <mightybyte> Although it would seem that stepcut has a fairly significant investment in happstack.
22:12:14 <mrsolo> user revolt is a powerful thing
22:12:16 <stepcut> :)
22:12:22 <mightybyte> As do I
22:14:31 <stepcut> right now I am mostly focused on refining the existing experience. Better documentation, serialization instances for Data.Text, smoother integration with HSP, etc
22:15:37 <stepcut> it would be nice if the WAI thing works out and happstack can be used more easily with hyena, etc
22:15:51 <mightybyte> I think of happstack as two conceptually separate parts.  A web server + associate haskel interfaces and a persistent state management system + IxSet.
22:16:01 <stepcut> that is three parts..
22:16:21 <mightybyte> Yeah, but IxSet is kind of a side note there...and related to the state management.
22:17:08 <stepcut> the 'web server' and 'associated haskell interfaces' should really be split apart.. more specifically,
22:17:54 <stepcut> the things like, RqData, dir, path, ServerMonad, FilterMonad, etc, should be separate from the lazy I/O web server that actually listens on the Socket and consumes/produces Repsonses/Requests
22:18:30 <mightybyte> Yeah, but all that comes in happstack-server
22:18:54 <stepcut> right, for now.. I would like to fix that..
22:18:57 <mrsolo> well i am start at ground 0 with happstack
22:19:43 <mrsolo> but it doesn't seem to have something simliar to ror scafold does it?
22:19:51 <stepcut> mrsolo: the current code seems to work quite reliably. The documentation could be greatly improved (though it is much better than it used to be)
22:20:20 <stepcut> mrsolo: not really. mae started something like that, but then his wife had a baby :)
22:21:14 <mrsolo> sure you gain some you miss some, such is the way with frameworks
22:22:19 <stepcut> it is also not clear to me yet, what it would mean for happstack to have scaffolding
22:23:16 <mightybyte> I would asses the server part of happstack as average.
22:24:00 <mightybyte> Decent reliability, but has been rumored to have some occasional stability problems/space leaks.
22:24:29 <mrsolo> oh? ouch
22:24:34 <mightybyte> Decent usability, albeit poor documentation and not much really noteworthy.
22:24:50 <mightybyte> mrsolo: I run a production site on happstack and it's uptime is perfectly fine.
22:24:52 <mrsolo> or steeper learning curse
22:25:13 <mightybyte> mrsolo: However, I run it inside of a while true bash script loop, so it gets restarted automatically if anything goes wrong.
22:25:25 <mrsolo> ah
22:25:32 <mrsolo> how often does that kick it?
22:25:48 <mightybyte> I haven't gathered statistics on restarts, but it gives me a reasonable uptime
22:26:38 <mrsolo> apple to orange comparesion, our RoR server has random thread timeout frequently very 20 hours
22:26:40 <mrsolo> 24 hours
22:26:48 <mightybyte> The state management system is what I consider to be happstack's significant contribution.
22:26:49 <mrsolo> usually when cpu is stressed
22:27:07 <stepcut> mrsolo: how big is your database?
22:27:16 <mrsolo> very small
22:27:20 <mrsolo> hmm 5-6 tables
22:27:29 <mrsolo> joine two tables at most
22:27:39 <mrsolo> biggest table is less than 200k rows that is my guess
22:27:44 <stepcut> how much data in MB ?
22:27:48 <stepcut> (or GB?)
22:27:51 <mrsolo> hmm
22:28:03 <mrsolo> 2gb to 4gb off top of my head
22:28:12 <stepcut> how many requests per second do you want to be able to handle?
22:28:36 <mrsolo> the bar is pretty low
22:28:50 <stepcut> less than 1?
22:28:56 <mrsolo> we do hmm 60 per min at peak
22:29:18 <mrsolo> less than 100 per hour normall i think
22:29:22 <stepcut> how much growth do you expect?
22:29:35 <mrsolo> hmm at most 3x in 12 months
22:29:39 <mightybyte> If your operations are database-intensive, then I think happstack will likely be a significant performance win for you if your server has enough RAM
22:29:52 <mrsolo> the bottleneck is computation
22:29:58 <mrsolo> right now
22:30:01 <mrsolo> yes
22:30:05 <stepcut> why so much CPU for so few requests?
22:30:28 <mrsolo> some number are crunched on the fly
22:30:36 <stepcut> floating point?
22:30:41 <mrsolo> just integer
22:30:48 <stepcut> hmm
22:30:51 <mrsolo> but the formula is pretty simple
22:31:04 <mrsolo> so my guess is basically ruby is just too slow :-)
22:31:13 <stepcut> yeah
22:31:33 <stepcut> with happstack you would get native code gen, and the ability to optimize those number crunching algorithms if required
22:31:52 <mrsolo> that i am not worry about happstack of course
22:32:05 <stepcut> the happstack platform itself is quite stable in my opinion/experience.
22:32:21 <mightybyte> I agree
22:32:27 <stepcut> however, since you are just using it as a library to build your own server, there is nothing to stop you from adding unstable parts :)
22:32:58 <mrsolo> i am more about for happstack is 1) ability for putting up front end for rapid development 2) documentations 3) glue to javascript frameworks etc
22:33:37 <mrsolo> have something similiar to capistrano will be nice as well
22:33:40 <stepcut> documentation is the weakest point right now, but an important goal going forward
22:33:49 <aavogt> what of the HSP stuff?
22:34:48 <stepcut> mrsolo: I use HSP for generating html+js 'templates'
22:36:08 <stepcut> mrsolo: what do you want in the way of gule for javascript? There are some (non-happstack specific) JSON libraries that you can use..
22:36:27 <stepcut> I am using jquery for my stuff at the moment
22:36:47 <mrsolo> i am using some minor javascript library and paying the price right now
22:36:53 <mrsolo> i like to switch to standard one
22:36:54 <mrsolo> jquery for one
22:37:17 <mrsolo> i know there are few others but jquery seems to be the biggest?
22:37:43 <mrsolo> but i don't have anything preset in mind right now
22:37:46 <stepcut> happstack is a rather open platform. For example, there are 3 different libraries commonly in use for generating HTML with happstack: Text.XHtml, HSP, and HStringTemplate. All three of those libraries are not part of happstack itself..
22:39:59 <mrsolo> is that plus or minus?
22:40:07 <stepcut> a plus
22:40:17 <stepcut> it means you are not locked into using a solution that is not right for you
22:41:47 <mrsolo> but i have to know all three to decide more or less
22:42:15 <stepcut> right. That is why you need good documentation which provides you with a solid starting foundation
22:42:42 <stepcut> and picks reasonable 'defaults' for you
22:42:59 <stepcut> but that is largely a documentation issue (though a very important one)
22:44:08 <stepcut> There wrote a small, but fairly featureful applications which I think covers a lot of things that people are going to want to do with happstack..
22:44:30 <stepcut> alas, it requires some patches and features that are not available in the mainline happstack yet
22:44:45 <stepcut> so I am working on getting those features into the mainline (for example, a Data instance for Data.Text)
22:44:52 <mrsolo> okay
22:45:21 <stepcut> once it builds against mainline I think we will be in a good position to start documenting a good method for building apps
22:45:41 <mrsolo> is happstack develpment community very small?
22:45:45 <stepcut> or at least, documenting the way *i* like to build apps :)
22:46:31 <stepcut> mrsolo: a lot of people follow the happstack project, but a much smaller number contribute -- which I think is true of most projects. happstack is certainly smaller than something like rails :)
22:47:00 <stepcut> I have talked to a lot of interested users, and it seems that documentation is by far the #1 obstacle to wider adoption at this point
22:47:23 <mightybyte> I think happstack is probably the biggest haskell web framework.
22:47:33 <stepcut> mightybyte: yeah, that is probably true
22:47:34 <aavogt> stepcut: while you're after missing instances, how about looking into Data instances for UTCTime (from Data.Time)?
22:47:36 <mightybyte> ...in terms of development community.
22:50:24 <stepcut> aavogt: that *may* be fixed in darcs head for time, but I can not find the time repo anywhere
22:51:42 <stepcut> aavogt: you might be better off using EpochMilli though for now. I am going to add some additional functions to happstack-util to convert EpochMilli to UTCTime, etc.
22:52:19 <aavogt> which brings us back to the need for documentation :)
22:52:45 <stepcut> aavogt: yes. I deal with this in the demo app, but until that builds against head and is documented, it does not help much :)
22:53:07 <mrsolo> stepcut, mintybte and all thanks a lot for the information. :-)
22:53:17 <stepcut> aavogt: but I submitted the darcs patch for Data for Data.Text.Text today, so I can move onto the next thing now hopefully
22:53:17 <sm> happstack has contributions from 24 developers
22:53:50 <stepcut> sm: is that happstack only? or happstack+happs?
22:53:59 <sm> the latter
22:54:27 <stepcut> sm: how did you calculate that?
22:54:38 <sm> darcs show authors | wc -l
22:54:51 <aavogt> any estimate how the work they did is distributed?
22:55:16 <aavogt> as in, are 3 of those core developers, and everybody else does something very minor?
22:55:19 <stepcut> sm: I wonder.. I think mae tossed the old history when he created the happstack repo, so that might only be happstack contributors?
22:55:27 <sm> oh, that's true
22:55:45 <sm> this is only since jan 2009
22:55:48 <sm> happstack only
22:55:55 <sm> I figure the number for happs was smaller
22:56:10 <stepcut> sm: yeah, and some of them contributed to happstack
22:56:33 <stepcut> sm: I think shapr and igloo contributed to happs but not happstack (yet)
22:57:29 <sm> aavogt: here's the detail: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=16888#a16888
22:57:54 <stepcut> sm: is that by number of commits?
22:58:00 <sm> yes
22:58:28 <stepcut> I'm listed twice :p
22:58:47 <stepcut> a few people are actually
22:58:49 <sm> yes, that can be cleaned up with a darcs .authorspellings file
22:58:54 <sm> then it will be more accurate
22:59:04 <sm> also, let's see.. http://joyful.com/repos/happstack/commits.png shows commit activity
23:00:04 <stepcut> needs more commits!
23:01:09 <sm> yes indeed
23:01:32 <stepcut> well, we have 1 commit for today so far!
23:01:39 <sm> nice!
23:01:44 <stepcut> actually, 2 commits
23:01:56 <sm> stepcut: want to tell me the issues in that author list, and I'll commit .authorspellings ?
23:02:15 <stepcut> sm: sure, jeremy@n-heptane.com and jeremy@seereason.com are both me
23:02:34 <stepcut> and 19 and 21 are the same person as well I would guess
23:02:39 <sm> ok.. so Full Name <primary@email> for you is.. Jeremy Shar <..seereason..> right ?
23:03:17 <stepcut> yeah
23:08:55 <sm> http://hpaste.org/fastcgi/hpaste.fcgi/view?id=16888#a16889 .. any other obvious full names I should add ?
23:11:26 <stepcut> simon.meier perhaps?
23:11:42 <sm> indeed, good idea
23:13:37 <quuxman> ok, I'm done with XML
23:13:53 <quuxman> writing XML docs by hand is absurd
23:16:26 <sm> ok, patch sent (to somewhere strange.. sseses@gmail.com)
23:17:34 <sm> after having to use some other vcs's lately I love the ease of contributing a patch with darcs, I must say
23:22:12 <stepcut> quuxman: XML is absurd no matter how you do it
23:22:22 <stepcut> sm: that is mae's email.. better off sending it to me..
23:22:38 <quuxman> what is the web standards community smoking?
23:22:45 <sm> ah ok.. maybe we could get that default send addr changed ?
23:22:53 <stepcut> sm: yeah, I wonder how?
23:22:58 <sm> I'll check
23:23:18 <stepcut> the default should be to send it to the mailing list I think..
23:23:36 <stepcut> maybe
23:23:38 <stepcut> or to me
23:24:17 <stepcut> sending to the mailing list is tricky, because only subscribers can do that.. don't want to have patches go missing from non-subscribers
23:25:33 <sm> The _darcs/prefs/email file is used to provide the e-mail address for your repository that others will use when they darcs send a patch back to you. The contents of the file should simply be an e-mail address.
23:25:36 <sm> right
23:26:52 <sm> stepcut: so someone with access needs to set that file in the trunk repo which I guess is still http://patch-tag.com/r/mae/happstack
23:27:55 <sm> someone wanting to join that repo is advised to contact mae, also.. should yours become the trunk instead ?
23:28:47 <sm> oh, I thought I saw a http://patch-tag.com/r/stepcut/happstack .. nm
23:28:48 <stepcut> sm: do I have to do that directory on the files on patch-tag.com ?
23:29:27 <sm> stepcut: yes I believe so, a patch can't do it
23:29:47 <stepcut> k
23:31:10 <sm> http://patch-tag.com/r/mae/happstack/darcscommands also recommends mae's address.. so the odd patch might still go direct to him
23:31:20 <sm> just fyi
23:32:15 <sm> gosh.. mightybyte what's this about snap framework