17:51:56 <donri> stepcut: i'm not sure i trust these ixset alternatives, so i'm thinking a type safer api on top of ixset should be easy. or are there other issues with ixset that warrant replacing it completely? otherwise, do you think we should break backwards compat and make the primary api type safer?
18:07:15 <srhb> Lemmih: Nice meeting you today. ;)
18:36:33 <KorriX> hello
18:37:07 <KorriX> i have compiled my webapplication and sent binary with object files to server without haskell
18:37:26 <KorriX> what i should install on server to have able start application ?
18:38:56 <donri> stepcut: http://hpaste.org/57688 is this because i'm using a binary release of ghc 7.4 perhaps?
18:39:37 <donri> although i'm using the i386 version ...
18:41:50 <KorriX> how to compile portable binary of happstack application ?
21:22:46 <Lemmih> srhb: Hi!
21:23:59 <srhb> Lemmih: Hi ;)
21:24:15 <Lemmih> srhb: Yeah, totally. You took me completely by surprise when you mentioned happstack.
21:24:35 <srhb> Lemmih: Sorry I tried to do the sales pitch on you. But in all honestly I had little chance to realize your part in those projects. ;)
21:24:39 <Lemmih> Glad I'm not the only Haskeller at DIKU.
21:24:41 <srhb> honesty*
21:24:58 <srhb> I'm afraid I'm a noobie. A very enthusiastic noobie, however!
21:25:05 <mightybyte> Hah, I had what sounds like a similar experience awhile back.
21:26:49 <srhb> mightybyte: What'd you do? ;)
21:27:51 <mightybyte> I met a guy in a bar and he was showing me a Haskell web app.
21:28:06 <mightybyte> (it was at a lisp meetup, so it wasn't completely out of the blue)
21:29:23 <mightybyte> So I asked the guy what libraries he used to build his app.
21:29:36 <mightybyte> And he says, "Have you ever heard of the Snap Framework?"
21:30:35 <srhb> Michael Snoyman? :P
21:30:45 <mightybyte> lol
21:31:24 <mightybyte> I'm like, "Well actually...I'm one of the authors."
21:31:29 <mightybyte> It was pretty funny.
21:31:37 <srhb> Haha. Definitely.
21:31:50 <donri> i've been recommended my own project once too
21:31:58 <srhb> Very similar. I was pitchin Happstack in a classroom. Ended up trying to sell it to Lemmih. Oh well! :-)
21:32:05 <mightybyte> NICE
21:32:23 <donri> but Lemmih doesn't do web dev does he
21:32:33 <Lemmih> Not if I can help it.
21:32:37 <donri> :)
21:32:39 <srhb> And acid-state..
21:32:43 <mightybyte> Just by proxy
21:32:45 <donri> ok then ;)
21:32:57 <srhb> Luckily most people didn't realize it was actually embarrasing. :-)
21:33:17 <donri> it's flattering for the "receiver" if anything
21:33:41 <srhb> :) I was surprised and a bit flustered. But he made a big mistake coming out like that - now I know who to pester. ;)
21:34:27 <mightybyte> Oh yeah, very flattering for the receiver.
21:34:27 <Lemmih> Major ego boost. Usually I have to travel far and wide to find people who have even heard of Haskell, much less actually used my software.
21:34:55 <srhb> I'm attempting to sell it to my group members. We'll see how it goes. ;)
21:36:33 <donri> i wonder if it's ever happened to say, steve jobs. "yea i usually buy mac instead of pc, you may have heard of it"
21:38:07 <srhb> Lemmih: By the way, you are _definitely_ not the only Haskeller at DIKU.
21:38:15 <srhb> Far from it.
21:40:30 <Lemmih> srhb: We should hang out and hack some code.
21:40:45 <Lemmih> Gotta go but I guess I'll see you in class.
21:40:49 <srhb> Lemmih: See you!
21:42:00 <donri> stepcut: looks like ghc 7.4 no likes new happstack-server, not on hackage either
21:45:00 <donri> ooh nice, beginnings of i18n stuff
21:45:19 <donri> looks like it builds fine on 7.0.4 though
21:47:29 <donri> "happstack-server-6.5.7 depends on sendfile-0.7.4 which failed to install." but on the other hand sendfile-0.7.4 has built fine on hackage and ghc 7.4
21:49:03 <donri> same error i had  Couldn't match expected type `Int32' with actual type `Int64'
22:08:26 <stepcut> donri: hmm
22:09:33 <stepcut> I have happstack-server/sendfile built with GHC 7.4rc1.. not sure if something changed, or if I have a patched version of sendfile and don't realize it
22:11:09 <stepcut> donri: as for IxSet. I think adding a more strongly type interface is both feasible and desirable. There are reasons to consider alternatives. IxSet does not do all that great if you want to change queries together (foo @> (Date lastWeek) @> (Score 50))
22:11:48 <stepcut> it will likely need to rebuilding the Score index after the Date query
22:13:12 <stepcut> IxSet is good if you mostly search on a single key per query (which turns out to often be the case). But compound keys or spatial searchs  (like searching for things within 5 miles of your house) do not work so well. Also, there are concerns about the space usage of IxSet -- though no one has have proven one way or another if there is a real problem or not
22:13:43 <stepcut> if you don't trust the alterantives -- that indicates a problem with their design and testing, which should be addressed
22:17:51 <donri> in deed, the alternative is to invest some time into making one of the alternative trustworthy
22:18:40 <donri> parts of the higgsset api seems weird or even ugly, but i might be missing the point
22:18:46 <donri> if not, perhaps those parts could be factored out
22:24:31 <donri> and otherwise i think i could whip up some template-haskell to hide those parts
22:58:57 <stepcut> I think for type safety, we just need and Index type class
22:59:12 <stepcut> class Index key a where ..
22:59:19 <stepcut> and then we can write things like:
22:59:49 <stepcut> (@=) :: (Index key a) => IxSet a -> key -> IxSet a
23:00:01 <stepcut> right now we have:
23:00:14 <stepcut> (@=) :: (Indexable a) => IxSet a -> key -> IxSet
23:00:19 <stepcut> where Indexable does not mention the key at all
23:07:41 <donri> i quite like how that is done in higgsset
23:08:03 <donri> the indices are all one type and then you have an associated type synonym in the Indexable class
23:08:30 <donri> but your idea might be more backwards compatible?
23:09:08 <stepcut> my idea is not backwards compatible
23:09:42 <stepcut> what is the advantage of the higgsset way? It allows for indexes that do not have a unique type?
23:09:51 <donri> no but might involve less incompatible changes
23:10:12 <donri> oh i don't know, i'm still a haskell newbie. just seemed elegant to me :D
23:10:17 <stepcut> :)
23:10:30 <stepcut> I do not have time to look at it until after 7 is released
23:10:50 <stepcut> I am halfway through splitting out the https code though.. which is the last major code change
23:11:08 <donri> cool