<lispy> I can’t even find the cabal trac <edwardk> in retrospect perhaps it would have been better had duncan and i looped him into the process earlier <lispy> Oh, the cabalinstall wiki is also down <lispy> Can we please just move this stuff to github or something if we’re not going to bring it back up? <lispy> Crucial haskell infrastructure and it’s rotting 
<byorgey> edwardk: so what’s the plan for moving forward with it? is there one? <byorgey> not that you are responsible for it, just wondering what your thoughts are <edwardk> byorgey: it basically stopped after that <mgsloan> This suggestion does not come with a detailed proposal, but I’ve been thinking that there needs to be more traditions for writing type-classes in a way that’s closer to having open functions. I realize this <mgsloan> There’s generally this problem with (+) :: a -> b -> c” being something you might want to define, but a more specific type is nicer for other situations. What if, in another module, you made a subclass of th<ddarius> elliott: People discover new ways of using type classes and parametric polymorphism and higher order functions after several decades of existence. <edwardk> well, the current situation is rather dysfunctional. the trick is trying to figure out a path forward that doesn’t leave us without a hackage maintainer <elliott> mgsloan: (+) really shouldn’t have the type a -> b -> c. <byorgey> edwardk: yeah <elliott> ddarius: well, indeed <lispy> edwardk: do we have a hackage maintainer right now? <elliott> ddarius: but extensions this rich don’t seem to come along that often <edwardk> ideally we’d be able to fund something for maintenance, etc. but the committee really doesn’t have enough money to pay anyone to do anything * cmccann proposes a vote appointing edwardk supreme overlord and granting him power to appoint a hackage maintainer by fiat <elliott> cmccann: what do you think his packages are for? <edwardk> lispy: as i understand it ross does it as needed <elliott> once you maintain over 50% of hackage you inherit the bloodline <cmccann> elliott, yes, just trying to speed things along <mgsloan> elliott: Or maybe AdditionOp1 a -> AdditionOp2 a -> AdditionResult a, bahaha * ddarius appoints cmccann Hackage maintainer. <elliott> all that needs to be done is for the current maintainer to die <ddarius> We’ll see how effective that is. <edwardk> right now i think heffalump is our duly elected supreme overlord <elliott> a bloody but effective method of administration <edwardk> elliott: well, thats one way for me to get control over transformers i suppose <edwardk> 
<cmccann> ddarius, ok. hm, I don’t think it’s working so far. <mgsloan> elliott: (+) not having that type isn’t really what I was trying to get at – this is a problem that afflicts all functions that have a potentially more polymorphic type <ddarius> elliott: The problem isn’t too many maintainers; it’s too few. Them dying doesn’t help. <elliott> mgsloan: I dunno, that sounds basically like bad typeclass design to me, if the type is too restrictive <elliott> Opening things up completely by convention seems wrong to me. <elliott> “create instances of the superclasses in terms of the subc” (got cut off there, btw) isn’t practical because typeclass instances can come from anywhere. <cmccann> edwardk, btw, where is the code for hackage2? kinda curious to see it. <elliott> ddarius: well, there is only ever one <elliott> the succession method is just a little complicated <edwardk> and what happens when i define a data type in one module, then define a comonad in one where comonads derive functor, and a monad in another where monad derives functor <elliott> if nobody controls over 50% of hackage, it defaults to spj <edwardk> cmccann: not sure honestly. i’d email gracenotes <elliott> who then appoints a new successor <edwardk> elliott: i’m sure he’d love that <edwardk> it’d be like after smerdyakov got banned <CodeWeaverX> The successor is alive, long live the successor? <edwardk> and freenode contacted him * elliott can’t even remember smerdyakov, let alone them being banned. <elliott> and I’m pretty sure I’ve been around long enough to <byorgey> hah, I hadn’t heard that part of the story <lispy> elliott: just visit #ocaml to get a taste * cmccann has no idea what this story is <elliott> (what did freenode contact them for?) <elliott> lispy: hmm, I think I’ll pass <lispy> He’s not a bad person, but he’s hard to communicate with. Especially if you’re a newbie. <byorgey> elliott: I thought edwardk meant that freenode contacted spj after smerdyakov was banned <edwardk> elliott: adam chlipala. prof over at mit now, at the time big into coq/ocaml (still is), scary smart guy, but able to answer almost any question in a way that makes you feel like an idiot <edwardk> elliott: yes <edwardk> er byorgey: i did <mgsloan> elliott – instances come from anywhere, sure, but there’s nothing really preventing you from doing something like $(mkSuperclass ”SuperAddition ”Num) <elliott> edwardk: oh, that’s smerdyakov? <edwardk> as i understand it because the simons are kind of the source of authority for haskell to be #haskell and not ##haskell or something, and he was raising a stink about being ‘denied his academic rights’ <elliott> ok, at least I know who they are now 
<mgsloan> which would get the info for superaddition – which functions it defines, and match them to a defintion out of Num <edwardk> elliott: yep <elliott> mgsloan: yes there is, because of the open world assumption <elliott> you can’t define a “default instance” like that <edwardk> anyways shapr banned him after a few attempts to socialize him didn’t go so well, it blew up in many ways <cmccann> edwardk, sounds hilarious <edwardk> for a while i had both him and shapr here in boston within a mile of each other. i kept waiting for one to get in the room with the other and for the gloves to come off, but i never did get them in the same s<mgsloan> elliott – I left something off of that – a concrete type <elliott> shame, I find (what I know of) his work interesting <mgsloan> ya, chipala style coq! <elliott> but I suppose “smart but hard to work with” applies to probably the majority of the most disruptive people in online communities <elliott> mgsloan: right, with a concrete type it’s easy enough <edwardk> elliott: part of it is just his online persona. he’s actually quite nice in person <tikhonjelvis> Largely because if they weren’t smart, they wouldn’t matter. <edwardk> but his style of humor doesn’t convey well over irc <edwardk> maybe i just need to send him an ircii script that randomly adds 
’s and ’s to his comments <cmccann> heh <lispy> heh <tikhonjelvis> The sad (happy?) thing is that that would probably work. <elliott> I just add to every completely serious thing I say. <edwardk> and deletes the random ‘but you should be using (ocaml|coq)’ asides <elliott> “i think you might be very confused ” <edwardk> tikhonjelvis: =) <elliott> “you should read a haskell tutorial or my brain is going to explode ” <mgsloan> I think there may still be the problem if you want Num to be a subclass of superaddition, though, because in order for the splice to reference the operation in the subclass, there needs to already be an insta<mgsloan> arggh <CodeWeaverX> I don’t think you’re being remotely serious. <Eduard_Munteanu> edwardk: I think I’ve seen him around recently, in a related channel <tikhonjelvis> Even when you know people are abusing the smilies, they still work 
<edwardk> Eduard_Munteanu: he’s active on ##logic, and a number of others * lispy is reminded of Tom Wait’s “big in japan” but with “logic” <lispy> I’m big ##logic <ion> I’m planning to murder everyone of you. <edwardk> i consciously went the opposite way of #haskell and didn’t ban him from there, which has been great for me, whenever i want to get my questions answered. perhaps not so great for the population count of the c<byorgey> haha <elliott> lol <Eduard_Munteanu> As long as you provide stimulating questions I guess * cmccann has a mental image of edwardk as a zookeeper with dangerous animals in cages <edwardk> considering it has about the same population it did when i started the channel ~3-4 years ago <lispy> I can only think of 2 bannings in my time here <lispy> keal and that one guy…what was his nick <elliott> we just need a variant of +q that lets certain people in the channel see things <ddarius> elliott: The Smerdyakov incident happened a -long- time ago, so I’d be surprised if you were there when it happen.d <tikhonjelvis> Ignoring the ranom racist forum advertisements? <lispy> (the emo guy) <elliott> edwardk gets his answers, channel population count is maintained <byorgey> dangerous animals that lay golden eggs, which edwardk carefully collects and distributes to the awestruck visitors <Eduard_Munteanu> lispy: wait, what? keal is a real person? <edwardk> it does peak in september though as people wind up stuck in philosophy classes, etc. <elliott> ddarius: fair enough, I thought I heard something about it being a couple of years ago <lispy> Eduard_Munteanu: well, there was a live keal. <ddarius> elliott: Where “a couple” is like 8. <cmccann> byorgey, yes exactly <mauke> preflex: seen smerdyakov <preflex> smerdyakov was last seen on ##logic 104 days, 8 hours, 44 minutes and 1 second ago, saying: Yes, having an expert outside the course staff check your answers should be considered cheating. <edwardk> thats my secret. really i don’t write anything. i just let others do the work and upload it to hackage behind their backs <Eduard_Munteanu> @keal <lambdabot> b*(Floor[v/b^p]/b-Floor[Floor[v/b^p]/b]) <elliott> ddarius: as is the definition of “couple” <mgsloan> was keal the #math one? <elliott> preflex: xseen protontorpedo <preflex> Sorry, I haven’t seen protontorpedo <byorgey> @remember edwardk thats my secret. really i don’t write anything. i just let others do the work and upload it to hackage behind their backs <lambdabot> It is stored. <lispy> Eduard_Munteanu: keal had this rant he kept revisiting about a computation he did that started to undo his reality <edwardk> =) * elliott wonders when protontorpedo was about <Eduard_Munteanu> Wha, heh… * cmccann recalls someone mentioning that keal had legit serious mental issues <Saizan> keal was kept for amusement i guess? <mauke> edwardk: reading your blog makes me depressed * Eduard_Munteanu kinda wants to meet the real vixen now <lispy> Saizan: yes <edwardk> mauke: ? <elliott> Eduard_Munteanu: Why not the real Nixon? <lispy> Eduard_Munteanu: heh <lispy> ?vixen are we still on for later tonight? <lambdabot> Any lady who is first lady likes being first lady. I don’t care what they say, they like it. <mauke> edwardk: it’s so much over my head <edwardk> don’t worry. i don’t understand it either. <edwardk> <Eduard_Munteanu> Mm, forgot about nixon. <elliott> occasionally i read one of edwardk’s posts and mostly understand it, and I worry that I’ve missed the point in an even more fundamental manner than expected <mauke> e.g. I think I’m using a codensity thing in code I pasted together but I don’t understand the code, ContT, Codensity, or Yoneda * ddarius huggles Yoneda. <edwardk> codensity is just Cont where you can’t callCC <mauke> “just” * cmccann takes a coinductive approach to understanding edwardk’s blog, every time I read an article again I get some finite amount of new knowledge <mauke> at least I think I know what’s wrong with my parsing lib now <cmccann> much easier than trying to understand the whole thing at once <edwardk> i really need to write up something on the polykinded category stuff <edwardk> mauke: ah found the parsing posts? <edwardk> if you look at the guts of trifecta i use a cps’d monad to make it work <mauke> class Stream c s | s -> c where { uncons :: s -> r -> (c -> s -> r) -> r } <mauke> edwardk: I’ve tried looking at the guts of trifecta <mauke> (f e -> n (a e’ -> eo (f a) (e <> e’)) ee (a e’ -> co (f a) (e <> e’)) ce) ee <mauke> then I ran away <edwardk> hahaha <elliott> the guts of trifecta, also known as the documentation <edwardk> Prim is a little scary =) <Eduard_Munteanu> ) <elliott> unparser :: (a -> ErrState e -> ErrLog e -> Bool -> Delta -> ByteString -> It Rope r) -> (ErrState e -> ErrLog e -> Bool -> Delta -> ByteString -> It Rope r) -> (a -> ErrState e -> ErrLog e -> Bool -> Delta -<edwardk> you can turn that inside out in your head <elliott> it’s scarier on the haddock page because all those identical functions are misaligned <ddarius> It’s just Church/Scott encoded for the most part. <edwardk> Prim is effectively the parser i put on my blog just cps’d <edwardk> elliott: if you read that as ErrLog e -> Bool -> Delta -> ByteString -> It (Result e a) — for some 4 constructor Result type it goes down a bit easier <ddarius> Either A B ~> (Either A B -> r) -> r ~> (A -> r, B -> r) -> r ~> (A -> r) -> (B -> r) -> r <cmccann> mauke, what’s that for, chewing through a stream lazily while being able to look at a “current” element using continuations? <edwardk> even easier if you tuple up ErrLog e, Bool, Delta, ByteString and make them into a state type <mgsloan> edwardk – I’m probably going to try to adopt trifecta’s text representation to be used as a buffer representation. Mainly, I’ve generalized the Highlight stuff to allow for other types of marks <mauke> cmccann: well, it’s supposed to abstract over lists <mauke> cmccann: so I can parse from e.g. arrays <edwardk> mgsloan: yeah i have that as well in a test branch <edwardk> mgsloan: the highlight stuff was mostly a ‘downpayment’ on the idea <mgsloan> no way! <mgsloan> well here’s mine: https://github.com/mgsloan/trifecta <edwardk> i also have a much simplified version of the diagnostics <ddarius> edwardk stole mgsloan code before mgsloan even wrote it. <mgsloan> hehehe <edwardk> that time machine does wonders for my productivity <edwardk> or at least the perception of it <mauke> edwardk: I had a look at your trifecta slides <edwardk> those are largely unrelated <mauke> is the utf-8 decoding code real? <edwardk> i just reused the name <edwardk> the monoid for it? <mauke> monoid? <edwardk> i have an old monoid for parallel utf-8 parsing <edwardk> http://hackage.haskell.org/packages/archive/monoids/0.1.36/doc/html/src/Data-Monoid-Lexical-UTF8-Decoder.html <edwardk> kinda horrid <mgsloan> one thing that I think would be very cool would be to have the Mark intervalmaps be stuck into the leafs of the rope, with local indices <edwardk> but it lets you take a char reducer and feed it bytes from either end, which it’ll break into utf8 character tokens <mauke> ah, that looks better <elliott> edwardk: that’s lovely <eyu100> I’m trying to make a 100x100x100 array (Data.Array), but it tells me I’m out of bounds <elliott> I have this weird bias towards monoid-based solutions. <edwardk> i actually consciously didn’t put the marks in the rope. <mgsloan> for reasons of simplicity? <edwardk> in general i find that annotated ropes work better as rope and annotation as separate structures for both performance and manipulation reasons <edwardk> performance, because most of the time the annotation doesn’t perfectly fit the ad hoc nature of how the rope was built in chunks <mauke> edwardk: I meant page 24 in http://comonad.com/reader/wp-content/uploads/2009/08/A-Parsing-Trifecta.pdf <edwardk> it may be a number of random locations along it <elliott> edwardk: the link in the header of that file is broken btw <mgsloan> I can buy that. It also potentially allows for multiple domains of annotation <edwardk> or it may be a value for every char, etc. <edwardk> yeah <ddarius> And it gives you a comonad … <edwardk> i have an OLD package lying around somewhere <edwardk> one sec. <edwardk> http://hackage.haskell.org/package/rope <edwardk> this used the separation concept. manipulating ropes as a comonad <elliott> july 2011 <elliott> SO OLD <edwardk> the bump was just to make it compile <elliott> oh <elliott> hmm, I wonder when hackage started actually gaining activity <edwardk> anyways my old c preprocessor used to work with annotated ropes <elliott> I never see packages before 2008 except iirc base and transformers <edwardk> 2007 or so <edwardk> shortly after i joined the community in 2006 <edwardk> because i was here just long enough to see the pain of the previous ecosystem <ddarius> edwardk: I didn’t think it was that old. It seems like it was just the other year when you were talking about them. <edwardk> and hackage showed up right as i was about to give up on this place <elliott> heh <edwardk> ddarius: i wrote the package and let the idea fester for a couple years <elliott> hmm <edwardk> there is an even older package as well <edwardk> not sure if thats up there <elliott> it’s weird because I remember Hackage being pretty active back when I got started <elliott> which was like 2008 <edwardk> ah, nope <elliott> I guess it took off quickly <edwardk> yeah <edwardk> people recognized a good thing <elliott> coinciding with haskell’s big boost in popularity I guess <elliott> what with RWH and everything <elliott> and dons :p <ddarius> Hackage and the whole LIP also seems like just the other year ago too. <elliott> LIP? <edwardk> hrmm, my old buffers package never made it to hackage <ion> “utf8 characters are at most 4 characters long†– /me cringes <elliott> ha <edwardk> ion: ouch <edwardk> i clearly meant bytes <ion> aye <edwardk> and i probably should have also said codepoints, etc. <ddarius> elliott: Library Infrastructure Project. <elliott> who remembers the old haddock style, and when hackage package pages used to look like the main site?! <elliott> that was WHOLE YEARS AGO! <edwardk> since character and glyph and codepoint are ambiguus <ddarius> elliott: The thing that defined Cabal and Hackage. <edwardk> er ambiguous <elliott> ddarius: Huh, I assumed Cabal was significantly older… but then I’ve seen pre-Cabal build systems, so I don’t know why I thought that. <elliott> It’s just so baked-in these days. <ddarius> Cabal is older than Hackage, but not by much. <edwardk> the old haddock stye wasn’t that long ago <elliott> yeah, it wasn’t <edwardk> elliott: meacham used to have one he supported, so did the guy behind happs <edwardk> (alternative build systems that is) <elliott> but every time I stumble on some ghc 6.12-era documentation from google i’m like “yow” <elliott> “how did we ever stand this style” <edwardk> then one of the snap guys with taste popped up <cmccann> even I remember the old haddock style <Eduard_Munteanu> The blue haddock? <cmccann> it wasn’t that long ago <Eduard_Munteanu> IIRC that’s pretty recent :/ <edwardk> its within the last 2 years, because snap hasn’t been around that long <elliott> it’s because so much has happened in the ghc 7 era, I guess <edwardk> ghc 7 has been an interesting rollercoaster of progress in some areas, and glacial in others <ddarius> http://www.haskell.org/ghc/docs/5.04.3/html/index.html <elliott> hmm, when was that circa? <ddarius> 2003 <elliott> wow, 2002 <edwardk> ddarius is old <elliott> hmm, http://www.haskell.org/ghc/docs/5.04.3/html/users_guide/license.html says 2002 <elliott> http://www.haskell.org/ghc/docs/0.29/users_guide/user_toc.html <elliott> wow, Concurrent and Parallel Haskell are that old? <elliott> Haskell 1.2 era? <ddarius> The whole point of GHC was parallel programming. <ddarius> GHC was(is?) part of the GRASP project. <edwardk> Concurrent and Parallel Haskell are Glasgow extensions to Haskell which let you structure your program as a group of independent `threads’. <elliott> http://www.haskell.org/ghc/docs/0.29/users_guide/user_74.html#SEC74 <elliott> “The constructor `Left’ is typically used for errors; it can be renamed to `Wrong’ on import.” <edwardk> i love the scare quotes <elliott> OK, who removed import renaming? <elliott> And where do they live? <CodeWeaverX> 223b Baker Street. <ddarius> I believe it happened between 1.3 and 1.4. <elliott> edwardk: If you can call that a `program’. <ddarius> Maybe between 1.4 and 98. <mauke> how do I programmed parallels
Feb 162012