November 2016 - Haskell-Cafe

Am I misunderstanding something about `StableName` and `hashStableName`?
by raichoo 20 Nov '16

20 Nov '16

Hi, I've been playing around with `System.Mem.StableName` and `hashStableName` and stumbled across a behavior that seems to be rather weird. First of all the documentation of `hashStableName` says the following: The Int returned is not necessarily unique; several StableNames may map to the same Int (in practice however, the chances of this are small, so the result of hashStableName makes a good hash key). OK, sounds fine, let's put this to the test. So I wrote a little program. module Main where import Control.Monad import System.Mem.StableName import System.Environment main :: IO () main = do args <- getArgs res <- forM [0..10000] $ \i -> do x <- makeStableName i let h' = hashStableName x unless (null args) $ print h' return h' putStrLn "---------------------------" print $ minimum res print $ maximum res Running this program without an argument only prints aus the minimum and maximum hash values that where produced. So surprises here: $ ./Test --------------------------- 1 10000 Here is how things get a bit weird when I make the progam print out the hash values while they are being produced I get a completely different result. Not only that but I get a whole lot of hash collisions, since the hashes produced only seem to be within the range of 1 and 260. $ ./Test print [...] 128 127 126 125 124 123 122 121 120 119 118 117 116 115 114 --------------------------- 1 260 Am I missing something fundamental here, because this behavior seems to be rather confusing. Kind regards, raichoo

3 2

Esqueleto maintainership
by Christopher Allen 20 Nov '16

20 Nov '16

Cross-posting my intent to take over maintainership of Esqueleto. Can read more here: https://github.com/haskell-infra/hackage-trustees/issues/84 This has gone on for awhile, but it's beginning to eat into maintenance of my work projects and it seems like there are others eager to contribute and help share the load where there wasn't 6 or 12 months ago. Cheers, Chris Allen

3 3

difficulty writing GEq instance
by Anders Papitto 20 Nov '16

20 Nov '16

Hello! I'm having a lot of trouble writing a Data.GADT.Compare.GEq instance - can anyone help me fill in the blank? For context - I'm generating Tag types automatically for using with DSum, and I need a GEq instance. That's part of an attempt to add efficient rendering of sum types to Reflex/Reflex-Dom (https://github.com/anderspapitto/reflex-sumtype- render/blob/master/src/ReflexHelpers.hs). I put this question on stackoverflow as well a day ago ( http://stackoverflow.com/questions/40698207/how-can-i- write-this-geq-instance). Here's the code (it's a full, standalone file - you can copy it into Foo.hs and run ghc to see the full error I'm facing). The error I get is that when I try to recursively call geq on the unwrapped x and y, I can't because ghc considers them to have different types - Quux a and Quux b. However, the whole point of why I'm trying to call geq is to see if a and b are the same, so I'm pretty confused. Note that I'm making use of the generics-sop library, which is where NP and NS and I come from. {-# LANGUAGE GADTs #-} {-# LANGUAGE DataKinds #-} {-# LANGUAGE TypeOperators #-} {-# LANGUAGE KindSignatures #-} {-# LANGUAGE RankNTypes #-} module Foo where import Data.GADT.Compare import Generics.SOP import qualified GHC.Generics as GHC data Quux i xs where Quux :: Quux (NP I xs) xs newtype GTag t i = GTag { unTag :: NS (Quux i) (Code t) } instance GEq (GTag t) where -- I don't know how to do this geq (GTag (S x)) (GTag (S y)) = let _ = x `geq` y in undefined

2 2

How to profile haskell web applications?
by Moritz Angermann 18 Nov '16

18 Nov '16

Hi, I’m looking to (perf, heap) profile a haskell web application. With ekg and the ekg wai middleware I can get some data on the rts, gc, and latency. What’s the approach towards performance (-p) and heap (-h) profiling for web applications? Cheers, moritz

1 0

GHC plugin: custom wired-in package, module, and type class
by J. Stutterheim 18 Nov '16

18 Nov '16

Hi all, I'm looking for a way to add my own wired-in package/module/type class for use in a core-to-core plugin. My first challenge is to identify the type class that I want to find. I have the following code: myPkgUnitId :: UnitId myPkgUnitId = fsToUnitId (fsLit "mypackage") mkMyModule :: FastString -> Module mkMyModule m = mkModule myPkgUnitId (mkModuleNameFS m) myModule :: Module myModule = mkMyModule (fsLit "My.Module") myClassName :: Name myClassName = clsQual myModule (fsLit "MyClass") myClassKey myClassKey :: Unique -- How do I get this? Here the missing piece is the myClassKey. Looking at how this is done for, e.g., the Monad class: monadClassKey :: Unique monadClassKey = mkPreludeClassUnique 8 we see that it has a hard-coded unique identifier. Is there a way to do the same for my own type class? Or to somehow find the Unique key dynamically? The second challenge is more or less the same in that I want to identify a member of MyClass: myFunMName :: Name myFunMName = varQual myModule (fsLit "myFun") myFunMClassOpKey myFunMClassOpKey :: Unique -- How do I get this? Again, the identifier is hard-coded for, e.g., ">>=". Any insights will be much appreciated. Best regards, Jurriën

2 1

4.4.2 Fixity Declarations - and what happens when they are missing
by Johannes Waldmann 18 Nov '16

18 Nov '16

Yeah, I know, it's bikeshedding, some 30 years after the fact, but - How on Earth was this allowed to happen? "Any operator lacking a fixity declaration is assumed to be infixl 9" https://www.haskell.org/onlinereport/haskell2010/haskellch4.html#x10-820061 If I don't specify associativity or precedence for an operator, then the parser should simply reject any expression whose meaning would depend on such declarations. So I'd have to write parentheses at each (ambiguous) use site - or write a fixity declaration. Both would serve to document the code. The current situation feels like "the type of identifiers without type declaration is int". Short of repairing the standard - I'd welcome GHC to print a huge warning for each definition and use (!) of "operator without fixity decl". - J. PS: to show that I did at least some research * there's no mention of this in 4.2 "Infix operators" in http://research.microsoft.com/~simonpj/papers/history-of-haskell/history.pdf * Scala also has default rules (6.12.3) "The precedence of an infix operator is determined by the operator’s first character." "Operators ending in a colon ‘ : ’ are right-associative. All other operators are left-associative." and why do I care at all? This: https://github.com/jwaldmann/haskell-obdd/issues/1

1 0

Preventing dereferencing an STRef after taking an action
by Timotej Tomandl 17 Nov '16

17 Nov '16

Hi, is it possible to prevent an STRef passed to one ST action from being passed further on to another ST action? I am trying to prevent these usages: invalid :: STRef s a -> STRef s b -> ST s () invalid x y=action1 x y >>=action2 x y invalid1 :: STRef s a -> STRef s b -> ST s () invalid1 x y=action1 x y >>=action2 y while still being able to do this: valid :: STRef s a -> STRef s b -> ST s () valid x y=action1 x y >>=action2 x where action1 will modify x and y inplace, but in the end only x will hold the useful information. In other words I am trying to make it impossible for any action after action1 to have a same argument as action1's second argument. timo

1 0

Re: [Haskell-cafe] commenting out a line of code
by Bernhard Herzog 17 Nov '16

17 Nov '16

On 15.11.2016, David Turner wrote: > The outer do block should be desugared first, and the `putStrLn "done"` has > the same indent as the `if` so it gets a semicolon, in that block, meaning > it looks like this: > > main = do > { if True > then do > putStrLn "A" > else do > -- putStrLn "B" > ; putStrLn "done" > } That's what one would expect. Except that GHC treats the 'putStrLn "done"' as part of the do-block in the else branch by default. This is due to the NondecreasingIndentation extension which is enabled by default (see https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/bugs.html#c…) With the option -XNoNondecreasingIndentation (or -XHaskell2010) GHC does produce an error message because of the empty do-block. Bernhard

2 1

[ANN] Ginger: Jinja-like HTML templates for Haskell
by Tobias Dammers 17 Nov '16

17 Nov '16

Dear Cafe, I've been working on this for a while now, and figured now is as good a time as any to announce it properly, so here it is. https://ginger.tobiasdammers.nl/ Ginger is a Haskell implementation of Jinja (http://jinja.pocoo.org/) the best-of-breed in the "DOM-agnostic HTML template engines for Python" arena (or at least I think it is). The reason I built this is because upon researching HTML templating solutions for Haskell, I noticed that almost all of them are typesafe compile-time solutions (either EDSLs or something TH/QQ), which is awesome but doesn't fit my use cases. I believe I am not alone here. Code is on github: https://github.com/tdammers/ginger And also on bitbucket: https://bitbucket.org/tdammers/ginger And of course Ginger is available from hackage: http://ginger.tobiasdammers.nl/ Any kind of feedback is very welcome. Thanks for your time! - Tobias -- Tobias Dammers - tdammers(a)gmail.com

3 3

CFP: Dutch Functional Programming Day 2017
by J. Stutterheim 17 Nov '16

17 Nov '16

Call for Participation: Dear Friends of Functional Programming, The next Dutch FP Day will be held on Friday, January 6 2017 at the Radboud University in Nijmegen. You are invited to some or all of: participate, give a presentation, join the dinner. For further information, please see the website. https://wiki.clean.cs.ru.nl/NL-FP_dag_2017 Feel free to forward this invitation to colleagues and students who might be interested. Best regards on behalf of Markus Klinik

1 0