August 2017 - Haskell-Cafe

Running ghci in a Cabal sandbox
by Emil Axelsson 04 Aug '17

04 Aug '17

Hi! I have a small file Test.hs alone in a directory: {-# LANGUAGE DeriveGeneric #-} module Test where import Data.Hashable import Data.Scientific import GHC.Generics data Sc = Sc Scientific deriving (Generic) instance Hashable Sc To be able to load this file, I set up a Cabal sandbox: $ ghc --numeric-version 8.0.2 $ cabal --numeric-version 1.24.0.2 $ cabal sandbox init ... $ cabal install hashable-1.2.6.0 scientific ... (Note: not the latest version of hashable.) Now, if I try to run GHCi and point it to the sandbox' package database I get this error: $ ghci -package-db=.cabal-sandbox/x86_64-linux-ghc-8.0.2-packages.conf.d Test.hs GHCi, version 8.0.2: http://www.haskell.org/ghc/ :? for help [1 of 1] Compiling Test ( Test.hs, interpreted ) Test.hs:12:14: error: • No instance for (Hashable Scientific) arising from a use of ‘hashable-1.2.6.1:Data.Hashable.Class.$dmhashWithSalt’ • In the expression: hashable-1.2.6.1:Data.Hashable.Class.$dmhashWithSalt @Sc In an equation for ‘hashWithSalt’: hashWithSalt = hashable-1.2.6.1:Data.Hashable.Class.$dmhashWithSalt @Sc In the instance declaration for ‘Hashable Sc’ Somehow it mixes in version 1.2.6.1 of hashable, even though this package isn't installed (neither in the sandbox nor the global database). It turns out that wrapping the command in `cabal exec` fixes the problem: $ cabal exec -- ghci -package-db=.cabal-sandbox/x86_64-linux-ghc-8.0.2-packages.conf.d Test.hs GHCi, version 8.0.2: http://www.haskell.org/ghc/ :? for help [1 of 1] Compiling Test ( Test.hs, interpreted ) Ok, modules loaded: Test. *Test> Any idea what's going on? / Emil

1 0

Question?
by Casey Newsome 03 Aug '17

03 Aug '17

Hello, I am a new user of Haskell. I was wondering if this is where I can ask a question on a program I am attempting to write. If this is the wrong place. I apologize. Sorry if there was an inconvenience, Casey :D

4 3

Low-level idea: Aligned tagged ByteArray#
by Zemyla 03 Aug '17

03 Aug '17

I had an idea for a way to attach small-but-necessary amounts of information to a ByteArray# or MutableByteArray# (and potentially also to Array# and ArrayArray#). The type would look something like "TaggedByteArray# 8#". This, for instance, would represent the address of a primitive array of bytes, with an alignment of 8, and with the last 3 bits having a number between 0 and 7 in them. Code to access the elements of the ByteArray# would disregard the last bits, and the garbage collector would simply ensure that the tag was retained after collection passes. "ByteArray#" itself would simply become an alias for "TaggedByteArray# 1#", and the existing functions would have their types changed to work with the new TaggedByteArray#s. The primary potential problems I see are how to have a KnownNat-style typeclass before Integer is defined, and how to store both the array and the tag in an ArrayArray#.

1 0

Haskell in Leipzig 2017: submission deadline shifted by two weeks
by Wolfgang Jeltsch 02 Aug '17

02 Aug '17

Hi! We have shifted the submission deadline for Haskell in Leipzig 2017 by two weeks. The new deadline is at 18 August 2017. Looking forward to your contribution. 😉 Event: Haskell in Leipzig 2017 Time: October 26–28, 2017 Place: HTWK Leipzig, Germany Homepage: https://hal2017.softbase.org/ About ===== Haskell is a modern functional programming language that allows rapid development of robust and correct software. It is renowned for its expressive type system, its unique approaches to concurrency and parallelism, and its excellent refactoring capabilities. Haskell is both the playing field of cutting-edge programming language research and a reliable base for commercial software development. The workshop series Haskell in Leipzig (HaL), now in its 12th year, brings together Haskell developers, Haskell researchers, Haskell enthusiasts, and Haskell beginners to listen to talks, take part in tutorials, join in interesting conversations, and hack together. To support the latter, HaL will include a one-day hackathon this year. The workshop will have a focus on functional reactive programming (FRP) this time, while continuing to be open to all aspects of Haskell. As in the previous year, the workshop will be in English. Contributions ============= Everything related to Haskell is on topic, whether it is about current research, practical applications, interesting ideas off the beaten track, education, or art, and topics may extend to functional programming in general and its connections to other programming paradigms. Contributions can take the form of * talks (about 30 minutes), * tutorials (about 90 minutes), * demonstrations, artistic performances, or other extraordinary things. Please submit an abstract that describes the content and form of your presentation, the intended audience, and required previous knowledge. We recommend a length of 2 pages, so that the program committee and the audience get a good idea of your contribution, but this is not a hard requirement. Please submit your abstract as a PDF document at https://easychair.org/conferences/?conf=hal2017 until Friday, August 18, 2017. You will be notified by Friday, September 8, 2017. Hacking Projects ================ Projects for the hackathon can be presented during the workshop. A prior submission is not needed for this. Invited Speaker =============== * Ivan Perez, University of Nottingham, UK Invited Performer ================= * Lennart Melzer, Robert-Schumann-Hochschule Düsseldorf, Germany Program Committee ================= * Edward Amsden, Plow Technologies, USA * Heinrich Apfelmus, Germany * Jurriaan Hage, Utrecht University, The Netherlands * Petra Hofstedt, BTU Cottbus-Senftenberg, Germany * Wolfgang Jeltsch, Tallinn University of Technology, Estonia (chair) * Andres Löh, Well-Typed LLP, Germany * Keiko Nakata, SAP SE, Germany * Henrik Nilsson, University of Nottingham, UK * Ertuğrul Söylemez, Intelego GmbH, Germany * Henning Thielemann, Germany * Niki Vazou, University of Maryland, USA * Johannes Waldmann, HTWK Leipzig, Germany Questions ========= If you have any questions, please do not hesitate to contact Wolfgang Jeltsch at wolfgang-it(a)jeltsch.info.

1 0

What code is generated for binary functions of data types with multiple constructors?
by Clinton Mead 02 Aug '17

02 Aug '17

Assume the following: 1. We have a type T with multiple constructors, but lets assume we're on a 64 bit machine and have at least 4 but no more than 7 constructors so we can still use pointer tagging. 2. And a function "f :: T -> T -> T" 3. Each combination of constructors (16 in the 4 constructor case, 49 in the 7 constructor case) has an implementation (naturally many of these will call helper functions). There's no wildcard matches. How will GHC generate the code for "f"? Will pushing through LLVM make any difference? Do we take the number of the first constructor, and then multiply the number of the second constructor by say 7, and then do one jump via a jump table? Is there a bunch of nested jump tables, each with 7 options? Is there a fall through if statement? Something else? I understand this probably a complex answer and perhaps doesn't have a simple answer, but I just wanted to get some sort of idea in my head of how GHC attempts to resolve code like this.

2 1

What can be UNPACK'ed?
by Tom Ellis 02 Aug '17

02 Aug '17

What types can be UNPACK'ed? From my vague understanding of what UNPACK does, it can only be types whose size is bounded by a known value, so whilst data Foo = Foo {-# UNPACK #-} !Float {-# UNPACK #-} !Int is fine, presumably data Foo = Foo {-# UNPACK #-} ![Float] is not fine. Or is it? Technically it could be possible to reserve enough space in Foo to store either a [] or a (:) Float [Float]. And if that is possible could we also UNPACK polymorphic fields? data Storable a => Foo a = Foo {-# UNPACK #-} a or data Foo a where Foo :: Storable a => {-# UNPACK #-} a -> Foo a if either DatatypeContexts worked or my imaginary GADT UNPACK syntax worked (I'm not sure if either does). What are the limits of what we could reasonably get to UNPACK in GHC? Thanks, Tom

5 9

Problems with Type family and Typeable
by Kai Zhang 01 Aug '17

01 Aug '17

I got an error when compiling the following codes: {-# LANGUAGE TypeOperators #-} {-# LANGUAGE DataKinds #-} {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE UndecidableInstances #-} import Data.Type.Bool import Data.Type.Equality import Data.Tagged (Tagged(..)) import Data.Typeable type family Elem x xs :: Bool where Elem _ '[] = 'False Elem a (x ': xs) = a == x || Elem a xs elemTag :: forall a s t ts . Proxy s -> Tagged (t ': ts) a -> Bool elemTag _ _ = if typeOf (Proxy :: Proxy (Elem s (t ': ts))) == typeOf (Proxy :: Proxy 'True) then True else False GHC says: • No instance for (Typeable (Data.Type.Equality.EqStar s t || Elem s ts)) arising from a use of ‘typeOf’ • In the first argument of ‘(==)’, namely ‘typeOf (Proxy :: Proxy (Elem s (t : ts)))’ In the expression: typeOf (Proxy :: Proxy (Elem s (t : ts))) == typeOf (Proxy :: Proxy True) In the expression: if typeOf (Proxy :: Proxy (Elem s (t : ts))) == typeOf (Proxy :: Proxy True) then True else False My question: why is ghc unable to deduce that "Data.Type.Equality.EqStar s t || Elem s ts" resolves to "Bool" which should be an instance of Typeable? How to fix this?

3 6

Many packages blocked on Stackage nightly for GHC 8.2.1
by Dan Burton 31 Jul '17

31 Jul '17

Please be advised that many packages were dropped from the Stackage nightly build plan in order to arrive at a successful build with the new ghc. (Note that this does not affect the build plans for LTS 9.) This is to be expected whenever ghc gets a major version bump. For details, see: https://github.com/fpco/stackage/blob/master/MAINTAINERS.md We have attempted to contact package authors whose packages were dropped. However, comprehensive notification is not fully automated, and there may be many who have not been directly contacted. You can check your section of build-constraints.yaml, as well as stackage.org/nightly, to see which of your packages are currently absent from the build plan. You may restore your packages to the nightly build plan (once they are buildable with ghc 8.2.1) in the same way they were added: by opening a PR, uncommenting your packages in build-constraints.yaml. Thank you for your efforts to keep the Haskell ecosystem up to date.

2 1