- Glasgow-haskell-users

Re: Local evidence and type class instances
by Max Bolingbroke 16 Oct '10

16 Oct '10

HI Antoine, I didn't know UHC already had this - thanks for the pointer! It seems they have read about implicit configurations too, as the example they use is very similar to the paper. In fact, they also have another extension to the concept that I was intentionally avoiding mentioning - they ensure that the lexically most closely bound instance is used to satisfy a demand for a type class operation. This means that you can sensibly have local instances that overlap global ones. I'm not sure how I feel about this feature, though I certainly see the motivation. On 16 October 2010 15:47, Antoine Latter <aslatter(a)gmail.com> wrote: > Substituting f for f' in this example will change the meaning of the > operation significantly, which is, in my mind, hard to explain and > reason about. That's the behaviour I had in mind. There is no doubt that local instances are a big complexity increase, but IMHO this behaviour is relatively easy to reason about as long as you know about the dictionary passing translation. Indeed, it's practically mandatory to use dictionary-passing to implement this behaviour. Local instances would probably never get added to a compiler (like JHC) that doesn't use that technique. RE your later point: > Along similar lines as my last post, would I be able to write the > following type signature: > > myFunct :: Ord Int => Int -> (...) > > and have it always use the local instance for Ord Int? Is this the > type that the type checker would infer for my function? To my mind, if you have that signature, then myFunct would use whatever Ord Int instance was available at the use site of myFunct (i.e. wherever it was mentioned syntactically). It is certainly interesting to note that adding a context like this might be useful if you were using local instances. Cheers, Max

2 1

Fwd: Local evidence and type class instances
by Antoine Latter 16 Oct '10

16 Oct '10

Forwarding to list. ---------- Forwarded message ---------- From: Antoine Latter <aslatter(a)gmail.com> Date: Sat, Oct 16, 2010 at 9:47 AM Subject: Re: Local evidence and type class instances To: Max Bolingbroke <batterseapower(a)hotmail.com> On Sat, Oct 16, 2010 at 7:11 AM, Max Bolingbroke <batterseapower(a)hotmail.com> wrote: > Hi GHC users, > > Now that the Glorious New type checker can handle local evidence > seamlessly, is it a big implementation burden to extend it to deal > with local *type class instances* in addition to local *equality > constraints*? > > For example, you could write this: > > """ > f :: Bool > f = id < id > where > instance Ord (a -> b) where > _ `compare` _ = LT > """ I think UHC has something like that: http://www.cs.uu.nl/wiki/bin/view/Ehc/UhcUserDocumentation#3_8_Local_instan… One thing that worries me is that the local instance only applies to type-expressions that consume constraints - if the constraint was satisfied before the local constraint was introduced then the local constraint would not apply. Am I making sense? I don't have much of a background in type-theory, so I may not be using the correct language, but I'm having trouble seeing how it wouldn't be confusing. What I'm thinking is that the following two functions would have different effect: f n = Set.insert n and: f' (n::Int) = Set.insert n When called like so: updateSet (n::Int) s = let n' = someComputation n s in f n' s where instance Ord s ... Substituting f for f' in this example will change the meaning of the operation significantly, which is, in my mind, hard to explain and reason about. Type-classes are engineered to be global in scope. It's nice that from a GHC type-checking perspective they need not be, but changing them to be non-global is also an engineering and design problem. Antoine > > Not only would this prevent instances from "leaking" into all > importing modules, but you could refer to *lexically scoped variables* > in the type class instance. This would let programmers implement e.g. > implicit configurations in the style of Kieslyov and Shan > (http://okmij.org/ftp/Haskell/types.html#Prepose) without any mad > hackery to tunnel pointers through the type system. > > Of course, it could lead to problems. For example: > > """ > newtype MyInt = MyInt { unMyInt :: Int } > > f :: [Int] -> S.Set Int > f xs = S.map unMyInt $ S.toList (map MyInt xs) > where > instance Ord MyInt where > x `compare` y = unMyInt x `compare` unMyInt y > > g :: [Int] -> S.Set Int > g xs = S.map unMyInt $ S.toList (map MyInt xs) > where > instance Ord MyInt where > x `compare` y = unMyInt y `compare` unMyInt x > > main = do > let s1 = f [1, 2] > s2 = g [1, 2] > print $ s1 == s2 -- prints False! > print $ S.toList s1 == s2 -- prints True! > """ > > I don't think this is a big problem. It would be up to the users of > this extension to do the Right Thing. > > As Kieslyov and Shan point out, this extension also kills the > principal types property - but so do GADTs. Just add more type > signatures! > > Are there big theoretical problems with this extension, or is it just > a lack of engineering effort that has prevented its implementation? > > Max > _______________________________________________ > Glasgow-haskell-users mailing list > Glasgow-haskell-users(a)haskell.org > http://www.haskell.org/mailman/listinfo/glasgow-haskell-users >

1 0

unused record fields
by Serge D. Mechveliani 15 Oct '10

15 Oct '10

Simon P. Jones wrote recently about that ghc-6.12 takes in account the elliplis in MkT {t1 = x, ..} when reporting about unused bindings. Now, here is the example: module TT where data T = T {t1, t2 :: Int} f d = x where T {t1 = x, ..} = d ghc-6.12.2 warns about unused value of t2. And for f (T {t1 = x, ..}) = x, it does not warn about this. Is this a bug? I use extensions: TypeSynonymInstances UndecidableInstances FlexibleContexts FlexibleInstances MultiParamTypeClasses OverlappingInstances RecordWildCards NamedFieldPuns ghc-options: -fno-warn-overlapping-patterns -fwarn-unused-binds -fwarn-unused-matches -fwarn-unused-imports -O0 , but the options can be simplified, of course. Regards, ----------------- Serge Mechveliani mechvel(a)botik.ru

1 0

Set inclusion
by Serge D. Mechveliani 14 Oct '10

14 Oct '10

I have the two notes on the GHC library. The docs show that 1. Map has the function for the Map inclusion relation, and Set does not have such for sets. 2. notMember looks unnecessary, because one can write not . Map.member k. Regards, ----------------- Serge Mechveliani mechvel(a)botik.ru

2 3

porting using unregisterised build: x86->amd64 solaris
by Karel Gardas 11 Oct '10

11 Oct '10

Hello, I'm attempting to make unregisterised build of recent GHC head on Solaris/x86 for Solaris/amd64 platform. I'm closely following: http://hackage.haskell.org/trac/ghc/wiki/Building/Porting#Cross-compilingto… with the only one issue: mentioned: echo compiler/main/Config.hs >> list seems to be incorrect due to missing compiler/main/Config.hs file which means gtar will fail on it later. However I've got to the: <T>$ make all_ghc_stage2 2>&1 | tee c.log step which fails with: gmake -r --no-print-directory -f ghc.mk phase=0 just-makefiles "" configure --with-ghc="/export/home/karel/vcs/ghc-target/" --with-ghc-pkg="/export/home/karel/vcs/ghc-target/" --with-gcc="/usr/bin/gcc" --configure-option=--with-cc="/usr/bin/gcc" --ghc-option=-DOMIT_NATIVE_CODEGEN --ghc-option=-DNO_REGS --flags=stage2 --ghc-option=-DSTAGE=2 --disable-library-for-ghci --with-hscolour="/usr/local/bin/HsColour" --configure-option=CFLAGS="-g -O0 -DNO_REGS -DUSE_MINIINTERPRETER -D__GLASGOW_HASKELL__=613 -Iincludes " --configure-option=LDFLAGS=" " -- stage2 compiler /bin/sh: line 1: : cannot execute [Is a directory] gmake[1]: *** [compiler/stage2/package-data.mk] Error 126 gmake: *** [all_ghc_stage2] Error 2 To me it looks like it's perhaps expecting to have `ghc' binary in /export/home/karel/vcs/ghc-target/ but there is only `ghc' subdirectory there hence failing. Also I'm not sure I've detected correctly mentioned "unimportant build failures in the RTS" as I've got one which looks like: gmake -r --no-print-directory -f ghc.mk all "inplace/bin/ghc-stage1" -optc-Wall -optc-Wextra -optc-Wstrict-prototypes -optc-Wmissing-prototypes -optc-Wmissing-declarations -optc-Winline -optc-Waggregate-return -optc-Wpointer-arith -optc-Wmissing-noreturn -optc-Wnested-externs -optc-Wredundant-decls -optc-Iincludes -optc-Irts -optc-DCOMPILING_RTS -optc-DNOSMP -optc-fno-strict-aliasing -optc-fno-common -optc-Ilibffi/build/include -optc-fomit-frame-pointer -optc-DRtsWay=\"rts_v\" -H32m -O -H32m -Iincludes -Irts -DCOMPILING_RTS -package-name rts -optc-DNOSMP -dcmm-lint -Ilibffi/build/include -i -irts -irts/dist/build -irts/dist/build/autogen -Irts/dist/build -Irts/dist/build/autogen -keep-hc-files -c rts/Linker.c -o rts/dist/build/Linker.o rts/Linker.c: In function `do_Elf_Rel_relocations': rts/Linker.c:3784:0: warning: cast from pointer to integer of different size rts/Linker.c:3785:0: warning: cast to pointer from integer of different size rts/Linker.c:3807:0: warning: cast from pointer to integer of different size rts/Linker.c:3815:0: warning: cast from pointer to integer of different size rts/Linker.c:3819:0: warning: cast from pointer to integer of different size rts/Linker.c: In function `do_Elf_Rela_relocations': rts/Linker.c:3866:0: warning: cast from pointer to integer of different size rts/Linker.c:3903:0: warning: cast from pointer to integer of different size rts/Linker.c:3915:0: warning: cast from pointer to integer of different size rts/Linker.c:4028:0: error: `R_X86_64_64' undeclared (first use in this function) rts/Linker.c:4028:0: error: (Each undeclared identifier is reported only once rts/Linker.c:4028:0: error: for each function it appears in.) rts/Linker.c:4029:0: warning: cast to pointer from integer of different size rts/Linker.c:4032:0: error: `R_X86_64_PC32' undeclared (first use in this function) rts/Linker.c:4038:0: warning: comparison between signed and unsigned rts/Linker.c:4048:0: warning: cast to pointer from integer of different size rts/Linker.c:4056:0: warning: cast to pointer from integer of different size rts/Linker.c:4060:0: error: `R_X86_64_32' undeclared (first use in this function) rts/Linker.c:4074:0: warning: cast to pointer from integer of different size rts/Linker.c:4078:0: error: `R_X86_64_32S' undeclared (first use in this function) rts/Linker.c:4082:0: warning: comparison between signed and unsigned rts/Linker.c:4092:0: warning: cast to pointer from integer of different size rts/Linker.c:4096:0: error: `R_X86_64_GOTPCREL' undeclared (first use in this function) rts/Linker.c:4100:0: warning: cast to pointer from integer of different size rts/Linker.c:4104:0: error: `R_X86_64_PLT32' undeclared (first use in this function) rts/Linker.c:4110:0: warning: comparison between signed and unsigned rts/Linker.c:4115:0: warning: cast to pointer from integer of different size gmake[1]: *** [rts/dist/build/Linker.o] Error 1 gmake: *** [all] Error 2 karel@silence:~/vcs/ghc-host$ I do have two questions: 1) is the url above describing cross-compiling GHC for porting up-to-date? 2) is there any trick how to get my last step (i.e. <T>$ make all_ghc_stage2 2>&1 | tee c.log) working? Thanks! Karel

3 7

Bringing back Monad Comprehensions (in style)
by George Giorgidze 07 Oct '10

07 Oct '10

Bringing back Monad Comprehensions (in style). Dear GHC users, My colleagues and I are working on Haskell embedded DSL for data-intensive and data-parallel applications [1]. The idea is to provide the Haskell list prelude combinators to manipulate database-resident data. The combinators are not executed in Haskell runtime, instead they are compiled down to SQL, executed on relational database systems and the results are marshalled back to Haskell for further in-heap processing or generation of new database-able embedded programs. Although programming with the standard list processing combinators is feasible, the embedded programs are much more concisely formulated using the list comprehension notation, especially, when extended with 'order by' and 'group by' constructs [2]. Unfortunately, in Haskell, the list comprehension notation is only available for processing lists. In order to support the list comprehension notation, we have built a quasiquter that desugars the list comprehension notation, but, instead of generating code using the Haskell list prelude combinators the quasiquter generates code that uses list processing combinators from our embedded language. Although the quasiquoting approach worked for us, it has a number of drawbacks: * Introduces extra syntactic noise * Error messages are hard to understand as they refer to generated code * Needs to be re-implemented for every list-based embedded language One way to address the aforementioned drawbacks is to define our queries as a monad (similar to list monad) and use the monad comprehension notation. The do notation can be used but it is less suited for query languages. Unfortunately monad comprehensions were removed from Haskell, prior to Haskell 98. However, I think that the notation is extremely useful not only for lists, but for other list like data structures, list-based query languages (see above), maybe even for wider range of EDSLs and monads. I think the feature deserves to be supported at least as a GHC language extension. Thus, I would like to propose to design and implement the monad comprehension notation as a GHC language extension. I am willing to invest some time and contribute to this effort. One can also look at how recently introduced 'order by' and 'group by' constructs generalise to monad comprehensions. If that works, one could implement even more "stylish" monad comprehension notation. Feedback from GHC users and developers would be very much appreciated. * Do you think that this is a good idea? * Would you use monad comprehensions (if available) for your library/EDSL/application? * Do you think that it would be hard to integrate this extension into current GHC codebase? * Have you already thought about how to generalise 'order by' and 'group by' to monad comprehensions? * Have you already thought about how to address the original objections to the monad comprehension notation? Cheers, George [1] http://www-db.informatik.uni-tuebingen.de/files/weijers/IFL2010complete.pdf [2] http://research.microsoft.com/en-us/um/people/simonpj/papers/list-comp/

6 8

Profiling: 0 entries but high costs
by Stefan Wehr 06 Oct '10

06 Oct '10

Hi all, here is some profiling output which shows strange behavior: COST CENTRE MODULE no. entries %time %alloc %time %alloc MAIN MAIN 1 0 0.0 0.0 100.0 100.0 ... CAF:runHttpNewsWatch_watchNews MobileGateway.DataClient.HttpNewsWatch 8685 1 0.0 0.0 90.2 95.7 watchNews MobileGateway.LubSync.NewsWatch 12239 14 0.0 0.0 90.2 95.7 watchNewsIteration MobileGateway.LubSync.NewsWatch 12283 4 0.0 0.0 90.2 95.7 ... getAndProcessNews MobileGateway.LubSync.NewsWatch 12284 156 0.0 0.0 90.2 95.7 ... handleNewsItem MobileGateway.LubSync.NewsWatch 12526 577 0.0 0.0 88.9 95.1 ... dsLoadObject_a7eq MobileGateway.DataClient.HttpDs 12648 144 0.0 0.0 84.0 91.5 loadObject MobileGateway.DataClient.HttpDs 12650 720 0.7 0.6 84.0 91.5 ... runNewsWatchTInIO MobileGateway.DataClient.HttpNewsWatch 12654 0 0.0 0.0 69.5 80.3 runHttpDsM MobileGateway.DataClient.HttpDs 12655 0 69.5 80.3 69.5 80.3 The cost centre runHttpDsM has yero entries but rather high costs. How can this be? According to the manual, the entries column specifies "the number of times this particular point in the call graph was entered". From this definition I would conclude that a cost centre with zero entries cannot have any costs. What am I doing wrong? Thanks for helping. -- Stefan

2 1

Attempt at building on OpenBSD/sparc64 - failure at "make all_ghc_stage2"
by Benjamin Jansen 05 Oct '10

05 Oct '10

Hello, I am attempting to build GHC 6.12.3 on OpenBSD4.4/sparc64 [1]. I've been following the instructions on the Porting page [2], using a Linux x86 system as the host machine. I've gotten to executing the following command on the target system: aogail@netra> gmake all_ghc_stage2 ===--- updating makefiles phase 0 gmake -r --no-print-directory -f ghc.mk phase=0 just-makefiles "" configure --with-ghc="/home/aogail/software/ghc-6.12.3/" --with-ghc-pkg="/home/aogail/software/ghc-6.12.3/" --with-gcc="/usr/bin/gcc" --configure-option=--with-cc="/usr/bin/gcc" --ghc-option=-DOMIT_NATIVE_CODEGEN --ld-options=-E --flags=stage2 --ghc-option=-DSTAGE=2 --ghc-options='-O2' --configure-option=CFLAGS="-g -O0 -DNO_REGS -DUSE_MINIINTERPRETER -D__GLASGOW_HASKELL__=612 -Iincludes " --configure-option=LDFLAGS=" " -- stage2 compiler /bin/sh: : cannot execute - Is a directory gmake[1]: *** [compiler/stage2/package-data.mk] Error 126 gmake: *** [all_ghc_stage2] Error 2 My experience with complicated make systems is limited so I'm not sure where to look at this point. Can I get some pointers? Thanks, Ben Jansen 1: http://hackage.haskell.org/trac/ghc/ticket/3511 2: http://hackage.haskell.org/trac/ghc/wiki/Building/Porting

2 2

Cabal constraint solver
by Christian Höner zu Siederdissen 05 Oct '10

05 Oct '10

Hi, does the Cabal constraint solver always try to solve the complete graph? example: (ghc-7.0.0-rc1) $ cabal install parsec-3.1.0 cabal: cannot configure syb-0.2.1. It requires base >=4.0 && <4.3 cd syb-0.2.1 * remove base<4.3 constraint from syb.cabal cabal install * syb-0.2.1 is now installed and works! $ cabal install parsec-3.1.0 cabal: cannot configure syb-0.2.1. It requires base >=4.0 && <4.3 This is rather annoying: the constraint is solved (eg. syb is installed) but we /still/ assume that we can not continue. So obviously, I would like to be able to have cabal assume constraints fullfilled if the package is installed. Am I missing something? (Otherwise, testing packages with 7-rc1 is becoming annoying as almost everything depends on syb) === And a second thing: cabal fetch How does one disable the solver for fetching packages? It is a bit annoying (here too) that 'cabal fetch parsec-3.1.0' fails. (It should maybe warn that constraints can not be fulfilled but the idea here is to get the packages to change them manually) === So did I miss anything or can I write a bug report? ;-) Gruss, Christian

5 5

haddock and associated data families
by Christian Höner zu Siederdissen 03 Oct '10

03 Oct '10

Hi, haddock seems to produce an error on associated data family decls.: http://hackage.haskell.org/packages/archive/PrimitiveArray/0.0.2.1/logs/fai… line 22, where the errors occurs is exactly this one: class PrimArrayOps a b where data PrimArray a b :: * -- ^ PrimArray data type I'll fix it by trying other methods to put comments there. Could someone enter this as a bug, if it is not done yet? (Assuming it is a bug ;-) Thanks, Christian

3 2