- Haskell-Cafe - Haskell.org

Re: [Haskell-cafe] Is there a name for this?
by Olaf Klinke 13 Mar '15

13 Mar '15

On 2015-03-09 11:26 AM, Ben Franksen wrote: >> Albert Y. C. Lai wrote: >>> Choose one: >>> >>> Foo' is a free Foo >>> >>> Foo' is a limit of the diagram of Foo >>> >>> Foo' is a terminal object of the [sub]category of Foo >> >> Interesting. Is the last variant how "free <whatever>" is usually defined? >> Or is it a coincidence that the two, well, coincide here? I remember I have >> seen other definitions that looked a lot less easy to understand. > > I have not checked, and too lazy to. > > The usual "free" is the less-easy-to-understand one: you need a > forgetful functor, then you need its left adjoint, and you call it your > free functor. Then the target objects hit by the free functor are the > free things. In general when talking about free things and adjunctions one must pay attention to what the functions are. The original definition of class Foo was coalgebraic: class Foo f where bar :: f -> Bool baz :: f -> Char -> Int meaning that the functions take an f and make something else rather than taking something and constructing an f (like e.g. the (:) of lists does). With the definition data Foo' = { bar' :: Bool, baz' :: Char -> Int} one can re-write class Foo f where barbaz :: f -> Foo’ Objects of our category Foo are instances of Foo, that are pairs <f,barbaz> where f is a Haskell type and barbaz is a function f -> Foo’. What are the morphisms? Well, we can consider Foo’ as a phantom type data F a = F Foo’ that is, a constant functor. Then instances of Foo are coalgebras for the functor F, and the right notion of function between coalgebras requires that some equations hold, namely: any h :: f -> f’ that wants to be a Foo-function must have barbaz . h = barbaz. Now we can see that Foo’ is indeed terminal: More precisely, the Foo-instance <Foo’,barbaz=id> is terminal. Indeed, if we have some h :: f -> Foo’ with id . h = barbaz then h must be barbaz, so there is precisely one Foo-function from the Foo-instance <f,barbaz> to <Foo’,id>. I can’t see how Foo’ would be the free Foo in this category, however. Things change if we let Foo-functions be arbitrary functions. Then the documentation of the free package tells us that the free Foo over type t is the instance <(Foo’,t),barbaz=fst>. Olaf

1 0

Deriving Show for non-regular data types
by Mark Laws 13 Mar '15

13 Mar '15

Hi, As part of studying Okasaki's PFDS book, I wanted to add Show support for each of the data structures, and some have proven to be challenging, as some of the types are non-regular and automatic derivation of Show doesn't work. I've been able to add some code that introduces a supplementary type class that serves as a way to pass "proof" that the wrapping data type supports traversal for Show-ability, but the solution seems unsatisfactory. I would greatly appreciate any suggestions for improvements. I've attached the code; the relevant bits are in BankersDeque.hs, Example.hs, NestedShowable.hs, and SimpleCatenableDeque.hs. The same code is available here: https://gist.github.com/drvink/30fb2a2b257fc99af281 Thanks, Mark Laws -- |v\ /\ |\ |< |_ /\ \^| //

3 5

Bad recursion? When can something be on both sides of an equation?
by Jeffrey Brown 13 Mar '15

13 Mar '15

Dear Haskellers, In another thread [1], I talked about computing what I've been lately calling the "total descendents" of a subset S of a graph: that is, the set of nodes t for which every maximal sequence of predecessors intersects S. I got it done, functionally [2]. But in the process I wrote something that hangs, and I can't figure out why. The problem arose in the following passage. (Glab stands for Graph Label, a synonym for Int.) totalDescendents :: S.Set Glab -> Graph -> S.Set Glab totalDescendents roots graph = b where (a,b,c,graph) = totalDescendentsCtrlr -- tricky (roots, S.empty, M.empty, graph) totalDescendentsCtrlr :: TotalDescendentsData -> TotalDescendentsData totalDescendentsCtrlr (a,b,c,gr) = if S.null a' -- fails, as does a == a' -- a == a' is one iteration slower but should have same effect then (a',b',c',gr') else totalDescendentsCtrlr (a',b',c',gr') where (a',b',c',gr') = visitUndetPrds $ visitTdSnvSucs (a,b,c,gr) Notice that in the equation marked "tricky", the object "graph" appears on both sides. I thought that was the problem, so I rewrote that line as the following: where (_,b,_,_) = totalDescendentsCtrlr and sure enough, the problem vanished. So I tried to distill that problem to something tiny, and came up with this: f (a) = a' where (a',b) = fBackend (a,b) -- tricky fBackend (a,b) = if a' < 1 then (a',b) else fBackend (a',b) where a' = a - 1 I see no substantive difference between the problem with the first body of code and the problem I expected f and fBackend to have -- but f and fBackend work fine! Any idea what might be happening? Thanks, Jeff [1] http://permalink.gmane.org/gmane.comp.lang.haskell.cafe/115353 [2] https://github.com/JeffreyBenjaminBrown/digraphs-on-text/blob/master/TotalD…

2 2

Layout rules for if then else
by Alan & Kim Zimmerman 13 Mar '15

13 Mar '15

I am working on ghc-exactprint, and need to flag points where layout must be preserved, and considering the `if` statement. My understanding of the layout rules for if then else is that the `then` has to start more to the right than the `if`. Using GHC 7.10 RC2, ghci cheerfully loads the following ``` f = if True then 1 else 2 ``` Is this valid? Have the rules been relaxed? Alan

5 7

Monadic/sequential migration implementation for data in acid-state
by Sybils Leaf 12 Mar '15

12 Mar '15

Full description of my problem is on Stack Overflow: http://stackoverflow.com/questions/29011479/how-can-a-monadic-sequential-mi… Any comments or suggestions would be greatly appreciated!

1 0

TH not compiling when writing the code directly does
by Michael Jones 12 Mar '15

12 Mar '15

I’m stuck getting a TH expansion to compile. When I take the generated code and compile it directly, it is fine. But when expanding and compiling, it fails. The problem seems related to instanceD, where I provide the type as: (appT (appT (conT devName') (varT $ mkName "Word8")) (varT $ mkName "BV”)) which expands to: IDENTIFICATION__MODEL_ID Word8 BV No matter what I put for the second type (BV), it generates the compile error. For example, Word8 or Word16 gives the same error. But if I take the resulting expression and just put it into the file and compile, it compiles fine. So something about expansion is causing the problem. Is there some more correct way to make the expression rather than nesting two appT? Any ideas? Note: the generation of the functions with (map (\f -> returnQ f) (concat funs)) is ugly in my opinion. I could not find the right way to express this. So all ideas welcome. Mike GEN DEC -------------- makeInstance :: String -> [String] -> Q [Dec] makeInstance devName regNames = do let devName' = mkName devName let regNames' = map mkName regNames let ctx = return [] :: Q [Pred] funs <- mapM (\regName -> [d| $(varP regName) = $(varE $ mkName "writeReadField") $(stringE $ nameBase regName)(get fi) (set fi) where fi = $(conE $ mkName "FieldInfo") 3 1 get info = $(varE $ mkName "getValue") info set info = $(varE $ mkName "setValue") info |]) regNames' instance_ <- instanceD ctx (appT (appT (conT devName') (varT $ mkName "Word8")) (varT $ mkName "BV")) (map (\f -> returnQ f) (concat funs)) return [instance_] COMPILE ERROR ------------------------- src/TestFields.hs:175:3: Illegal type variable name: ‘BV’ When splicing a TH declaration: instance IDENTIFICATION__MODEL_ID Word8 BV where id = writeReadField "id" (get_0 fi_1) (set_2 fi_1) where fi_1 = FieldInfo 3 1 get_0 info_3 = getValue info_3 set_2 info_4 = setValue info_4 ids = writeReadField "ids" (get_5 fi_6) (set_7 fi_6) where fi_6 = FieldInfo 3 1 get_5 info_8 = getValue info_8 set_7 info_9 = setValue info_9

2 2

David Turner quote on lisp and FP
by Rustom Mody 12 Mar '15

12 Mar '15

There is this quote: *It needs to be said very firmly that LISP is not a functional language at all. My suspicion is that the success of Lisp set back the development of a properly functional style of programming by at least ten years.* David Turner found here and there on the net eg http://dis.4chan.org/read/prog/1376090701 Does anyone have/know the original reference? Thanks Rusi -- http://blog.languager.org

4 5

Multiple compiles with cabal
by Ernesto Rodriguez 12 Mar '15

12 Mar '15

Dear Cafe, I have a program [1] which half of the code is meant to be compiled with GHCJS and the other half with GHC (Or any other Haskell compiler). Currently I do the compilation separately and simply include the compiled JS as a static resource for the regular project. Has anyone run into a similar scenario? Can Cabal handle this? Is there some way of having multiple compiles w/o requiring a shell script to initialize them? Thank you & Cheers N. [1] https://github.com/netogallo/Cryptographer (annoying bc Github now believes my project is like 99% JS since the GHCJS runtime is bundled as a static resource) -- Ernesto Rodriguez Masters Student Computer Science Utrecht University www.netowork.me github.com/netogallo

2 1

Mathematical functions with multiple arguments
by Sumit Sahrawat, Maths & Computing, IIT (BHU) 12 Mar '15

12 Mar '15

Hi everybody, I have a function of type plot :: ([Double] -> Double) -- A function to plot -> [(Double, Double)] -- Range for all arguments -> IO () I want to enforce the fact that ranges for all arguments should be provided. Is there a way to make the type system enforce it? -- Regards Sumit Sahrawat

3 3

Is there a name for this?
by Eric 12 Mar '15

12 Mar '15

Given a class Foo: > class Foo f where> bar :: f -> Bool> baz :: f -> Char -> Int does the record type Foo' > data Foo' = Foo' {> bar' :: Bool,> baz' :: Char -> Int> }> > instance Foo Foo' where> bar = bar'> baz = baz' where the fields of the record Foo' are just implementations of the functions of Foo have any special name? In some sense it's a canonical instance of Foo; we could trivially write a universal > toFoo' :: Foo f => f -> Foo' function. I ran across this pattern and wondered whether it was useful enough to have received a name. It vaguely resembles a typeclass dictionary or a vtable in an OOP language. --Eric

3 3