October 2011 - Glasgow-haskell-users

Unwanted eta-expansion
by Roman Cheplyaka 12 Oct '11

12 Oct '11

Suppose I want a foldl which is evaluated many times on the same list but with different folding functions. I would write something like this, to perform pattern-matching on the list only once: module F where myFoldl :: [a] -> (b -> a -> b) -> b -> b myFoldl [] = \f a -> a myFoldl (x:xs) = let y = myFoldl xs in \f a -> y f (f a x) However, for some reason GHC eta-expands it back. Here's what I see in the core: % ghc -O2 -ddump-simpl -fforce-recomp -dsuppress-module-prefixes \ -dsuppress-uniques -dsuppress-coercions F.hs ==================== Tidy Core ==================== Rec { myFoldl [Occ=LoopBreaker] :: forall a b. [a] -> (b -> a -> b) -> b -> b [GblId, Arity=3, Caf=NoCafRefs, Str=DmdType SLL] myFoldl = \ (@ a) (@ b) (ds :: [a]) (eta :: b -> a -> b) (eta1 :: b) -> case ds of _ { [] -> eta1; : x xs -> myFoldl @ a @ b xs eta (eta eta1 x) } end Rec } Why does it happen and can it be suppressed? This is GHC 7.0.4. -- Roman I. Cheplyaka :: http://ro-che.info/

4 7

GHC infinite loop when building vector program
by Bas van Dijk 12 Oct '11

12 Oct '11

Hello, When benchmarking my new vector-bytestring[1] package I discovered that building the following program causes GHC to go into, what seems to be, an infinite loop: import Data.Vector (Vector) import qualified Data.Vector.Generic as VG main = print $ VG.foldl f z (VG.fromList [] :: Vector Int) f = flip (:) z = [] I build it with: $ ghc --make vectorGHCloop.hs -O2 It compiles fine without the -O2 or if you specify -fno-enable-rewrite-rules. So it's probably a loop in a rule somewhere. Note that the program also builds fine when I change the 'f' and 'z' to: f = flip (:) z = [] I use vector-0.9 and ghc-7.2.1. Regards, Bas [1] https://github.com/basvandijk/vector-bytestring

2 3

Making a small GHC installation
by Iavor Diatchki 12 Oct '11

12 Oct '11

Hello, Does anyone have any advice about how I might make a smallish GHC installation? The context is that I need to make a demo VM, which has a limited amount of space, and I'd like to have GHC installed on the system but the default GHC installation (~700MB) does not fit. The installation does not need to be complete---I don't need documentation, or profiling, or Template Haskell---and I only need to install a fairly limited set of libraries, just enough to build my project. I'd be happy to build a custom version of GHC, if that's the easiest way to achieve the goal. So, if you have experience doing something similar, or you know of what might be the best way to approach the problem, advice would be most welcome! -Iavor

4 3

foldr oddity
by Frodo Chao 12 Oct '11

12 Oct '11

Hi, I came upon this when playing with foldr and filter. Compare the two definitions: testr n = foldr (\x y -> x && y) True [t | (_, t) <- zip [1 .. n] [True, True ..]] testr' n = foldr (\x y -> y && x) True [t | (_, t) <- zip [1 .. n] [True, True ..]] I tried these functions on ghci (The Glorious Glasgow Haskell Compilation System, version 7.0.3), and get the following results (with :set +s): testr 1000000 => True (0.01 secs, 7920832 bytes) testr' 1000000 => True (8.72 secs, 446585344 bytes) This bizarre (at least to me) behavior also applies to ||. Should I mind the orderings of the parameters (especially the accumulator) in the function passed to foldr? Thak you for reading. Sincerely yours, Frodo Chao

4 5

Fwd: Two Proposals
by Philip Wadler 12 Oct '11

12 Oct '11

George, Thanks very much for this. I like your suggestion, which fits the logical structure perfectly; and you've suggested a neat way around the ugliness of 'group groupBy'. I also note that if we aren't so worried about not introducing new keywords, that 'then group' could become 'group'. Yours, -- P On Mon, Oct 10, 2011 at 11:21 PM, George Giorgidze <giorgidze(a)gmail.com> wrote: > A quick thought that came to me after hoogling [a] -> [[a]]. > The first four functions in the search result are named after what they > return (noun in plural) rather than what they do (verb). I am talking about > inits, permutations, subsequence and tails. > So I thought the following syntax might work as well if (as it is already > common) grouping functions are named after what they return. > then f > then f by e > then group f > then group f by e > For example the following code fragments read well: > then group inits > then group permutations > then group subsequences > then group tails > Here we use the special identifier group as a verb. > I have not told you about the fifth result of the hoogling, the groupWith > function. The following really looks ugly: > then group groupWith by e > But following the aforementioned naming convention the groupWith function > could as well be named as equals. Now this reads well: > then group equals by e > Cheers, George -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

1 0

Re: Two Proposals
by Roman Leshchinskiy 11 Oct '11

11 Oct '11

Manuel M T Chakravarty wrote: > Roman Leshchinskiy: >> >> What data structures other than lists do we want to construct using list >> literals? I'm not really sure what the use cases are. > > Parallel arrays! (I want to get rid of our custom syntax.) Why? Don't you think it is useful to have a visual indication of which data structure you are using and what is going to be evaluated in parallel? In any case, if we want to get rid of the parallel array syntax, we have to overload list literals, enumerations and list comprehensions. We have the generic monadic desugaring for the latter but recovering an efficient DPH program from that sn't trivial. Roman

4 4

Fwd: Two Proposals
by George Giorgidze 11 Oct '11

11 Oct '11

For some reasons Philip's email was rejected by the mailing list. I am reposting his message. See below. Cheers, George Begin forwarded message: > From: Philip Wadler <wadler(a)inf.ed.ac.uk> > Date: 2011-October-11 11:48:31 GMT+02:00 > To: Simon Peyton-Jones <simonpj(a)microsoft.com>, George Giorgidze <george.giorgidze(a)uni-tuebingen.de> > Subject: Fwd: Two Proposals > > FYI. -- P > > > ---------- Forwarded message ---------- > From: <glasgow-haskell-users-owner(a)haskell.org> > Date: Tue, Oct 11, 2011 at 9:28 AM > Subject: Re: Two Proposals > To: wadler(a)inf.ed.ac.uk > > > You are not allowed to post to this mailing list, and your message has > been automatically rejected. If you think that your messages are > being rejected in error, contact the mailing list owner at > glasgow-haskell-users-owner(a)haskell.org. > > > > ---------- Forwarded message ---------- > From: Philip Wadler <wadler(a)inf.ed.ac.uk> > To: George Giorgidze <giorgidze(a)gmail.com> > Date: Tue, 11 Oct 2011 09:28:05 +0100 > Subject: Re: Two Proposals > George, Thanks very much for this. I like your suggestion, which > fits the logical structure perfectly; and you've suggested a neat way > around the ugliness of 'group groupBy'. I also note that if we aren't > so worried about not introducing new keywords, that 'then group' could > become 'group'. Yours, -- P > > On Mon, Oct 10, 2011 at 11:21 PM, George Giorgidze <giorgidze(a)gmail.com> wrote: >> A quick thought that came to me after hoogling [a] -> [[a]]. >> The first four functions in the search result are named after what they >> return (noun in plural) rather than what they do (verb). I am talking about >> inits, permutations, subsequence and tails. >> So I thought the following syntax might work as well if (as it is already >> common) grouping functions are named after what they return. >> then f >> then f by e >> then group f >> then group f by e >> For example the following code fragments read well: >> then group inits >> then group permutations >> then group subsequences >> then group tails >> Here we use the special identifier group as a verb. >> I have not told you about the fifth result of the hoogling, the groupWith >> function. The following really looks ugly: >> then group groupWith by e >> But following the aforementioned naming convention the groupWith function >> could as well be named as equals. Now this reads well: >> then group equals by e >> Cheers, George >> >> On 2011-Oct-5, at 09:14 , Simon Peyton-Jones wrote: >> >> [adding ghc-users] >> >> It's not easy, Phil. Do you have any ideas? >> >> For the 'then' case the name of the function serves as the verb. One might >> say >> >> then take 4 >> or >> then takeWhile by salary > 40 >> >> For grouping one might like to say the same thing, such as >> then groupBy by salary >> but the typing rule is quite different, so we really need a different >> keyword. We chose the compound keyword "then group" to avoid needing a >> whole new keyword ("group" is treated specially only in tthis context). So >> you write >> then group by salary using groupBy >> >> Using this order of the pieces for the sorting case is harder. What would >> one say? "then process"? Like this? >> then process by salary > 40 using takeWhile >> Not very nice. >> >> One could use a new keyword for grouping "theng" say, thus: >> theng groupBy by salary >> But that is hardly beautiful either. >> >> So the current story is not great, but it's the best I could think of. >> Improvements welcome. >> >> Simon >> >> | -----Original Message----- >> | From: Philip Wadler [mailto:wadler@inf.ed.ac.uk] >> | Sent: 04 October 2011 18:15 >> | To: Simon Peyton-Jones; George Giorgidze >> | Subject: Re: FW: Two Proposals >> | >> | George, >> | >> | Nice proposal. I like the idea of symmetry, but don't at all like the >> | idea that f comes before e for 'then' but f comes after e for 'then >> | group'. Can you rethink it and come up with something even more >> | symmetric? >> | >> | Yours, -- P >> | >> | >> | On Tue, Oct 4, 2011 at 9:23 AM, Simon Peyton-Jones >> | <simonpj(a)microsoft.com> wrote: >> | > FYI >> | > >> | > -----Original Message----- >> | > From: glasgow-haskell-users-bounces(a)haskell.org >> [mailto:glasgow-haskell- >> | users-bounces(a)haskell.org] On Behalf Of George Giorgidze >> | > Sent: 30 September 2011 18:28 >> | > To: glasgow-haskell-users(a)haskell.org >> | > Subject: Two Proposals >> | > >> | > GHC Users, >> | > >> | > I would like to make to the following two proposals: >> | > * Eliminate the default grouping close from SQL-like comprehensions >> | > * Introduce a GHC extension for list literal overloading >> | > >> | > OK, let me start with the first proposal. >> | > >> | > Currently, the SQL-like comprehension notation (both in its list >> comprehension >> | and monad comprehension variants) features the following five clauses: >> | > >> | > then f >> | > then f by e >> | > then group by e >> | > then group using f >> | > then group by e using f >> | > >> | > The first two clauses are used for specifying transformations of type >> [a] -> [a] >> | (or Monad m => m a-> m a for monad comprehensions). The following three >> | clauses are used for specifying transformations of type [a] -> [[a]] (or >> Monad m, >> | Functor f => m a -> m (f a) for monad comprehensions). See [1] for >> further >> | details. >> | > >> | > Note that the third clause does not mention which function is used for >> grouping. >> | In this case GHC.Exts.groupWith function is used as a default for list >> | comprehensions and the mgroupWith function from the MonadGroup class is >> used >> | as a default for monad comprehensions. >> | > >> | > I would like to suggest to remove the third clause for the following >> reasons: >> | > * Currently the syntax is asymmetrical. Note that there is the default >> case for >> | the 'then group' clause and not for the 'then' clause. >> | > * In the current notation it is not clear which grouping function is >> used in the >> | default case >> | > * For many monads including lists it is not clear which function should >> be >> | selected as a default (e.g., the groupWith function also does sorting and >> it is not >> | clear to me why this should be the default) >> | > * Gets rid of the MonadGroup class. Currently the sole purpose of this >> class is to >> | introduce a default grouping function for monad comprehensions. >> | > * Explicit mention of the grouping function would make monad/list >> | comprehensions much easier to read by making it immediately apparent >> which >> | function is used for grouping. >> | > >> | > My second proposal is to introduce the OverloadedLists extension that >> overloads >> | list literals. See Section 5.2 in [1] for details. >> | > >> | > Basically the idea is to treat list literals like: >> | > >> | > [1,2,3] >> | > >> | > as >> | > >> | > fromList [1,2,3] >> | > >> | > where >> | > >> | > class IsList l where >> | > type Item l >> | > fromList :: [Item l] -> l >> | > >> | > In the following I give useful instances of the IsList class. >> | > >> | > instance IsList [a] where >> | > type Item [a] = a >> | > fromList = id >> | > >> | > instance (Ord a) => IsList (Set a) where >> | > type Item (Set a) = a >> | > fromList = Set.fromList >> | > >> | > instance (Ord k) => IsList (Map k v) where >> | > type Item (Map k v) = (k,v) >> | > fromList = Map.fromList >> | > >> | > instance IsList (IntMap v) where >> | > type Item (IntMap v) = (Int,v) >> | > fromList = IntMap.fromList >> | > >> | > instance IsList Text where >> | > type Item Text = Char >> | > fromList = Text.pack >> | > >> | > As you can see the extension would allow list literals to be used for >> sets, maps >> | and integer maps. In addition the suggested OverloadedLists extension >> would >> | subsume OverloadedStrings extension (see the instance for Text, for >> example). >> | Having said that, for now, I am not suggesting to remove the >> OverloadedStrings >> | extension as it appears to be widely used. >> | > >> | > This extension could also be used for giving data-parallel array >> literals instead of >> | the special syntax used currently. >> | > >> | > Unless there is a vocal opposition to the aforementioned two proposals, >> I would >> | like to implement them in GHC. Both changes appear to be straightforward >> to >> | implement. >> | > >> | > Thanks in advance for your feedback. >> | > >> | > Cheers, George >> | > >> | > [1] >> http://www-db.informatik.uni-tuebingen.de/files/giorgidze/haskell2011.pdf >> | > _______________________________________________ >> | > Glasgow-haskell-users mailing list >> | > Glasgow-haskell-users(a)haskell.org >> | > http://www.haskell.org/mailman/listinfo/glasgow-haskell-users >> | > >> | > >> | > >> | >> | >> | >> | -- >> | .\ Philip Wadler, Professor of Theoretical Computer Science >> | ./\ School of Informatics, University of Edinburgh >> | / \ http://homepages.inf.ed.ac.uk/wadler/ >> | >> | The University of Edinburgh is a charitable body, registered in >> | Scotland, with registration number SC005336. >> | >> >> >> > > > > -- > .\ Philip Wadler, Professor of Theoretical Computer Science > ./\ School of Informatics, University of Edinburgh > / \ http://homepages.inf.ed.ac.uk/wadler/ > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > > > > > > -- > .\ Philip Wadler, Professor of Theoretical Computer Science > ./\ School of Informatics, University of Edinburgh > / \ http://homepages.inf.ed.ac.uk/wadler/

1 0

"case" of an empty type should have no branches
by Roman Beslik 11 Oct '11

11 Oct '11

Hi. Why the following code does not work? > data Empty > quodlibet :: Empty -> a > quodlibet x = case x of "parse error (possibly incorrect indentation)" This works in Coq, for instance. Demand for empty types is not big, but they are useful for generating finite types: > Empty ≅ {} > Maybe Empty ≅ {0} > Maybe (Maybe Empty) ≅ {0, 1} > Maybe (Maybe (Maybe Empty)) ≅ {0, 1, 2} etc. Number of 'Maybe's = number of elements. I can replace @Maybe Empty@ with @()@, but this adds some complexity.

5 5

Two Proposals
by George Giorgidze 10 Oct '11

10 Oct '11

GHC Users, I would like to make to the following two proposals: * Eliminate the default grouping close from SQL-like comprehensions * Introduce a GHC extension for list literal overloading OK, let me start with the first proposal. Currently, the SQL-like comprehension notation (both in its list comprehension and monad comprehension variants) features the following five clauses: then f then f by e then group by e then group using f then group by e using f The first two clauses are used for specifying transformations of type [a] -> [a] (or Monad m => m a-> m a for monad comprehensions). The following three clauses are used for specifying transformations of type [a] -> [[a]] (or Monad m, Functor f => m a -> m (f a) for monad comprehensions). See [1] for further details. Note that the third clause does not mention which function is used for grouping. In this case GHC.Exts.groupWith function is used as a default for list comprehensions and the mgroupWith function from the MonadGroup class is used as a default for monad comprehensions. I would like to suggest to remove the third clause for the following reasons: * Currently the syntax is asymmetrical. Note that there is the default case for the 'then group' clause and not for the 'then' clause. * In the current notation it is not clear which grouping function is used in the default case * For many monads including lists it is not clear which function should be selected as a default (e.g., the groupWith function also does sorting and it is not clear to me why this should be the default) * Gets rid of the MonadGroup class. Currently the sole purpose of this class is to introduce a default grouping function for monad comprehensions. * Explicit mention of the grouping function would make monad/list comprehensions much easier to read by making it immediately apparent which function is used for grouping. My second proposal is to introduce the OverloadedLists extension that overloads list literals. See Section 5.2 in [1] for details. Basically the idea is to treat list literals like: [1,2,3] as fromList [1,2,3] where class IsList l where type Item l fromList :: [Item l] -> l In the following I give useful instances of the IsList class. instance IsList [a] where type Item [a] = a fromList = id instance (Ord a) => IsList (Set a) where type Item (Set a) = a fromList = Set.fromList instance (Ord k) => IsList (Map k v) where type Item (Map k v) = (k,v) fromList = Map.fromList instance IsList (IntMap v) where type Item (IntMap v) = (Int,v) fromList = IntMap.fromList instance IsList Text where type Item Text = Char fromList = Text.pack As you can see the extension would allow list literals to be used for sets, maps and integer maps. In addition the suggested OverloadedLists extension would subsume OverloadedStrings extension (see the instance for Text, for example). Having said that, for now, I am not suggesting to remove the OverloadedStrings extension as it appears to be widely used. This extension could also be used for giving data-parallel array literals instead of the special syntax used currently. Unless there is a vocal opposition to the aforementioned two proposals, I would like to implement them in GHC. Both changes appear to be straightforward to implement. Thanks in advance for your feedback. Cheers, George [1] http://www-db.informatik.uni-tuebingen.de/files/giorgidze/haskell2011.pdf

12 19

log time instead of linear for case matching
by Greg Weber 10 Oct '11

10 Oct '11

We have a couple use cases in Yesod that can potentially match many different patterns. Routing connects the url of an http request to a Haskell function. The current code just uses a pattern match, which scales linearly. So if a site has a hundred different routes (urls), it could take 100 comparisons to finally match the url to a function. Michael Snoyman is writing some code to make this issue obsolete. One of the things it does is use a Map lookup instead of a case match. More worrying is our system for internationalization of a website. A user is supposed to make a sum type with every custom message as a constructor in that sum type. data Msg = Model | Year | Please -- Rendering function for English. renderEnglish Model = "Model" renderEnglish Year = "Year" renderEnglish Please = "Please fill in your car's details" -- Rendering function for Swedish. renderSwedish Model = "Modell" renderSwedish Year = "Årgång" renderSwedish Please = "Vänligen fyll i uppgifterna för din bil" So if there are 100 custom messages on a site, that will setup a case match with potentially 100 comparisons. Note that we are using this technique for type safety- switching to a map here would be difficult. We want to know at compile time that a translation exists for every message. So first of all I am wondering if a sum type comparison does in fact scale linearly or if there are optimizations in place to make the lookup constant or logarithmic. Second, I as wondering (for the routing case) if Haskell can make a string case comparison logarithmic so that users can use case comparisons instead of maps for string collections that are known at compile time and won't change.

3 3