August 2003 - Haskell

RE: container for different types, avoiding boiler plate
by Hal Daume 21 Aug '03

21 Aug '03

Yes, this is essentially what DynamicMap is....The url i posted before was wrong...it should have been: www.isi.edu/~hdaume/haskell/DynamicMap.hs -- Hal Daume III | hdaume(a)isi.edu "Arrest this man, he talks in maths." | www.isi.edu/~hdaume > -----Original Message----- > From: Markus.Schnell(a)infineon.com > [mailto:Markus.Schnell@infineon.com] > Sent: Thursday, August 21, 2003 1:58 AM > To: haskell(a)haskell.org > Cc: Hal Daume; ddarius(a)hotpop.com > Subject: Re: container for different types, avoiding boiler plate > > > After taking a look at Data.Dynamic I came up with: > > > data Gender = Masc | Fem | Neutr deriving (Typeable, Show) > > data Number = First | Second | Third deriving (Typeable, Show) > > > > type Attrs = [Dynamic] > > > > attrs = [toDyn Masc, toDyn Second] > > > > gattr :: (Typeable a) => Attrs -> Maybe a > > gattr [] = Nothing > > gattr (x:xs) = case fromDynamic x of > > Nothing -> gattr xs > > Just x -> Just x > > > > type GetAttr a = Attrs -> Maybe a > > ggender = gattr :: GetAttr Gender > > gnumber = gattr :: GetAttr Number > > which already is much more readable. But I'm sure > problems are just around the corner ... > > Thank you. > Markus > >

1 0

RE: overlapping instances and functional dependencies
by Simon Peyton-Jones 21 Aug '03

21 Aug '03

| With overlapping instances, I'm allowed | | class OK x y | | instance Functor f => OK (f x) (f y) | | instance Functor f => OK x (f y) | | but I'm not allowed | | class Bad x y z | x y -> z | | instance Functor f => Bad (f x) (f y) Bool | | instance Functor f => Bad x (f y) Int | | I don't quite see why. Naively, I imagine that if the OK instances are | effectively prioritized, then Bad's rules for computing z from x and y | could be similarly prioritized. Can anyone explain why this naive | imagining is incorrect? I don't think anyone has really worked through the interaction of overlapping instances (already swampy) with functional dependencies. What you say makes sense, but it's not what GHC or Hugs implement. Maybe they should Simon

1 0

RE: overlapping instances and functional dependencies
by Simon Peyton-Jones 21 Aug '03

21 Aug '03

| class C a b c | a b -> c where | f :: a -> b -> c | | instance C a b c => C a (x,y,b) c where | f a (_,_,b) = f a b | | instance C a (a,c,b) c where | f _ (_,c,_) = c | ghci -fglasgow-exts -fallow-overlapping-instances compiles it without | complaint but hugs -98 +o says: | ERROR "ClassProblem.hs":7 - Instances are not consistent with dependencies | *** This instance : C a (a,b,c) b | *** Conflicts with : C a (b,c,d) e | *** For class : C a b c | *** Under dependency : a b -> c There are several things going on here. 1. The functional dependency (a b -> c), together with an instance declaration instance C <ta> <tb> <tc> means that if you have a constraint (C <sa> <sb> <sc>), and <sa> matches <ta> and <sb> matches <tb> then <sc> must look like <tc> Here "matches" means "is a substitution instance of". Mark Jones's original paper specified that the free variables of <tc> must be a subset of the free variable of <ta> and <tb> Reason: then the matching of <sa> and <sb> generates a substitution that covers all the free variables of <tc>, and then "<sc> must look like <tc>" means exactly that "<sc> must be the same as <tc> instantiated by the matches of <ta>/<sa> and <tb>/<sb>". However, it seems that neither Hugs nor GHC make this test, which is very naughty of them 2. Having omitted this test, Hugs is complaining that a constraint C Int (Int,Bool,Bool) <sc> would match both instance declarations, so it's not sure which <tc> to make <sc> equal to. I could check why GHC does not complain, but it may be that the answer is the same in both cases (namely the unbound 'c'). 3. None of this has to do with overlapping instances per se. I'm working with Martin Sulzmann on some ideas for loosening the restriction (1) above, but it's all a bit vacuous in your example. What is the fundep is buying you? It doesn't convey any information to say that "<sc> must look like c", if c is completely unbound! Simon

1 0

Re: container for different types, avoiding boiler plate
by Markus.Schnell＠infineon.com 21 Aug '03

21 Aug '03

After taking a look at Data.Dynamic I came up with: > data Gender = Masc | Fem | Neutr deriving (Typeable, Show) > data Number = First | Second | Third deriving (Typeable, Show) > > type Attrs = [Dynamic] > > attrs = [toDyn Masc, toDyn Second] > > gattr :: (Typeable a) => Attrs -> Maybe a > gattr [] = Nothing > gattr (x:xs) = case fromDynamic x of > Nothing -> gattr xs > Just x -> Just x > > type GetAttr a = Attrs -> Maybe a > ggender = gattr :: GetAttr Gender > gnumber = gattr :: GetAttr Number which already is much more readable. But I'm sure problems are just around the corner ... Thank you. Markus

1 0

RE: more unsafePerformIO questions (is it safe to use with ReadMode Handles)?
by Simon Marlow 20 Aug '03

20 Aug '03

> >If the databases *do* change over time, then there are two > >possibilities: > > > > 1. the contents change due to external factors only > > 2. the contents change because this program doing the writing > > > >in (1), you can still pretend the interface is pure, by > imagining that > >all the changes happened earlier. This works as long as you > only read > >the external data once. > > Isn't there the possibility of inlining causing a read to > happen twice even if it only appears to happen once? In theory that would be a valid transformation, but in practice no compiler would duplicate arbitrary computations. GHC certainly doesn't. Cheers, Simon

5 5

RE: container for different types, avoiding boiler plate
by Hal Daume 20 Aug '03

20 Aug '03

I use my 'DynamicMap' type to handle this sort of thing. However, I don't really recommend this approach unless you're very careful. You basically lose out on all nice type checking properties and enter a world of dynamic typing (more or less). Anyway, you can find it at: http://www.isi.edu/~hdaume/DynamicMap.hs it uses "NLP.FiniteMap", but you can replace this with "Data.FiniteMap". You can then do things like: data Gender = Masc | Fem | Neutr deriving Typeable data Number = First | Second | Third deriving Typeable let dm = addToDM (addToDM emptyDM Masc) Second case lookupDM dm of Just Masc -> "is a guy" Just _ -> "is not a guy" _ -> "i don't know gender" case lookupDM dm of Just First -> "is first" Just Second -> "is second" _ -> "either i don't know or is third" of course 'deriving Typeable' means you need GHC6; otherwise you can write the instances by hand. -- Hal Daume III | hdaume(a)isi.edu "Arrest this man, he talks in maths." | www.isi.edu/~hdaume > -----Original Message----- > From: haskell-admin(a)haskell.org > [mailto:haskell-admin@haskell.org] On Behalf Of > Markus.Schnell(a)infineon.com > Sent: Wednesday, August 20, 2003 5:01 AM > To: haskell(a)haskell.org > Subject: container for different types, avoiding boiler plate > > > I think similar things have been asked before, but I couldn't > find anything > specific. > I have a data type with attributes. These attributes have > different types. > Right now I'm using a lot of boilerplate like that: > > > > data Gender = Masc | Fem | Neutr > > ... > > data Attr = Gender Gender | Cat Cat | Graph Graph | ... > > data Type = TypeCat | TypeGender | ... deriving Eq > > > > myTypeOf (Gender _) = TypeGender > > myTypeOf (Cat _) = TypeCat > > ... > > myTypeOf _ = TypeError > > > > data Segment = Seg { attrs :: [Attr] } > > > > attr f seg = seg { attrs = f (attrs seg) } > > > > gattr :: Type -> [Attr] -> Maybe Attr > > gattr theType [] = fail "attribute not found" > > gattr theType (a:as) = if myTypeOf a == theType then return > a else gattr > theType as > > > > cat :: Cat -> Segment -> Segment > > cat c = attr ((Cat c):) -- set value > > > > gcat :: Segment -> Maybe Cat -- get value > > gcat = deCat . gattr TypeCat . attrs > > where deCat (Just (Cat c)) = c > > deCat x = x > > ... > > Does anyone have some suggestions for making this more concise? > Generic Haskell? Tricky type classes? > > Thanks, > Markus > > -- > Markus Schnell > Infineon Technologies AG, CPR ET > Tel +49 (89) 234-20875 > > _______________________________________________ > Haskell mailing list > Haskell(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell >

2 1

'Forest inversion'?
by Marnix Klooster 20 Aug '03

20 Aug '03

Hi all, Recently I've discovered an inversion operation on forests that transforms 'wide' forests into 'deep' onces and vice versa. I'm sure that this operation is already known, however, I could not find any information on it. (Largely because I don't know under what name to look for it.) You Haskell guys know about data structures, so you should know more. :-) Example (use a fixed with font to view): the single-tree forest A /|\ B C D | | |\ E F G H | | I J | K is inverted to the forest A B E /| |\ C F I K / \ D G / \ H J and vice versa. In an implementation where every node has two pointers, one to its first child and one to its right-hand sibling, the inversion means that in every node those two pointers are swapped. In Haskell the algorithm looks like this: > data Forest a = Forest {trees :: [(a, Forest a)]} deriving (Show, Eq) > > inv :: Forest a -> Forest a > inv (Forest []) = Forest [] > inv (Forest ((a,f) : g)) = Forest ((a, inv (Forest g)) : trees (inv f)) and it is easy to prove that for every f :: Forest a, inv (inv f) = f. What would I want to do with it? Well, deep trees are difficult to navigate using existing tree controls. Example: Suppose I want to let the user play a text adventure, but I allow 'undo'. Then I can show him the tree of moves that he already tried, and let him backup to a previous state, and try something else at that point. This results often in a deep tree. So showing a wide tree instead (using the above inversion) might help the user. Especially if the children of a node are 'ordered' in the sense that the first child node is most important. So: is this 'forest inversion' known, under which name, what are the known uses, etc.? Thanks in advance! Groetjes, <>< Marnix -- Marnix Klooster mklooster(a)baan.nl

2 1

GPCE 03 Call for Participation
by Akos Ledeczi 20 Aug '03

20 Aug '03

GPCE'03: 2nd International Conference on Generative Programming and Component Engineering In Cooperation with ACM SIGPLAN and SIGSOFT and co-located at NetObjectDays'03 September 22-25, 2003, Erfurt, Germany http://gpce.org/GPCE03 GPCE is the premier forum for dissemination and discussion of ideas on domain-specific languages, generative programming, model-based development, and component engineering. It brings together software engineering and programming language researchers and practitioners interested in key technologies for automating program development. This second conference will be co-located with NetObjectDays'03, one of the largest main-stream software development conferences in Germany. GPCE'03 program features 2 invited talks, by Olivier Danvy and Peri Tarr 21 papers featuring latest research in - Domain-Specific Languages - Staged Programming - Modeling-Based Development - Aspect-Orientation - Meta-Programming and Language Extension - Domain-Specific Code Generation 4 workshops - Reflectively Extensible Programming Languages and Systems - Product Line Engineering - The early steps: Planning, Managing, and Modeling - Component Engineering Methodology - GPCE Young Researchers Workshop 6 tutorials - Executable UML - Domain-Specific Languages - Multi-Staged Programming - Generatrive Programming - Model-Driven Development - Model-Driven Architecture 6 tool demonstrations of tools for - Metamodeling - Model-Driven Architecture - Generative development GPCE'03 attendees will have access to all NetObjectDays'03 events including Keynote Addresses by Steve Mellor, Jim Odell, and Dave Weiss. Poster submissions and submissions to the Product Line Engineering Workshop are still being solicited. For details of the conference program please refer to http://gpce.org/GPCE03.

1 0

Weighted Automata: Workshop Announcement
by Sandra Grossmann 20 Aug '03

20 Aug '03

[apologies if you receive multiple copies of this message] ----------------------------------------------------------- Weighted Automata: Theory and Applications Dresden University of Technology, June 1 - 5, 2004 The workshop will cover all aspects of weighted automata, ranging from the theory of formal power series to max-plus- algebra and applications as e.g. in digital image processing or algebraic optimization of networks. The aim is to present tutorials and survey lectures by outstanding scientists in this area. Moreover, we encourage everybody to participate in this workshop and present their own technical contribution in this area. Tutorials: P. Butkovic (Birmingham) U. Montanari (Pisa) J. Kari (Iowa City) J. Sakarovitch (Paris) Survey lectures: J. Albert (Wuerzburg) F. Katritzke (Siegen) S. Bozapalidis (Thessaloniki) W. Kuich (Wien) Z. Esik (Szeged) M. Mohri (AT&T, New Jersey) P. Gastin (Paris) L. Staiger (Halle) All lecturers are asked to give a survey on their field of interest followed by recent research results. Submission: If you wish to present a technical contribution, please send us an abstract (preferably of at most one page) before February 15, 2004. Authors will be informed about acceptance of their submission before March 15, 2004. There will be a special issue of the Journal of Automata, Languages, and Combinatorics (JALC) on the topic of this workshop. The deadline for submission of papers is July 15, 2004. All submissions will be refereed according to the usual journal standards. Registration: Interested participants are most welcome. There are no conference fees. Participants from central/eastern european countries can apply for financial support. Important Dates: deadline for submission of abstracts February 15, 2004 notification of acceptance March 15, 2004 registration April 1, 2004 submission of papers July 15, 2004 Further Information: http://www.orchid.inf.tu-dresden.de/gk-spezifikation/wata4.html --------------------------------------------------------------- Manfred Droste Heiko Vogler Dresden University of Technology Dresden University of Technology Institute of Algebra Inst. of Theor. Computer Science D-01062 Dresden, Germany D-01062 Dresden, Germany Tel.: +49-351-463-33908 Tel.: +49-351-463-38232 Fax: +49-351-463-34235 Fax: +49-351-463-37959 E-Mail: droste(a)math.tu-dresden.de E-Mail: vogler(a)inf.tu-dresden.de

1 0

container for different types, avoiding boiler plate
by Markus.Schnell＠infineon.com 20 Aug '03

20 Aug '03

I think similar things have been asked before, but I couldn't find anything specific. I have a data type with attributes. These attributes have different types. Right now I'm using a lot of boilerplate like that: > data Gender = Masc | Fem | Neutr > ... > data Attr = Gender Gender | Cat Cat | Graph Graph | ... > data Type = TypeCat | TypeGender | ... deriving Eq > > myTypeOf (Gender _) = TypeGender > myTypeOf (Cat _) = TypeCat > ... > myTypeOf _ = TypeError > > data Segment = Seg { attrs :: [Attr] } > > attr f seg = seg { attrs = f (attrs seg) } > > gattr :: Type -> [Attr] -> Maybe Attr > gattr theType [] = fail "attribute not found" > gattr theType (a:as) = if myTypeOf a == theType then return a else gattr theType as > > cat :: Cat -> Segment -> Segment > cat c = attr ((Cat c):) -- set value > > gcat :: Segment -> Maybe Cat -- get value > gcat = deCat . gattr TypeCat . attrs > where deCat (Just (Cat c)) = c > deCat x = x > ... Does anyone have some suggestions for making this more concise? Generic Haskell? Tricky type classes? Thanks, Markus -- Markus Schnell Infineon Technologies AG, CPR ET Tel +49 (89) 234-20875

1 0