- Beginners - Haskell.org

A rigid type and a better pattern
by Javier M Mora 16 Aug '09

16 Aug '09

(I had problems sending this email and I don't know if earlier arrived to mail list. I'm going to send again. Sorry if you get it duplicate) Hi All, this is my first email to beginners list. I'm ending my Master Degree in Robotics (in Spain). One of my assignment is about Motion Planning and I thought solve with Haskell. This is my problem: * Every robot has a local state. I can get the position of the robot from the state. The state can be rect coords or polar coords or distance moved on a fixed path or whatever. * I have a global state witch keep the local state of each robot. * Every state has a different system to action over it. So, in distance over fixed path we can "stop" or "go". In polar coords we can increase and decrease the radius and phi. And in rect coords we can move in x and y direction * I have several robots at same time. But every robot has the same state and the same action. So, all robots in the run of my program has same local state and action. So, I split the algorithm in a common part (used in all states and actions) and other for each specific state-action. This is the structure (abstract): class Robot a where ... class State a where getpos a -> [Position] ... class Action a where ... later I have instances for concrete cases. In phase 1, local state is a Int for each robot and action is a Bool (stop or run). data State1 = State1 [Int] instance State State1 where getpos (State1 a) = ... ... data Action1 = Action1 [Bool] instance Action Action1 where ... I have a function "newstate" to calculate the new state from an old state and the action. This is the type signature: newstate:: (State a,Action b) => a -> b -> a and my instantiation in phase 1 is newstate (State1 s) (Action1 u) = State1 $ zipWith (\x y -> if y then succ x else x) s u I want to use polymorphism in newstation because I need use newstate (State2 s) (Action2 u) in the future (phase 2) so I tried include newstate in a typeclass. But if I write: class State a where newstate:: (Action b)=> a -> b -> a ... or class Action a where newstate:: (State b) => b -> a -> b ... (and move newstate definition to correct instance block) I get similar errors in both cases: (this is the first case): Couldn't match expected type `b' against inferred type `Action1' `b' is a rigid type variable bound by the type signature for `newstate' at Tipos.hs:17:24 In the pattern: Action1 u In the definition of `newstate': newstate (State1 s) (Action1 u) = State1 $ zipWith (\ x y -> if y then succ x else x) s u In the instance declaration for `State State1' My questions: * How can I fix this error? * the point is associate every instance of Robot with one instance of State and with one instance of Action. So when I select the Robot I'm goint to use I get "automagically" the correct instances of newstate and others. Is there a pattern to this behaviour? My ideal is associate every function related with a Robot with a general typeclass "Robot" (I don't tested this) : class Robot a where initstate:: (State a) => a newstate:: (State a,Action b) => a-> b-> a And the instance: instance Robot Robot1 where initstate = State1 [0,0,0] newstate (State1 a) (Action1 b) = ... and instance Robot Robot2 where initstate = State2 [(0,0),(1,1),(2,2)] -- i.e. newstate (State2 a) (Action2 b) = ... and so on. But I think It can't work because neither initstate or newstate use Robot1. Any tips? Javier M Mora.

2 1

Circular programming
by Maciej Piechotka 16 Aug '09

16 Aug '09

I'm not understending circular programming. What it is used for? Where feedback variable came from (yes - I know it's wrong question as Haskell is lazy)? I've read http://www.haskell.org/haskellwiki/Circular_programming but I didn't understend the concept and pattern. Regards

2 3

List combination function?
by Ian Duncan 16 Aug '09

16 Aug '09

Hello all, I'm trying to build a function that takes a string such as "123" and gives me permutations including permutations with lesser list lengths. I'm not sure how to phrase it, but here is what the output could look like: foo "123" => ["123","213","321","231","312","132", "12", "13", "21", "23", "31", "32", "1", "2", "3", ""] The ordering doesn't matter, and that null list at the end doesn't particularly matter, but I don't really know the mathematical phrasing of what I'm asking for. I'm trying to build a scrabble helper that can find the optimal score given a set of letters to work with. Thanks for your help, Ian Duncan

2 1

monad transformers
by Michael P Mossey 15 Aug '09

15 Aug '09

In Martin Grabmuller's tutorial "Monad Transformers Step by Step", found here http://user.cs.tu-berlin.de/~magr/pub/Transformers.pdf he gives an example of composing ErrorT, StateT, ReaderT, and WriterT. Early in the paper, where he composes just ErrorT and ReaderT type Eval3 a = ReaderT Env (ErrorT String Identity) a he uses 'ask' and 'throwError' in some example code. I notice that he doesn't have to lift throwError into the ErrorT monad. Why is this? Do I misunderstand something about monad transformers? Is it a convenience definition of throwError? Same thing the rest of the paper. I don't see anywhere he lifts anything. Thanks, Mike

3 3

some terminology
by Michael P Mossey 14 Aug '09

14 Aug '09

I was looking at some code, saw a variable x, and said to myself, "Ah that variable is a monad." Then I realized "Monad" is the name of a type class. So maybe x should be called "an instance of a Monad." I think the word "instance" in this case is OO-like; but in Haskell "instance" refers to a type that is an instance of a type class. Or maybe it can refer to both? And Monad is a type class, not a type. Maybe I need the phrase "monadic type" to refer to an instance of a type class. So maybe x is just "a variable of a monadic type"? Thanks, Mike

4 3

function parameter types after flip
by Robert Ziemba 14 Aug '09

14 Aug '09

Can anyone help me understand the behavior of the following expressions in ghci? I was trying to construct a flipped function 'elem' that would accept a list of characters first but I ran into this problem with the type signature. Thank you. Prelude Data.List Data.Char> :t elem elem :: (Eq a) => a -> [a] -> Bool -- OK Prelude Data.List Data.Char> :t (flip elem) (flip elem) :: (Eq a) => [a] -> a -> Bool -- OK this is what I want Prelude Data.List Data.Char> let fElem = (flip elem) Prelude Data.List Data.Char> :t fElem fElem :: [()] -> () -> Bool -- ?? Function will not accept a [Char]

3 2

brief example of ZipList?
by Michael P Mossey 11 Aug '09

11 Aug '09

Can someone give me a brief example using ZipList? I want to do something like z1 :: [Int -> Int] z1 = [succ,succ] z2 :: [Int] z2 = [1,2] z3 = z1 <*> z2 But don't know to get it to regard these as ZipLists and not regular lists. This is purely for learning purposes. No application in mind. -Mike

3 2

some insights into functional programming
by Michael Mossey 11 Aug '09

11 Aug '09

I'm starting to figure out a few things that I didn't "get" about functional programming and monads. I wanted to explain them. I'm not looking for a particular response to this post, other than any elaboration that seems natural. There is an exercise here working with the trivial monad W: <http://blog.sigfpe.com/2007/04/trivial-monad.html> Write a function g :: W a -> W a -> W a such that g (W x) (W y) = W (x+y) except don't use pattern matching, but >>= instead. The answer is g mx my = mx >>= (\x -> my >>= \y -> W (x+y)) There are a couple things here that threw me off. One is that I didn't expect 'my' to be available inside the first lambda. I somehow thought of lambda as isolated, sealed-off from the rest of the universe. But they aren't. I believe this is the concept of closures, or related to it? Secondly, I didn't expect >>= to be available inside the lambda. This is related to the mistaken conception of >>= as a procedural statement rather than an expression. In Python, where I have previously encountered lambdas, no statements are allowed inside lambdas. Of course, >>= is actually an expression and you can put any expression to the right of a lambda ->. Maybe these are typical beginner misconceptions, or maybe they have more to do with coming from Python and complete beginners actually find it more natural. Mike

5 4

creating a complement for a given given test generator.
by Srikanth K 11 Aug '09

11 Aug '09

Hi, I am trying to use the quickcheck to generate some test-data to test an api, along the lines of http://www.haskell.org/haskellwiki/QuickCheck_as_Test_Set_Generator. For the sake of example, I choose the following function to test data Result = Valid | Invalid api_under_test :: (Int,Int) -> Result api_under_test (x,y) | (x == 1) = Valid | otherwise = Invalid I had the following valid-generator which worked quite easily(trivial) validCombinations= [ (1,1), (1,2) ] validGen = elements validCombinations prop_valid_api_under_test = forAll validGen $ \xs -> (api_under_test xs) == (Valid) Now, I want to have a complement to state: forall tuples not in validCombinations, the api_under_test must return "Invalid". (i.e.) prop_invalid_api_under_test = forAll invalidGen $ \xs -> (api_under_test xs) == (Invalid) However, inspite of all googling, and reading the various docs including quickcheck, I am at loss on how I can elegantly define the "invalidGen" generator. One possible way I can think is to have a customized generator, that would generate two random numbers and then look up the tuple generated against the validCombinations list. However, I feel there just might be a better way of solving this. Any suggestion on how I should be trying to solve this. - Srikanth

2 1

unit tests
by Michael Mossey 10 Aug '09

10 Aug '09

As I write Haskell code I find myself needing to write test code. If this test code were organized and documented, I could probably call it a set of unit tests. I'm trying to find some convention for naming functions and variables used in unit testing, deciding on location (in same module or a second module? same directory or sub-directory?), etc. Are there any Haskell community conventions? Thanks, Mike

3 2