February 2008 - Haskell-Cafe

Anyone mind proofing a short monad transformers explanation?
by Jack Kelly 22 Apr '08

22 Apr '08

I sent this to a friend who asked me about monad transformers and he reckons it should be saved somewhere. If anyone feels up to giving it a quick fact-check, that'd be great. I'm not sure what to do with it, though. It doesn't seem to be made of the right stuff for a tutorial but I'm not sure what to do with it. --- SNIP --- Basically it's like making a double, triple, quadruple, ... monad by wrapping around existing monads that provide wanted functionality. You have an innermost monad (usually Identity or IO but you can use any monad). You then wrap monad transformers around this monad to make bigger, better monads. Concrete example: Suppose I was writing a server. Each client thread must be of type IO () (that's because forkIO :: IO () -> IO ThreadID)). Suppose also that this server has some configuration that (in an imperative program) would be global that the clients all need to query. > data Config = Config Foo Bar Baz One way of doing this is to use currying and making all the client threads of type Config -> IO (). Not too nice because any functions they call have to be passed the Config parameter manually. The Reader monad solves this problem but we've already got one monad. We need to wrap IO in a ReaderT. The type constructor for ReaderT is ReaderT r m a, with r the shared environment to read from, m the inner monad and a the return type. Our client_func becomes: > client_func :: ReaderT Config IO () We can then use the ask, asks and local functions as if Reader was the only Monad: (these examples are inside do blocks) > p <- asks port (Assuming some function port :: Config -> Int or similar.) To do stuff in IO (or in the general case any inner monad) the liftIO function is used to make an IO function work in this wrapped space: (given h :: Handle, the client's handle) > liftIO $ hPutStrLn h "You lose" > liftIO $ hFlush h IO is once again special. For other inner monads, the lift function does the same thing. Note also that IO has no transformer and must therefore always be the innermost monad. This is all well and good, but the client_func now has type ReaderT Config IO () and forkIO needs a function of type IO (). The escape function for Reader is runReader :: Reader r a -> r -> a and similarly for ReaderT the escape function is runReaderT :: ReaderT r m a -> r -> m a: (Given some c :: Config that's been assembled from config files or the like) > forkIO (runReaderT client_func c) Will do the trick. Monad transformers are like onions. They make you cry but you learn to appreciate them. Like onions, they're also made of layers. Each layer is the functionality of a new monad, you lift monadic functions to get into the inner monads and you have transformerised functions to unwrap each layer. They're also like a present in that regard: in this example we unwrapped the outer wrapping paper to get to the present: an object of type IO (), which lets us make haskell do something.

3 2

Separate a string into a list of strings
by Sara Kenedy 18 Apr '08

18 Apr '08

Hi all, I want to write a function to separate a string into a list of strings separated by commas. Example: separate :: String -> [String] separate "Haskell, Haskell, and Haskell" = ["Haskell", "Haskell", "and Haskell"] If anyone has some ideas, please share with me. Thanks. S.

14 35

Libraries need a new owner
by Adrian Hey 27 Mar '08

27 Mar '08

Hello Folks, As some of you will be aware, I have been working on various Map implementations that currently live here.. http://code.haskell.org/collections/collections-ghc6.8 The libs in question being Data.Tree.AVL, Data.Trie.General and a few other bits like Data.COrdering and the AVL based Data.Map/Set clones. Well, I have decided to stop work on these. So they need a new owner if they're going to go anywhere. If anyone is interested in the job then I suggest they contact myself or Jean-Philippe Bernardy. Of course I will be happy to provide any help or advise anyone who takes over these libs may feel they need from me. I might even contribute a few patches from time to time myself :-) Thanks -- Adrian Hey

3 7

Implementing fixed-sized vectors (using datatype algebra?)
by Alfonso Acosta 15 Mar '08

15 Mar '08

Hi, The EDSL implementation (system design) I'm working on would really benefit from an implementation of fixed-sized vectors. I thought this would be a generally desired ADT but it turned out I wasn't able to find an implementation. I was thinking about using datatype algebra plus GADTs to implement a type-level parameter indicating the size of the vector. I'm a total noob with regard to GADTs and type-level algebra trickery. So my questions are: 1) Do you think it is feasible? Can you think about a better approach? 2) An implementation of type-level Naturals would really help. What has already been done? Thanks in advance, Fons

13 84

A question about "monad laws"
by Deokjae Lee 14 Mar '08

14 Mar '08

Tutorials about monad mention the "monad axioms" or "monad laws". The tutorial "All About Monads" says that "It is up to the programmer to ensure that any Monad instance he creates satisfies the monad laws". The following is one of the laws. (x >>= f) >>= g == x >>= (\v -> f v >>= g) However, this seems to me a kind of mathematical identity. If it is mathematical identity, a programmer need not care about this law to implement a monad. Can anyone give me an example implementation of monad that violate this law ?

28 58

Where does ~> come from?
by Steve Lihn 11 Mar '08

11 Mar '08

While I am reading TypeCompose module, I am having trouble finding where ~> comes from? (And its son ~~> and grandson ~~~>, etc.) This is my immediate question. Help is appreciated. A bigger (higher-order?) issue I encountered and I think other people could also have is to look up all these special "symbols" during the learning curve. As things are getting higher and higher order in Haskell, more inventive symbols are appearing. It becomes harder to remember the exact definitions are. The difficulties are: 1) I used to see a page on haskell.org that lists many symbols, but I could not find that page any more. 2) You would hope there is a quick way to search those symbols. But most search engines do not treate symbols friendly, often just ignore them. I typed ~> in Hoogle, it also returned nothing. 3) If the module defining the symbol is not in standard library, it is not possible to look up the symbol in the core library index. The only way to find the symbol seems to go to the source code, check every module being imported and if I am lucky I will find it in the same package. If I am not lucky, I will have to search that module on the Internet, and hunt it down recursively. I wonder if anybody has the same problem. If enough crowd have the same problem, maybe there could be a better way to handle this kind of documentation/learning issue! Thanks, Steve

5 4

Re: [Haskell-cafe] Doubting Haskell
by Cale Gibbard 06 Mar '08

06 Mar '08

(I'm copying the list on this, since my reply contains a tutorial which might be of use to other beginners.) On 19/02/2008, Alan Carter <alangcarter(a)gmail.com> wrote: > Hi Cale, > > On Feb 19, 2008 3:48 PM, Cale Gibbard <cgibbard(a)gmail.com> wrote: > > Just checking up, since you haven't replied on the list. Was my > > information useful? Did I miss any questions you might have had? If > > you'd like, I posted some examples of using catch here: > > Thanks for your enquiry! My experiment continues. I did put a progress > report on the list - your examples together with a similar long an > short pair got me over the file opening problem, and taught me some > things about active whitespace :-) I couldn't get withFile working > (says out of scope, maybe 'cos I'm ghc 6.6 on my Mac) Make sure to put: import System.IO at the top of your source file, if you haven't been. This should import everything documented here: http://www.haskell.org/ghc/docs/latest/html/libraries/base/System-IO.html > but it turned out the line I was looking for (collapsed from the examples) > was: > > text <- readFile "data.txt" `catch` \_ -> return "" > > This ensures the program never loses control, crashing or becoming > unpredictable by attempting to use an invalid resource, by yielding an > empty String if for any reason the file read fails. Then an empty > String makes it very quickly through parsing. I guess that's quite > "functiony" :-) > > Amazing how easy once I knew how. Even stranger that I couldn't find a > "bread and butter" example of it. > > Then I was going very quickly for a while. My file is dumped from a > WordPress MySql table. Well formed lines have 4 tab separated fields > (I'm using pipes for tabs here): > > line id | record id | property | value > > Line IDs are unique and don't matter. All lines with the same record > ID give a value to a property in the same record, similar to this: > > 1|1|name|arthur > 2|1|quest|seek holy grail > 3|1|colour|blue > 4|2|name|robin > 5|2|quest|run away > 6|2|colour|yellow > > Organizing that was a joy. It took minutes: let cutUp = tail (filter (\fields -> (length fields) == 4) (map (\x -> split x '\t') (lines text))) This should almost certainly be a function of text: cutUp text = tail (filter (\fields -> (length fields) == 4) (map (\x -> split x '\t') (lines text))) > I found a split on someone's blog (looking for a library tokenizer), > but I can understand it just fine. I even get to chuck out ill-formed > lines and remove the very first (which contains MySql column names) on > the way through! Sadly, there's no general library function for doing this. We have words and lines (and words would work here, if your fields never have spaces), but nobody's bothered to put anything more general for simple splitting into the base libraries (though I'm sure there's plenty on hackage -- MissingH has a Data.String.Utils module which contains split and a bunch of others, for example). However, for anything more complicated, there are also libraries like Parsec, which are generally really effective, so I highly recommend looking at that at some point. > I then made a record to put things in, and wrote some lines to play > with it (these are the real property names): > > data Entry = Entry > { occupation :: String > , iEnjoyMyJob :: Int > , myJobIsWellDefined :: Int > , myCoworkersAreCooperative :: Int > , myWorkplaceIsStressful :: Int > , myJobIsStressful :: Int > , moraleIsGoodWhereIWork :: Int > , iGetFrustratedAtWork :: Int > } > ... > let e = Entry{occupation = "", iEnjoyMyJob = 0} > let f = e {occupation = "alan"} > let g = f {iEnjoyMyJob = 47} > putStrLn ((occupation g) ++ " " ++ (show (iEnjoyMyJob g))) > > Then I ran into another quagmire. I think I have to use Data.Map to > build a collection of records keyed by record id, and fill them in by > working through the list of 4 item lists called cutUp. As with the > file opening problem I can find a few examples that convert a list of > tuples to a Data.Map, one to one. I found a very complex example that > convinced me a map from Int to a record is possible, but gave me no > understanding of how to do it. I spent a while trying to use foldl > before I decided it can't be appropriate (I need to pass more values). > So I tried a couple of recursive functions, something like: > > type Entries = M.Map Int Entry > ... > let entries = loadEntries cutUp > ... > loadEntries :: [[String]] -> Entries > loadEntries [] = M.empty Entries > loadEntries [x : xs] = loadEntry (loadEntries xs) x -- Possible common beginner error here: [x:xs] means the list with one element which is a list whose first element is x and whose tail is xs. Your type signature and the type of cutUp seems to confirm that this is the right type, but you don't seem to have a case to handle a longer list of lists. If you want just a list with first entry x, and with tail xs, that's just (x:xs). If you want to handle lists of lists recursively, you'll generally need two cases: ([]:xss) and ((x:xs):xss). We'll end up doing something different instead of recursion in a moment. > > loadEntry entries _ rid fld val = entries > > Trying to create an empty map at the bottom of the recursion so later > I can try to fiddle about checking if the key is present and crating a > new record otherwise, then updating the record with a changed one (a > big case would be needed deep in to do each property update). If I'm > on the right track it's not good enough to get better, so now I'm just > throwing bits of forest animals into the pot at random again :-( > > So I certainly would be grateful for a clue! The bits I can do (I got > a non-trivial wxHaskell frame sorted out quite easily, the tokenizing > and record bit were OK) I think show I'm not *totally* stupid at this, > I'm putting loads of time investment in (it's an experiement in > itself) but there do seem to be certain specific things that would be > ubiquitous patterns in any production or scripting environment, which > are not discussed at all and far from obvious. The more I see of > Haskell the more I suspect this issue is the gating one for popular > uptake. > > I couldn't help thinking of this bit, from the Wikipedia entry on the > Cocteau Twins: > > "The band's seventh LP, Four-Calendar Café, was released in late 1993. > It was a departure from the heavily-processed, complex and layered > sounds of Blue Bell Knoll and Heaven or Las Vegas, featuring clearer > and more minimalistic arrangements. This, along with the record's > unusually comprehensible lyrics, led to mixed reviews for the album: > Some critics accused the group of selling out and producing an > 'accessible album,' while others praised the new direction as a > felicitous development worthy of comparison with Heaven or Las Vegas." > > Best wishes, > > Alan I woke up rather early, and haven't much to do, so I'll turn this into a tutorial. :) Okay. The most common ways to build a map are by using the fromList, fromListWith, or fromListWithKey functions. You can see them in the documentation here: http://www.haskell.org/ghc/docs/latest/html/libraries/containers/Data-Map.h… The types are: fromList :: (Ord k) => [(k,a)] -> Map k a fromListWith :: (Ord k) => (a -> a -> a) -> [(k,a)] -> Map k a fromListWithKey :: (Ord k) => (k -> a -> a -> a) -> [(k,a)] -> Map k a They take a list of (key,value) pairs, and build a map from it. Additionally, the fromListWith function takes a function which specifies how the values should be combined if their keys collide. There is also a fromListWithKey function which allows the means of combination to depend on the key as well. At this point we realise something interesting about the way the data is being represented: if there is a field in someone's record with no row in the database, what should the resulting field contain? In C, they often use some integer which is out of range, like -1 for this. How about for a missing occupation field? Well, that's a String, you could use some generic failure string, or an empty string, but I'll show you another possibility that just might be convenient. If t is any type, then the type (Maybe t) consists of the values Nothing, and Just x, whenever x is a value of type t. This is another convenient way to represent the idea that a computation might fail. Let's start by changing your record type so that each field is a Maybe value, that is, either the value Nothing, or the value Just x, where x is the value it would have been. data Entry = Entry { occupation :: Maybe String , iEnjoyMyJob :: Maybe Int , myJobIsWellDefined :: Maybe Int , myCoworkersAreCooperative :: Maybe Int , myWorkplaceIsStressful :: Maybe Int , myJobIsStressful :: Maybe Int , moraleIsGoodWhereIWork :: Maybe Int , iGetFrustratedAtWork :: Maybe Int } There's a very general function in the module Control.Monad which I'd like to use just for the Maybe type here. It's called mplus, and for Maybe, it works like this: mplus (Just x) _ = Just x mplus Nothing y = y So if the first parameter isn't Nothing, that's what you get, otherwise, you get the second parameter. Of course, this operation has an identity element which is Nothing. So this lets you combine partial information expressed by Maybe types, in a left-biased way. It's about to become obvious that record types are less convenient than perhaps they could be in Haskell, and this is absolutely true -- I'd actually probably use a somewhat different representation myself (possibly something involving a Map from Strings (field names) to Int values), but I can't really be sure what you intend with this data, and how much type safety you want. I'll elide the field names just because I can here. It's not necessarily good style. combine :: Entry -> Entry -> Entry combine (Entry a1 a2 a3 a4 a5 a6 a7 a8) (Entry b1 b2 b3 b4 b5 b6 b7 b8) = Entry (a1 `mplus` b1) (a2 `mplus` b2) (a3 `mplus` b3) (a4 `mplus` b4) (a5 `mplus` b5) (a6 `mplus` b6) (a7 `mplus` b7) (a8 `mplus` b8) Even with all the shorthand, this is pretty ugly (and I'll show how I'd represent the data in a moment), but what this does is to combine two partial entries, favouring the information in the first, but filling the holes in the first with data from the second. This operation has an identity element, which is: emptyEntry = Entry Nothing Nothing Nothing Nothing Nothing Nothing Nothing Nothing Let's try a different representation, which is a little more flexible, but expresses less in the type system. data Entry = Entry { occupation :: Maybe String, survey :: M.Map String Int } deriving (Eq, Ord, Show) So now, instead of a bunch of separate Maybe Int fields, we have just one Map from String to Int. If we don't have information for a field, we simply won't have that key in the Map. Of course, this means we'll have to use strings for field labels. If that seems unhappy, you could always define a type like: data SurveyQuestion = IEnjoyMyJob | MyJobIsWellDefined | MyCoworkersAreCooperative | MyWorkplaceIsStressful | MyJobIsStressful | MoraleIsGoodWhereIWork | IGetFrustratedAtWork deriving (Eq, Ord, Show) to be used in place of the String type. Let's see how combine will look now: combine :: Entry -> Entry -> Entry combine (Entry o1 s1) (Entry o2 s2) = Entry (o1 `mplus` o2) (s1 `M.union` s2) Or, using the record syntax more: combine :: Entry -> Entry -> Entry combine e1 e2 = Entry { occupation = (occupation e1 `mplus` occupation e2), survey = (survey e1 `M.union` survey e2) } Again, this new version has an identity with respect to combine, which is: emptyEntry = Entry {occupation = Nothing, survey = (M.empty)} Now, we just need a way to take one of your rows, and turn it into a (key,value) pair, where the value is a partial entry. readRow :: [String] -> (Int, Entry) readRow [n, k, "occupation", v] = (read k, emptyEntry { occupation = Just v }) readRow [n, k, f, v] = (read k, emptyEntry { survey = M.singleton f (read v) }) readRow xs = error "failure case, should never happen!" There is actually a failure case that I'm not handling here, which is what happens when the value or key fails to parse as an Int. For that we'd use reads instead of read, but let's ignore it for now. We can then map this function over our cut up rows, something along the lines of: map readRow (cutUp text) at which point we'll have a list of (Int, Entry) pairs. We then want to fill up our Entries Map with those, and we want to combine them as we go using the combine function: entryMap text = M.fromListWith combine (map readRow (cutUp text)) Some final changes we could consider would be putting more of the error handling into readRow itself: if it was to return a singleton Map rather than an (Int, Entry) pair, then it could return the empty Map on failure, and the results would then be combined using the function M.unionsWith combine. We could move the length 4 test out of cutUp then, and just make it the fall-through case in readRow. I'll also use reads, which returns a list of (parse,rest-of-string) pairs, to handle the failure cases where the numbers don't parse, by just treating those rows as nonexistent: readRow :: [String] -> M.Map Int Entry readRow [n, k, f, v] = case reads k of [] -> M.empty -- the key didn't parse (k',_):_ -> if f == "occupation" then M.singleton k' (emptyEntry { occupation = Just v }) else case reads v of [] -> M.empty -- the value didn't parse (v',_):_ -> M.singleton k' (emptyEntry { survey = M.singleton f v' }) readRow xs = M.empty -- this handles the case when the list is any length but 4 cutUp text = tail (map (\x -> split x '\t') (lines text)) -- which allows cutUp to be simpler entryMap text = M.unionsWith combine (map readRow (cutUp text)) Anyway, I hope this tutorial gives some idea of how things progress, and what sort of thinking is generally involved. Note that the focus here was more on finding the right combining operations, and then using pre-existing higher-order functions to collapse the structure, than it was on recursion. Indeed, the final program there doesn't contain any explicit recursion at all! This is generally something to aim for. Using explicit recursion is generally a means of last resort. Higher order functions on data structures are our control structures. - Cale

14 15

Connection helpers: for people interested in network code
by Adam Langley 05 Mar '08

05 Mar '08

I generally find that I'm wrapping sockets in the same functions a lot and now I'm looking writings code which works with both Sockets and SSL connections. So I wrote a module, presumptuously called Network.Connection, although I'm not actually going to try and take that name (even in Hackage) unless I get a general agreement that this is a good thing. So, any comments on the interface, similar things that I should look at etc? http://www.imperialviolet.org/binary/network-connection/Network-Connection.… I made the BaseConnection an ADT, rather than a class because I wanted to avoid hitting the monomorphism restriction in code. That might have been a mistake, I'm not sure yet. If it doesn't excite anyone enough to reply, I'll change the name and put it in Hackage, mostly as is. Then I'll tie HsOpenSSL into it so that SSL connections work transparently. Cheers, AGL -- Adam Langley agl(a)imperialviolet.org http://www.imperialviolet.org

4 5

"Wrong kind" when attempting to build a monad for a circular list of functions
by Aaron Altman 02 Mar '08

02 Mar '08

Hi everybody. I'm working towards a better understanding of Haskell monads the only way I know how: by working through an example. I am working on an AI agent that will perform a finite series of actions before starting the sequence over again. I figured a circular list of functions that shifts as you apply them would be the way to do it. Here is my test code: ----------------------------- import Control.Monad runActionAndIterate :: [a -> a] -> a -> (a, [a -> a]) runActionAndIterate (currentAction:actionList) actionInput = (currentAction actionInput, concat [actionList, [currentAction]]) shiftActionList :: [a -> a] -> [a -> a] shiftActionList (currentHead:rest) = concat [rest, [currentHead]] newtype CircularFuncList funcList arg = CircularFuncList (funcList -> arg -> (arg, funcList)) instance Monad (CircularFuncList funcList arg) where return a = CircularFuncList (\funcList a -> (a, funcList)) CircularFuncList currentFuncList currentArg >>= argTransform = let result = argTransform $ (head currentFuncList) currentArg newFuncList = map argTransform $ shiftActionList currentFuncList in CircularFuncList newFuncList result ----------------------------- I get an error that CircularFuncList funcList arg has kind * while Monad is looking for * -> *. This is a first attempt so it may be I'm a ways off from a monad that implements a circular list of functions. That is the goal though. What advice can you offer? Thanks, Aaron Altman

5 8

Re: static constants -- ideas?
by Don Stewart 29 Feb '08

29 Feb '08

jay: > Don Stewart dons(a)galois.com: > > >jay: > >> Don Stewart dons(a)galois.com: > >> >jay: > >> >> Don Stewart dons(a)galois.com: > >> >> >jay: > >> >> >> I also have constants that are too large to compile. I am resigned to > >> >> >> loading them from data files--other solutions seem even worse. > >> >> ... > >> >> >> Data.Binary eases the irritation somewhat. > >> >> > > >> >> >Did you try bytestring literals (and maybe parsing them in-memory with > >> >> >Data.Binary)? > >> > >> I finally squeezed enough time to try it, and it didn't work for me. > > > >> > >> -- > >> ghc Overflow.hs > >> [1 of 1] Compiling Overflow ( Overflow.hs, Overflow.o ) > > > >Enable optimisations! Compile with ghc -O2. You need this to avoid > >having a very slow pack call at runtime. > > Yes, I tried basic variations like that. The result is the same with -O1 > or with -O2, and with Data.ByteString or Data.ByteString.Lazy . Ok, hmm, that really shouldn't be the case. Do you have the example available somewhere? It's just a 40M inline bytestring? -- Don

1 0