February 2015 - Beginners

abstracting parametrized record types
by Alia 30 Apr '15

30 Apr '15

Hi Folks, Here's a problem that's frustrating me: I have defined a parametrized record type which looks like this: data Job a = Job { jobID :: String , jobName :: String , jobTable :: String , jobSource :: IO [a] , jobProcessors :: [a -> a] } where a is a data model type that maps on to a database table, for example: data Person = Person { name :: String , age :: Int } deriving (Show) data Car = Car { brand :: String , value :: Double , year :: Int } deriving (Show) I would like to define a higher-level record type which contains, for configuration purposes, a certain set of jobs for execution. Let's say we call it a JobSet and which could possibly look like this: data JobSet = JobSet { jobsetID :: String , jobsetName :: String , jobs :: [Job] <-- yes I know this is not legal } Is there a legal haskell way to achieve the above objective without having to do something like this which hardcodes the model type into the jobset schema? data JobSet = JobSet { jobsetID :: String , jobsetName :: String , personJobs :: [Job Person] , carJobs :: [Job Car] } Many thanks for any enlightenment on this front. Alia

2 1

Avoiding stack space overflow
by Hans Georg Schaathun 11 Mar '15

11 Mar '15

Hi, can someone give some hints on how to get around a stack space overflow? My problem is with the training function for a neural network: trainNetwork :: Double -> Samples -> Int -> Network -> Network trainNetwork _ _ 0 n = n trainNetwork eta samples c n = trainNetwork eta samples (c-1) $! epoch eta n samples epoch :: Double -> Network -> Samples -> Network So trainNetwork runs epoch c times, each time taking a Network in and modifying the Network as output. Clearly, space complexity can be made constant in c, but I get stack overflow if and only if c is too large. As you can see, I have tried to make the epoch evaluation strict ($!). I have also tried bang patterns on the input parameter n, and I have tried rewriting with foldr/foldl/foldl', and I have tried switchin the inner and outer calls (epoch vs. trainNetwork), all to no avail. I reckon this loop like pattern should be fairly common ... does it have a common solution too? TIA -- :-- Hans Georg

5 9

FFI Problems
by Thomas Jakway 04 Mar '15

04 Mar '15

I'm very new to Haskell and am trying to write a "real" program to motivate myself to learn it better (so far I've only gotten through Project Euler problems after reading LYAH and most of RWH). I'm using Taglib (https://github.com/taglib/taglib) to read the metadata from a music file and print it. I have a struct C-side (with C linkage) serving as the bridge between Taglib's C++ and Haskell's FFI. A small demo program (compiled with gcc and linked against the C++ object files) gives the correct results, but Haskell is weirdly only getting /some /of it right. Specifically, the C string fields are working but ints are not. The output from the C demo (what Haskell should be printing): music_metadata title: It's My Life, artist: Bon Jovi, album: Bon Jovi Greatest Hits - The Ultimate Collection comment: , genre: Rock, track: 3, length: 224, bitrate: 256, channels: 2, codec: 768 The output from Haskell: MusicMetadata {codec = UNKNOWN, length = 1099511628000, bitrate = 8589934848, channels = 12884901890, track = 8589934848, title = "It's My Life", artist = "Bon Jovi", album = "Bon Jovi Greatest Hits - The Ultimate Collection", comment = "", genre = "Rock"} I would have expected it to work or not work at all, but did not anticipate getting only some of it right. I was going to include snippets from my hsc file but given how new I am to Haskell I don't trust myself to assume where the problem is, so sorry if this is way too long: {-# LANGUAGE CPP, ForeignFunctionInterface #-} module MusicReader ( Codec, MusicMetadata, readMusicMetadata ) where import Control.Monad import Foreign import Foreign.C.Types import Foreign.C.String import System.IO.Unsafe as Unsafe #include "CodecDefines.h" #include "MusicReader.h" constantToCodec code | code == mp3 = MP3 | code == flac = FLAC | code == ogg = OGG_VORBIS | code == mp4 = MP4 | code == mpeg = MPEG | code == none = NONE | code == unknown = UNKNOWN | otherwise = UNKNOWN where mp3 = #const MP3_CODEC flac = #const FLAC_CODEC ogg = #const OGG_VORBIS_CODEC mp4 = #const MP4_CODEC mpeg = #const MPEG_CODEC none = #const NO_EXTENSION unknown = #const UNKNOWN_EXTENSION data Codec = MP3 | FLAC | OGG_VORBIS | MP4 | MPEG | NONE | UNKNOWN deriving (Show) data MusicMetadata = MusicMetadata { codec :: Codec, length :: Int, bitrate :: Int, channels :: Int, track :: Int, title :: String, artist :: String, album :: String, comment :: String, genre :: String } deriving (Show) instance Storable MusicMetadata where sizeOf _ = (#size struct music_metadata) alignment _ = alignment (undefined::CDouble) peek a = do codec <- liftM constantToCodec $ (((#peek struct music_metadata, codec) a) :: IO Int) length <- ((#peek struct music_metadata, length) a) :: IO Int bitrate <- ((#peek struct music_metadata, bitrate) a) :: IO Int channels <- ((#peek struct music_metadata, channels) a) :: IO Int track <- ((#peek struct music_metadata, bitrate) a) :: IO Int title <- ((#peek struct music_metadata, title) a) :: IO CString artist <- ((#peek struct music_metadata, artist) a) :: IO CString album <- ((#peek struct music_metadata, album) a) :: IO CString comment <- ((#peek struct music_metadata, comment) a) :: IO CString genre <- ((#peek struct music_metadata, genre) a) :: IO CString --FIXME: find replacement for temporary names marshalledTitle <- peekCString title marshalledArtist <- peekCString artist marshalledAlbum <- peekCString album marshalledComment <- peekCString comment marshalledGenre <- peekCString genre return (MusicMetadata codec length bitrate channels track marshalledTitle marshalledArtist marshalledAlbum marshalledComment marshalledGenre) poke a = undefined --This is the "primitive" FFI call--calls the C function and gets a pointer --in return --TODO: write a higher level function this module should export that calls --primReadMusicMetadata and converts the C Pointer into the Haskell data --MusicMetadata foreign import ccall unsafe "read_metadata" primReadMusicMetadata :: CString -> IO (Ptr MusicMetadata) --convert the Haskell string to a CString, call into the FFI then --dereference the resulting pointer readMusicMetadata a = join $ withCString a $ \cs -> ((liftM peek) $ primReadMusicMetadata cs) Here's the struct in MusicReader.h (in an extern C block): struct music_metadata { int codec; int length, bitrate, channels; int track; char *title, *artist, *album, *comment, *genre; }; with the corresponding call: struct music_metadata* read_metadata(char*); I've tried playing around with the alignment but it didn't do anything. I also tried declaring the struct's fields as int32_t which also did nothing. The C demo in question is a very simple: #include <stdio.h> #include "MusicReader.h" #define FILENAME "Its_My_Life.m4a" int main() { struct music_metadata* metadata = read_metadata(FILENAME); printf("music_metadata\ntitle: %s,\tartist: %s,\talbum: %s\n", metadata->title, metadata->artist, metadata->album); printf("comment: %s,\tgenre: %s,\ttrack: %d,\n", metadata->comment, metadata->genre, metadata->track); printf("length: %d,\tbitrate: %d,\tchannels: %d,\n", metadata->length, metadata->bitrate, metadata->channels); printf("codec: %d\n"); } It just reads the metadata into the struct and prints the fields. I've gotten the impression of the Haskell FFI being very beginner-unfriendly, which isn't surprising but still disappointing because it would be a great opportunity to replace some projects with Haskell. Any help is appreciated, including general feedback on my code!

3 4

trees on Haskell : Do I understand it right ?
by Roelof Wobben 02 Mar '15

02 Mar '15

Hello, Suppose we have this definition of a tree : data MessageTree = Leaf | Node MessageTree LogMessage MessageTree deriving (Show, Eq) let Message = LogMessage "E 1 this is a test error" let Message = LogMessage "e 2 this is the second test error " As I understand it right I can make the first entry like this : first_entry = Node Messagetree Message Messagetree And the second one like this second_entry = Node Message Messagetree Message2 Messagetree ?? Roelof

4 10

Using findIndex and then splitAt in Data.List
by Geoffrey Bays 01 Mar '15

01 Mar '15

Hi. An elementary question here about two functions in Data.List: how to use a value from findIndex which returns a Maybe Int, and then use that result in splitAt which takes a regular Int? This is in an IO() do Like so: (This is in an IO() do block) let gradeItemIndex = findIndex (\g -> (itemName g) == (itemName gradeItemP)) gradeItemList let twoListsTuple = splitAt gradeItemIndex gradeItemList // does not compile obviously Many Thanks, Geoffrey

4 8

Bounded STM Queues with load shedding
by Alex Hammel 28 Feb '15

28 Feb '15

Hi list, It would be handy for me to have a bounded STM Queue along the lines of TBQueue <https://hackage.haskell.org/package/stm-2.4/docs/Control-Concurrent-STM-TBQ…> with the difference that when the queue is full, a call to writeTBQueue is a non-blocking no-op. This is in the context of an asynchronous logging system. If the load (somehow) gets heavy enough that the log message queue is taking up a lot of memory, It's better to lose a few log messages than to have processes block until the logger catches up. Are there any off-the-shelf solutions? Cheers, Allex <https://hackage.haskell.org/package/stm-2.4/docs/Control-Concurrent-STM-TBQ…>

2 1

Testing Yesod applications.
by Njagi Mwaniki 27 Feb '15

27 Feb '15

Could I get help with testing yesod applications when it comes to situations requiring authentication and authorization? Here is my full question on stack overflow. http://stackoverflow.com/questions/28764807/performing-authentication-durin…

1 0

Random variable holding a function
by Ondrej Nekola 27 Feb '15

27 Feb '15

Hi I try to simulate some processes, that contain randomness and I would like do it idiomatic way and keep as much code not to know about randomness as possible. I have a following code (simplified to make email more readable): import Data.List import Data.Functor import Genes import Data.Random.RVar import Data.Random.Extras data Individual = Individual (Int, Int) deriving (Eq, Show) -- this part is a bit more complicated... data Population = Population { individuals :: [Individual] } deriving (Eq, Show) type Selection = Population -> Population so far, it's easy: allSurvive :: Selection allSurvive = id extinction :: Selection extinction _ = Population [] The issue comes, when I want to have some randomness in the process. I am able to write a "pick some individuals into next generation" sort of "Selection" fairChance :: Int -> Population -> RVar Population fairChance newSize p = Population <$> (Data.Random.Extras.sample newSize $ individuals p) But it obviously doesn't fit into the "Selection" type. I would like to have something like fairChance :: Int -> RVar Selection (e.g. fairChance :: Int -> RVar (Population -> Population)) Is there a way to do this? Or should I give up this way and try to give up some purity and go withtype Selection :: RVar Population -> RVar population? Thanks Ondrej

3 4

Re: [Haskell-beginners] Relatively simple undirected graph library?
by Stefan Höck 27 Feb '15

27 Feb '15

Dear list Sorry that it took so long. The usual excuses apply: Two little children, other priorities at work, had to write two more monad tutorials. For better or worse, here is a first draft of a labeled graph library. https://github.com/stefan-hoeck/labeled-graph This is all highly experimental. No unit tests so far and lots of stuff is still missing, so feel free to contribute. In the next couple of days, I'll add some more type class instances to LGraphs (Foldable, Traversable, Bifunctor, Bifoldable, Bitraversable). The implementation might well still change. One implementation detail that is different compared to other graph libraries: Edges are represented as Ints internally, so the maximal vertex is limited by the square of the largest Int, which is rather low on 32 bit systems. This is a typical case of bad preliminary optimisation, since I hope that using this implementation we can use efficient IntMaps to store edge labels. It's well possible, that this will change in the future. Please note also that I will use the library mainly in a cheminformatics project, where graphs typically are small and sparse (vertices are almost always of degree four or lower). Grüsse aus Zürich Stefan On Wed, Feb 25, 2015 at 11:30:54PM +0100, Torsten Otto wrote: > I second that it would be terrific to see this posted. > > Regards from Hamburg, > Torsten >

1 0

Re: [Haskell-beginners] how to do this ?
by Roelof Wobben 25 Feb '15

25 Feb '15

Yes, that is one of the books I have to read. For this exercises chapter 2 and 3 . Roelof Francesco Ariis schreef op 24-2-2015 om 10:37: > off list: try to give a more meaningful title to your posts > (i.e. "cis194 help exercise 2" instead of "how to do this"). > It might lead to more answers! > > Also, have you tried "Learn you a haskell for great good"? > > > On Tue, Feb 24, 2015 at 10:28:28AM +0100, Roelof Wobben wrote: >> Hello, >> >> Im trying to solve exercise 2 of this course : >> http://www.seas.upenn.edu/~cis194/spring13/hw/02-ADTs.pdf >> >> I have this till so far : >> >> -- | Main entry point to the application. >> {-# OPTIONS_GHC -Wall #-} >> >> module LogAnalysis where >> >> import Data.Char (isDigit, isLetter) >> import Log >> >> -- | checks if the String is of the format char int >> isValid :: String -> Bool >> isValid s = >> case words s of >> [a]:b:_ -> isLetter a && all isDigit b >> _ -> False >> >> -- | Parse the String to make the right logmessage >> parse_line :: String -> LogMessage >> parse_line s = >> case words s of >> ("I":time:text) -> LogMessage Info (read time) >> (unwords text) >> ("W":time:text) -> LogMessage Warning (read time) >> (unwords text) >> ("E":errorcode:time:text) -> LogMessage (Error (read >> errorcode)) (read time) (unwords text) >> _ -> Unknown "This is not in the >> right format" >> >> >> -- | here the actual function which uses isValid and parse to make >> the right log messages >> parseMessage :: String -> LogMessage >> parseMessage s = >> if isValid(s) then parse_line(s) else Unknown "This is not the >> right format" >> >> >> parse :: String -> [logMessage] >> parse s = parse_line s >> >> >> -- | The main entry point. >> main :: IO () >> main = do >> print $ show (parseMessage "I 4681 ehci 0xf43d000:15: regista14: >> [0xbffff 0xfed nosabled 00-02] Zonseres: brips byted nored)") >> print $ show (parseMessage "W 3654 e8] PGTT ASF! 00f00000003.2: >> 0x000 - 0000: 00009dbfffec00000: Pround/f1743colled") >> print $ show (parseMessage "E 47 1034 'What a pity it wouldn't >> stay!' sighed the Lory, as soon as it was quite") >> >> But I do not see how I can get the output of the parse_line into a List. >> >> Can anyone give me a tip ? >> >> Roelof >> >> >> >> _______________________________________________ >> Beginners mailing list >> Beginners(a)haskell.org >> http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners

2 1