- Haskell-Cafe - Haskell.org

Rigid type variable error
by Darryn 27 Jun '09

27 Jun '09

From: Darryn <djreid(a)aapt.net.au> To: beginners(a)haskell.org Subject: Rigid type variables match error Date: Sat, 27 Jun 2009 10:18:13 +0930 Hi, I wonder if anyone can explain what is going on here and what to do about it. I'm fairly new to Haskell, so apologies in advance if my question seems naive. I've cut my code down to a minimum that reproduces the problem. I receive an error in GHCI in the following code, complaining that it cannot match the rigid type variables for the instance definition for Ainst for the function a3. Can anyone advise about what to do about it? ------------------------------------------------ class A a where a1 :: a a2 :: a -> a a3 :: (B b) => b -> a class B b where b1 :: Int -> b data (B b) => Ainst b = I | J (Ainst b) | K b instance (B b) => A (Ainst b) where a1 = I -- a2 :: (B b, A a) => a -> a a2 = J -- a3 :: (B b, A a) => b -> a a3 = K -- Error! -- a3 = K `asTypeOf` a3 -- Error even with this! data Binst = Val Int instance B Binst where b1 = Val ------------------------------------------------ Test5.hs:17:9: Couldn't match expected type `b1' against inferred type `b' `b1' is a rigid type variable bound by the instance declaration at Test5.hs:12:12 `b' is a rigid type variable bound by the type signature for `a3' at Test5.hs:5:13 Expected type: b -> Ainst b1 Inferred type: b -> Ainst b In the expression: K `asTypeOf` a3 In the definition of `a3': a3 = K `asTypeOf` a3 Failed, modules loaded: none. Thanks in advance for any help anyone can provide.

2 1

System.Vacuum.Ubigraph
by Cetin Sert 26 Jun '09

26 Jun '09

Hi, The package seems to have a connection problem on my machine. Prelude> import System.Vacuum.Ubigraph Prelude System.Vacuum.Ubigraph> view (0,1) Loading package syb ... linking ... done. Loading package base-3.0.3.1 ... linking ... done. Loading package parsec-2.1.0.1 ... linking ... done. Loading package network-2.2.1.3 ... linking ... done. Loading package array-0.2.0.0 ... linking ... done. Loading package containers-0.2.0.1 ... linking ... done. Loading package bytestring-0.9.1.4 ... linking ... done. Loading package old-locale-1.0.0.1 ... linking ... done. Loading package old-time-1.0.0.2 ... linking ... done. Loading package pretty-1.0.1.0 ... linking ... done. Loading package filepath-1.1.0.2 ... linking ... done. Loading package unix-2.3.2.0 ... linking ... done. Loading package directory-1.0.0.3 ... linking ... done. Loading package process-1.0.1.1 ... linking ... done. Loading package random-1.0.0.1 ... linking ... done. Loading package haskell98 ... linking ... done. Loading package mtl-1.1.0.2 ... linking ... done. Loading package time-1.1.3 ... linking ... done. Loading package vacuum-0.0.95 ... linking ... done. Loading package HTTP-4000.0.7 ... linking ... done. Loading package HaXml-1.13.3 ... linking ... done. Loading package dataenc-0.13.0.0 ... linking ... done. Loading package haxr-3000.2.1 ... linking ... done. Loading package vacuum-ubigraph-0.1.0.3 ... linking ... done. *** Exception: user error (Error calling ubigraph.clear: �: openFile: does not exist (No such file or directory)) In ubigraph.clear: �: openFile: the part between clear and openFile is not always the same character or sequence of characters! I guess it's a problem with Network.XmlRpc.Client but couldn't locate it. Anyone else that has (had) this problem? Best Regards, Cetin Sert

1 0

download, download-curl ...
by Keith Sheppard 26 Jun '09

26 Jun '09

Hello Cafe, I'm looking to make it possible for people to use urls directly in my haskell program (TxtSushi) and I'd like your suggestions if you have any. I really like the API's for download and download-curl, but I'm wondering what the practical differences are between the two? Is download-curl a replacement for download? Will it build OK on any platform or do I have to use something like the HTTP library if I want it to work for people with Windows? (I really like the fact that the "download" APIs support FTP too though) Thank You Keith -- keithsheppard.name

2 1

Multi-module scripts
by Maurício 26 Jun '09

26 Jun '09

Is it possible to write a Haskell script that uses a module that is also going to be interpreted? Like, say: #!/usr/bin/runhaskell module Main (main) where import OtherModule main = (...) ---- where OtherModule is available in, say, OtherModule.hs and not in the already compiled and installed library? Thanks, Maurício

2 1

Why is Bool no instance of Num and Bits?
by Stephan Friedrichs 26 Jun '09

26 Jun '09

Hi! When looking for an xor function, I found one in Data.Bits but couldn't use it for Bool, because Bool is no instance of Bits and of Num (which would be necessary, because it's "class (Num b) => Bits b"). My question is: Why not? We could declare instance Num Bool where (+) False = id (+) True = not (*) True True = True (*) _ _ = False (-) = (+) negate = id abs = id signum = const True fromInteger = not . even which basically implements the field with 2 elements and instance Bits Bool where bitSize = const 1 isSigned = const False (.&.) = (&&) (.|.) = (||) xor = (+) complement = not shift = const shiftL = const shiftR = const rotate = const rotateL = const rotateR = const bit = (==0) setBit _ 0 = True setBit b _ = b clearBit _ 0 = False clearBit b _ = b complementBit b 0 = not b complementBit b _ = b testBit b 0 = b testBit _ _ = False quite trivial... Why is this not part of base? Or am I missing something? //Stephan -- Früher hieß es ja: Ich denke, also bin ich. Heute weiß man: Es geht auch so. - Dieter Nuhr

12 16

Hack (web) and apache configuration
by Henry Laxen 26 Jun '09

26 Jun '09

Dear Group, I am posting this here even though it probably belongs on the apache list because I suspect other haskell users will be able to find it here more easily. I am playing around with hack, and am having trouble with configuring apache with fastcgi to make things work. My understanding of the "hack" concept is that it provides a stardardized interface that lets you glue together web "Applications". It also provides several front-ends, such as happs, and fastcgi etc. Now based on looking at the Middleware supplied with hack, it seems to be trying to dispatch based on the contents of the pathInfo field of the Env record. So, my question is, how do we configure Apache2 with the fastcgi handler so that something appears in the pathInfo field? I have tried several things, the most recent being: RewriteEngine on RewriteRule ^/(.*)$ /hackTest?input=$1 [T=application/x-httpd-cgi] <Location /> SetHandler fastcgi-script Options ExecCGI FollowSymLinks </Location> but the pathInfo field is always null. Env {requestMethod = GET, scriptName = "/lambda", pathInfo = "", queryString = "input=lambda", serverName = "127.0.0.1", serverPort = 80, http = [("FCGI_ROLE","RESPONDER"),("SCRIPT_URL","/lambda"), ("SCRIPT_URI","http://127.0.0.1/lambda"),("User-Agent","curl/7.18.2 (x86_64-pc-linux-gnu) libcurl/7.18.2 OpenSSL/0.9.8g zlib/1.2.3.3 libidn/1.8 libssh2/0.18"), ("Host","127.0.0.1"),("Accept","*/*"),("PATH","/usr/local/bin:/usr/bin:/bin"), ("SERVER_SIGNATURE","<address>Apache/2.2.9 (Debian) mod_fastcgi/2.4.6 proxy_html/3.0.0 mod_apreq2-20051231/2.6.0 mod_perl/2.0.4 Perl/v5.10.0 Server at 127.0.0.1 Port 80</address>\n"), ("SERVER_SOFTWARE","Apache/2.2.9 (Debian) mod_fastcgi/2.4.6 proxy_html/3.0.0 mod_apreq2-20051231/2.6.0 mod_perl/2.0.4 Perl/v5.10.0"), ("SERVER_NAME","127.0.0.1"),("SERVER_ADDR","127.0.0.1"), ("SERVER_PORT","80"),("REMOTE_ADDR","127.0.0.1"), ("DOCUMENT_ROOT","/home/henry/maztrave2/www/fcgi"), ("SERVER_ADMIN","[no address given]"), ("SCRIPT_FILENAME","/home/henry/maztrave2/www/fcgi/hackTest"), ("REMOTE_PORT","44936"), ("GATEWAY_INTERFACE","CGI/1.1"),("SERVER_PROTOCOL","HTTP/1.1"), ("REQUEST_METHOD","GET"),("QUERY_STRING","input=lambda"), ("REQUEST_URI","/lambda"),("SCRIPT_NAME","/lambda")], hackVersion = [2009,5,19], hackUrlScheme = HTTP, hackInput = Empty, hackErrors = HackErrors, hackHeaders = []}% I think what I want is to have all URLS, such as: http://127.0.0.1/lambda be dispatched though my hackTest executable, without having to go through the rewrite, but I can't convice apache to do that. In the interest of completeness, my hackTest.hs file is the following: import Hack import Hack.Handler.FastCGI import Data.ByteString.Lazy.Char8 (pack) import Hack.Contrib.Middleware.Lambda app :: Application app = \env -> return $ Response { status = 200 , headers = [ ("Content-Type", "text/plain") ] , body = pack $ show env } main = runFastCGIorCGI $ lambda app ------------------------------------------------------ One final comment to the authors of hack. Would you please consider renaming this project. hack is such a common word that has nothing to do with this project that it make searching the web with google, etc. almost useless. I realize it is a clever respelling of the ruby version "rack", but please consider naming it something more unique, while it is still relatively new on the web. Thanks in advance for your help. Best wishes, Henry Laxen

4 4

Generic Graph Class
by Ivan Lazar Miljenovic 26 Jun '09

26 Jun '09

At the moment, there are at least three ways people use graph data-structures in Haskell: * Data.Graph from containers * Data.Graph.Inductive from FGL * A custom job (usually using something like IntMap). If we look on Hackage, there are a number of graph-related packages there, each of which uses one of the above approaches. However, for the more "generic" packages that operate _on_ graphs (rather than using graphs as an internal data structure), there is (as far as I can tell, anyway) usually no reason why at least most of that package cannot work on _any_ of these types of graphs. Examples include (in no particular order; found by looking for "graph" on the main hackage package list page): * hgal {uses Data.Graph} * vacuum and related packages {custom, [(a, [a])]} * PkgGraph {custom, a -> [b] AFAICT} * cabalgraph {custom, a -> b AFAICT} * graphviz {uses FGL} * SceneGraph {FGL} * data-reify {custom} * hsc3 and related packages {custom, not sure how} * dotgen {custom} * fenfire {custom} * GraphSCC {Data.Graph} * Graphalyze {FGL} * SourceGraph {FGL via Graphalyze} Just to give you an idea, out of these the following implement some form of graph -> .dot function (for GraphViz): * vacuum * PkgGraph * cabalgraph * graphviz (well, that's the point of this package... used by Graphalyze and SourceGraph) * SceneGraph * hsc3-dot * dotgen (From its description: "This package provides a simple interface for building .dot graph files...") Why do we duplicate so much work and effort? Not just in terms of generating Dot representations for use in GraphViz, but also algorithm development? For example, I can't see any reason why hgal shouldn't also work for any generic graph type, possibly even GraphSCC (note that sometimes specialisation isn't always avoidable: my Graphalyze library uses the inductive nature of FGL for its algorithms). I thus propose that we work out a generic graph class that can be used by the various libraries we have and use, to avoid this duplication of effort (I have already proposed that I intended to add such functionality to the graphviz library, but I'm throwing open the design of such a class to the general community). This means that even if you have to use some custom graph-like data structure in your program, you can take advantage of one of the libraries available (e.g. graphviz) without having to write your own graph functions for common tasks. Here is my current thinking on how such a class could be defined. Note that I'm defining it for _directed_ graphs, as all graph definitions used seem to match this mould (and we can emulate undirected graphs using directed graphs but not the other way around). ====================================================================== type Vertex = Int class GenericGraph g where vertices :: g -> [Vertex] edges :: g -> [(Vertex, Vertex)] order :: g -> Int order = length . vertices size :: g -> Int size = length . edges -- All vertices v' such that there's an edge (v',v) adjacentIn :: g -> Vertex -> [Vertex] adjacentIn g v = map fst . filter ((==) v . snd) $ edges g degIn :: g -> Vertex -> Int degIn g = length . adjacentIn g -- All vertices v' such that there's an edge (v,v') adjacentOut :: g -> Vertex -> [Vertex] adjacentOut g v = map snd . filter ((==) v . fst) $ edges g degOut :: g -> Vertex -> Int degOut g = length . adjacentIn g adjacent :: g -> Vertex -> [Vertex] adjacent g v = adjacentIn g v ++ adjacentOut g v degree :: g -> Vertex -> [Vertex] degree g = length . adjacent g -- For this function, maybe use as a default something based on the -- hidden isUndir function in graphviz's Data.GraphViz isDirected :: g -> Bool ====================================================================== I'm not saying that the above class is perfect (though I can't think of any other generic graph-related functions we might want to include in there off the top of my head). However, I do think it's a good starting point. Once we have defined such a class, how should we package it? In particular, how should we define instances for pre-packaged (i.e. Data.Graph and FGL) graph data structures? Method 1: Define the instances in the same package. --------------------------------------------------- If we include the instances in the same package, then all libraries that use this class will require FGL, etc. to be installed (e.g. if you want to use the graphviz library for your own custom class, then thanks to the static linking used by GHC you will still have an unwanted FGL dependency in your program). Method 2: Have seperate instance packages. ------------------------------------------ No unwanted dependencies, but this means we have what are basically useless packages containing nothing but an orphan instance. Method 3: Define the instance in Data.Graph and FGL --------------------------------------------------- I'm not sure if the maintainers of these packages (let alone the GHC maintainers who require containers) will appreciate us giving them another dependency for their packages. Furthermore, this limits users of libraries that use this class to using versions of the data structures contained in libraries new enough to have this instance defined in them. Method 4: Give a default instance in the documentation. ------------------------------------------------------- Users can copy/paste the instance definition from the documentation and put it in their code. This will allow them to tweak the definitions if they so desire, but will lead to code duplication. Please tell me any ideas/criticisms you have of this proposal. My main reason for wanting this at the moment is to define it in graphviz, though I can think of some other uses I might have of it. If we all agree to define this type of class in a custom package (rather than me just defining it in graphviz and leaving it up to other users to define instances). I'd like to have this resolved sooner rather than later, as I want to work on graphviz soonish so I can start working on Graphalyze and SourceGraph again. -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com IvanMiljenovic.wordpress.com

3 3

combining monads with IO
by Richard Silverman 25 Jun '09

25 Jun '09

Hi all, I'm puzzled by something. Suppose I have some code that does lots of IO, and also occasionally refers to some global state. No problem, use ReaderT for the state, combining with the IO monad. Except... since IO is on the bottom, simple uses of do-notation such as "foo <- ask" work in the Reader monad, and to access the IO monad, I need to lift, e.g. (bar <- liftIO getLine). If my code does lots of IO, this is *very* ugly -- the code is littered with lift functions! Is there no cleaner way to do this? Thanks, - Richard

5 6

Re: [Haskell-cafe] GHCi infers a type but refuses it as type signature
by papa.eric＠free.fr 25 Jun '09

25 Jun '09

> Simple: the definition of MonadState uses those extensions. Thanks, yes it helps and explains all. :^) I suppose then that if -XFlexibleContexts is indeed required by the standard libraries, it is a "safe" extension, meaning supported by all compilers? Are many such extensions de-facto standard that anyone can enable by default?

6 8

[ANN] full-sessions: yet another implementation of session types
by Keigo Imai 25 Jun '09

25 Jun '09

Hello haskell-cafe, I would like to announce the pre-release of full-sessions, yet another implementation of session types in Haskell. Session types are used to statically check the safe and consistent use of communication channels according to protocols. Our work is quite similar to the Session monad in Haskell'08 paper of Pucella and Tov, however, a notable advantage of our implementation is that it requires almost no type annotation or term annotation in a source code, and at the same time provides full functionality of session types including channel-generation and channel-passing. It can be obtained from the website. A few examples are also provided. http://www.agusa.i.is.nagoya-u.ac.jp/person/sydney/full-sessions.html Any comments are much appreciated. Cheers, - Keigo IMAI

2 2