April 2010 - Haskell-Cafe

Using the Clipboard under Windows
by Günther Schmidt 30 Apr '10

30 Apr '10

Hello, is there some sample code on how to use the Graphics.Win32.GDI.Clip? Günther

3 5

maling-list manager in haskell?
by Günther Schmidt 30 Apr '10

30 Apr '10

Hi, is there a mailing-list manager written in haskell? Best regards Günther

2 1

REMINDER: Haskell Communities and Activities Report
by Janis Voigtländer 30 Apr '10

30 Apr '10

Dear all, It is not yet too late to contribute to the 18th edition of HCAR. If you haven't already, please write an entry for your new project or update your old entry, and send to me this weekend or early next week. More information about format etc. can be found in the original call: http://www.haskell.org/pipermail/haskell/2010-April/022012.html I am looking forward to receiving your contributions, Janis (current editor) -- Jun.-Prof. Dr. Janis Voigtländer http://www.iai.uni-bonn.de/~jv/ mailto:jv@iai.uni-bonn.de

1 0

Rank-2 polymorphism and overloading
by Thomas van Noort 30 Apr '10

30 Apr '10

Hello all, I'm having difficulties understanding rank-2 polymorphism in combination with overloading. Consider the following contrived definition: f :: (forall a . Eq a => a -> a -> Bool) -> Bool f eq = eq True True Then, we pass f both an overloaded function and a regular polymorphic function: x :: forall a . Eq => a -> a -> Bool x = \x y -> x == y y :: forall a . a -> a -> Bool y = \x y -> True g :: (Bool, Bool) g = (f x, f y) Could someone explain to me, or point me to some reading material, why g is correctly typed? I understand that x's type is what f expects, but why does y's polymorphic type fulfill the overloaded type of f's argument? I can imagine that it is justified since f's argument type is more restrictive than y's type. However, it requires y to throw away the provided dictionary under the hood, which seems counter intuitive to me. Regards, Thomas

4 5

MySQL and Haskell on Windows
by gladstein＠gladstein.com 30 Apr '10

30 Apr '10

cabal install hdbc-mysql almost works on windows, but errors because the mysql_config command isn't part of the windows MySQL install. Has anyone been able to get a Haskell interface to MySql going on Windows, either directly or via odbc? cabal install haskelldb-hsql-odbc fails because it can't find an 'odbc' library. There are any number of odbc dll's installed by the MySQL odbc connector, but they all have slightly different names. As always, a precompiled binary would be preferable but I'm happy to compile stuff under Cygwin or MINGW if need be. Thanks in advance for your help.

2 1

Cabal package with BSD3 library and LGPL executable?
by Stephen Tetley 29 Apr '10

29 Apr '10

Hello all How do I mark a Cabal package as containing a BSD3 library and an LGPL executable? I'm using CppHs (LGPL) from the exe but have no dependencies on it in the library; as the library might be independently useful I want it to be BSD3. As a caveat - I don't want two packages... Thanks Stephen

4 5

Problem about pattern matching.
by Magicloud Magiclouds 29 Apr '10

29 Apr '10

Hi, I have code as below. How come "case" version works wrong and gives me "overlap" compiling warning? Thanks. if dayOfMonth == firstDayOfMonth then v day (x, y) else if dayOfMonth == lastDayOfMonth then not_ $ v day (x, y) else Mider day (x, y) case dayOfMonth of firstDayOfMonth -> v day (x, y) lastDay -> not_ $ v day (x, y) _ -> Mider day (x, y) -- 竹密岂妨流水过山高哪阻野云飞

4 5

GSoC: Improving Cabal's Test Support
by Thomas Tuegel 29 Apr '10

29 Apr '10

Hello community! I've been working on a proposal for Google Summer of Code 2010 to work on improving Cabal's test support, as described on the Haskell SoC Trac [1]. Today I'm looking for feedback to see if what I intend is what people want/need. As you read this, I kindly ask that you consider: 1) Would you use the facility I describe, were it available? and 2) What additional features would you like to see? There have been two separate suggestions (of which I am aware) of ways to integrate tests into Cabal. One is to build the tests into their own executable which uses an error code on exit to indicate test failure. The second is to have package authors write modules which Cabal will load (dynamically?) and run the tests from. The former method has the advantage of being simpler to implement, but is probably too granular. Although the second suggestion avoids some security concerns, it seems to me that a malicious party could simply put nefarious code into their Setup.hs file anyway, or even in the library being tested. I propose to build a test suite as its own executable, but to avoid the problem of granularity by producing an output file detailing the success or failure of individual tests and any relevant error messages. The format of the file would be standardized through library routines I propose to write; these routines would run tests with HUnit or QuickCheck and process them into a common format. Cabal, or any other utility, could read this file to determine the state of the test suite. Perhaps Cabal could even warn the user about installing packages with failing tests. Under this proposal, a test suite would look something like this (suppose I am writing a test suite for a module Foo, which has an existing test suite in QuickCheck): > module Main where > > import Foo > import Test.QuickCheck > import Distribution.Test -- This module is part of the project I propose > > main = runTests > [ ("testBar", wrap $ testBar), ("testBaz", wrap $ testBaz) ] -- (name, test) 'runTests' and 'wrap' would be provided by 'Distribution.Test'. 'wrap' would standardize the output of test routines. For QuickCheck tests, it would probably look like: > wrap :: Testable a => a -> IO (Bool, String) where the Bool indicates success and the String can be an error message the test produced. 'runTests' would take the list of tests, format their results, and write the output to a file: > runTests :: [(String, IO (Bool, String))] -> IO () I would probably gather the test results into a value of type '[(String, Bool, String)]' -- the name, status, and messages associated with each test -- and use 'show' to produce a nice, human-readable, machine-parsable file. The test suite would be included in the package description file with a stanza such as: > Test > main-is: Test.hs > build-depends: foo, QuickCheck, Cabal This would take all the same options as an 'Executable' stanza, but would tell Cabal to run this executable when './Setup test' is invoked. This of course requires Cabal to support building executables that depend on the library in the same package. Since version 1.8, Cabal supposedly supports this, but my experiments indicate the support is a little broken. (GHC is invoked with the '-package-id' option, but Cabal only gives it the package name. Fixing this would naturally be on the agenda for this project.) At this point, the package author need only run: $ ./Setup configure $ ./Setup build $ ./Setup test to produce a file detailing the results of the test suite. I apologize for taking up your time with a such a lengthy message, and eagerly await your feedback! Thanks! -- Thomas Tuegel [1] http://hackage.haskell.org/trac/summer-of-code/ticket/1581

10 40

Re: [Haskell-cafe] ANN: CPSA - Cryptographic Protocol Shapes Analyzer
by John D. Ramsdell 29 Apr '10

29 Apr '10

>> We are working towards a version of CPSA with the property that >> whenever it successfully terminates, every possible execution is >> described by its output. However, the current implementation >> occasionally fails to find some executions. > > That is concerning - is it due to ... We have formally specified the algorithm, but we haven't been able to prove the algorithm finds all solutions. In fact, we now have an example in the test suite that shows it misses an answer. The example is in the source distribution in tst/missing-contraction.scm. There should be two answers, but CPSA only finds one. We have a fix you can try out in 2.0.4. In src/CPSA/Lib/Strand.hs, you can change the flag useDisplacement from False to True, and CPSA will find both answers. We don't know if this fix is all one needs to ensure CPSA finds every answer. By the way, the algorithm is in doc/cpsaspec.pdf, which can be build from a source distribution of CPSA. > What is the current status of development both wrt openness and future direction? Our highest priority is to resolve the correctness issue you just raised. As for openness, the sources are available to you. We haven't thought about other forms of openness. > I'm wondering if you plan to add MQV or other non-trivial primitives. I don't know what MQV is, so I can't say. Thanks for your interest, Thomas. John

1 0

Lazy Parsing (ANN: vcd-0.1.4)
by Tom Hawkins 29 Apr '10

29 Apr '10

I had been using Parsec to parse VCD files, but needed to lazily parse streaming data. After stumbling on this thread below, I switch to polyparse. What a great library! I was able to migrate from a strict to a semi-lazy parser and many of my parse reductions didn't even need to change. Thanks Malcolm! In addition to lazy VCD parsing, this version of vcd [1] also includes step', which forces a step regardless if variables have changed or not -- helpful for realtime simulation. (BTW, parsec is a great library too.) -Tom [1] http://hackage.haskell.org/package/vcd-0.1.4 On Sun, May 31, 2009 at 6:41 AM, Malcolm Wallace <malcolm.wallace(a)cs.york.ac.uk> wrote: > > I don't know whether you will be willing to change over to polyparse > library, but here are some hints about how you might use it. > > Given that you want the input to be a simple character stream, rather than > use a more elaborate lexer, the first thing to do is to specialise the > parser type for your purposes: > >> type TextParser a = Parser Char a > > Now, to recognise a "mere digit", > >> digit :: TextParser Char >> digit = satisfy Char.isDigit > > and for a sequence of digits forming an unsigned integer: > >> integer :: TextParser Integer >> integer = do ds <- many1 digit >> return (foldl1 (\n d-> n*10+d) >> (map (fromIntegral.digitToInt) ds)) >> `adjustErr` (++("expected one or more digits")) > >> I mean I'd like to be able to turn "12.05.2009" into something like (12, >> 5, 2009) and got no clue what the code would have to look like. I do know >> almost every variation what the code must not look like :). > >> date = do a <- integer >> satisfy (=='.') >> b <- integer >> satisfy (=='.') >> c <- integer >> return (a,b,c) > > Of course, that is just the standard (strict) monadic interface used by many > combinator libraries. Your original desire was for lazy parsing, and to > achieve that, you must move over to the applicative interface. The key > difference is that you cannot name intermediate values, but must construct > larger values directly from smaller ones by something like function > application. > >> lazydate = return (,,) `apply` integer `discard` dot >> `apply` integer `discard` dot >> `apply` integer >> where dot = satisfy (=='.') > > The (,,) is the constructor function for triples. The `discard` combinator > ensures that its second argument parses OK, but throws away its result, > keeping only the result of its first argument. > > Apart from lazy space behaviour, the main observable difference between > "date" and "lazydate" is when errors are reported on incorrect input. For > instance: > > > fst $ runParser date "12.05..2009" > *** Exception: In a sequence: > Parse.satisfy: failed > expected one or more digits > > > fst $ runParser lazydate "12.05..2009" > (12,5,*** Exception: In a sequence: > Parse.satisfy: failed > expected one or more digits > > Notice how the lazy parser managed to build the first two elements of the > triple, whilst the strict parser gave no value at all. > > I know that the error messages shown here are not entirely satisfactory, but > they can be improved significantly just by making greater use of the > `adjustErr` combinator in lots more places (it is rather like Parsec's <?>). > Errors containing positional information about the input can be constructed > by introducing a separate lexical tokenizer, which is also not difficult. > > Regards, > Malcolm > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe >

3 4