March 2012 - Haskell-Cafe

Conduits: Is Source a valid instance of Monad?
by Aristid Breitkreuz 04 Apr '12

04 Apr '12

Hi all, As you may have noticed, Michael Snoyman has been working on an alternative approach to I/O, called conduits. You can find it here: https://github.com/snoyberg/conduit When looking at the Source type (explained here: http://www.yesodweb.com/blog/2011/12/conduits) I noticed that they seem to behave "like lists", and naturally wondered if I could write a Monad instance for them. But first, let's have a brief look at the definition of Source: data SourceResult a = Open a | Closed data PreparedSource m a = PreparedSource { sourcePull :: ResourceT m (SourceResult a) , sourceClose :: ResourceT m () } newtype Source m a = Source { prepareSource :: ResourceT m (PreparedSource m a) } ResourceT deals with resource acquisition and releasing (making sure that all resources are released), and provides a an abstraction over IORef/STRef. For our purposes here, ResourceT is probably close enough to IO. So now the question again is, can we write a Monad instance for this? I have been able to write join (concatenate) and return (a source with a single non-repeated) element. https://gist.github.com/1525471 I _think_ it behaves properly like a Monad, but I'm not quite sure, and neither was Michael. Greg Weber then suggested bringing the question to this forum. What made the question difficult for me is that this would be a stateful Monad transformer, so I'm not quite sure how to test the Monad laws properly. There's a second part to this question: If Source turns out not to be a Monad, is it possibly a ZipList-like Applicative? And either way, which is more useful: The list-like or the ziplist-like instances (of Applicative/Monad)? Thank you, Aristid

4 4

ANNOUNCE: fast-tags-0.0.1
by Evan Laforge 04 Apr '12

04 Apr '12

A while back I was complaining about the profusion of poorly documented tags generators. Well, there is still a profusion of poorly documented tags generators... I was able to find 5 of them. So, that said, here's my contribution to the problem: fast-tags, haskell tag generator #6. Why not use one of the other 5? Two of them use haskell-src which means they can't parse my code. Two more use haskell-src-exts, which is slow and fragile, breaks on partially edited source, and doesn't understand hsc. Then there's the venerable hasktags, but it's buggy and the source is a mess. I fixed a bug where it doesn't actually strip comments so it makes tags to things inside comments, but then decided it would be easier to just write my own. fast-tags is fast because it has a parser that's just smart enough to pick out the tags. It can tagify my entire 300 module program in about a second. But it's also incremental, so it only needs to do that the first time. I have vim's BufWrite autocommand bound to updating the tags every time a file is written, and it's fast enough that I've never noticed the delay. It understands hsc directly (that's trivial, just ignore the # lines) so there's no need to run hsc2hs before tagifying. The result is tags which are automatically up to date all the time, which is nice. If people care about lhs and emacs tags then it wouldn't be hard to support those too, and at that point I could replace hasktags and we'd be back down to 5 again. But I'm not even sure anyone uses hasktags, since surely someone would have noticed that comment bug. Anyway, it's been working great for me for a couple weeks so I uploaded it to hackage: http://hackage.haskell.org/package/fast-tags

5 9

Re: [Haskell-cafe] hint and type synonyms
by Claude Heiland-Allen 01 Apr '12

01 Apr '12

Hi Daniel, cafe, On 31/03/12 17:47, Daniel Gorín wrote: > Could you provide a short example of the code you'd like to write but gives you problems? I'm not able to infer it from your workaround alone... This problem originally came up on #haskell, where Rc43 had a problem making a library with a common module that re-exports several other modules: http://hpaste.org/66281 My personal interest is somewhat secondary, having not yet used hint in a real project, but code I would like to write at some point in the future is much like the 'failure' below, unrolled it looks like: main = (print =<<) . runInterpreter $ do setImports ["Prelude"] interpret "1/5" (as :: Rational) -- fails Having to remember that Rational is defined as type Rational = Ratio Integer and that Ratio is defined in Data.Ratio and then to add that to the setImports list is a little inconvenient: main = (print =<<) . runInterpreter $ do setImports ["Prelude", "Data.Ratio" ] interpret "1/5" (as :: Rational) -- works But for my own purposes this is probably much saner in the long run than my newtype wrapping approach below. However, this is not always possible: supposing Ratio itself was defined as a type synonym of Ratio2, and Ratio2 was not exported. Perhaps this is what Rc43 was experiencing, but I shouldn't speculate, as this is all getting a bit theoretical - I should try out hint in the real world to see if this problem makes things impractical for me - sorry for the noise! Thanks, Claude > Thanks, > Daniel > > > On Mar 31, 2012, at 6:19 PM, Claude Heiland-Allen wrote: > >> Hi all, >> >> What's the recommended way to get hint[0] to play nice with type synonyms[1]? >> >> A problem occurs with in scope type synonyms involving types not in scope. >> >> I came up with this after looking at the source[2], but it makes me feel ill: >> >> --8<-- >> -- hint and type synonyms don't play nice >> module Main where >> >> import Language.Haskell.Interpreter >> >> import Data.Typeable as T >> import Data.Typeable.Internal >> import GHC.Fingerprint.Type >> >> main = failure>> success >> >> test t = (print =<<) . runInterpreter $ do >> setImports [ "Prelude" ] >> interpret "1/5" t >> >> failure = test (as :: Rational) >> -- Left (WontCompile [GhcError {errMsg = "Not in scope: type constructor or class `Ratio'"}]) >> >> success = test (as :: Q) >> -- Right (1 % 5) >> >> newtype Q = Q Rational >> >> instance Show Q where >> show (Q a) = show a >> showsPrec n (Q a) = showsPrec n a >> >> instance Typeable Q where >> typeOf _ = TypeRep (Fingerprint 0 0) (T.mkTyCon "Rational") [] >> --8<-- >> >> Thanks, >> >> >> Claude >> >> [0] http://hackage.haskell.org/package/hint >> [1] http://www.haskell.org/onlinereport/decls.html#type-synonym-decls >> [2] http://hackage.haskell.org/packages/archive/hint/0.3.3.4/doc/html/src/Hint-… >> >> _______________________________________________ >> Haskell-Cafe mailing list >> Haskell-Cafe(a)haskell.org >> http://www.haskell.org/mailman/listinfo/haskell-cafe >

2 1

Re: [Haskell-cafe] Google Summer of Code - Lock-free data
by John Lato 01 Apr '12

01 Apr '12

Slightly related: I think it would be interesting to compare a Disruptor-based concurrency communications mechanism and compare it to e.g. TChans 1. Disruptor - http://code.google.com/p/disruptor/ > From: Ryan Newton <rrnewton(a)gmail.com> > >> I think so. Atomically reading and writing a single memory location >>> (which CAS does) is just a very simple transaction. But using a CAS >>> instruction should be more efficient, since STM has to maintain a >>> transaction log and commit transactions, which creates some overhead. >>> >> >> Ah, I see. In that case, it may be worthwhile to implement the CAS >> instruction in terms of STM as well and measure the performance difference >> this makes for the final data structure. After all, STM is a lot more >> compositional than CAS, so I'd like to know whether the loss of >> expressiveness is worth the gain in performance. > > > There's one annoying hitch with doing apples-to-apples comparisons here. > > The problem is that CAS falls short of being a hardware-accelerated version > of a simple transaction (read TVar, (==) against expected value, > conditionally update TVar), because CAS is based on pointer equality rather > than true equality. (eq? rather than equal? for any Schemers out there.) > > For this reason our "Fake" version of CAS -- for older GHCs and for > performance comparison -- has to use reallyUnsafePointerEquality#: > > > http://hackage.haskell.org/packages/archive/IORefCAS/0.2/doc/html/Data-CAS-… > > But we do provide a "CASable" type class in that package which is precisely > for comparing the performance of: > > - Hardware CAS > - Fake CAS -- atomicModifyIORef + ptrEquality > - Foreign CAS -- gcc intrinsic + function call

3 2

get a string representation (show) of a function argument
by TP 31 Mar '12

31 Mar '12

Hi, I don't known the advanced features and extensions of GHC at all. Look at the following program: ---------------- f :: Integer -> Integer -> IO Integer f a b = do print $ "first argument=" ++ (show a) print $ "second argument=" ++ (show b) print $ a+b return (a+b) main = do k <- f 3 5 f 2 k ---------------- It yields: ---------------- "first argument=3" "second argument=5" 8 "first argument=2" "second argument=8" 10 10 ---------------- I am wondering if there is any means to get "f 3 5" instead of "8" in the output of this program. My idea would be to write some generic test function: I give it an expression (probably a function call) and the expected result (as a string), and then: * it should check that the obtained result is identical to the expected one. * it should also print the expression that yielded the first argument of the test function (i.e. the function call), under an understandable form (show). This is the feature that is missing to my mind in some test suites: for example in HUnit, it is not automatic: in the first argument of assertEqual, we may give a string corresponding to the executed command, see the documentation of Test.HUnit: test1 = TestCase (assertEqual "for (foo 3)," (1,2) (foo 3)) > runTestTT tests # Failure in: 0:test1 for (foo 3), expected: (1,2) but got: (1,3) Cases: 2 Tried: 2 Errors: 0 Failures: 1 So, we have duplication of code: "foo 3". On the other hand, I guess that the needed features to do that are not possible in Haskell (related to some kind of metaprogramming), but I prefer to ask people about that. Thanks in advance, TP

2 1

ANNOUNCE: fix-imports 1.0.0
by Evan Laforge 31 Mar '12

31 Mar '12

I just uploaded a new version to hackage. This improves the heuristics for guessing the right module quite a bit, and adds a local config file where you can prioritize or de-prioritize packages or module prefixes. It's also a bit faster. >From the hackage description: A small standalone program to manage the import block of a haskell program. It will try to add import lines for qualified names with no corresponding import, remove unused import lines, and sort the import block according to some heuristic you can define. This only works for qualified imports! Unqualified imports are left untouched. It's most convenient if bound to an editor key. Recent major changes: version 1.0.0 Change name from FixImports to fix-imports, which is more unixy. Change ghc-pkg parsing from String to Text. It's noticeably faster. Add a more flexible system for prioritizing imports. When there are several possibilities for a module name, they are all given to a single function to decide. The config file moved from fix-imports-priority to .fix-imports and can now specify sort orders for packages and modules by prefix. Make -i includes for non-existent dirs ignored instead of causing an error.

1 0

hint and type synonyms
by Claude Heiland-Allen 31 Mar '12

31 Mar '12

Hi all, What's the recommended way to get hint[0] to play nice with type synonyms[1]? A problem occurs with in scope type synonyms involving types not in scope. I came up with this after looking at the source[2], but it makes me feel ill: --8<-- -- hint and type synonyms don't play nice module Main where import Language.Haskell.Interpreter import Data.Typeable as T import Data.Typeable.Internal import GHC.Fingerprint.Type main = failure >> success test t = (print =<<) . runInterpreter $ do setImports [ "Prelude" ] interpret "1/5" t failure = test (as :: Rational) -- Left (WontCompile [GhcError {errMsg = "Not in scope: type constructor or class `Ratio'"}]) success = test (as :: Q) -- Right (1 % 5) newtype Q = Q Rational instance Show Q where show (Q a) = show a showsPrec n (Q a) = showsPrec n a instance Typeable Q where typeOf _ = TypeRep (Fingerprint 0 0) (T.mkTyCon "Rational") [] --8<-- Thanks, Claude [0] http://hackage.haskell.org/package/hint [1] http://www.haskell.org/onlinereport/decls.html#type-synonym-decls [2] http://hackage.haskell.org/packages/archive/hint/0.3.3.4/doc/html/src/Hint-…

1 0

New in haskell for old-timers?
by Mats Rauhala 31 Mar '12

31 Mar '12

I ended up discussing in university with the person next to me about Haskell. He's an old-timer who's even been to conferences, but it's a long time since he's looked at Haskell. He asked me to write "what's new in Haskell", but I'm not too experienced myself. I didn't realize asking how long it's been for him, but he mentioned about new fronts in optimizing compilers and specifically mentioned a compiler that could compile to legible C. I've been following Haskell only for a couple of years, so my intuition tells me that he means `-fvia-c`, and I mentioned to him that it's about to be deprecated and replaced with the llvm back-end. I also mentioned hackage, but he hadn't heard of it. He mentioned that he'd like to use Haskell for his master's thesis, but we got a interrupted when he said what it was about. Formal something or other. I trust you fellow Haskellers have better understanding of the time frame, and can think of something interesting new features, libraries, communities etc. that's since been emerged. I ask you to give me something interesting to give him and reinstate his interest. -- Mats Rauhala MasseR

6 6

for = flip map
by Johannes Waldmann 30 Mar '12

30 Mar '12

Good: we have mapM, and we have forM ( = flip mapM ) . Sure this is just a convenience, and indeed "forM xs $ \ x -> do ..." is quite handy, especially if "xs" is really small, and "..." is some larger expression. Bad: we have map, but we are missing: for ( = flip map ) . The function is very convenient, for the same reasons as above. I can't remember how often I typed "for = flip map" in a source file. I never put this definition in a module either, since the import statement would be longer than the definition. So, I'm all for "for" . In Data.List? In the Prelude? (Should put it right next to "map".) - J.W.

8 9

ANN: Happstack 7
by Jeremy Shaw 30 Mar '12

30 Mar '12

We are pleased to announce the release of Happstack 7! Happstack is a fast, modern, web application framework written in Haskell. Please check out the brand new happstack.com website to read about what is new in Happstack 7, and what we are planning for Happstack 8, and what makes Happstack great! http://www.happstack.com/ - jeremy

3 3