- Haskell-Cafe - Haskell.org

How to correctly benchmark code with Criterion?
by Janek S. 19 Oct '12

19 Oct '12

Dear list, during past few days I spent a lot of time trying to figure out how to write Criterion benchmarks, so that results don't get skewed by lazy evaluation. I want to benchmark different versions of an algorithm doing numerical computations on a vector. For that I need to create an input vector containing a few thousand elements. I decided to create random data, but that really doesn't matter - I could have as well use infinite lists instead of random ones. My problem is that I am not certain if I am creating my benchmark correctly. I wrote a function that creates data like this: dataBuild :: RandomGen g => g -> ([Double], [Double]) dataBuild gen = (take 6 $ randoms gen, take 2048 $ randoms gen) And I create benchmark like this: bench "Lists" $ nf L.benchThisFunction (L.dataBuild gen) The question is how to generate data so that its evaluation won't be included in the benchmark. I already asked this question on StackOverflow ( http://stackoverflow.com/questions/12896235/how-to-create-data-for-criterio… ) and got answer to use evaluate + force. After spending one day on testing this approach I came to conclusion that doing this does not seem to influence results of a benchmark in any way (I did stuf like unsagePerformIO + delayThread). On the other hand I looked into sources of criterion and I see that the benchmark code is run like this: evaluate (rnf (f x)) I am a Haskell newbie and perhaps don't interpret this correctly, but to me it looks as though criterion did not evaluate the possibly non-evaluated parameter x before running the benchmark, but instead evaluates the final result. Can someone provide an explanation on how this exactly works and how should I write my benchmarks so that results are correct? Janek

5 9

ANNOUNCE: apphointments - A simple functional calendar
by Steffen Schuldenzucker 18 Oct '12

18 Oct '12

Dear Café, I just wrote myself a small (~ 200 LOC) calendar application today. https://bitbucket.org/rainmaker/apphointments Comments/Patches welcome. From the readme: apphointments - A simple functional calendar ============================================ This is a "UI = Code" calendar application: You create an event by writing code that defines it. To see what's up next, you create a "report", i.e. a summary of the events in a certain time range such as 'thisWeek'. This is done by haskell functions again from GHCi or ghc -e. See example.hs. Using just haskell instead of our own language or GUI allows great flexibility in both defining events and generating reports. See e.g. the 'lecture' combinator from Apphointments.Util. Status ------ Works for me, but has no features yet. See TODO. Credits ------- Created by Steffen Schuldenzucker <sschuldenzucker(a)uni-bonn.de>

1 0

Fwd: hdbc parametrized select
by Wilfried van Asten 18 Oct '12

18 Oct '12

This was not sent to the cafe since I used the wrong from address. ---------- Forwarded message ---------- From: Asten, W.G.G. van (Wilfried, Student M-CSC) <w.g.g.vanasten(a)student.utwente.nl> Date: Thu, Oct 18, 2012 at 12:27 PM Subject: Re: [Haskell-cafe] hdbc parametrized select To: s9gf4ult(a)gmail.com Cc: haskell-cafe(a)haskell.org To my knowledge you can combine execute and the fetch functions like this: find conn id = do stmt <- prepare conn "SELECT * FROM table WHERE id = ? LIMIT 1" excute stmt [id] row <- fetchRow stmt On Thu, Oct 18, 2012 at 10:51 AM, <s9gf4ult(a)gmail.com> wrote: > does hdbc have parametrized selects ? > There is execute > execute :: Statement -> [SqlValue] -> IO Integer > and some other functions which get list of parameters but return IO Integer or > IO () > and there is other couple of functions > fetchRow :: Statement -> IO (Maybe [SqlValue]) > which return values of select but do not get parameters > > How can i execute parametrized select with hdbc ? > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe

1 0

hdbc parametrized select
by s9gf4ult＠gmail.com 18 Oct '12

18 Oct '12

does hdbc have parametrized selects ? There is execute execute :: Statement -> [SqlValue] -> IO Integer and some other functions which get list of parameters but return IO Integer or IO () and there is other couple of functions fetchRow :: Statement -> IO (Maybe [SqlValue]) which return values of select but do not get parameters How can i execute parametrized select with hdbc ?

1 0

Re: [Haskell-cafe] Safe Haskell and instance coherence
by Mikhail Glushenkov 17 Oct '12

17 Oct '12

Hello David, On Thu, Oct 18, 2012 at 6:21 AM, David Mazieres expires 2013-01-15 PST <mazieres-f97tx4hr7aw7n8sqs9kpcvuwd2(a)temporary-address.scs.stanford.edu> wrote: > > What's interesting is that these examples doesn't violate the goals of > Safe Haskell--i.e., they don't violate type safety, encapsulation, or > semantic consistency as we defined it--but they are worrisome. I think that my example violates encapsulation in a similar way to how GND violates encapsulation in the MinList example from the Safe Haskell paper. IIUC, the problem is that all instances are assumed to be coherent (that's why it's an error to have two Monoid Int instances in the same module), but this assumption isn't actually enforced across module boundaries, even though it is reflected in the type system. GHC does print a warning after having encountered orphan instances, but that's all. > This is essentially the same thing as my Monoid example. You've got > two different dictionaries for the same type, and can pass either one > around depending on what module you imported. Are "dictionaries" something that is a part of Haskell semantics or an artefact of the implementation? > It's > not hard to cook up a contrived example where some sort of security > monitor hands over the keys to the kingdom if ever it encounters > duplicate items in a set. Auditing the code, that might seem fine > unless you understand the implementation of Set, which makes reasoning > about security a lot harder. Agreed. > What we really want is for the dictionary to be associated with the > data structure at the time of creation, not passed in as an argument > for each operation. But that's not even implementable without the > existential types extension, and also would require re-writing all the > containers, which is absolutely not what we want. This is what Scala does. Unfortunately, this can make some operations (e.g. set union) asymptotically less efficient (as it's now impossible to rely on the fact that both sets use the same associated dictionary). > Failing that, I guess we could disallow overlapping instances even > when they don't overlap in a single module. This is a whole-program > check similar to what type families requires, so could possibly be > implemented in a similar way. I'm really interested in the link-time check that is performed for type families. Is it described somewhere? > However, as with type families, making > it work with dynamic loading is going to be kind of hard. Essentially > there would have to be some run-time inspectable information about all > instances defined. I think that's why Rust chose to just disallow orphan instances :-) Even without dynamic loading, supporting all GHC extensions (e.g. FlexibleInstances) will probably be non-trivial. -- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments

1 0

Re: [Haskell-cafe] Safe Haskell and instance coherence
by Mikhail Glushenkov 17 Oct '12

17 Oct '12

Hello David, On Wed, Oct 17, 2012 at 6:02 PM, David Mazieres expires 2013-01-15 PST <mazieres-sebg4mvthwuzk9r27pqmr8jndi(a)temporary-address.scs.stanford.edu> wrote: > Can you elaborate on how this can be used to break the data structure > invariant? If in safe Haskell you import two modules that have > overlapping instances, you will not be able to use the two instances. > Modules that import only one instance will be able to use that > instance. Please take a look at the code example I provided: https://gist.github.com/3854294 I don't use overlapping instances or any other extensions besides Safe Haskell. By defining two orphan Ord instances for U I'm able to construct a value of type Set U that contains two equal elements: > test fromList [X,Y,X] This is what I meant by "breaking the data structure invariant". This shouldn't normally be possible: the documentation for Data.Set.insert says: "If the set already contains an element equal to the given value, it is replaced with the new value." Regarding your Monoid example, it will still be possible to make it work even if instance coherence is enforced by using a newtype wrapper (in fact, Data.Monoid already includes Sum and Product newtype wrappers that provide this functionality). -- () ascii ribbon campaign - against html e-mail /\ www.asciiribbon.org - against proprietary attachments

1 0

generalizing the writer monad
by Petr P 17 Oct '12

17 Oct '12

Hi, (this is a literate Haskell post.) lately I was playing with the Writer monad and it seems to me that it is too tightly coupled with monoids. Currently, MonadWriter makes the following assumptions: (1) The written value can be read again later. (2) For that to be possible it has to be monoid so that multiple (or zero) values can be combined. I fell say that this is a bit restricting. Sometimes, the written value can be lost - either used to compute something else or for example sent out using some IO action to a file, network etc. For example, I'd like to create an IO-based writer monad whose `tell` logs its argument somewhere - prints it, stores to a file etc. So what I'm suggesting is to have another type class between Monad and MonadWriter, let's say MonadTell, which only allows to write values, not to retrieve them later: > {-# LANGUAGE FlexibleInstances, MultiParamTypeClasses, FunctionalDependencies #-} > import Control.Monad > import Control.Monad.Trans > import qualified Control.Monad.Writer as W > import qualified Control.Monad.Reader as R > import Data.Monoid > > class Monad m => MonadTell w m where > tell :: w -> m () > tell w = writer ((), w) > writer :: (a, w) -> m a > writer ~(a, w) = tell w >> return a (We don't need fun.deps. here, they're needed in MonadWriter because of `listen`. IDK if it'd be still better to add fun.dep. just to eliminate typing problems?) And MonadWriter would be defined by inheriting from MonadTell: > class (MonadTell w m, Monoid w) => MonadWriter' w m | m -> w where > listen :: m a -> m (a, w) > pass :: m (a, w -> w) -> m a Now we could use MonadWriter as before, but we could also make more generic writers like: > newtype Log = Log String deriving Show > -- Prints logs to stdout. > instance MonadTell Log IO where > tell (Log s) = putStrLn s > > -- Collects the length of written logs. > instance Monad m => MonadTell Log (W.WriterT (Sum Int) m) where > tell (Log s) = W.tell (Sum $ length s) > > > main = do > let l = Log "Hello world" > tell l > print . getSum . W.execWriter $ (tell l :: W.Writer (Sum Int) ()) The same applies to MonadReader. We could make another type class between Monad and MonadReader just with `ask`: > class Monad m => MonadAsk r m | m -> r where > ask :: m r This would allow us to write instances like > instance MonadAsk Log IO where > ask = liftM Log getLine Does it make sense? Best regards, Petr Pudlak

2 1

poor performance when generating random text
by Dmitry Vyal 17 Oct '12

17 Oct '12

Hello anyone I've written a snippet which generates a file full of random strings. When compiled with -O2 on ghc-7.6, the generation speed is about 2Mb per second which is on par with interpreted php. That's the fact I find rather disappointing. Maybe I've missed something trivial? Any suggestions and explanations are welcome. :) % cat ext_sort.hs import qualified Data.Text as T import System.Random import Control.Exception import Control.Monad import System.IO import qualified Data.Text.IO as TI gen_string g = let (len, g') = randomR (50, 450) g in T.unfoldrN len rand_text (len, g') where rand_text (0,_) = Nothing rand_text (k,g) = let (c, g') = randomR ('a','z') g in Just (c, ((k-1), g')) write_corpus file = bracket (openFile file WriteMode) hClose $ \h -> do let size = 100000 sequence $ replicate size $ do g <- newStdGen let text = gen_string g TI.hPutStrLn h text main = do putStrLn "generating text corpus" write_corpus "test.txt" % cat ext_sort.prof Wed Oct 17 10:59 2012 Time and Allocation Profiling Report (Final) ext_sort +RTS -p -RTS total time = 32.56 secs (32558 ticks @ 1000 us, 1 processor) total alloc = 12,742,917,332 bytes (excludes profiling overheads) COST CENTRE MODULE %time %alloc gen_string.rand_text.(...) Main 70.7 69.8 gen_string Main 17.6 15.8 gen_string.rand_text Main 5.4 13.3 write_corpus.\ Main 4.3 0.8 individual inherited COST CENTRE MODULE no. entries %time %alloc %time %alloc MAIN MAIN 67 0 0.0 0.0 100.0 100.0 main Main 135 0 0.0 0.0 100.0 100.0 write_corpus Main 137 0 0.0 0.0 100.0 100.0 write_corpus.\ Main 138 1 4.3 0.8 100.0 100.0 write_corpus.\.text Main 140 100000 0.0 0.0 95.7 99.2 gen_string Main 141 100000 17.6 15.8 95.7 99.2 gen_string.g' Main 147 100000 0.0 0.0 0.0 0.0 gen_string.rand_text Main 144 25109743 5.4 13.3 77.5 83.2 gen_string.rand_text.g' Main 148 24909743 0.6 0.0 0.6 0.0 gen_string.rand_text.(...) Main 146 25009743 70.7 69.8 70.7 69.8 gen_string.rand_text.c Main 145 25009743 0.8 0.0 0.8 0.0 gen_string.len Main 143 100000 0.0 0.0 0.0 0.0 gen_string.(...) Main 142 100000 0.6 0.3 0.6 0.3

3 4

License of CloudHaskell code write by Haskell language.
by Chatsiri Ratana 17 Oct '12

17 Oct '12

Hello All, License of term of CloudHaskell[1]. It isn't GNU & GPL public license. I not sure to include it's to my project. Who know MPI write by haskell pure and not dependent on C/C++ wrapper such as Haskell-mpi[2]? I need to idea for implement OpenMPI standard by Haskell language. Who have an idea for implementing MPI by Haskell? [1]https://github.com/jepst/CloudHaskell/blob/master/LICENSE [2]http://hackage.haskell.org/package/haskell-mpi Thank in advance, Chatsiri Rattana. -- :--------------------------------------------------------

3 5

Is there a tool to see reduction steps?
by Daryoush Mehrtash 17 Oct '12

17 Oct '12

I have been given a piece of code that uses Tie-ing the Knot concept to label a tree of nodes in breath first manner. It seems to work fine, but I am having trouble expanding the code on my own to see the evaluation process. I like to know if there is a tools to use to see the reduction steps. > data Tree = Leaf | Node Tree Int Tree deriving Show > > label (Node ln _ rn) ((h:r):rest) = (Node lr h rr, r:r2) where > (lr, r1) = label ln rest > (rr, r2) = label rn r1 > label _ _ = (Leaf, []) > lt t = let (r, unused) = label t ([1..]:unused) > in r -- Daryoush Weblog: http://onfp.blogspot.com/

3 2