March 2017 - Haskell-Cafe

Semigroup and Monoid instances for ReaderT
by Olaf Klinke 21 Mar '17

21 Mar '17

Maybe I missed something, but aren't the transformers a red herring here? instance (Semigroup a, Applicative f) => Semigroup (f a) where (<>) = liftA2 (<>) My first guess was that you could add this instance to your code without the need for changes to the transformers package. The problem with the semigroup instance above is that it conflicts with semigroup instances for f. For example, there is a semigroup instance for [a] when a is not a semigroup. For the declaration above, ghci 7.8.3 gave me: Overlapping instance declarations: instance (Semigroup a, Applicative f) => Semigroup (f a) -- Defined at <interactive>:4:10 instance (Semigroup a, Semigroup b) => Semigroup (a, b) -- Defined in ‘Data.Semigroup’ Hence I guess you must use a newtype wrapper to declare when you want the semigroup instance of a and when you want the instance of f. Olaf

1 0

Registration Open: Compose 2017, New York, May 18-19
by Gershom B 20 Mar '17

20 Mar '17

http://www.composeconference.org/2017 Compose is a conference for typed functional programmers, focused specifically on Haskell, OCaml, F#, SML, and related technologies. It will be held in New York on Thursday and Friday, May 18-19, 2017. Early Bird registration is now open at eventbrite: https://www.eventbrite.com/e/cmpse-conference-2017-tickets-32106159387 Additionally, we're happy to announce our keynote speaker as Emily Riehl, who is a professor of mathematics at Johns Hopkins University, and the author of the book Category Theory in Context (as well as much else). To get a sense of Compose, you can check out the great talks from past conferences: https://www.youtube.com/channel/UC0pEknZxL7Q1j0Ok8qImWdQ Finally, as a reminder, the Call for Presentations closes on March 23, so hurry up and get those talk proposals submitted! http://www.composeconference.org/2017/cfp/ Cheers, Gershom

1 0

Has anyone used Graql in Haskell?
by Jeffrey Brown 20 Mar '17

20 Mar '17

Graql[1] is a (and stands for) graph query language. It is unusually expressive, allowing relationships to be members of other relationships[2] -- effectively, allowing one edge to be an endpoint of another edge. Someone named Felix posted a Haskell library for generating Graql queries[3]. Does anyone have a Haskell program that not only generates but uses such a query? [1] https://hackage.haskell.org/package/graql [2] https://discuss.grakn.ai/t/nested-relationships/ [3] https://discuss.grakn.ai/t/graql-haskell-how-to-run-the-test-program/ -- Jeff Brown | Jeffrey Benjamin Brown Website <https://msu.edu/~brown202/> | Facebook <https://www.facebook.com/mejeff.younotjeff> | LinkedIn <https://www.linkedin.com/in/jeffreybenjaminbrown>(spammy, so I often miss messages here) | Github <https://github.com/jeffreybenjaminbrown>

4 5

Best practices for linting wrt performance
by Adam Bergmark 19 Mar '17

19 Mar '17

I'd like to make aeson hlint clean. aeson contains a lot of performance optimizations which made me wonder; Which linting refactorings should you think twice about when it comes to performance? The best way to find an answer to this for aeson is of course to run the benchmarks, but are there any rules-of-thumb that you take into account when linting? Here's the default hlint output for the current aeson master: https://gist.github.com/bergmark/5dd63c24fe6bdcf67eae0e88495a24e7 Cheers, Adam

3 6

Re: [Haskell-cafe] To [] Or Not To []
by Olaf Klinke 19 Mar '17

19 Mar '17

Just adding to the programming contest anecdotes. Here is what happens if you issue an applied math assignment without specifying the programming language to be used. The first student solves it in bash, to demonstrate how inefficient the solution can be. The second student writes a Turing machine simulator in Maple (I can't recall whether it was the brainf*ck dialect), then programs that Turing machine. The third student writes the solution in postscript and makes the printer solve the differential equation. All were languages available in the setting. I have no idea whether any of the above submissions were inspected by the tutor, but received a mark. Olaf

2 1

Loop with the StateT monad
by p75213＠gmail.com 19 Mar '17

19 Mar '17

This is an exercise to learn the StateT monad. The program implements the game Morra. The two players are the computer and a person. The state accumulates the score of the computer and player. The program works for one iteration of function morra only. However I am at a loss how to loop it. I have tried a few things but nothing seems to work. I have narrowed the problem down to the "p <- liftIO getChar" statement. The second time through the loop it isn't executed. I have bracketed the statement with "liftIO $ putStrLn "before" and liftIO $ putStrLn "after". This is the program: module Morra where import Control.Monad.Trans.State.Lazy import Control.Monad.IO.Class import Data.Char (isDigit, digitToInt) import System.Random (randomRIO) import Control.Monad (when) morra :: StateT (Int, Int) IO () morra = do liftIO $ putStrLn "before" p <- liftIO getChar liftIO $ putStrLn "after" when (isDigit p) $ do let p' = digitToInt p c <- liftIO $ randomRIO (1, 2) liftIO $ putStrLn ("P: " ++ [p]) liftIO $ putStrLn ("C: " ++ show c) (pt, ct) <- get if even (c + p') then do liftIO $ putStrLn "Computer Wins" put (pt, ct + 1) else do liftIO $ putStrLn "Player Wins" put (pt + 1, ct) morra main :: IO () main = do putStrLn "-- p is Player" putStrLn "-- c is Computer" putStrLn "-- Player is odds, Computer is evens." (personS,compS) <- execStateT morra (0,0) putStrLn ("Person Score: " ++ show personS) putStrLn ("Computer Score: " ++ show compS) if personS > compS then putStrLn "Winner is Person" else putStrLn "Winner is Computer" and this is the output: *Morra> main -- p is Player -- c is Computer -- Player is odds, Computer is evens. before 1 after P: 1 C: 2 Player Wins before after Person Score: 1 Computer Score: 0 Winner is Person *Morra>

3 8

Re: [Haskell-cafe] benchmarking a function that returns an unlifted type
by David Feuer 18 Mar '17

18 Mar '17

The critical feature of whnf (as I understand it) is that GHC does not use the static argument transformation. The function passed to "go" therefore doesn't inline into it. So you'll have to write your version in some fashion that similarly prevents undesirable inlining. Could you just copy the code and specialize it to the desired type? On Mar 18, 2017 1:32 PM, "Richard Eisenberg" <rae(a)cs.brynmawr.edu> wrote: Apologies for a newbie question, but I am seeking advice on how to benchmark a pure function that returns an unlifted type. I've heard tell that `criterion` is good for benchmarking, and so I'm using that library (but have no particular allegiance to it, if something else is better). Specifically, I'm comparing a function over unlifted types with the same function written over lifted types, trying to observe how much we pay for the boxing/unboxing. To benchmark the lifted version, I just use criterion's whnf function. That seems to be working splendidly. Of course, whnf doesn't work over unlifted types, so I found its source code and inlined it. This compiles. But it doesn't work! No matter what I do, the unlifted version measures in the nanoseconds. I'm guessing that GHC "cleverly" is avoiding recomputation. How can I stop this? Here is my code: > main :: IO () > main = do > initializeTime > let lifted = whnf fastSumPrimes 100000 > > unlifted = Benchmarkable go > where > go n > | n <= 0 = return () > | otherwise = let x = fastSumPrimes# 100000# in go (n-1) > > benchmark lifted > benchmark unlifted I've tried increasing the argument in the unlifted case, to no avail. I've also tried using `case` instead of `let`. -ddump-simpl suggests that fastSumPrimes# is really being called, but I'm dubious. Can anyone offer advice? Thanks! Richard _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

1 0

benchmarking a function that returns an unlifted type
by Richard Eisenberg 18 Mar '17

18 Mar '17

Apologies for a newbie question, but I am seeking advice on how to benchmark a pure function that returns an unlifted type. I've heard tell that `criterion` is good for benchmarking, and so I'm using that library (but have no particular allegiance to it, if something else is better). Specifically, I'm comparing a function over unlifted types with the same function written over lifted types, trying to observe how much we pay for the boxing/unboxing. To benchmark the lifted version, I just use criterion's whnf function. That seems to be working splendidly. Of course, whnf doesn't work over unlifted types, so I found its source code and inlined it. This compiles. But it doesn't work! No matter what I do, the unlifted version measures in the nanoseconds. I'm guessing that GHC "cleverly" is avoiding recomputation. How can I stop this? Here is my code: > main :: IO () > main = do > initializeTime > let lifted = whnf fastSumPrimes 100000 > > unlifted = Benchmarkable go > where > go n > | n <= 0 = return () > | otherwise = let x = fastSumPrimes# 100000# in go (n-1) > > benchmark lifted > benchmark unlifted I've tried increasing the argument in the unlifted case, to no avail. I've also tried using `case` instead of `let`. -ddump-simpl suggests that fastSumPrimes# is really being called, but I'm dubious. Can anyone offer advice? Thanks! Richard

2 1

Tools to assist with development cycle of hackage packages
by Clinton Mead 17 Mar '17

17 Mar '17

I've only just started uploading packages to hackage (my package listing is here <https://hackage.haskell.org/user/clinton>) but currently there's a lot of repetitive activities for creating and updating packages. For example, here's the steps in creating a package: 1. Initialise a repository on github 2. Initialise git repository locally 3. Set github repository as remote 4. Add a LICENSE file 5. Add a standard ".gitignore file" 6. Create a cabal file with the appropriate files that hackage requires including: 1. Git repository source 2. Issues page 3. Licence 4. Licence file 7. Run multi-ghc-travis <https://github.com/hvr/multi-ghc-travis> to create a ".travis.yml" file 8. Make an initial commit and push 9. Refresh travis-ci.org's repository list so it detects the new repository There's also the stack stuff, but the GUI I'm using, IntelliJ with a Haskell plugin, handles most of that. It also creates a cabal file, but it's missing a number of key fields as mentioned above. When I actually want to upload the package I go though these steps: 1. Push to github 2. Wait for Travis-CI to compile the package (this is a test to ensure it builds in a clean remote environment). 3. Run "cabal sdist 2>&1" 4. Parse the output of sdist to see where the dist file is. 5. Run "hup <https://hackage.haskell.org/package/hup> packup fileFromSDist" to upload the package, putting in my hackage user/pass 6. Run "hup docboth", to both build and upload the documentation. 7. Tag the commit as a release Currently, I've got two scripts with help with a lot of this, but it's a bit adhoc, and it's not fully automated (for example, I still have to manually ensure all the correct fields are in the cabal file, usually by copy/pasting from another package and modifying). Are there any tools that I haven't found that make this process a bit more painless? I'm a bit new to this area, and I've only started using git recently as a prelude to uploading my Haskell packages, so admittedly I may have missed something obvious or perhaps I'm just doing it all wrong. But if other people do find it painful like me, perhaps I'll put some effort into rewriting my perl scripts into nice haskell packages and executables for others to use.

6 9

Package takeover: dbus
by Andrey Sverdlichenko 17 Mar '17

17 Mar '17

I'd like to take over maintenance of the package dbus. It depends on several other packaged with upper bounds too restrictive for the latest versions. I tried to contact owner a few times weeks ago, but to no avail. My hackage username is blaze.

1 0