- Haskell-Cafe - Haskell.org

Re: [Haskell-cafe] Version constraints and cabal.config files
by Miëtek Bak 26 Mar '15

26 Mar '15

On 2015-03-25, at 12:16, Michael Snoyman <michael(a)fpcomplete.com> wrote: > Trying to understand the problem: currently, with your approach, if the project depends on a library not in the LTS Haskell release, then the cabal-install dependency solver won't be able to find it. Instead, you'd like to be able to use the dependency solver to track down those extra dependencies. Is that correct? > > If so, why not take a multi-pass approach: download the cabal.config from stackage.org (which creates an inclusive snapshot), and then use --dry-run, which will tell you all the packages to be used. You’re correct. However, this will have to be `cabal install --dependencies-only --dry-run`, and not `cabal freeze --dry-run`, because `cabal freeze` always completely ignores any existing version constraints, whether local or global (https://github.com/haskell/cabal/issues/2265) I’m planning to switch away from `cabal freeze` soon, but it’s not going to be a drop-in replacement (https://github.com/mietek/halcyon/issues/52) This is a good moment to carefully consider the meaning of a per-application `cabal.config` file, and whether the `cabal freeze` flavour should be completely replaced by separate constraints files. Perhaps, as mentioned in the original `cabal freeze` proposal, we could also have separate constraints sets for different GHC versions. I’m hoping Cabal developers could chime in on the discussion. -- Miëtek https://mietek.io

4 16

Do we have idiom for lifting a state monad into pair of states?
by Ki Yung Ahn 26 Mar '15

26 Mar '15

Consider you have developed library routines that act on (State s1 a). For some reason, you need more than one state simultaneously. Let's say two side by side for simple example, that is (State (s1,s2) a). To use library functions on one state monad in a two state monad, we need to wrapper that lifts actions of (State s1 a) to (State (s1,s2) a). It is not difficult to write a lifter for such purposes as below. It is kind of like doing liftA in Applicative libarary, but instead of the last argument 'a' but on the fist argument 's' of (State s a). This seems like an idiom that can often come up. So, I tried some searching in some applicative related libraries and monad transformer related libraries but haven't found this idiom yet. If you had a need for idioms like below, what do you call it? Or, is there a way more general way I've missed that makes this a very special case of it. > import Control.Monad.State > import Control.Monad.Error > import Control.Applicative > > -- lift an action over a state into a pair of states > -- > do1st :: State s1 a -> State (s1,s2) a > do1st m1 = do (s1, s2) <- get > let (a, s1') = runState m1 s1 > put (s1',s2) > return a > > do2nd :: State s2 a -> State (s1,s2) a > do2nd m2 = do (s1, s2) <- get > let (a, s2') = runState m2 s2 > put (s1,s2') > return a > > > -- lift an action over a state with error > -- into a pair of states with error > -- > do1 :: Error e => ErrorT e (State s1) a -> ErrorT e (State (s1,s2)) a > do1 m1 = do (s1, s2) <- lift get > let (ma,s1') = (runState . runErrorT) m1 s1 > case ma of > Left e -> throwError e > Right a -> do lift $ put (s1',s2) > return a > > > do2 :: Error e => ErrorT e (State s2) a -> ErrorT e (State (s1,s2)) a > do2 m2 = do (s1, s2) <- lift get > let (ma,s2') = (runState . runErrorT) m2 s2 > case ma of > Left e -> throwError e > Right a -> do lift $ put (s1,s2') > return a

4 5

Re: [Haskell-cafe] Version constraints and cabal.config files
by Miëtek Bak 25 Mar '15

25 Mar '15

On 2015-03-25, at 11:40, Michael Snoyman <michael(a)fpcomplete.com> wrote: > Every Stackage Nightly and LTS Haskell release has a YAML file[1] that provides a lot of information, included the exact versions of packages and package dependencies. The stackage-types[2] package provides the datatypes necessary to parse that YAML file. Putting those two together, it should be possible to write something that looks at a project's list of shallow package dependencies, and construct a list of precisely which package-version combinations are necessary to build the entire project, which I think is what you're saying you get from cabal freeze but not from the current cabal.config on stackage.org. If I understand correctly, the result should be equivalent to configuring `cabal-remote-repo` to point to a particular Stackage snapshot, running `cabal freeze`, and parsing its output. Halcyon already does this — the following command will ignore any constraints declared by the application, and instead use the newest versions of all dependencies available in the specified Cabal repository: ``` $ halcyon install http://github.com/mietek/hello-yesod --ignore-all-constraints --cabal-remote-repo=stackage-lts-latest:http://www.stackage.org/lts ``` However, this means we’re using Stackage in “exclusive” mode. I’d like to support using Stackage in “inclusive” mode, with `cabal-remote-repo` pointing to Hackage, and a Stackage-flavoured `cabal.config`. -- Miëtek https://mietek.io

1 0

Version constraints and cabal.config files
by Miëtek Bak 25 Mar '15

25 Mar '15

On 2015-03-25, at 09:59, Chris Done <chrisdone(a)fpcomplete.com> wrote: >> Halcyon wants a constraints file for a particular app to list all Cabal packages which are the app’s direct or indirect dependencies. > > Why is that? What does it do with the cabal.config file? There are two flavours of `cabal.config` files: (1) `cabal freeze`-flavoured, declaring a complete set of dependencies for a particular application, and (2) the Stackage flavour, declaring a set of packages which are guaranteed to build together. It’s worth noting that the first flavour was originally proposed to be a separate file with a simpler syntax, named `cabal.freeze` (https://github.com/haskell/cabal/issues/1502) Unfortunately, this didn’t make it into cabal-install, due to lack of time (https://github.com/haskell/cabal/pull/1519) Halcyon supports `cabal freeze`-flavoured `cabal.config` files, and also accepts separate constraints files, named `.halcyon/constraints`. The declared version constraints are used to calculate a sandbox hash. When building an application, Halcyon uses the sandbox hash to try and locate a previously-built sandbox archive. If available in local cache or remote storage, this archive is restored, completely side-stepping the need to `cabal install --dependencies-only`. Otherwise, a new sandbox is built, archived, and uploaded to remote storage. Moreover, if a full match for the sandbox hash isn’t located, Halcyon attempts to locate a partial match. Each previously-built sandbox is assigned a score, which reflects the number of required dependencies contained within. Sandboxes with extraneous dependencies are rejected. The highest-scoring partial match is used as a base for building the new sandbox. While inferior to the package-level granularity provided by Nix and proposed for future versions of Cabal, the approach implemented in Halcyon has the benefit of working with regular Cabal packages and repositories, and being used in production since June 2014. -- Miëtek https://mietek.io

1 0

cabal test and tests that overlap in covarage
by Anatoly Yakovenko 25 Mar '15

25 Mar '15

is this a cabal error or hpc error? i see this when running 'cabal test' with tests that overlap in coverage 5 of 5 test suites (5 of 5 test cases) passed. hpc: found 2 instances of RBM.List in ["./.hpc","./dist/hpc/vanilla/mix/proto","./dist/hpc/vanilla/mix/perf-repa-RBM","./dist/hpc/vanilla/mix/test-repa-RBM","./dist/hpc/vanilla/mix/perf-list-RBM","./dist/hpc/vanilla/mix/test-list-RBM","./dist/hpc/vanilla/mix/rbm-0.0"] Thanks, Anatoly

2 1

hpc - ignore some expressions
by Ozgur Akgun 24 Mar '15

24 Mar '15

Hi, I am generating code coverage reports using hpc, it works great. The reports would be much more helpful if I could somehow tell hpc to treat certain expressions as "covered" though. As an example, let's think of "impossible" cases of a function. f x | check x = result x | otherwise = error "impossible" I know there are techniques to get rid of (or minimise the number of) the impossible cases. I also can see how this information can be relevant, i.e. seeing whether `check` always holds or not in the report. However, at least for one version of the report, I want the error call like the above example to be treated as covered. Or maybe a different colour could be used to annotate "ignored"? Did anyone need/want anything like this? Is it possible? My guess is that this is somewhat possible using manually generated .tix files and merging them with the .tix file that is generated in the usual way. However this feels too adhoc, fragile and seems to be too much effort to reach a simple goal. Best, Ozgur

4 7

GSOC 2015 proposal feedback
by Nikita Kartashov 24 Mar '15

24 Mar '15

Hello, everyone! As someone on #haskell-gsoc IRC suggested, I would like to post my proposal. Please, if you have time, read it and send me your thoughts. Thanks! Here is the link: https://gist.github.com/nkartashov/e46fd146b1df2d79aaf3 With regards, Nikita Kartashov

1 0

GSoC Proposal: Concurrent Bag.
by Alex Semin 24 Mar '15

24 Mar '15

Hello everyone! I have written a proposal for this GSoC. It is mostly related to this ticket: https://ghc.haskell.org/trac/summer-of-code/ticket/1608 Here is link to the proposal: https://gist.github.com/Alllex/009d9842f97963aa9955 I will be glad to get any feedback and comments. Regards, Alex Semin

1 0

Getting Haskeline to quit early
by G Philip 23 Mar '15

23 Mar '15

(I posted this to Stack Overflow [1], but no luck there so far.) Hi all, I am trying to use Haskeline [2] to write a program which asks the user a sequence of questions, each one optionally with a default value in [brackets], and reads in their responses. I want the user to be able to 1. Press Enter to submit the [default] value; 2. Type in a string, edit it if needed, and then press Enter to submit this value; 3. Press Ctrl-C to reset all values to the defaults and start over; and, 4. Press Ctrl-D or enter "quit" to quit, in which case all the values which they submitted are lost. I have been able to get points 1-3 working, but I cannot get point 4 to work: pressing Ctrl-D (or entering "quit") just brings up the next prompt instead of making the program quit the questioning. Looking at my program (please see below) I understand why this happens, but I am not able to figure out how to fix this so that Ctrl-D (or "quit") actually makes the questioning stop. How do I fix the program to make this happen? I did see this question [2] over at Stack Overflow which seems to ask something similar, but I could not get much from there; I am not even sure that they are asking the same question as I am. As a secondary question: my current program has quite a few `case` statements which switch on `Maybe` values. In particular, I currently check for `Nothing` two or three levels deep so that I can correctly return a `Nothing` when the user presses Ctrl-D. I have a feeling that this could be simplified using (something like) the monadic `>>=` operator, but I am unable to figure out how to do this in this case. Is my hunch right? Is there a way to do away with all this pattern matching which looks for `Nothing`? Also: please tell me anything else which could improve my code below. I am quite new to this, so it is very likely that I am missing many obvious things here. Thanks in advance! Regards, Philip ---------------- The program --------------- My program asks the user about the composition of a fruit basket. The information associated with a fruit basket consists of the name of the owner of the fruit basket and the names of the different kinds of fruit in the basket. To be able to ask for the latter, I first ask for the _number_ of different kind of fruit in the basket, and then ask for the name of each kind. We start with a default fruit basket whose information is then modified based on what the user tells us. module Main where import System.Console.Haskeline type PersonName = String type FruitName = String data FruitBasket = FruitBasket { ownerName :: PersonName, fruitCount :: Int, fruitNames :: [FruitName] } deriving Show defaultBasket = FruitBasket "Mary" 2 ["Apple", "Peach"] main :: IO () main = do basket <- getBasketData defaultBasket putStrLn $ "Got: " ++ show(basket) -- Prompt the user for information about a fruit basket, and -- return a FruitBasket instance containing this information. The -- first argument is an instance of FruitBasket from which we get -- the default values for the various prompts. The return value -- has a Maybe type because the user may abort the questioning, in -- which case we get nothing from them. getBasketData :: FruitBasket -> IO (Maybe FruitBasket) getBasketData basket = runInputT defaultSettings $ withInterrupt $ getData basket where getData :: FruitBasket -> InputT IO (Maybe FruitBasket) getData initialBasket = handleInterrupt f $ do outputStrLn banner input <- getInputLine $ "Who owns this basket? [" ++ defaultOwner ++ "] : " basket <- case input of Nothing -> return Nothing -- User pressed Ctrl-D with the input being empty Just "" -> return (Just initialBasket) -- User pressed Enter with the input being empty Just "quit" -> return Nothing -- User typed in "quit" and pressed Enter Just newOwner -> return (Just initialBasket{ownerName = newOwner}) input <- getInputLine $ "Number of kinds of fruit in the basket? [" ++ show defaultCount ++ "] : " basket' <- case input of Nothing -> return Nothing Just "" -> return basket Just "quit" -> return Nothing Just count -> return $ updateFruitCount basket (read count) where updateFruitCount Nothing _ = Nothing updateFruitCount (Just realBasket) newCount = Just $ realBasket{fruitCount = newCount} let defaultFruitNames = pruneOrPadNames basket' newNames <- getFruitNames defaultFruitNames 1 case newNames of Nothing -> return (Just defaultBasket) Just newSetOfNames -> return $ updateFruitNames basket' newSetOfNames where updateFruitNames Nothing _ = Nothing updateFruitNames (Just realBasket) realNewNames = Just $ realBasket{fruitNames = realNewNames} where f = (outputStrLn "Press Ctrl-D or enter \"quit\" to quit." >> getData initialBasket) defaultOwner = ownerName initialBasket defaultCount = fruitCount initialBasket banner :: String banner = "Please enter details of the fruit basket below. At each prompt you can do one of the following:\n\ \\t (a) Press Enter to submit the [default] value;\n\ \\t (b) Type in a string, edit it if needed, and then press Enter to submit this value;\n\ \\t (c) Press Ctrl-C to reset all values to the defaults and start over;\n\ \\t (d) Press Ctrl-D or enter \"quit\" to quit; all the values you submitted will be lost." pruneOrPadNames :: Maybe FruitBasket -> Maybe [String] pruneOrPadNames Nothing = Nothing pruneOrPadNames (Just basket) = Just $ pruneOrPad (fruitNames basket) (fruitCount basket) -- When requiredLength is not larger than (length inputList), -- (pruneOrPad inputList requiredLength) is the prefix of -- inputList of length requiredLength. Otherwise, it is inputList -- padded with as many empty strings as required to make the total -- length equal to requiredLength. pruneOrPad :: [String] -> Int -> [String] pruneOrPad inputList requiredLength | requiredLength <= inputLength = take requiredLength inputList | otherwise = inputList ++ (replicate difference "") where inputLength = length inputList difference = requiredLength - inputLength getFruitNames Nothing _ = return Nothing getFruitNames (Just []) _ = return $ Just [""] getFruitNames (Just (name:names)) count = do input <- getInputLine $ "Name of fruit " ++ (show count) ++ " [" ++ name ++ "] : " newNames <- case input of Nothing -> return Nothing Just "" -> do -- Keep the default name for this fruit ... newNames' <- getFruitNames (Just names) (count + 1) case newNames' of Nothing -> return Nothing -- ... unless the user chose to quit -- while entering a name Just [""] -> return $ Just [name] -- At this point names = [] so it is -- already time to stop asking for -- more names. Just furtherNames -> return $ Just (name : furtherNames) Just "quit" -> return Nothing Just name' -> do newNames' <- getFruitNames (Just names) (count + 1) case newNames' of Nothing -> return Nothing Just [""] -> return $ Just [name'] Just furtherNames -> return $ Just (name' : furtherNames) return newNames References ------------- [1] http://stackoverflow.com/questions/29189428/getting-haskeline-to-quit-early [2]: https://hackage.haskell.org/package/haskeline-0.7.1.3/docs/System-Console-H… [3]: http://stackoverflow.com/questions/4771199/haskell-best-practise-early-term…

2 2

ANN: Magic Cookies! Android board game in Haskell
by Ivan Perez 23 Mar '15

23 Mar '15

Dear Café This is a just a quick announcement: we have just released Magic Cookies, an Android board game written in Haskell. It is now available on Google Play [1]. The game uses the Functional Reactive Programming library Yampa and SDL2 for multimedia. If you like the game, please share the link on facebook/twitter/etc and let all of your friends know. We have gone through dozens of iterations during beta testing, but we cannot guarantee that the game works 100% perfectly on all devices. If you find a bug, or anything else you don't like, please do not rate low. Instead, open an issue [2] or send us an email (support(a)keera.co.uk) Suggestions are also welcome. We will continue working to make the game lighter, faster, and more fun. See [3] for more details, and to sign up as a beta-tester for other games we are already working on. We certainly enjoyed making it, and we hope you all enjoy playing it. All the best Ivan, Fass & Jorge [1] https://play.google.com/store/apps/details?id=uk.co.keera.games.magiccookies [2] https://github.com/keera-studios/magic-cookies [3] http://keera.co.uk/blog/2015/03/19/magic-cookies-released-google-play/

3 2