Awesome

Repline

Slightly higher level wrapper for creating GHCi-like REPL monads that are composable with normal MTL transformers. Mostly exists because I got tired of implementing the same interface for simple shells over and over and decided to canonize the giant pile of hacks that I use to make Haskeline work.

See Documentation for more detailed usage.

Examples

Migration from 0.3.x

This release adds two parameters to the ReplOpts constructor and evalRepl function.

finaliser
multilineCommand

The finaliser function is a function run when the Repl monad is is exited.

-- | Decide whether to exit the REPL or not
data ExitDecision
  = Continue -- | Keep the REPL open
  | Exit     -- | Close the REPL and exit

For example:

final :: Repl ExitDecision
final = do
  liftIO $ putStrLn "Goodbye!"
  return Exit

The multilineCommand argument takes a command which invokes a multiline edit mode in which the user can paste/enter text across multiple lines terminating with a Ctrl-D / EOF. This can be used in conjunction with a customBanner function to indicate the entry mode.

customBanner :: MultiLine -> Repl String
customBanner SingleLine = pure ">>> "
customBanner MultiLine = pure "| "

See Multiline for a complete example.

Migration from 0.2.x

The underlying haskeline library that provides readline support had a breaking API change in 0.8.0.0 which removed the bespoke System.Console.Haskeline.MonadException module in favour of using the exceptions package. This is a much better design and I strongly encourage upgrading. To migrate simply add the following bounds to your Cabal file.

build-depends:
  repline   >= 0.3.0.0
  haskeline >= 0.8.0.0

You may also need to add the following to your stack.yaml file if using Stack.

resolver: lts-15.0
packages:
  - .
extra-deps:
  - haskeline-0.8.0.0
  - repline-0.3.0.0

Usage

type Repl a = HaskelineT IO a

-- Evaluation : handle each line user inputs
cmd :: String -> Repl ()
cmd input = liftIO $ print input

-- Tab Completion: return a completion for partial words entered
completer :: Monad m => WordCompleter m
completer n = do
  let names = ["kirk", "spock", "mccoy"]
  return $ filter (isPrefixOf n) names

-- Commands
help :: [String] -> Repl ()
help args = liftIO $ print $ "Help: " ++ show args

say :: [String] -> Repl ()
say args = do
  _ <- liftIO $ system $ "cowsay" ++ " " ++ (unwords args)
  return ()

options :: [(String, [String] -> Repl ())]
options = [
    ("help", help)  -- :help
  , ("say", say)    -- :say
  ]

ini :: Repl ()
ini = liftIO $ putStrLn "Welcome!"

repl :: IO ()
repl = evalRepl (pure ">>> ") cmd options Nothing (Word completer) ini

Trying it out:

$ stack repl Simple.hs
Prelude> main

Welcome!
>>> <TAB>
kirk spock mccoy

>>> k<TAB>
kirk

>>> spam
"spam"

>>> :say Hello Haskell
 _______________
< Hello Haskell >
 ---------------
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||

Stateful Tab Completion

Quite often tab completion is dependent on the internal state of the Repl so we'd like to query state of the interpreter for tab completions based on actions performed themselves within the Repl, this is modeleted naturally as a monad transformer stack with StateT on top of HaskelineT.

type IState = Set.Set String
type Repl a = HaskelineT (StateT IState IO) a

-- Evaluation
cmd :: String -> Repl ()
cmd input = modify $ \s -> Set.insert input s

-- Completion
comp :: (Monad m, MonadState IState m) => WordCompleter m
comp n = do
  ns <- get
  return  $ filter (isPrefixOf n) (Set.toList ns)

-- Commands
help :: [String] -> Repl ()
help args = liftIO $ print $ "Help!" ++ show args

puts :: [String] -> Repl ()
puts args = modify $ \s -> Set.union s (Set.fromList args)

opts :: [(String, [String] -> Repl ())]
opts = [
    ("help", help) -- :help
  , ("puts", puts) -- :puts
  ]

ini :: Repl ()
ini = return ()

-- Tab completion inside of StateT
repl :: IO ()
repl = flip evalStateT Set.empty
     $ evalRepl (pure ">>> ") cmd opts Nothing (Word comp) ini

Prefix Completion

Just as GHCi will provide different tab completion for kind-level vs type-level symbols based on which prefix the user has entered, we can also set up a provide this as a first-level construct using a Prefix tab completer which takes care of the string matching behind the API.

type Repl a = HaskelineT IO a

-- Evaluation
cmd :: String -> Repl ()
cmd input = liftIO $ print input

-- Prefix tab completeter
defaultMatcher :: MonadIO m => [(String, CompletionFunc m)]
defaultMatcher = [
    (":file"    , fileCompleter)
  , (":holiday" , listCompleter ["christmas", "thanksgiving", "festivus"])
  ]

-- Default tab completer
byWord :: Monad m => WordCompleter m
byWord n = do
  let names = ["picard", "riker", "data", ":file", ":holiday"]
  return $ filter (isPrefixOf n) names

files :: [String] -> Repl ()
files args = liftIO $ do
  contents <- readFile (unwords args)
  putStrLn contents

holidays :: [String] -> Repl ()
holidays [] = liftIO $ putStrLn "Enter a holiday."
holidays xs = liftIO $ do
  putStrLn $ "Happy " ++ unwords xs ++ "!"

opts :: [(String, [String] -> Repl ())]
opts = [
    ("file", files)
  , ("holiday", holidays)
  ]

init :: Repl ()
init = return ()

repl :: IO ()
repl = evalRepl (pure ">>> ") cmd opts Nothing (Prefix (wordCompleter byWord) defaultMatcher) init