Documentation

Differences when using H from compiled Haskell modules

There are two ways to use inline-r. The simplest is at an interactive prompt, such as H or IHaskell, for interacting with R in the small. But we support equally well writing full blown programs that interact with R in elaborate and intricate ways, possibly even multiple instances of R, using the inline-r library.

For simplicity, at the H interactive prompt, every function in the R library is either pure or lifted to the IO monad. This is because the prompt itself is an instance of the IO monad. However, in large projects, one would like to enforce static guarantees about the R interpreter being properly initialized before attempting to invoke any of its internal functions. Hence, in .hs source files inline-r instead lifts all functions to the R monad, which provides stronger static guarantees than the IO monad. This parametricity over the underlying monad is achieved by introducing the MonadR class, as explained in previous sections.

To avoid having multiple instances of MonadR lying around, it is important NOT to import Language.R.Instance.Interactive in compiled code - that module should only be loaded in an interactive session.

“Hello World!” from source

Another major difference between interactive sessions and compiled programs linked against inline-r directly is that one needs to handle R initialization and configuration explicitly in compiled programs, while H --interactive takes care of this for us. Here is a template small program using the inline-r library:

{-# LANGUAGE QuasiQuotes #-}
module Main where

import qualified Foreign.R as R
import Foreign.R (SEXP, SEXPTYPE)
import Language.R.Instance as R
import Language.R.QQ

hello :: String -> R s ()
hello name = do
    [r| print(s_hs) |]
    return ()
  where
    s = "Hello, " ++ name ++ "!"

main :: IO ()
main = do
    putStrLn "Name?"
    name <- getLine
    R.withEmbeddedR R.defaultConfig $ R.runRegion $ hello name