@hackage bidirectionalization-combined0.1

Prototype Implementation of Combining Syntatic and Semantic Bidirectionalization (ICFP'10)

b18n-combined: Combining Syntactic and Semantic Bidirectionalization

What's This?

This program, b18n-combined, is a prototype implementation of the paper "Combining Syntactic and Semantic Bidirectionalization" written by Janis Voigtländer, Zhenjiang Hu, Kazutaka Matsuda, and Meng Wang, which is accepted at ICFP'10.

The program takes input program describing a view function that constructs a view from a source, and generates a put-back function that reflects updates on the view to the source. Thanks to combination of syntactic and semantic bidirectionalization, the program can generates effective put-back functions in the sense that the generated put-back functions can put back more updates than those obtained by either of syntactic/semantic bidirectionalization.

Online Version

A CGI version of the program is provided for your convenience.

Example

$ cat example/init.txt 
init []          = []
init [a]         = []
init (a:b:x)     = a:initWork b x
initWork a []    = []
initWork a (b:x) = a:initWork b x
$ ./dist/build/b18n-combined/b18n-combined example/sieve.txt -syn -hs
import Control.Monad
import BUtil
data Cmpl t0 t1
    = Csieve_Cmpl_1
    | Csieve_Cmpl_2 t0
    | Csieve_Cmpl_3 t1 (Cmpl t0 t1)
sieve ([]) = []
sieve (a : []) = []
sieve (a : (b : x)) = b : sieve (x)
sieve_B s v = head (sieve_T_I v (sieve_Cmpl s))
sieve_Cmpl ([]) = Csieve_Cmpl_1
sieve_Cmpl (a : []) = Csieve_Cmpl_2 a
sieve_Cmpl (a : (b : x)) = Csieve_Cmpl_3 a (sieve_Cmpl (x))
sieve_T_I x1 x2 = mplus (sieve_T_I_1 x1 x2) (mplus (sieve_T_I_2 x1 x2) (mplus (sieve_T_I_3 x1 x2) mzero))
sieve_T_I_1 [] (Csieve_Cmpl_1) = do return ([])
sieve_T_I_1 _ _ = mzero
sieve_T_I_2 [] (Csieve_Cmpl_2 a) = do return (a : [])
sieve_T_I_2 _ _ = mzero
sieve_T_I_3 (b : tv1) (Csieve_Cmpl_3 a
                                     tc1) = do (x) <- sieve_T_I tv1 tc1
                                               return (a : (b : x))
sieve_T_I_3 _ _ = mzero
$ ./b18n-combined example/sieve.txt -sem
import Data.Bff
import BUtil
sieve_B s v
    = bff Main.sieve s v
sieve ([]) = []
sieve (a : []) = []
sieve (a : (b : x)) = b : sieve (x)
$ ./b18n-combined example/sieve.txt -comb 
import Control.Monad
import BUtil
sieve_Bb bias s v
    = gen_put_bias bias Main.sieve(\x y -> castError $ (sieve_22_B $! x) $! y) s v
sieve_Bbd = withDefaultBias sieve_Bb
sieve_Bd = withDefault sieve_B
sieve_B s v = sieve_Bb rear s v
data Cmpl
    = Csieve_22_Cmpl_1
    | Csieve_22_Cmpl_2
sieve ([]) = []
sieve (a : []) = []
sieve (a : (b : x)) = b : sieve (x)
sieve_22_B s v = head (sieve_22_T_I v (sieve_22_Cmpl s))
sieve_22_Cmpl (Z) = Csieve_22_Cmpl_1
sieve_22_Cmpl (S (Z)) = Csieve_22_Cmpl_2
sieve_22_Cmpl (S (S x)) = sieve_22_Cmpl (x)
sieve_22_T_I x1 x2 = mplus (sieve_22_T_I_1 x1 x2) (mplus (sieve_22_T_I_2 x1 x2) (mplus (sieve_22_T_I_3 x1 x2) mzero))
sieve_22_T_I_1 (Z) (Csieve_22_Cmpl_1) = do return (Z)
sieve_22_T_I_1 _ _ = mzero
sieve_22_T_I_2 (Z) (Csieve_22_Cmpl_2) = do return (S Z)
sieve_22_T_I_2 _ _ = mzero
sieve_22_T_I_3 (S tv1) tc1 = do (x) <- sieve_22_T_I tv1 tc1
                                return (S (S x))
sieve_22_T_I_3 _ _ = mzero

How to Build Using cabal-install

If you have cabal-install (part of the Haskell Platform) available on your system, you can install b18n-combined using the command cabal install bidirectionalization-combined

How to Build Manually

  1. Install GHC and make sures that following Haskell packages are installed.
    • mtl, template-haskell, containers, pretty, parsec (for web interface: directory, xhtml, cgi, utf8-string, bytestring, unix, hint >= 0.3.2)
  2. Get the source tarball from ./bidirectionalization-combined-0.1.tar.gz or from hackage.
  3. Unfold the source tarball
    • tar zxvf bidirectionalization-combined-0.1.tar.gz
  4. Build the executables by
    1. runhaskell Setup.hs configure
    2. runhaskell Setup.hs build

Note that the source code is also available via darcs.

 darcs get http://www.kb.ecei.tohoku.ac.jp/~kztk/darcs/sem_syn/

Limitation

Current implementation can handle lists only.