Merge pull request #359 from github/ffi-ffs

FFI, ffs

semantic-diff.cabal

@@ -36,8 +36,9 @@ library
                      , bytestring
                      , blaze-html
                      , tree-sitter-parsers
+                     , c-storable-deriving
   default-language:    Haskell2010
-  default-extensions:  DeriveFunctor, FlexibleInstances, DeriveFoldable, DeriveTraversable
+  default-extensions:  DeriveFunctor, FlexibleInstances, DeriveFoldable, DeriveTraversable, DeriveGeneric
   ghc-options:        -Wall -fno-warn-name-shadowing
 
 executable semantic-diff-exe

src/Diff.hs

@@ -8,16 +8,15 @@ import Term
 import Range
 import Categorizable
 
-data Annotated a annotation f = Annotated annotation (Syntax a f)
+data Annotated a annotation f = Annotated !annotation !(Syntax a f)
   deriving (Functor, Eq, Show, Foldable)
 
-
 type Category = String
-data Info = Info { characterRange :: Range, categories :: (Set Category) }
+data Info = Info { characterRange :: !Range, categories :: !(Set Category) }
   deriving (Eq, Show)
 
 instance Categorizable Info where
-  categories info = Diff.categories info
+  categories = Diff.categories
 
 type Diff a annotation = Free (Annotated a (annotation, annotation)) (Patch (Term a annotation))
 

src/Range.hs

@@ -2,8 +2,9 @@ module Range where
 
 import Control.Applicative ((<|>))
 import qualified Data.Char as Char
+import Data.Maybe (fromMaybe)
 
-data Range = Range { start :: Int, end :: Int }
+data Range = Range { start :: !Int, end :: !Int }
   deriving (Eq, Show)
 
 substring :: Range -> String -> String
@@ -17,7 +18,7 @@ offsetRange i (Range start end) = Range (i + start) (i + end)
 
 rangesAndWordsFrom :: Int -> String -> [(Range, String)]
 rangesAndWordsFrom _ "" = []
-rangesAndWordsFrom startIndex string = maybe [] id $ takeAndContinue <$> (word <|> punctuation) <|> skipAndContinue <$> space
+rangesAndWordsFrom startIndex string = fromMaybe [] $ takeAndContinue <$> (word <|> punctuation) <|> skipAndContinue <$> space
   where
     word = parse isWord string
     punctuation = parse (not . isWordOrSpace) string

src/TreeSitter.hs

@@ -3,11 +3,12 @@ module TreeSitter where
 import Diff
 import Range
 import Parser
-import Term
 import qualified Data.Set as Set
 import Foreign
 import Foreign.C
 import Foreign.C.Types
+import Foreign.CStorable
+import qualified GHC.Generics as Generics
 
 data TSLanguage = TsLanguage deriving (Show, Eq)
 foreign import ccall "prototype/doubt-difftool/doubt-difftool-Bridging-Header.h ts_language_c" ts_language_c :: Ptr TSLanguage
@@ -20,17 +21,15 @@ foreign import ccall "prototype/External/tree-sitter/include/tree_sitter/runtime
 foreign import ccall "prototype/External/tree-sitter/include/tree_sitter/runtime.h ts_document_parse" ts_document_parse :: Ptr TSDocument -> IO ()
 foreign import ccall "prototype/External/tree-sitter/include/tree_sitter/runtime.h ts_document_free" ts_document_free :: Ptr TSDocument -> IO ()
 
-data TSLength = TsLength { bytes :: CSize, chars :: CSize }
-  deriving (Show, Eq)
-
-data TSNode = TsNode { _data :: Ptr (), offset :: TSLength }
-  deriving (Show, Eq)
+data TSNode = TsNode { _data :: Ptr (), offset0 :: CSize, offset1 :: CSize, offset2 :: CSize }
+  deriving (Show, Eq, Generics.Generic)
 
+instance CStorable TSNode
 instance Storable TSNode where
-  alignment _ = 32
-  sizeOf _ = 32
-  peek _ = error "Haskell code should never read TSNode values directly."
-  poke _ _ = error "Haskell code should never write TSNode values directly."
+  alignment = cAlignment
+  sizeOf = cSizeOf
+  peek = cPeek
+  poke = cPoke
 
 foreign import ccall "app/bridge.h ts_document_root_node_p" ts_document_root_node_p :: Ptr TSDocument -> Ptr TSNode -> IO ()
 foreign import ccall "app/bridge.h ts_node_p_name" ts_node_p_name :: Ptr TSNode -> Ptr TSDocument -> IO CString
@@ -55,7 +54,7 @@ parseTreeSitterFile :: Language -> Parser
 parseTreeSitterFile (Language language constructor) contents = do
   document <- ts_document_make
   ts_document_set_language document language
-  withCString contents (\source -> do
+  withCString contents (\ source -> do
     ts_document_set_input_string document source
     ts_document_parse document
     term <- documentToTerm constructor document contents
@@ -63,25 +62,18 @@ parseTreeSitterFile (Language language constructor) contents = do
     return term)
 
 documentToTerm :: Constructor -> Ptr TSDocument -> Parser
-documentToTerm constructor document contents = alloca $ \root -> do
+documentToTerm constructor document contents = alloca $ \ root -> do
   ts_document_root_node_p document root
-  snd <$> toTerm root where
-    toTerm :: Ptr TSNode -> IO (String, Term String Info)
-    toTerm node = do
-      name <- ts_node_p_name node document
-      name <- peekCString name
-      children <- withNamedChildren node toTerm
-      return (name, constructor contents (Info (range node) $ Set.singleton name) children)
-
-withNamedChildren :: Ptr TSNode -> (Ptr TSNode -> IO (String, a)) -> IO [(String, a)]
-withNamedChildren node transformNode = do
-  count <- ts_node_p_named_child_count node
-  if count == 0
-    then return []
-    else mapM (alloca . getChild) [0..pred count] where
-      getChild n out = do
-        _ <- ts_node_p_named_child node n out
-        transformNode out
-
-range :: Ptr TSNode -> Range
-range node = Range { start = fromIntegral $ ts_node_p_start_char node, end = fromIntegral $ ts_node_p_end_char node }
+  (_, term) <- toTerm root
+  return term
+  where toTerm node = do
+          name <- ts_node_p_name node document
+          name <- peekCString name
+          count <- ts_node_p_named_child_count node
+          children <- mapM (alloca . getChild node) $ take (fromIntegral count) [0..]
+          -- Note: The strict application here is semantically important. Without it, we may not evaluate the range until after we’ve exited the scope that `node` was allocated within, meaning `alloca` will free it & other stack data may overwrite it.
+          range <- return $! Range { start = fromIntegral $ ts_node_p_start_char node, end = fromIntegral $ ts_node_p_end_char node }
+          return (name, constructor contents (Info range (Set.singleton name)) children)
+        getChild node n out = do
+          _ <- ts_node_p_named_child node n out
+          toTerm out