{-# LANGUAGE BangPatterns               #-}
{-# LANGUAGE DeriveAnyClass             #-}
{-# LANGUAGE DeriveDataTypeable         #-}
{-# LANGUAGE DeriveGeneric              #-}
{-# LANGUAGE GeneralizedNewtypeDeriving #-}
{-# LANGUAGE CPP                        #-}
-- | DOM parser and API for XML.

module Xeno.DOM
  ( parse
  , Node
  , Content(..)
  , name
  , attributes
  , contents
  , children
  ) where

import           Control.Monad.ST
import           Control.Spork
import           Data.ByteString          (ByteString)
#if MIN_VERSION_bytestring(0,11,0)
import           Data.ByteString.Internal (ByteString(BS))
#else
import           Data.ByteString.Internal (ByteString(PS))
#endif
import qualified Data.ByteString as S
import           Data.Mutable
import           Data.STRef
import qualified Data.Vector.Unboxed as UV
import qualified Data.Vector.Unboxed.Mutable as UMV
#if MIN_VERSION_bytestring(0,11,0)
import           Foreign.Ptr (minusPtr)
import           Foreign.ForeignPtr (ForeignPtr, withForeignPtr)
import           System.IO.Unsafe (unsafeDupablePerformIO)
#endif
import           Xeno.SAX
import           Xeno.Types
import           Xeno.DOM.Internal

-- | Parse a complete Nodes document.
parse :: ByteString -> Either XenoException Node
parse :: ByteString -> Either XenoException Node
parse ByteString
str =
  Either XenoException Node
-> (Node -> Either XenoException Node)
-> Maybe Node
-> Either XenoException Node
forall b a. b -> (a -> b) -> Maybe a -> b
maybe (XenoException -> Either XenoException Node
forall a b. a -> Either a b
Left XenoException
XenoExpectRootNode) Node -> Either XenoException Node
forall a b. b -> Either a b
Right (Maybe Node -> Either XenoException Node)
-> (Vector Int -> Maybe Node)
-> Vector Int
-> Either XenoException Node
forall b c a. (b -> c) -> (a -> b) -> a -> c
. Vector Int -> Maybe Node
findRootNode (Vector Int -> Either XenoException Node)
-> Either XenoException (Vector Int) -> Either XenoException Node
forall (m :: * -> *) a b. Monad m => (a -> m b) -> m a -> m b
=<< Vector Int -> Either XenoException (Vector Int)
forall e a. Exception e => a -> Either e a
spork Vector Int
node
  where
    findRootNode :: Vector Int -> Maybe Node
findRootNode Vector Int
r = Int -> Maybe Node
go Int
0
      where
        go :: Int -> Maybe Node
go Int
n = case Vector Int
r Vector Int -> Int -> Maybe Int
forall a. Unbox a => Vector a -> Int -> Maybe a
UV.!? Int
n of
          Just Int
0x0 -> Node -> Maybe Node
forall a. a -> Maybe a
Just (ByteString -> Int -> Vector Int -> Node
Node ByteString
str Int
n Vector Int
r)
          -- skipping text assuming that it contains only white space
          -- characters
          Just Int
0x1 -> Int -> Maybe Node
go (Int
nInt -> Int -> Int
forall a. Num a => a -> a -> a
+Int
3)
          Maybe Int
_        -> Maybe Node
forall a. Maybe a
Nothing
#if MIN_VERSION_bytestring(0,11,0)
    BS ForeignPtr Word8
offset0 Int
_ = ByteString
str
#else
    PS _ offset0 _ = str
#endif
    node :: Vector Int
node =
      let !initialSize :: Int
initialSize = Int -> Int -> Int
forall a. Ord a => a -> a -> a
max Int
1000 (ByteString -> Int
S.length ByteString
str Int -> Int -> Int
forall a. Integral a => a -> a -> a
`div` Int
8) in
      (forall s. ST s (Vector Int)) -> Vector Int
forall a. (forall s. ST s a) -> a
runST
        (do nil <- Int -> ST s (MVector (PrimState (ST s)) Int)
forall (m :: * -> *) a.
(PrimMonad m, Unbox a) =>
Int -> m (MVector (PrimState m) a)
UMV.unsafeNew Int
initialSize
            vecRef <- newSTRef nil
            sizeRef <- fmap asURef (newRef 0)
            parentRef <- fmap asURef (newRef 0)
            process Process {
#if MIN_VERSION_bytestring(0,11,0)
                openF = \(BS ForeignPtr Word8
name_start Int
name_len) -> do
#else
                openF = \(PS _ name_start name_len) -> do
#endif
                     let tag :: Int
tag = Int
0x00
                         tag_end :: Int
tag_end = -Int
1
                     index <- URef s Int -> ST s (RefElement (URef s Int))
forall c (m :: * -> *).
(MutableRef c, PrimMonad m, PrimState m ~ MCState c) =>
c -> m (RefElement c)
forall (m :: * -> *).
(PrimMonad m, PrimState m ~ MCState (URef s Int)) =>
URef s Int -> m (RefElement (URef s Int))
readRef URef s Int
sizeRef
                     v' <-
                       do v <- readSTRef vecRef
                          if index + 5 < UMV.length v
                            then pure v
                            else do
                              v' <- UMV.unsafeGrow v (predictGrowSize name_start name_len (index + 5) (UMV.length v))
                              writeSTRef vecRef v'
                              return v'
                     tag_parent <- readRef parentRef
                     do writeRef parentRef index
                        writeRef sizeRef (index + 5)
                     do UMV.unsafeWrite v' index tag
                        UMV.unsafeWrite v' (index + 1) tag_parent
                        UMV.unsafeWrite v' (index + 2) (distance name_start offset0)
                        UMV.unsafeWrite v' (index + 3) name_len
                        UMV.unsafeWrite v' (index + 4) tag_end
#if MIN_VERSION_bytestring(0,11,0)
              , attrF = \(BS ForeignPtr Word8
key_start Int
key_len) (BS ForeignPtr Word8
value_start Int
value_len) -> do
#else
              , attrF = \(PS _ key_start key_len) (PS _ value_start value_len) -> do
#endif
                     index <- URef s Int -> ST s (RefElement (URef s Int))
forall c (m :: * -> *).
(MutableRef c, PrimMonad m, PrimState m ~ MCState c) =>
c -> m (RefElement c)
forall (m :: * -> *).
(PrimMonad m, PrimState m ~ MCState (URef s Int)) =>
URef s Int -> m (RefElement (URef s Int))
readRef URef s Int
sizeRef
                     v' <-
                       do v <- readSTRef vecRef
                          if index + 5 < UMV.length v
                            then pure v
                            else do
                              v' <- UMV.unsafeGrow v (predictGrowSize value_start value_len (index + 5) (UMV.length v))
                              writeSTRef vecRef v'
                              return v'
                     let tag = Int
0x02
                     do writeRef sizeRef (index + 5)
                     do UMV.unsafeWrite v' index tag
                        UMV.unsafeWrite v' (index + 1) (distance key_start offset0)
                        UMV.unsafeWrite v' (index + 2) key_len
                        UMV.unsafeWrite v' (index + 3) (distance value_start offset0)
                        UMV.unsafeWrite v' (index + 4) value_len
              , endOpenF = \ByteString
_ -> () -> ST s ()
forall a. a -> ST s a
forall (m :: * -> *) a. Monad m => a -> m a
return ()
#if MIN_VERSION_bytestring(0,11,0)
              , textF = \(BS ForeignPtr Word8
text_start Int
text_len) -> do
#else
              , textF = \(PS _ text_start text_len) -> do
#endif
                     let tag :: Int
tag = Int
0x01
                     index <- URef s Int -> ST s (RefElement (URef s Int))
forall c (m :: * -> *).
(MutableRef c, PrimMonad m, PrimState m ~ MCState c) =>
c -> m (RefElement c)
forall (m :: * -> *).
(PrimMonad m, PrimState m ~ MCState (URef s Int)) =>
URef s Int -> m (RefElement (URef s Int))
readRef URef s Int
sizeRef
                     v' <-
                       do v <- readSTRef vecRef
                          if index + 3 < UMV.length v
                            then pure v
                            else do
                              v' <- UMV.unsafeGrow v (predictGrowSize text_start text_len (index + 3) (UMV.length v))
                              writeSTRef vecRef v'
                              return v'
                     do writeRef sizeRef (index + 3)
                     do UMV.unsafeWrite v' index tag
                        UMV.unsafeWrite v' (index + 1) (distance text_start offset0)
                        UMV.unsafeWrite v' (index + 2) text_len
              , closeF = \ByteString
_ -> do
                     v <- STRef s (MVector s Int) -> ST s (MVector s Int)
forall s a. STRef s a -> ST s a
readSTRef STRef s (MVector s Int)
vecRef
                     -- Set the tag_end slot of the parent.
                     parent <- readRef parentRef
                     index <- readRef sizeRef
                     UMV.unsafeWrite v (parent + 4) index
                     -- Pop the stack and return to the parent element.
                     previousParent <- UMV.unsafeRead v (parent + 1)
                     writeRef parentRef previousParent
#if MIN_VERSION_bytestring(0,11,0)
              , cdataF = \(BS ForeignPtr Word8
cdata_start Int
cdata_len) -> do
#else
              , cdataF = \(PS _ cdata_start cdata_len) -> do
#endif
                     let tag :: Int
tag = Int
0x03
                     index <- URef s Int -> ST s (RefElement (URef s Int))
forall c (m :: * -> *).
(MutableRef c, PrimMonad m, PrimState m ~ MCState c) =>
c -> m (RefElement c)
forall (m :: * -> *).
(PrimMonad m, PrimState m ~ MCState (URef s Int)) =>
URef s Int -> m (RefElement (URef s Int))
readRef URef s Int
sizeRef
                     v' <- do
                       v <- readSTRef vecRef
                       if index + 3 < UMV.length v
                         then pure v
                         else do
                           v' <- UMV.unsafeGrow v (predictGrowSize cdata_start cdata_len (index + 3) (UMV.length v))
                           writeSTRef vecRef v'
                           return v'
                     writeRef sizeRef (index + 3)
                     UMV.unsafeWrite v' index tag
                     UMV.unsafeWrite v' (index + 1) (distance cdata_start offset0)
                     UMV.unsafeWrite v' (index + 2) cdata_len
              } str
            wet <- readSTRef vecRef
            arr <- UV.unsafeFreeze wet
            size <- readRef sizeRef
            return (UV.unsafeSlice 0 size arr))
            where
                -- Growing a large vector is slow, so we need to do it less times.
                -- We can predict final array size after processing some part (i.e. 1/4) of input XML.
                --
                -- predictGrowSize _bsStart _bsLen _index vecLen = round $ fromIntegral vecLen * (1.25 :: Double)
                predictGrowSize :: ForeignPtr a -> Int -> p -> b -> b
predictGrowSize ForeignPtr a
bsStart Int
bsLen p
index b
vecLen =
                    let -- at least 1 so we don't divide by zero below and end up with 
                        -- a negative grow size if (bsStart + bsLen - offset0) == 0
                        processedLen :: Int
processedLen = Int -> Int -> Int
forall a. Ord a => a -> a -> a
max Int
1 (ForeignPtr a -> ForeignPtr Word8 -> Int
forall a b. ForeignPtr a -> ForeignPtr b -> Int
distance ForeignPtr a
bsStart ForeignPtr Word8
offset0 Int -> Int -> Int
forall a. Num a => a -> a -> a
+ Int
bsLen)
                        -- 1. Using integral operations, such as
                        --    "predictedTotalSize = (index * S.length str) `div` processedLen"
                        --    cause overflow, so we use float.
                        -- 2. Slightly enlarge predicted size to compensite copy on vector grow
                        --    if prediction is incorrect
                        k :: Double
k = (Double
1.25 :: Double) Double -> Double -> Double
forall a. Num a => a -> a -> a
* Int -> Double
forall a b. (Integral a, Num b) => a -> b
fromIntegral (ByteString -> Int
S.length ByteString
str) Double -> Double -> Double
forall a. Fractional a => a -> a -> a
/ Int -> Double
forall a b. (Integral a, Num b) => a -> b
fromIntegral Int
processedLen
                        predictedTotalSize :: b
predictedTotalSize = Double -> b
forall b. Integral b => Double -> b
forall a b. (RealFrac a, Integral b) => a -> b
round (Double -> b) -> Double -> b
forall a b. (a -> b) -> a -> b
$ p -> Double
forall a b. (Integral a, Num b) => a -> b
fromIntegral p
index Double -> Double -> Double
forall a. Num a => a -> a -> a
* Double
k
                        growSize :: b
growSize = b
predictedTotalSize b -> b -> b
forall a. Num a => a -> a -> a
- b
vecLen
                    in b
growSize

#if MIN_VERSION_bytestring(0,11,0)
minusForeignPtr :: ForeignPtr a -> ForeignPtr b -> Int
minusForeignPtr :: forall a b. ForeignPtr a -> ForeignPtr b -> Int
minusForeignPtr ForeignPtr a
fpA ForeignPtr b
fpB = IO Int -> Int
forall a. IO a -> a
unsafeDupablePerformIO (IO Int -> Int) -> IO Int -> Int
forall a b. (a -> b) -> a -> b
$
  ForeignPtr a -> (Ptr a -> IO Int) -> IO Int
forall a b. ForeignPtr a -> (Ptr a -> IO b) -> IO b
withForeignPtr ForeignPtr a
fpA ((Ptr a -> IO Int) -> IO Int) -> (Ptr a -> IO Int) -> IO Int
forall a b. (a -> b) -> a -> b
$ \Ptr a
ptrA -> ForeignPtr b -> (Ptr b -> IO Int) -> IO Int
forall a b. ForeignPtr a -> (Ptr a -> IO b) -> IO b
withForeignPtr ForeignPtr b
fpB ((Ptr b -> IO Int) -> IO Int) -> (Ptr b -> IO Int) -> IO Int
forall a b. (a -> b) -> a -> b
$ \Ptr b
ptrB ->
    Int -> IO Int
forall a. a -> IO a
forall (f :: * -> *) a. Applicative f => a -> f a
pure (Ptr a -> Ptr b -> Int
forall a b. Ptr a -> Ptr b -> Int
minusPtr Ptr a
ptrA Ptr b
ptrB)

distance :: ForeignPtr a -> ForeignPtr b -> Int
distance :: forall a b. ForeignPtr a -> ForeignPtr b -> Int
distance = ForeignPtr a -> ForeignPtr b -> Int
forall a b. ForeignPtr a -> ForeignPtr b -> Int
minusForeignPtr
#else
distance :: Int -> Int -> Int
distance a b = a - b
#endif