[project @ 2005-02-18 15:06:45 by simonmar]

[ghc-base.git] / Data / HashTable.hs

diff --git a/Data/HashTable.hs b/Data/HashTable.hs
index 37ddb5e..0f31af5 100644 (file)
--- a/Data/HashTable.hs
+++ b/Data/HashTable.hs
@@ -1,4 +1,4 @@
-{-# OPTIONS -fno-implicit-prelude #-}
+{-# OPTIONS_GHC -fno-implicit-prelude #-}
 
 -----------------------------------------------------------------------------
 -- |
@@ -19,7 +19,7 @@
 
 module Data.HashTable (
        -- * Basic hash table operations
-       HashTable, new, insert, delete, lookup,
+       HashTable, new, insert, delete, lookup, update,
        -- * Converting to and from lists
        fromList, toList,
        -- * Hash functions
@@ -187,7 +187,7 @@ new
 
 new cmp hash_fn = do
   -- make a new hash table with a single, empty, segment
-  dir     <- newIOArray (0,dIR_SIZE) undefined
+  dir     <- newIOArray (0,dIR_SIZE-1) undefined
   segment <- newIOArray (0,sEGMENT_SIZE-1) []
   myWriteArray dir 0 segment
 
@@ -209,7 +209,14 @@ new cmp hash_fn = do
 -- -----------------------------------------------------------------------------
 -- Inserting a key\/value pair into the hash table
 
--- | Inserts an key\/value mapping into the hash table.
+-- | Inserts an key\/value mapping into the hash table.  
+--
+-- Note that 'insert' doesn't remove the old entry from the table -
+-- the behaviour is like an association list, where 'lookup' returns
+-- the most-recently-inserted mapping for a key in the table.  The
+-- reason for this is to keep 'insert' as efficient as possible.  If
+-- you need to update a mapping, then we provide 'update'.
+--
 insert :: HashTable key val -> key -> val -> IO ()
 
 insert (HashTable ref) key val = do
@@ -264,11 +271,17 @@ expandHashTable
       newsegment = newbucket `shiftR` sEGMENT_SHIFT
       newindex   = newbucket .&. sEGMENT_MASK
   --
-  when (newindex == 0) $
+  if newsegment >= dIR_SIZE    -- make sure we don't overflow the table.
+       then return table
+       else do
+  --
+   when (newindex == 0) $
        do segment <- newIOArray (0,sEGMENT_SIZE-1) []
-          myWriteArray dir newsegment segment
+          writeIOArray dir newsegment segment
+          -- doesn't happen very often, so we might as well use a safe
+          -- array index here.
   --
-  let table' =
+   let table' =
        if (split+1) < max
            then table{ split = split+1,
                        bcount = bcount+1 }
@@ -279,7 +292,7 @@ expandHashTable
                        max_bucket = max * 2,
                        mask1 = mask2,
                        mask2 = mask2 `shiftL` 1 .|. 1 }
-  let
+   let
     split_bucket old new [] = do
        segment <- myReadArray dir oldsegment
        myWriteArray segment oldindex old
@@ -291,10 +304,10 @@ expandHashTable
                then split_bucket old ((k,v):new) xs
                else split_bucket ((k,v):old) new xs
   --
-  segment <- myReadArray dir oldsegment
-  bucket <- myReadArray segment oldindex
-  split_bucket [] [] bucket
-  return table'
+   segment <- myReadArray dir oldsegment
+   bucket <- myReadArray segment oldindex
+   split_bucket [] [] bucket
+   return table'
 
 -- -----------------------------------------------------------------------------
 -- Deleting a mapping from the hash table
@@ -311,6 +324,43 @@ delete (HashTable ref) key = do
   return ()
 
 -- -----------------------------------------------------------------------------
+-- Deleting a mapping from the hash table
+
+-- | Updates an entry in the hash table, returning 'True' if there was
+-- already an entry for this key, or 'False' otherwise.  After 'update'
+-- there will always be exactly one entry for the given key in the table.
+--
+-- 'insert' is more efficient than 'update' if you don't care about
+-- multiple entries, or you know for sure that multiple entries can't
+-- occur.  However, 'update' is more efficient than 'delete' followed
+-- by 'insert'.
+update :: HashTable key val -> key -> val -> IO Bool
+
+update (HashTable ref) key val = do
+  table@HT{ kcount=k, bcount=b, dir=dir, cmp=cmp } <- readIORef ref
+  let table1 = table{ kcount = k+1 }
+  -- optimistically expand the table
+  table2 <-
+       if (k > hLOAD * b)
+          then expandHashTable table1
+          else return table1
+  writeIORef ref table2
+  (segment_index,segment_offset) <- tableLocation table2 key
+  segment <- myReadArray dir segment_index
+  bucket <- myReadArray segment segment_offset
+  let 
+    (deleted,bucket') = foldr filt (0::Int32,[]) bucket
+    filt pair@(k,v) (deleted,bucket)
+       | key `cmp` k = (deleted+1, bucket)
+       | otherwise   = (deleted,   pair:bucket)
+  -- in  
+  myWriteArray segment segment_offset ((key,val):bucket')
+  -- update the table load, taking into account the number of
+  -- items we just deleted.
+  writeIORef ref table2{ kcount = kcount table2 - deleted }
+  return (deleted /= 0)
+
+-- -----------------------------------------------------------------------------
 -- Looking up an entry in the hash table
 
 -- | Looks up the value of a key in the hash table.