 | Haskell Hierarchical Libraries (base package) | Source code | Contents | Index |
| Data.HashTable |
|
| Contents |
| data HashTable key val |
| new :: (key -> key -> Bool) -> (key -> Int32) -> IO (HashTable key val) |
| insert :: HashTable key val -> key -> val -> IO () |
| delete :: HashTable key val -> key -> IO () |
| lookup :: HashTable key val -> key -> IO (Maybe val) |
| update :: HashTable key val -> key -> val -> IO Bool |
| fromList :: Eq key => (key -> Int32) -> [(key, val)] -> IO (HashTable key val) |
| toList :: HashTable key val -> IO [(key, val)] |
| hashInt :: Int -> Int32 |
| hashString :: String -> Int32 |
| prime :: Int32 |
| longestChain :: HashTable key val -> IO [(key, val)] |
| :: (key -> key -> Bool) | eq: An equality comparison on keys |
| -> (key -> Int32) | hash: A hash function on keys |
| -> IO (HashTable key val) | Returns: an empty hash table |
Creates a new hash table. The following property should hold for the eq and hash functions passed to new: eq A B => hash A == hash B | |
Inserts a 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.
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.
This implementation of hash tables uses the low-order n bits of the hash value for a key, where n varies as the hash table grows. A good hash function therefore will give an even distribution regardless of n.
If your keyspace is integrals such that the low-order bits between keys are highly variable, then you could get away with using id as the hash function.
We provide some sample hash functions for Int and String below.
A sample (and useful) hash function for Int and Int32, implemented by extracting the uppermost 32 bits of the 64-bit result of multiplying by a 32-bit constant. The constant is from Knuth, derived from the golden ratio:
golden = round ((sqrt 5 - 1) * 2^31) :: Int
A sample hash function for Strings. We keep multiplying by the golden ratio and adding. The implementation is:
hashString = foldl' f 0 where f m c = fromIntegral (fromEnum c + 1) * golden + mulHi m golden
Note that this has not been extensively tested for reasonability, but Knuth argues that repeated multiplication by the golden ratio will minimize gaps in the hash space.