Chapter11onHashTables-Part2.pptVIP

* * * * * * * Comparing Open Addressing Schemes Linear Probe: Research: At load factor of .5, successful searches take 1.5 comparison and unsuccessful searches take 2.5 probes. At load factor of .67, numbers are 2.0 and 5.0!! Clearly, there are tradeoffs between memory efficiency and speed of access. Quadratic Probe and Double Hashing: Load factor of .5: both successful and unsuccessful searches require an average of 2 probes. (better than linear) Load factor of .67, quadratic probe is 2.37 and double hashing is at 3.0 (better than linear probe for double) Load factor of .80: quadratic probe: 2.90 double hashing: 5.0 ? Somewhat higher load factors are possible for quadratic probing and double hashing as compared to linear probes. * Consider Separate Chaining Easier than for open addressing but… Here, each element of the hash table is the potential start of a linked list (size of hash table = arraySize) Searching: We must hash to the appropriate list and then search the linked list for the desired item. On average, half items must be searched whether items in linked list are ordered or not. In unsuccessful search, all items are searched assuming linked list is unordered. In unsuccessful search, half items (average) are search if linked list is ordered. Would you recommend sorting the list? (based on a key?) * Consider Separate Chaining Insertion If linked lists are unordered, insertion is immediate. Still need to run the hash function. Still will need to allocate node for data if home address is full. If linked lists are ordered Still must search half items in linked list Still need to allocate a node (unless we have dupe processing) In separate chaining, typically we may use a load factor of 1 (number of items equals the arraySize. Note: additional operations (searching, inserting, etc.) all take time. Thus increasing the load factor beyond 2 is not a good idea. * Compare Open Addressing : Separate Chaining If using Open Addre