Java Set and Map

Sets

A set is a collection which represents a mathematical set. There are two important properties of a set:

• The same element value may only occur once in a set
• The order in which the elements of a set appear (when iterating through the elements) is typically different than the order in which the elements were added

The Set<E> interface, which is a subclass of Collection<E>, describes the important operations. Here are are the three most important set operations:

boolean add(E elt);         // add an element to the set
boolean contains(Object o); // does the set contain given object?
boolean remove(Object o);   // remove given object from the set

Concrete set implementations typically provide efficient (O(1) or O(log N)) implementations of these methods.

There are two important implementations of the Set<E> interface: HashSet<E> and TreeSet<E>.

HashSet<E> implements the set operations in O(N) worst case, O(1) average case. The element type E must provide working implementations of the boolean equals(Object) and int hashCode() methods inherited from java.lang.Object. When iterating, the order of the elements of a hash set is arbitrary, and seemingly random.

TreeSet<E> implements the set operations in O(log N) worst case and average case. The element type E must either implement the Comparable<E> interface, or an explicit comparator object (implementing Comparator<E>) must be passed to the TreeSet constructor when the tree set object is created. When iterating, the elements of a tree set appear in sorted order: the first element returned is the smallest, etc.

Maps

A map is a dictionary data structure. A map contains a set of keys, each key being a member of the map’s key type. Each key is associated with an arbitrary value which is a member of the map’s value type. Duplicate keys are not allowed.

A good analogy is a phone book, which is a map from names (the key type) to phone numbers (the value type).

The Map<K, V> interface defines the map operations. (K is the type parameter specifying the key type, and V is the type parameter specifying the value type.) The most important map operations are:

public boolean put(K key, V val);       // store an association with given key,value pair
public V get(Object key);               // get the value associated with given key
public boolean remove(Object key);      // remove key,value pair for given key
public boolean containsKey(Object key); // determine whether key exists in Map
public Set<K> keySet();                 // return the set of keys

An implementation of the map interface is expected to define efficient (O(1) or O(log N)) implementations of these methods.

There are two built-in map implementations in Java: HashMap<K, V> and TreeMap<K, V>.

For HashMap<K, V>, the put, get, and removeKey methods all have O(N) worst case, O(1) average case running time. The key type K must correctly implement the public boolean equals(Object) and public int hashCode() methods.

For TreeMap<K, V>, the put, get, and remove methods all have O(log N) worst case and average case running times. The key type K must either implement Comparable<K>, or an explicit comparator object implementing Comparator<K> must be passed to the tree map object’s constructor when the tree map is created.

Data Structures

This section is a high-level summary of the underlying data structures used by the built-in set and map implementations provided in Java.

A full treatment of these data structures will be provided in CS 350, Data Structures.

Hash tables

HashSet<E> and HashMap<K, V> both use a hash table as the underlying data structure. A hash table works by using a hash function to assign hash codes to each key, based on the data contained in the key. The goal of the hash function is to uniformly distribute hash codes to keys, so that two different key values are unlikely to have the same hash code.

The hash table itself is typically an array of linked lists. Each linked list is a hash bucket. Hash buckets have the property that all of the keys contained in the bucket have hash codes such that

HashCode % NumBuckets == BucketIndex

where HashCode is the hash code of a key in the bucket, NumBuckets is the number of buckets in the hash table, and BucketIndex is the index of the array element storing the linked list.

Assuming that the hash function does a good job at distributing keys, any particular key is equally likely to be stored in any bucket. This means that the chain of nodes in each bucket is, on average,

N / P

where N is the total number of keys in the hash table, and P is the number of buckets.

By growing the array of buckets as more keys are added to the hash table, we can ensure that N/P is bounded by a constant. This is why, on average, set and map operations implemented using a hash table are O(1).

It is unlikely, but possible, that all keys in a hash table could be placed in the same bucket. In this case, the hash table devolves into a linked list, and the set and map operations take O(N) time.

When you write a class that will be used as a key in a HashSet or HashMap

There are two things that you must do when you write a class that will either

• be stored in a HashSet
• be used as a key in a HashMap

They are:

1. You must implement the public boolean equals(Object) method
2. You must implement the public int hashCode() method

We have seen how to write an equals method in Lecture 9.

The public int hashCode() method implements a hash function for the class.

To fully discuss how to write a hash function would require several lectures. The idea is that a hash function should compute an integer hash code that is a “summary” of the data contained in the object. There is one simple rule to follow:

Equal objects must have equal hash codes.

Built-in classes (String, Integer, etc.)

More or less all of the “built-in” classes in Java, such as String, Integer, etc., have correct and efficient implementations of both equals and hashCode. So, you can use instances of these classes in instances of HashSet and as keys in instances of HashMap.

Balanced Search Trees

TreeSet and TreeMap use a balanced search tree as the underlying data structure.

A full discussion of search trees would take several lectures. The link above discusses binary search trees, which are the basis for most balanced search trees.