java-treemap-vs-hashmap
Java TreeMap vs HashMap
1. Introduction
Both implementations form an integral part of the Java Collections Framework and store data as key-value pairs.
2. Differences
We’ll first talk about the HashMap which is a hashtable-based implementation. It extends the AbstractMap class and implements the Map interface. A HashMap works on the principle of hashing.
This Map implementation usually acts as a bucketed hash table, but when buckets get too large, they get transformed into nodes of TreeNodes, each structured similarly to those in java.util.TreeMap.
You can find more on the HashMap’s internals in the article focused on it.
On the other hand, TreeMap extends AbstractMap class and implements NavigableMap interface. A TreeMap stores map elements in a Red-Black tree, which is a Self-Balancing Binary Search Tree.
And, you can also find more on the TreeMap’s internals in the article focused on it here.
2.2. Order
It means, we can’t assume any order while iterating over keys and values of a HashMap:
@Test
public void whenInsertObjectsHashMap_thenRandomOrder() {
Map<Integer, String> hashmap = new HashMap<>();
hashmap.put(3, "TreeMap");
hashmap.put(2, "vs");
hashmap.put(1, "HashMap");
assertThat(hashmap.keySet(), containsInAnyOrder(1, 2, 3));
}However, items in a TreeMap are sorted according to their natural order.
If TreeMap objects cannot be sorted according to natural order then we may make use of a Comparator or Comparable to define the order in which the elements are arranged within the Map:
@Test
public void whenInsertObjectsTreeMap_thenNaturalOrder() {
Map<Integer, String> treemap = new TreeMap<>();
treemap.put(3, "TreeMap");
treemap.put(2, "vs");
treemap.put(1, "HashMap");
assertThat(treemap.keySet(), contains(1, 2, 3));
}2.3. Null Values
Let’s see an example:
@Test
public void whenInsertNullInHashMap_thenInsertsNull() {
Map<Integer, String> hashmap = new HashMap<>();
hashmap.put(null, null);
assertNull(hashmap.get(null));
}However, TreeMap doesn’t allow a null key but may contain many null values.
A null key isn’t allowed because the compareTo() or the compare() method throws a NullPointerException:
@Test(expected = NullPointerException.class)
public void whenInsertNullInTreeMap_thenException() {
Map<Integer, String> treemap = new TreeMap<>();
treemap.put(null, "NullPointerException");
}If we’re using a TreeMap with a user-defined Comparator, then it depends on the implementation of the compare() method how null values get handled.
3. Performance Analysis
Performance is the most critical metric that helps us understand the suitability of a data-structure given a use-case.
In this section, we’ll provide a comprehensive analysis of performance for HashMap and TreeMap.
3.1. HashMap
HashMap, being a hashtable-based implementation, internally uses an array-based data structure to organize its elements according to the hash function.
HashMap provides expected constant-time performance O(1) for most operations like add(), remove() and contains(). Therefore, it’s significantly faster than a TreeMap.
The average time to search for an element under the reasonable assumption, in a hash table is O(1). But, an improper implementation of the hash function may lead to a poor distribution of values in buckets which results in:
-
Memory Overhead – many buckets remain unused
-
Performance Degradation – the higher the number of collisions, the lower the performance
Before Java 8, Separate Chaining was the only preferred way to handle collisions. It’s usually implemented using linked lists, i.e., if there is any collision or two different elements have same hash value then store both the items in the same linked list.
Therefore, searching for an element in a HashMap, in the worst case could have taken as long as searching for an element in a linked list i.e. O(n) time.
However, with JEP 180 coming into the picture, there’s been a subtle change in the implementation of the way the elements are arranged in a HashMap.
According to the specification, when buckets get too large and contain enough nodes they get transformed into modes of TreeNodes, each structured similarly to those in TreeMap.
Hence, in the event of high hash collisions, the worst-case performance will improve from O(n) to O(log n).
The code performing this transformation has been illustrated below:
if(binCount >= TREEIFY_THRESHOLD - 1) {
treeifyBin(tab, hash);
}The value for TREEIFY_THRESHOLD is eight which effectively denotes the threshold count for using a tree rather than a linked list for a bucket.
It is evident that:
-
A HashMap requires way more memory than is needed to hold its data
-
A HashMap shouldn’t be more than 70% – 75% full. If it gets close, it gets resized and entries rehashed
-
Rehashing requires n operations which is costly wherein our constant time insert becomes of order O(n)
-
It’s the hashing algorithm which determines the order of inserting the objects in the HashMap
The performance of a HashMap can be tuned by setting the custom initial capacity and the load factor, at the time of HashMap object creation itself.
However, we should choose a HashMap if:
-
we know approximately how many items to maintain in our collection
-
we don’t want to extract items in a natural order
Under the above circumstances, HashMap is our best choice because it offers constant time insertion, search, and deletion.
3.2. TreeMap
A TreeMap stores its data in a hierarchical tree with the ability to sort the elements with the help of a custom Comparator.
A summary of its performance:
-
TreeMap provides a performance of O(log(n)) for most operations like add(), remove() and contains()
-
A Treemap can save memory (in comparison to HashMap) because it only uses the amount of memory needed to hold its items, unlike a HashMap which uses contiguous region of memory
-
A tree should maintain its balance in order to keep its intended performance, this requires a considerable amount of effort, hence complicates the implementation
We should go for a TreeMap whenever:
-
memory limitations have to be taken into consideration
-
we don’t know how many items have to be stored in memory
-
we want to extract objects in a natural order
-
if items will be consistently added and removed
-
we’re willing to accept O(log n) search time
4. Similarities
Both TreeMap and HashMap don’t support duplicate keys. If added, it overrides the previous element (without an error or an exception):
@Test
public void givenHashMapAndTreeMap_whenputDuplicates_thenOnlyUnique() {
Map<Integer, String> treeMap = new HashMap<>();
treeMap.put(1, "Baeldung");
treeMap.put(1, "Baeldung");
assertTrue(treeMap.size() == 1);
Map<Integer, String> treeMap2 = new TreeMap<>();
treeMap2.put(1, "Baeldung");
treeMap2.put(1, "Baeldung");
assertTrue(treeMap2.size() == 1);
}4.2. Concurrent Access
Both must be synchronized externally whenever multiple threads access them concurrently and at least one of the threads modifies them.
We have to explicitly use Collections.synchronizedMap(mapName) to obtain a synchronized view of a provided map.
4.3. Fail-Fast Iterators
The Iterator throws a ConcurrentModificationException if the Map gets modified in any way and at any time once the iterator has been created.
Additionally, we can use the iterator’s remove method to alter the Map during iteration.
Let’s see an example:
@Test
public void whenModifyMapDuringIteration_thenThrowExecption() {
Map<Integer, String> hashmap = new HashMap<>();
hashmap.put(1, "One");
hashmap.put(2, "Two");
Executable executable = () -> hashmap
.forEach((key,value) -> hashmap.remove(1));
assertThrows(ConcurrentModificationException.class, executable);
}5. Which Implementation to Use?
In general, both implementations have their respective pros and cons, however, it’s about understanding the underlying expectation and requirement which must govern our choice regarding the same.
Summarizing:
-
We should use a TreeMap if we want to keep our entries sorted
-
We should use a HashMap if we prioritize performance over memory consumption
-
Since a TreeMap has a more significant locality, we might consider it if we want to access objects that are relatively close to each other according to their natural ordering
-
HashMap can be tuned using the initialCapacity and loadFactor, which isn’t possible for the TreeMap
-
We can use the LinkedHashMap if we want to preserve insertion order while benefiting from constant time access
6. Conclusion
As always, the code examples for this article are available over on GitHub.
