ConcurrentHashMap JDK 8 when to use computeIfPresent

Question

The new version of Concurrent Hash Map of jdk 8 has two new Methods.

computeIfAbsent

computeIfPresent

putIfAbsent - Old method

I understand the use cases of putIfAbsent and computeIfAbsent. But i am not sure of the scenarios when i will use computeIfPresent. Also why do i need putIfAbsent if i have computeIfPresent now. putIfAbsent do create atleast one extra instance of the value.

Is the reason is only to have backward compatability?

Solution 1

As mentioned in the other answer: Methods will always be kept for backward compatibility, even if there are new, more "powerful" methods introduced.

Concerning the use case for computeIfPresent: It may be hard to find an example that is small enough to not look contrived and still be convincing. In general, the intention of this method is to update an existing value in any form.

One example could be a (constrained) word count: For a given set of words, one stores an initial count of 0 in the map. Then, a sequence of words is processed: Whenever one finds a word from the initial set, its count is increased by 1:

import java.util.LinkedHashMap;
import java.util.Map;

public class ComputeIfPresentExample 
{
    public static void main(String[] args) 
    {
        Map<String, Integer> wordCounts = new LinkedHashMap<String, Integer>();

        String s = 
            "Lorem ipsum dolor sit amet consetetur iam nonumy sadipscing " + 
            "elitr, sed diam nonumy eirmod tempor invidunt ut erat sed " + 
            "labore et dolore magna dolor sit amet aliquyam erat sed diam";

        wordCounts.put("sed", 0);
        wordCounts.put("erat", 0);

        for (String t : s.split(" "))
        {
            wordCounts.computeIfPresent(t, (k,v) -> v+1);
        }
        System.out.println(wordCounts);
    }
}

(Of course, things like this could be solved differently, but this is a rather frequent task in one or the other form, and the new method allows a rather concise and elegant solution)

Solution 2

A common use case are maps with collections, like

Map<String, Collection<String>> strings = new HashMap<>();

computeIfAbsent and computeIfPresent are very handy operations for adding and removing elements to/from the collection. Not least because unlike put(), the compute*() methods return the current value (whether it was just created or not). Here's an example that groups strings by their first char. Note that both the keys and the collections are created when necessary and cleaned it up when the collection becomes empty:

void addString(String a) {
    String index = a.substring(0, 1);
    strings.computeIfAbsent(index, ign -> new HashSet<>()).add(a);
}

void removeString(String a) {
    String index = a.substring(0, 1);
    strings.computeIfPresent(index, (k, c) -> {
        c.remove(a);
        return c.isEmpty() ? null : c;
    });
}

Example:

                         // {}
addString("a1");         // {a=[a1]}      <-- collection dynamically created
addString("a2");         // {a=[a1, a2]}
removeString("a1");      // {a=[a2]}
removeString("a2");      // {}            <-- both key and collection removed

This is extremely powerful in multithreading environments as ConcurrentMaps perform these operations atomically.

The remove operation can be a one-liner:

void removeString(String a) {
    String index = a.substring(0, 1);
    strings.computeIfPresent(index, (i, c) -> c.remove(a) && c.isEmpty() ? null : c);
}

So once again in short:

Map<String, Set<String>> map = new ConcurrentHashMap<>();
map.computeIfAbsent(key, i -> ConcurrentHashMap.newKeySet()).add(value);
map.computeIfPresent(key, (i, s) -> s.remove(value) && s.isEmpty() ? null : s);