字符串或对象比较的Java哈希集

Question

I developed a Java module that checks if any component on a JFrame form has changed its value(text) and puts the component it into a HashSet for the JFrame.

Currently, I put the component name(String type) into the HashSet when its value/text changed.

I have a second thought that if I put Object(the component with changed value) into the HashSet thinking it might save time compared to using component name(String).

Here, my question is which is faster/efficent between

• using String(component's name) to HashSet, or
• using Object(component itself) to HashSet?

In short, maybe, which is cheaper? String comparision or Object comparision?

Thank you,

JB

Answer 1

Martin's answer is correct, stop worrying about performance unless you know for sure that you have a significant performance problem which has a practical negative effect in your product.

But to answer the question:

If you use a HashSet , then performance will depend on whether the components implement hashCode() , and if so, how expensive their hashCode() computation is. Most chances are that they do not even implement hashCode() , so their hash codes will essentially be their Identity Hash Codes , (look it up,) so putting objects in the map instead of strings will be infinitesimally faster. We are talking about clock cycles here.

The same applies to comparisons: the components probably do not implement equals() either, so they will be compared by reference, which is going to be slightly faster than comparing strings.

If "probably" is not good enough for you, and if you cannot look at the source code of these components, then you can use an IdentityHashSet instead of a HashSet , to guarantee that only identity hash codes and reference comparisons will be used.

But who cares. The best approach by far is to do whatever is simpler, more understandable, and more maintainable.

Answer 2

Cheapness of the insertion should not be an overriding concern at the development stage. equals methods are generally not expensive to execute for objects. You should start with whatever gives you the least code complexity and only start targeting efficiency only if performance becomes a concern. Also, the comparison won't even take place unless there is a collision on the hashcode for items inserted into the set.

Answer 3

I just tested my own two choices for the element type of HashSet in Java. The result says everything I think.

HashSet element type: Object, experiment count: 1000, Elapsed time: 16ms
HashSet element type: Object, experiment count: 1000, Elapsed time: 8ms
HashSet element type: Object, experiment count: 1000, Elapsed time: 8ms
HashSet element type: String, experiment count: 1000, Elapsed time: 124ms
HashSet element type: String, experiment count: 1000, Elapsed time: 110ms
HashSet element type: String, experiment count: 1000, Elapsed time: 130ms

You know, Java still has some bad reputation on its execution speed. What kind of HashSet element type would you use?

Here is my code that showed above result.

boolean StringUsed = true;

changedComponents = new HashSet<Component>() {};
changedComponentNames = new HashSet<String>() {};

private void changeValueManyTimes() {
    long start = System.currentTimeMillis();

    for (int i = 0; i<count; i++) {
        recordStatisticsCheckBox.setSelected
            (!recordStatisticsCheckBox.isSelected());
    }
    long stop = System.currentTimeMillis();
    System.out.print("HashSet element type: " +
        (StringUsed ? "String" : "Object") + ", experiment count: " + count);
        System.out.println(", Elapsed time: " + (stop - start) + "ms");
        System.out.println("");
}

public static int count = 1000;      
private void jButton1ActionPerformed(java.awt.event.ActionEvent evt) {                                         
    StringUsed = false;
    changeValueManyTimes();
}                                        

private void recordStatisticsCheckBoxStateChanged     
    (javax.swing.event.ChangeEvent evt) {                                                      
    recordStatisticsCheckBoxActionPerformed(null);
}                                                     

private void jButton2ActionPerformed(java.awt.event.ActionEvent evt) {                                         
    StringUsed = true;
    changeValueManyTimes();        
}    

private void recordStatisticsCheckBoxActionPerformed 
    (java.awt.event.ActionEvent evt) {                                                         
    if (StringUsed) {
        if (recordStatisticsCheckBox.isSelected() == origSelection) {
            changedComponents.remove(recordStatisticsCheckBox);
        } else {
            changedComponents.add(recordStatisticsCheckBox);     
        }
        changeSaveEnabledIfNeeded();
    } else {
        if (recordStatisticsCheckBox.isSelected() == origSelection) {
           changedComponentNames.remove(recordStatisticsCheckBox.getName());
        } else {
           changedComponentNames.add(recordStatisticsCheckBox.getName());     
        }
        changeSaveEnabledIfNeeded();
    }
}

private void changeSaveEnabledIfNeeded() {
    if (changedComponents.size() == 0)
        saveButton.setEnabled(false);
    else
        saveButton.setEnabled(true);    
}