java中LinkedHashMap中的addEntry方法

Question

Recently I read the source code of LinkedHashmap in java，I know that the linked hashmap use doubly link to restore entry. When addEntry ，it will remove the eldest key of the list, but when I read the code, I have some problem

426    addEntry(int hash, K key, V value, int bucketIndex) {
427        super.addEntry(hash, key, value, bucketIndex);
428
429        // Remove eldest entry if instructed
430        Entry<K,V> eldest = header.after;
431        if (removeEldestEntry(eldest)) {
432            removeEntryForKey(eldest.key);
433        }
434    }

in line 430， eldest = header.after ，but I think the end of the list is not header.after , because header.after is the second entry.

Answer 1

The code for LinkedHashMap has changed over time, and it's completely different code for android, so I'm not exactly sure which version that is. However, in my view it makes sense to have "dummy" nodes before the first node and after the last node (so that even an empty LinkedHashMap has 2 nodes). This way you can remove a node X from the map by getting the nodes before and after X :

X.before -> X -> X.after

and setting the new node following X.before to be X.after and the new node coming before X.after to be X.before .

X.before ------> X.after     (X has gone.)

The code for this can simply be:

X.before.after = X.after;
X.after.before = X.before;

If you don't have dummy nodes, either or both of X.before or X.after could be null , so it makes this process slightly more awkward to code.

I have seen versions of LinkedHashMap that use these 2 extra nodes, and other versions that don't. In the version you're looking at , the first and last "dummy" nodes (called header ) are actually the same object reused.