本地哈希图的并发问题

Question

I have a method that receives an Object ob as parameter.
This method has a local HashMap<String, String> mp which is filled with the fields of ob , it also calls other methods passing ob and mp as parameters.

The problem is that sometimes mp gets filled with fields of distinct ob objects.

I think mp is experiencing concurrency problems,
What is the best way to synchronize mp knowing that it's being filled locally and being passed as parameter to various methods to get filled as well.

I don't have or control threads, the application is deployed in weblogic server, so it gets called through webservices, so a lot of calls

Answer 1

For such cases, you have

java.util.concurrent.ConcurrentHashMap<K,V>

Details here .

In short, ConcurrentHashMap works by having separated blocks that would be locked independently, so no other threads can access it until work is done (but can access other blocks at that time). Default block size is 16. More about it here .

Now, you must understand map sizes in order to talk about filling it up. Default size is 16. But, there is a load factor of 0.75, that means, when map is 75% full, it will resize to double of previous size, that means, creating new array new Set of keys and then copy everything to it (when I say new Set, I mean same logic for resize in it). Now, if you fill it up fast, that means a lot of copying and for larger sizes, it takes time, so its better to give it a large initial size at the beginning, like:

ConcurrentHashMap<String, Object> mp = new ConcurrentHashMapy<>(someLargeSize);

Let someLargeSize be the order of something you expect or at least near it, so you decrease resizing to minimum. Dont touch load factor, it will just decrease performance, 75% is ok.

And btw., max size of HashMap and ConcurrentHashMap (and HashSet, and ArrayList, or any array based Data Structure) in Java is 1,073,741,824 , since the array[] is in background, DS in Java are mostly just wrappers with specific functions for it, but I doubt you will ever reach this limit, since for objects that contains some strings or anything else, you need a lot of GB of RAM for it.