How to avoid ConcurrentModificationException when

问题:

I've got a map containing some keys (Strings) and values (POJOs)

I want to iterate through this map and alter some of the data in the POJO.

The current code I've inherited removes the given entry, and adds it back in after making some changes to the POJO.

This doesn't work well, since you shouldn't be modifying a map whilst your iterating through it (method is synchronised, but ConcurrentModificationException still appears)

My question is, if I need to iterate over a map and change values, what are the best practices/methods I can use for doing so? To create a separate map and build that up as I go, then return the copy?

回答1:

Two options:

Option 1

The current code I've inherited removes the given entry, and adds it back in after making some changes to the POJO.

Are you changing the reference to the POJO? E.g., so the entry points to something else entirely? Because if not, there's no need to remove it from the map at all, you can just change it.

Option 2

If you do need to actually change the reference to the POJO (e.g., the value of the entry), you can still do that in place by iterating over the Map.Entry instances from entrySet(). You can use setValue on the entry, which doesn't modify what you're iterating over.

Example:

Map<String,String>                  map;
Map.Entry<String,String>            entry;
Iterator<Map.Entry<String,String>>  it;

// Create the map
map = new HashMap<String,String>();
map.put("one", "uno");
map.put("two", "due");
map.put("three", "tre");

// Iterate through the entries, changing one of them
it = map.entrySet().iterator();
while (it.hasNext())
{
    entry = it.next();
    System.out.println("Visiting " + entry.getKey());
    if (entry.getKey().equals("two"))
    {
        System.out.println("Modifying it");
        entry.setValue("DUE");
    }
}

// Show the result
it = map.entrySet().iterator();
while (it.hasNext())
{
    entry = it.next();
    System.out.println(entry.getKey() + "=" + entry.getValue());
}

The output (in no particular order) is:

Visiting two
Modifying it
Visiting one
Visiting three
two=DUE
one=uno
three=tre

...without any modification exception. You will probably want to synchronize this in case something else is also looking at / mucking with that entry.

回答2:

Iterating over a Map and adding entries at the same time will result in a ConcurrentModificationException for most Map classes. And for the Map classes that don't (e.g. ConcurrentHashMap) there is no guarantee that an iteration will visit all entries.

Depending on exactly what it is you are doing, you may be able to do the following while iterating:

• use the Iterator.remove() method to remove the current entry, or
• use the Map.Entry.setValue() method to modify the current entry's value.

For other types of change, you may need to:

• create a new Map from the entries in the current Map, or
• build a separate data structure containing changes to be made, then applied to the Map.

And finally, the Google Collections and Apache Commons Collections libraries have utility classes for "transforming" maps.

回答3:

For such purposes you should use the collection views a map exposes:

• keySet() lets you iterate over keys. That won't help you, as keys are usually immutable.
• values() is what you need if you just want to access the map values. If they are mutable objects, you can change directly, no need to put them back into the map.
• entrySet() the most powerful version, lets you change an entry's key and / or value.

Example: convert the values of all keys that contain an upperscore to uppercase

for(Map.Entry<String, String> entry:map.entrySet()){
    if(entry.getKey().contains("_"))
        entry.setValue(entry.getValue().toUpperCase());
}

Actually, if you just want to edit the value objects, do it using the values collection. I assume your map is of type <String, Object>:

for(Object o: map.values()){
    if(o instanceof MyBean){
        ((Mybean)o).doStuff();
    }
}

回答4:

Create a new map (mapNew). Then iterate over the existing map (mapOld), and add all changed and transformed entries into mapNew. After the iteration is complete, put all values from mapNew to mapOld. This might not be good enough if the amount of data is large though.

Or just use Google collections - they have Maps.transformValues() and Maps.transformEntries().

回答5:

In order to provide a proper answer, you should explain a bit more, what you are trying to achieve.

Still, some (possibly useful) advice:

• make your POJOs thread-safe and do data updates on POJOs directly. Then you do not need to manipulate the map.
• use ConcurrentHashMap
• keep on using simple HashMap, but build a new map on each modification and switch maps behind the scenes (synchronizing the switch operation or using AtomicReference)

Which approach is best depends heavily on your application, it is difficult to give you any "best practice". As always, make your own benchmark with realistic data.

回答6:

Try using ConcurrentHashMap.

From JavaDoc,

A hash table supporting full concurrency of retrievals and adjustable expected concurrency for updates.

For ConcurrentModificationException to occur, generally:

it is not generally permissible for one thread to modify a Collection while another thread is iterating over it.

回答7:

Another approach, somewhat tortured, is to use java.util.concurrent.atomic.AtomicReference as your map's value type. In your case, that would mean declaring your map of type

Map<String, AtomicReference<POJO>>

You certainly don't need the atomic nature of the reference, but it's a cheap way to make the value slots rebindable without having to replace the entire Map.Entry via Map#put().

Still, having read some of the other responses here, I too recommend use of Map.Entry#setValue(), which I had never needed nor noticed until today.