I want to return two objects from a Java method and was wondering what could be a good way of doing so?
The possible ways I can think of are: return a HashMap (since the two Objects are related) or return an ArrayList of Object objects.
To be more precise, the two objects I want to return are (a) List of objects and (b) comma separated names of the same.
I want to return these two Objects from one method because I dont want to iterate through the list of objects to get the comma separated names (which I can do in the same loop in this method).
Somehow, returning a HashMap does not look a very elegant way of doing so.
If you want to return two objects you usually want to return a single object that encapsulates the two objects instead.
You could return a List of NamedObject objects like this:
public class NamedObject<T> {
public final String name;
public final T object;
public NamedObject(String name, T object) {
this.name = name;
this.object = object;
}
}
Then you can easily return a List<NamedObject<WhateverTypeYouWant>>.
Also: Why would you want to return a comma-separated list of names instead of a List<String>? Or better yet, return a Map<String,TheObjectType> with the keys being the names and the values the objects (unless your objects have specified order, in which case a NavigableMap might be what you want.
If you know you are going to return two objects, you can also use a generic pair:
public class Pair<A,B> {
public final A a;
public final B b;
public Pair(A a, B b) {
this.a = a;
this.b = b;
}
};
Edit A more fully formed implementation of the above:
package util;
public class Pair<A,B> {
public static <P, Q> Pair<P, Q> makePair(P p, Q q) {
return new Pair<P, Q>(p, q);
}
public final A a;
public final B b;
public Pair(A a, B b) {
this.a = a;
this.b = b;
}
#Override
public int hashCode() {
final int prime = 31;
int result = 1;
result = prime * result + ((a == null) ? 0 : a.hashCode());
result = prime * result + ((b == null) ? 0 : b.hashCode());
return result;
}
#Override
public boolean equals(Object obj) {
if (this == obj) {
return true;
}
if (obj == null) {
return false;
}
if (getClass() != obj.getClass()) {
return false;
}
#SuppressWarnings("rawtypes")
Pair other = (Pair) obj;
if (a == null) {
if (other.a != null) {
return false;
}
} else if (!a.equals(other.a)) {
return false;
}
if (b == null) {
if (other.b != null) {
return false;
}
} else if (!b.equals(other.b)) {
return false;
}
return true;
}
public boolean isInstance(Class<?> classA, Class<?> classB) {
return classA.isInstance(a) && classB.isInstance(b);
}
#SuppressWarnings("unchecked")
public static <P, Q> Pair<P, Q> cast(Pair<?, ?> pair, Class<P> pClass, Class<Q> qClass) {
if (pair.isInstance(pClass, qClass)) {
return (Pair<P, Q>) pair;
}
throw new ClassCastException();
}
}
Notes, mainly around rustiness with Java & generics:
both a and b are immutable.
makePair static method helps you with boiler plate typing, which the diamond operator in Java 7 will make less annoying. There's some work to make this really nice re: generics, but it should be ok-ish now. (c.f. PECS)
hashcode and equals are generated by eclipse.
the compile time casting in the cast method is ok, but doesn't seem quite right.
I'm not sure if the wildcards in isInstance are necessary.
I've just written this in response to comments, for illustration purposes only.
In the event the method you're calling is private, or called from one location, try
return new Object[]{value1, value2};
The caller looks like:
Object[] temp=myMethod(parameters);
Type1 value1=(Type1)temp[0]; //For code clarity: temp[0] is not descriptive
Type2 value2=(Type2)temp[1];
The Pair example by David Hanak has no syntactic benefit, and is limited to two values.
return new Pair<Type1,Type2>(value1, value2);
And the caller looks like:
Pair<Type1, Type2> temp=myMethod(parameters);
Type1 value1=temp.a; //For code clarity: temp.a is not descriptive
Type2 value2=temp.b;
You may use any of following ways:
private static final int RETURN_COUNT = 2;
private static final int VALUE_A = 0;
private static final int VALUE_B = 1;
private static final String A = "a";
private static final String B = "b";
1) Using Array
private static String[] methodWithArrayResult() {
//...
return new String[]{"valueA", "valueB"};
}
private static void usingArrayResultTest() {
String[] result = methodWithArrayResult();
System.out.println();
System.out.println("A = " + result[VALUE_A]);
System.out.println("B = " + result[VALUE_B]);
}
2) Using ArrayList
private static List<String> methodWithListResult() {
//...
return Arrays.asList("valueA", "valueB");
}
private static void usingListResultTest() {
List<String> result = methodWithListResult();
System.out.println();
System.out.println("A = " + result.get(VALUE_A));
System.out.println("B = " + result.get(VALUE_B));
}
3) Using HashMap
private static Map<String, String> methodWithMapResult() {
Map<String, String> result = new HashMap<>(RETURN_COUNT);
result.put(A, "valueA");
result.put(B, "valueB");
//...
return result;
}
private static void usingMapResultTest() {
Map<String, String> result = methodWithMapResult();
System.out.println();
System.out.println("A = " + result.get(A));
System.out.println("B = " + result.get(B));
}
4) Using your custom container class
private static class MyContainer<M,N> {
private final M first;
private final N second;
public MyContainer(M first, N second) {
this.first = first;
this.second = second;
}
public M getFirst() {
return first;
}
public N getSecond() {
return second;
}
// + hashcode, equals, toString if need
}
private static MyContainer<String, String> methodWithContainerResult() {
//...
return new MyContainer("valueA", "valueB");
}
private static void usingContainerResultTest() {
MyContainer<String, String> result = methodWithContainerResult();
System.out.println();
System.out.println("A = " + result.getFirst());
System.out.println("B = " + result.getSecond());
}
5) Using AbstractMap.simpleEntry
private static AbstractMap.SimpleEntry<String, String> methodWithAbstractMapSimpleEntryResult() {
//...
return new AbstractMap.SimpleEntry<>("valueA", "valueB");
}
private static void usingAbstractMapSimpleResultTest() {
AbstractMap.SimpleEntry<String, String> result = methodWithAbstractMapSimpleEntryResult();
System.out.println();
System.out.println("A = " + result.getKey());
System.out.println("B = " + result.getValue());
}
6) Using Pair of Apache Commons
private static Pair<String, String> methodWithPairResult() {
//...
return new ImmutablePair<>("valueA", "valueB");
}
private static void usingPairResultTest() {
Pair<String, String> result = methodWithPairResult();
System.out.println();
System.out.println("A = " + result.getKey());
System.out.println("B = " + result.getValue());
}
I almost always end up defining n-Tuple classes when I code in Java. For instance:
public class Tuple2<T1,T2> {
private T1 f1;
private T2 f2;
public Tuple2(T1 f1, T2 f2) {
this.f1 = f1; this.f2 = f2;
}
public T1 getF1() {return f1;}
public T2 getF2() {return f2;}
}
I know it's a bit ugly, but it works, and you just have to define your tuple types once. Tuples are something Java really lacks.
EDIT: David Hanak's example is more elegant, as it avoids defining getters and still keeps the object immutable.
Before Java 5, I would kind of agree that the Map solution isn't ideal. It wouldn't give you compile time type checking so can cause issues at runtime. However, with Java 5, we have Generic Types.
So your method could look like this:
public Map<String, MyType> doStuff();
MyType of course being the type of object you are returning.
Basically I think that returning a Map is the right solution in this case because that's exactly what you want to return - a mapping of a string to an object.
Apache Commons has tuple and triple for this:
ImmutablePair<L,R> An immutable pair consisting of two Object
elements.
ImmutableTriple<L,M,R> An immutable triple consisting of
three Object elements.
MutablePair<L,R> A mutable pair consisting of
two Object elements.
MutableTriple<L,M,R> A mutable triple
consisting of three Object elements.
Pair<L,R> A pair consisting of
two elements.
Triple<L,M,R> A triple consisting of three elements.
Source: https://commons.apache.org/proper/commons-lang/apidocs/org/apache/commons/lang3/tuple/package-summary.html
Alternatively, in situations where I want to return a number of things from a method I will sometimes use a callback mechanism instead of a container. This works very well in situations where I cannot specify ahead of time just how many objects will be generated.
With your particular problem, it would look something like this:
public class ResultsConsumer implements ResultsGenerator.ResultsCallback
{
public void handleResult( String name, Object value )
{
...
}
}
public class ResultsGenerator
{
public interface ResultsCallback
{
void handleResult( String aName, Object aValue );
}
public void generateResults( ResultsGenerator.ResultsCallback aCallback )
{
Object value = null;
String name = null;
...
aCallback.handleResult( name, value );
}
}
While in your case, the comment may be a good way to go, in Android, you can use Pair . Simply
return new Pair<>(yourList, yourCommaSeparatedValues);
Use of following Entry object
Example :
public Entry<A,B> methodname(arg)
{
.......
return new AbstractMap.simpleEntry<A,B>(instanceOfA,instanceOfB);
}
Regarding the issue about multiple return values in general I usually use a small helper class that wraps a single return value and is passed as parameter to the method:
public class ReturnParameter<T> {
private T value;
public ReturnParameter() { this.value = null; }
public ReturnParameter(T initialValue) { this.value = initialValue; }
public void set(T value) { this.value = value; }
public T get() { return this.value; }
}
(for primitive datatypes I use minor variations to directly store the value)
A method that wants to return multiple values would then be declared as follows:
public void methodThatReturnsTwoValues(ReturnParameter<ClassA> nameForFirstValueToReturn, ReturnParameter<ClassB> nameForSecondValueToReturn) {
//...
nameForFirstValueToReturn.set("...");
nameForSecondValueToReturn.set("...");
//...
}
Maybe the major drawback is that the caller has to prepare the return objects in advance in case he wants to use them (and the method should check for null pointers)
ReturnParameter<ClassA> nameForFirstValue = new ReturnParameter<ClassA>();
ReturnParameter<ClassB> nameForSecondValue = new ReturnParameter<ClassB>();
methodThatReturnsTwoValues(nameForFirstValue, nameForSecondValue);
Advantages (in comparison to other solutions proposed):
You do not have to create a special class declaration for individual methods and its return types
The parameters get a name and therefore are easier to differentiate when looking at the method signature
Type safety for each parameter
All possible solutions will be a kludge (like container objects, your HashMap idea, “multiple return values” as realized via arrays). I recommend regenerating the comma-separated list from the returned List. The code will end up being a lot cleaner.
Keep it simple and create a class for multiple result situation. This example accepts an ArrayList and a message text from a databasehelper getInfo.
Where you call the routine that returns multiple values you code:
multResult res = mydb.getInfo();
In the routine getInfo you code:
ArrayList<String> list= new ArrayList<String>();
add values to the list...
return new multResult("the message", list);
and define a class multResult with:
public class multResult {
public String message; // or create a getter if you don't like public
public ArrayList<String> list;
multResult(String m, ArrayList<String> l){
message = m;
list= l;
}
}
As I see it there are really three choices here and the solution depends on the context. You can choose to implement the construction of the name in the method that produces the list. This is the choice you've chosen, but I don't think it is the best one. You are creating a coupling in the producer method to the consuming method that doesn't need to exist. Other callers may not need the extra information and you would be calculating extra information for these callers.
Alternatively, you could have the calling method calculate the name. If there is only one caller that needs this information, you can stop there. You have no extra dependencies and while there is a little extra calculation involved, you've avoided making your construction method too specific. This is a good trade-off.
Lastly, you could have the list itself be responsible for creating the name. This is the route I would go if the calculation needs to be done by more than one caller. I think this puts the responsibility for the creation of the names with the class that is most closely related to the objects themselves.
In the latter case, my solution would be to create a specialized List class that returns a comma-separated string of the names of objects that it contains. Make the class smart enough that it constructs the name string on the fly as objects are added and removed from it. Then return an instance of this list and call the name generation method as needed. Although it may be almost as efficient (and simpler) to simply delay calculation of the names until the first time the method is called and store it then (lazy loading). If you add/remove an object, you need only remove the calculated value and have it get recalculated on the next call.
Can do some thing like a tuple in dynamic language (Python)
public class Tuple {
private Object[] multiReturns;
private Tuple(Object... multiReturns) {
this.multiReturns = multiReturns;
}
public static Tuple _t(Object... multiReturns){
return new Tuple(multiReturns);
}
public <T> T at(int index, Class<T> someClass) {
return someClass.cast(multiReturns[index]);
}
}
and use like this
public Tuple returnMultiValues(){
return Tuple._t(new ArrayList(),new HashMap())
}
Tuple t = returnMultiValues();
ArrayList list = t.at(0,ArrayList.class);
I followed a similar approach than the described in the other answers with a few tweaks based on the requirement I had, basically I created the following classes(Just in case, everything is Java):
public class Pair<L, R> {
final L left;
final R right;
public Pair(L left, R right) {
this.left = left;
this.right = right;
}
public <T> T get(Class<T> param) {
return (T) (param == this.left.getClass() ? this.left : this.right);
}
public static <L, R> Pair<L, R> of(L left, R right) {
return new Pair<L, R>(left, right);
}
}
Then, my requirement was simple, in the repository Class that reaches the DB, for the Get Methods than retrieve data from the DB, I need to check if it failed or succeed, then, if succeed, I needed to play with the returning list, if failed, stop the execution and notify the error.
So, for example, my methods are like this:
public Pair<ResultMessage, List<Customer>> getCustomers() {
List<Customer> list = new ArrayList<Customer>();
try {
/*
* Do some work to get the list of Customers from the DB
* */
} catch (SQLException e) {
return Pair.of(
new ResultMessage(e.getErrorCode(), e.getMessage()), // Left
null); // Right
}
return Pair.of(
new ResultMessage(0, "SUCCESS"), // Left
list); // Right
}
Where ResultMessage is just a class with two fields (code/message) and Customer is any class with a bunch of fields that comes from the DB.
Then, to check the result I just do this:
void doSomething(){
Pair<ResultMessage, List<Customer>> customerResult = _repository.getCustomers();
if (customerResult.get(ResultMessage.class).getCode() == 0) {
List<Customer> listOfCustomers = customerResult.get(List.class);
System.out.println("do SOMETHING with the list ;) ");
}else {
System.out.println("Raised Error... do nothing!");
}
}
In C++ (STL) there is a pair class for bundling two objects. In Java Generics a pair class isn't available, although there is some demand for it. You could easily implement it yourself though.
I agree however with some other answers that if you need to return two or more objects from a method, it would be better to encapsulate them in a class.
Why not create a WhateverFunctionResult object that contains your results, and the logic required to parse these results, iterate over then etc. It seems to me that either:
These results objects are intimately tied together/related and belong together, or:
they are unrelated, in which case your function isn't well defined in terms of what it's trying to do (i.e. doing two different things)
I see this sort of issue crop up again and again. Don't be afraid to create your own container/result classes that contain the data and the associated functionality to handle this. If you simply pass the stuff around in a HashMap or similar, then your clients have to pull this map apart and grok the contents each time they want to use the results.
public class MultipleReturnValues {
public MultipleReturnValues() {
}
public static void functionWithSeveralReturnValues(final String[] returnValues) {
returnValues[0] = "return value 1";
returnValues[1] = "return value 2";
}
public static void main(String[] args) {
String[] returnValues = new String[2];
functionWithSeveralReturnValues(returnValues);
System.out.println("returnValues[0] = " + returnValues[0]);
System.out.println("returnValues[1] = " + returnValues[1]);
}
}
This is not exactly answering the question, but since every of the solution given here has some drawbacks, I suggest to try to refactor your code a little bit so you need to return only one value.
Case one.
You need something inside as well as outside of your method. Why not calculate it outside and pass it to the method?
Instead of:
[thingA, thingB] = createThings(...); // just a conceptual syntax of method returning two values, not valid in Java
Try:
thingA = createThingA(...);
thingB = createThingB(thingA, ...);
This should cover most of your needs, since in most situations one value is created before the other and you can split creating them in two methods. The drawback is that method createThingsB has an extra parameter comparing to createThings, and possibly you are passing exactly the same list of parameters twice to different methods.
Case two.
Most obvious solution ever and a simplified version of case one. It's not always possible, but maybe both of the values can be created independently of each other?
Instead of:
[thingA, thingB] = createThings(...); // see above
Try:
thingA = createThingA(...);
thingB = createThingB(...);
To make it more useful, these two methods can share some common logic:
public ThingA createThingA(...) {
doCommonThings(); // common logic
// create thing A
}
public ThingB createThingB(...) {
doCommonThings(); // common logic
// create thing B
}
Pass a list to your method and populate it, then return the String with the names, like this:
public String buildList(List<?> list) {
list.add(1);
list.add(2);
list.add(3);
return "something,something,something,dark side";
}
Then call it like this:
List<?> values = new ArrayList<?>();
String names = buildList(values);
You can utilize a HashMap<String, Object> as follows
public HashMap<String, Object> yourMethod()
{
.... different logic here
HashMap<String, Object> returnHashMap = new HashMap<String, Object>();
returnHashMap.put("objectA", objectAValue);
returnHashMap.put("myString", myStringValue);
returnHashMap.put("myBoolean", myBooleanValue);
return returnHashMap;
}
Then when calling the method in a different scope, you can cast each object back to its initial type:
// call the method
HashMap<String, Object> resultMap = yourMethod();
// fetch the results and cast them
ObjectA objectA = (ObjectA) resultMap.get("objectA");
String myString = (String) resultMap.get("myString");
Boolean myBoolean = (Boolean) resultMap.get("myBoolean");
I noticed there is no no-custom class, n-length, no-cast, type-safe answers yet to returning multiple values.
Here is my go:
import java.util.Objects;
public final class NTuple<V, T extends NTuple<?, ?>> {
private final V value;
private final T next;
private NTuple(V value, T next) {
this.value = value;
this.next = next;
}
public static <V> NTuple<V, ?> of(V value) {
return new NTuple<>(value, null);
}
public static <V, T extends NTuple<?, ?>> NTuple<V, T> of(V value, T next) {
return new NTuple<>(value, next);
}
public V value() {
return value;
}
public T next() {
return next;
}
public static <V> V unpack0(NTuple<V, ?> tuple) {
return Objects.requireNonNull(tuple, "0").value();
}
public static <V, T extends NTuple<V, ?>> V unpack1(NTuple<?, T> tuple) {
NTuple<?, T> tuple0 = Objects.requireNonNull(tuple, "0");
NTuple<V, ?> tuple1 = Objects.requireNonNull(tuple0.next(), "1");
return tuple1.value();
}
public static <V, T extends NTuple<?, NTuple<V, ?>>> V unpack2(NTuple<?, T> tuple) {
NTuple<?, T> tuple0 = Objects.requireNonNull(tuple, "0");
NTuple<?, NTuple<V, ?>> tuple1 = Objects.requireNonNull(tuple0.next(), "1");
NTuple<V, ?> tuple2 = Objects.requireNonNull(tuple1.next(), "2");
return tuple2.value();
}
}
Sample use:
public static void main(String[] args) {
// pre-java 10 without lombok - use lombok's var or java 10's var if you can
NTuple<String, NTuple<Integer, NTuple<Integer, ?>>> multiple = wordCount("hello world");
String original = NTuple.unpack0(multiple);
Integer wordCount = NTuple.unpack1(multiple);
Integer characterCount = NTuple.unpack2(multiple);
System.out.println(original + ": " + wordCount + " words " + characterCount + " chars");
}
private static NTuple<String, NTuple<Integer, NTuple<Integer, ?>>> wordCount(String s) {
int nWords = s.split(" ").length;
int nChars = s.length();
return NTuple.of(s, NTuple.of(nWords, NTuple.of(nChars)));
}
Pros:
no-custom container class - no need to write a class just for a return type
n-length - can handle any number of return values
no-cast - no need to cast from Object
type-safe - the types are checked via Java's generics
Cons:
inefficient for large numbers of return values
according to my experience with python's multiple return values, this should not happen in practice
heavy type declarations
can be alleviated by lombok/Java 10 var
In C, you would do it by passing pointers to placeholders for the results as arguments:
void getShoeAndWaistSizes(int *shoeSize, int *waistSize) {
*shoeSize = 36;
*waistSize = 45;
}
...
int shoeSize, waistSize;
getShoeAndWaistSize(&shoeSize, &waistSize);
int i = shoeSize + waistSize;
Let's try something similar, in Java.
void getShoeAndWaistSizes(List<Integer> shoeSize, List<Integer> waistSize) {
shoeSize.add(36);
waistSize.add(45);
}
...
List<Integer> shoeSize = new List<>();
List<Integer> waistSize = new List<>();
getShoeAndWaistSizes(shoeSize, waistSize);
int i = shoeSize.get(0) + waistSize.get(0);
PASS A HASH INTO THE METHOD AND POPULATE IT......
public void buildResponse(String data, Map response);
I have to use a map which stores keys of type Integer, String and Long only.
One solution: To store type Object and in put method check with instanceof operator. Is there any better solution, maybe with enum
You can use a map and storing Long as String into it
or you can use two different hashmap and duplicate put/get methods. If you have two types, it is probably for two different things, and having two different map should probably be the correct answer
Create a class that has a map as a member and add methods that will store and retrieve int and long as Strings.
class MyMap {
private Map mabObject = Map<String, Object>;
public void add(long key, Object value) {
mapObject.put(Long.toString(key),value);
}
public void add(String key, Object value) {
mapObject.put(key, value);
}
public Object get(long key) {
return mapObject.get(Long.toString(key));
}
public Object get(String key) {
return mapObject.get(key);
}
}
I agree with Paul Boddington's comment, and the need of such trick shows that code smells.
Just for a funny excercise (not for production code) I've made an example that shows what we can do in compile time for limiting types of keys in a map.
For example we can create a wrapper allowing only values of specific classes.
common/map/Wrap.java
package common.map;
import java.util.Arrays;
import java.util.List;
public class Wrap<T> {
private T value;
private Wrap(T value){
this.value = value;
}
public T get() {
return this.value;
}
/*
* it's important to implement this method
* if we intend to use Wrap instances as map's key
*
* and it's needed to see that hash codes are computing differently in different classes,
* and depending on `allowedClasses` contents we can face some unexpected collisions
* so if you care of performance - test your maps usage accurately
*/
public int hashCode() {
return this.value.hashCode();
}
/*
* static
*/
private static List<Class> allowedClasses = Arrays.asList(Long.class, String.class);
public static <T> Wrap<T> create(Class<? extends T> clazz, T value) {
if (!allowedClasses.contains(clazz)) {
throw new IllegalArgumentException("Unexpected class " + clazz);
}
return new Wrap<>(value);
}
public static <T> Wrap<T> create(AllowedClasses allowedClass, T value) {
return create(allowedClass.clazz, value);
}
public enum AllowedClasses {
LONG(Long.class),
STRING(String.class);
private Class clazz;
AllowedClasses(Class clazz) {
this.clazz = clazz;
}
}
}
And let's run it
common/map/Example.java
package common.map;
import common.map.Wrap.AllowedClasses;
import java.util.HashMap;
import java.util.Map;
public class Example {
public static void main(String... args) {
Map<Wrap, Object> map = new HashMap<>();
// next two lines create wrappers for values of types we added to enum AllowedClasses
// but since enums cannot have type parameters, we are not able to check
// if the second parameter type is compatible with a type associated with given enum value
// so I think usage of enum is useless for your purpose
Wrap<?> valLong0 = Wrap.create(AllowedClasses.LONG, "the string in place of Long is OK");
Wrap<?> valString0 = Wrap.create(AllowedClasses.STRING, 12345);
// from the next lines you can see how we can use the Wrap class to keep
// only allowed types to be associated with the map keys
Wrap<Long> valLong = Wrap.create(Long.class, 1L); // legal
Wrap<String> valString = Wrap.create(String.class, "abc"); // legal
Wrap<String> valWrong = Wrap.create(String.class, 123); // doesn't compile
Wrap<Object> valWrong2 = Wrap.create(Object.class, 123); // compiles but throws exception in runtime
Object obj = ThirdParty.getObjectOfUnknownClass();
Wrap<?> valDynamic = Wrap.create(obj.getClass(), obj); // compiles but MAYBE throws exception in runtime
// so we get to this point only if all the wrappers are legal,
// and we can add them as keys to the map
map.put(valLong, new Object());
map.put(valString, new Object());
map.put(valDynamic, new Object());
}
}
HashMap<DataType1,DataType2>hm = new HashMap<DataType1,DataType2>();
or
Map<DataType1,DataType2> m = new HashMap<DataType1,DataType2>();
m.put(key, value);
Instead of DataType1 & DataType2 you can add Integer,String,Long ,etc. and use the put(key,value) method to enter key and values into the HashMap.
I'm trying to implement some kind of accumulation logic in a multi threaded environment; I’m wondering is there any better/faster way to do it without the lock and synchronized keyword? The following is my current code:
public class ConcurrentHashMapTest {
private static final int NB_THREADS = 1_000;
private final Map<String, Integer> cnts = new HashMap<>();
private static final Lock RWLOCK = new ReentrantLock(true);
private static final String[] NAMES = {
"A", "B"
};
public void testIt() {
ExecutorService executor =
Executors.newFixedThreadPool(NB_THREADS);
for (int i = 0; i < NB_THREADS; i++) {
Runnable task = new WorkerThread();
executor.submit(task);
}
executor.shutdown();
try {
executor.awaitTermination(Long.MAX_VALUE, TimeUnit.SECONDS);
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
}
System.out.println(cnts);
}
private void accumulate(String name) {
RWLOCK.lock();
try {
Integer cnt = cnts.get(name);
if (cnt == null) {
cnts.put(name, 1);
} else {
cnts.put(name, cnt + 1);
}
} finally {
RWLOCK.unlock();
}
}
private class WorkerThread implements Runnable {
#Override
public void run() {
accumulate(NAMES[ThreadLocalRandom.current().nextInt(0, NAMES.length)]);
}
}
}
Java 8:
private final Map<String, AtomicInteger> cnts =
new ConcurrentHashMap<>();
private void accumulate(String name) {
cnts.computeIfAbsent(name, k -> new AtomicInteger()).incrementAndGet();
}
The ConcurrentHashMap can be freely accessed from multiple threads. The computeIfAbsent method takes a lambda to evaluate to get a value for the key if the key is not present in the map, and adds it if and only if there is no such mapping, and then returns that value. It's effectively putIfAbsent followed by get. The value is a new AtomicInteger with the value 0. Whether there was an existing value, or whether a new one with value 0 was just added, in either case increment it.
Java 7:
private final ConcurrentMap<String, AtomicInteger> cnts =
new ConcurrentHashMap<>();
private void accumulate(String name) {
cnts.putIfAbsent(name, new AtomicInteger());
cnts.get(name).incrementAndGet();
}
For Java 7, there is no computeIfAbsent method, but that effectively just does a putIfAbsent followed by a get, so the same effect is achieved by calling those methods. There is no concern that the value already existed in the map; a new, zero AtomicInteger is added if and only if the map had no value for that key. Even if another thread got in there before us and added a zero, both threads would then see and increment that same AtomicInteger instance.
use a concurrent hash map with String and AtomicInteger. Both are thread safe and thus can be used freely.
I'd be wary of using fairness on your ReentrantLock in this case, as there's no benefit to your accumulator if longer waiting threads get access first. Take a look at Brian Goetz's 'Java Concurrency in Practice'
Why wouldn't we want to make all locks fair? After all, fairness is good and unfairness is bad, right? (It's not accidental that whenever kids want to appeal a decision, "that's not fair" almost certainly comes up. We think fairness is pretty important, and they know it.) In reality, the fairness guarantee for locks is a very strong one, and comes at a significant performance cost. The bookkeeping and synchronization required to ensure fairness mean that contended fair locks will have much lower throughput than unfair locks. As a default, you should set fair to false unless it is critical to the correctness of your algorithm that threads be serviced in exactly the order they queued up.
You could use a Map of name to AtomicInteger and use double-check locking when there is no counter in the map at all. Be aware that you need to use the volatile keyword for effective double-check locking.
This way you will only lock the whole map for actually adding brand new entries, the rest of the processing can happen in parallel.
You risk massively over-complicating your program here though and possibly even reducing performance in real-world cases. Is contention on this map really a performance bottle-neck?
According to Oracle Java 7 API : implementation of HashMap is not synchronized.
You can use Hashtable implementation or declare : private final Map<String, Integer> cnts = Collections.synchronizedMap(new HashMap<String, Integer>());
I think what you are looking for is a Multiton:
/**
* Holds a thread-safe map of unique create-once items.
*
* Contract:
*
* Only one object will be made for each key presented.
*
* Thread safe.
*
* #author OldCurmudgeon
* #param <K>
* #param <V>
*/
public class Multiton<K, V> {
// Map from the key to the futures of the items.
private final ConcurrentMap<K, Future<V>> multitons = new ConcurrentHashMap<>();
// The creator can create an item of type V.
private final Creator<K, V> creator;
public Multiton(Creator<K, V> creator) {
this.creator = creator;
}
/**
* There can be only one.
*
* Use a FutureTask to do the creation to ensure only one construction.
*
* #param key
* #return
* #throws InterruptedException
* #throws ExecutionException
*/
public V get(final K key) throws InterruptedException, ExecutionException {
// Already made?
Future<V> f = multitons.get(key);
if (f == null) {
// Plan the future but do not create as yet.
FutureTask<V> ft = new FutureTask<>(() -> creator.create(key));
// Store it.
f = multitons.putIfAbsent(key, ft);
if (f == null) {
// It was successfully stored - it is the first (and only)
f = ft;
// Make it happen.
ft.run();
}
}
// Wait for it to finish construction and return the constructed.
return f.get();
}
/**
* Returns a Map indicating the current state.
*
* #return a Map which should reflect the current state.
*
* #throws java.lang.InterruptedException
* #throws java.util.concurrent.ExecutionException
*/
public Map<K, V> getMap() throws InterruptedException, ExecutionException {
Map<K, V> map = new HashMap<>();
for (Map.Entry<K, Future<V>> e : multitons.entrySet()) {
map.put(e.getKey(), e.getValue().get());
}
return map;
}
/**
* User provides one of these to do the construction.
*
* #param <K>
* #param <V>
*/
public abstract static class Creator<K, V> {
// Return a new item under the key.
abstract V create(K key) throws ExecutionException;
}
}
Usage - for demonstration - adds up all integers up to 999, keying on their first digit:
Multiton<String, AtomicInteger> counts = new Multiton<>(
new Creator<String, AtomicInteger>() {
#Override
AtomicInteger create(String key) throws ExecutionException {
return new AtomicInteger();
}
}
);
public void test() throws InterruptedException, ExecutionException {
for (int i = 0; i < 1000; i++) {
counts.get(Integer.toString(i).substring(0, 1)).addAndGet(i);
}
System.out.println(counts.getMap());
}
Prints:
{0=0, 1=15096, 2=25197, 3=35298, 4=45399, 5=55500, 6=65601, 7=75702, 8=85803, 9=95904}
Java < 8 version:
/**
* Holds a thread-safe map of unique create-once items.
*
* Contract:
*
* Only one object will be made for each key presented.
*
* Thread safe.
*
* #author OldCurmudgeon
* #param <K>
* #param <V>
*/
public class Multiton<K, V> {
// Map from the key to the futures of the items.
private final ConcurrentMap<K, Future<V>> multitons = new ConcurrentHashMap<>();
// The creator can create an item of type V.
private final Creator<K, V> creator;
public Multiton(Creator<K, V> creator) {
this.creator = creator;
}
/**
* There can be only one.
*
* Use a FutureTask to do the creation to ensure only one construction.
*
* #param key
* #return
* #throws InterruptedException
* #throws ExecutionException
*/
public V get(final K key) throws InterruptedException, ExecutionException {
// Already made?
Future<V> f = multitons.get(key);
if (f == null) {
// Plan the future but do not create as yet.
FutureTask<V> ft = new FutureTask<>(new Callable<V>() {
#Override
public V call() throws Exception {
// Doing this inline may be a little contrived but it maintains the linkage with the Java-8 version.
return creator.create(key);
}
}
);
// Store it.
f = multitons.putIfAbsent(key, ft);
if (f == null) {
// It was successfully stored - it is the first (and only)
f = ft;
// Make it happen.
ft.run();
}
}
// Wait for it to finish construction and return the constructed.
return f.get();
}
/**
* Returns a Map indicating the current state.
*
* #return a Map which should reflect the current state.
*
* #throws java.lang.InterruptedException
* #throws java.util.concurrent.ExecutionException
*/
public Map<K, V> getMap() throws InterruptedException, ExecutionException {
Map<K, V> map = new HashMap<>();
for (Map.Entry<K, Future<V>> e : multitons.entrySet()) {
map.put(e.getKey(), e.getValue().get());
}
return map;
}
/**
* User provides one of these to do the construction.
*
* #param <K>
* #param <V>
*/
public abstract static class Creator<K, V> {
// Return a new item under the key.
abstract V create(K key) throws ExecutionException;
}
}
Is there such a collection implemention can let us know what objects are newly added, modified or deleted comparing to a specific point?
I want to use such a collection to hold objects loaded from database, and bind it to user interface ,so user can add new object to it, delete items in it or modify some ones. when user click a save button , I need to persist changes to database, so I need to know the changed objects.
Here is my own solution, but I am not sure about whether it is very bad, please give me some advice.
interface :
import java.util.Collection;
/**
* #author ggfan#amarsoft
*
*/
public interface DataObjectsMonitor {
/**
* take a snapshot for comparing
*/
public void snapshot();
/**
*
* #return Objects that are modified comparing to those ones before {#link #snapshot()} last called
*/
public Collection<?> getmodifiedObjects();
/**
*
* #return Objects that are deleted comparing to those ones before {#link #snapshot()} last called
*/
public Collection<?> getDeletedObjects();
/**
*
* #return Objects that are added comparing to those ones before {#link #snapshot()} last called
*/
public Collection<?> getAddedObjects();
}
Model Class must be extended from such a abstract class :
public abstract class DataObject {
public abstract int dataHashCode();
}
implemention class :
public class DataObjectListMonitor<T extends DataObject> extends ArrayList<T> implements DataObjectsMonitor {
private static final long serialVersionUID = 1L;
private Map<T, Integer> oldVersion = new HashMap<T, Integer>();
public void snapshot() {
oldVersion.clear();
for(T t : this){
oldVersion.put(t, new Integer(t.dataHashCode()));
}
}
public Collection<T> getmodifiedObjects() {
ArrayList<T> modified = new ArrayList<T>();
for(T t : oldVersion.keySet()){
if(this.contains(t) && t.dataHashCode() != oldVersion.get(t)){
modified.add(t);
}
}
return modified;
}
public Collection<T> getDeletedObjects() {
ArrayList<T> deleted = new ArrayList<T>();
for(T t : oldVersion.keySet()){
if(!this.contains(t)){
deleted.add(t);
}
}
return deleted;
}
public Collection<T> getAddedObjects() {
ArrayList<T> added = new ArrayList<T>();
for(T t : this){
if(!oldVersion.keySet().contains(t)){
added.add(t);
}
}
return added;
}
}
test :
public class Model extends DataObject {
private String id;
private String name;
public String toString() {
return "Model [id=" + id + ", name=" + name + "]";
}
public Model(String id, String name) {
super();
this.id = id;
this.name = name;
}
public String getId() {
return id;
}
public void setId(String id) {
this.id = id;
}
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
public int dataHashCode() {
int dataHashCode = 0;
if(id != null){
dataHashCode += id.hashCode();
}
if(name != null){
dataHashCode += name.hashCode();
}
return dataHashCode;
}
public static void main(String[] args){
DataObjectListMonitor<Model> data = new DataObjectListMonitor<Model>();
Model m1 = new Model("m1", "model 1");
Model m2 = new Model("m2", "model 2");
Model m3 = new Model("m3", "model 3");
Model m4 = new Model("m4", "model 4");
Model m5 = new Model("m5", "model 5");
data.add(m1);
data.add(m2);
data.add(m3);
data.add(m4);
data.add(m5);
data.snapshot();
Model m6 = new Model("m6", "model 6");
data.add(m6);
m3.setName("model 3 changed");
m3.setName("model 3");
data.remove(m5);
m1.setName("model 1 chaned");
for(Model m : data.getAddedObjects()){
System.out.println("added : " + m);
}
for(Model m : data.getDeletedObjects()){
System.out.println("deleted : " + m);
}
for(Model m : data.getmodifiedObjects()){
System.out.println("modified : " + m);
}
}
}
output :
added : Model [id=m6, name=model 6]
deleted : Model [id=m5, name=model 5]
modified : Model [id=m1, name=model 1 chaned]
edit: using hashCode is totally wrong, but maybe we can use MD5 or CRC32.
You can implement a custom collection that can easily track and log add, replace and delete operations but tracking and logging modifications on collection items is quite impossible..
The general idea: implement the List interface, add a delegate (a real list) and delegate all method calls to the internal list (or set, map, queue, ...):
public class LoggingList implements List {
private List delegate = new ArrayList();
private List<ChangeEvent> history = new ArrayList<ChangeEvent>();
// ...
#Override
public boolean add(Object item) {
boolean success = delegate.add(item);
if (success) {
history.add(new ChangeEvent(Type.ADD, item));
}
return success;
}
}
But - the list can't track if someone uses a reference to item to modify its state after it has been added.
One way is to create your own collection implementing the Map interface and add a dirty flag to it. So that when you add or remove elements from the collection make the isDirty flag true and do the operation according to it.
if the existing object is changed then you need to have some different logic.
Tracking modifications means there has to be some ID that you can attach to each item, which also means your better choice is a Map instead of a Collection.
One example to do this is to subclass HashMap and intercept the put (for add/modify operations) and remove (for delete operations).
This is the general idea -- bottom line is you are most likely on your own with the implementation, unless somebody else can recommend a third-party API for this:
Map<K, V> items = new HashMap<K, V>() {
public V put(K key, V value) {
if (containsKey(key))
modified.put(key, value);
else
created.put(key, value);
return super.put(key, value);
}
public V remove(Object key) {
if (containsKey(key))
deleted.put((K) key, get(key));
return super.remove(key);
}
};
In my opinion you should not track changes in the collection but in the changed object itself. So have a changed flag as an instance field inside your object and set it to true when something changed and set it to false after you've written it to the database.
The problem in tracking inside the collection is that it is hard to keep track of modifications to objects already in the collection. Then you need something like a wrapper around each and every object that "informs" your collection when changes to that object happen. Too much in my opinion when this can be solved by tracking inside the objects itself.
This objects' states can afterwards be used to filter the collection or whatever...