MXBean

Annotation to mark an interface explicitly as being an MXBean interface, or as not being an MXBean interface. By default, an interface is an MXBean interface if its name ends with {@code MXBean}, as in {@code SomethingMXBean}. The following interfaces are MXBean interfaces:

public interface WhatsitMXBean {}

@MXBean
public interface Whatsit1Interface {}

@MXBean(true)
public interface Whatsit2Interface {}

The following interfaces are not MXBean interfaces:

public interface Whatsit3Interface{}

@MXBean(false)
public interface MisleadingMXBean {}

MXBean specification

The MXBean concept provides a simple way to code an MBean that only references a predefined set of types, the ones defined by {@link javax.management.openmbean}. In this way, you can be sure that your MBean will be usable by any client, including remote clients, without any requirement that the client have access to model-specific classes representing the types of your MBeans.

The concepts are easier to understand by comparison with the Standard MBean concept. Here is how a managed object might be represented as a Standard MBean, and as an MXBean:

public interface MemoryPoolMBean {
String getName();
MemoryUsage getUsage();
// ...
}

public interface MemoryPoolMXBean {
String getName();
MemoryUsage getUsage();
// ...
}

As you can see, the definitions are very similar. The only difference is that the convention for naming the interface is to use SomethingMXBean for MXBeans, rather than SomethingMBean for Standard MBeans.

In this managed object, there is an attribute called Usage of type {@link MemoryUsage}. The point of an attribute like this is that it gives a coherent snapshot of a set of data items. For example, it might include the current amount of used memory in the memory pool, and the current maximum of the memory pool. If these were separate items, obtained with separate {@link MBeanServer#getAttribute getAttribute} calls, then we could get values seen at different times that were not consistent. We might get a used value that was greater than the max value.

So, we might define MemoryUsage like this:

public class MemoryUsage implements Serializable {
// standard JavaBean conventions with getters

public MemoryUsage(long init, long used,
long committed, long max) {...}
long getInit() {...}
long getUsed() {...}
long getCommitted() {...}
long getMax() {...}
}

public class MemoryUsage {
// standard JavaBean conventions with getters
@ConstructorProperties({"init", "used", "committed", "max"})
public MemoryUsage(long init, long used,
long committed, long max) {...}
long getInit() {...}
long getUsed() {...}
long getCommitted() {...}
long getMax() {...}
}

The definitions are the same in the two cases, except that with the MXBean, MemoryUsage no longer needs to be marked Serializable (though it can be). On the other hand, we have added a {@code @ConstructorProperties} annotation to link the constructor parameters to the corresponding getters. We will see more about this below.

MemoryUsage is a model-specific class. With Standard MBeans, a client of the MBean Server cannot access the Usage attribute if it does not know the class MemoryUsage. Suppose the client is a generic console based on JMX technology. Then the console would have to be configured with the model-specific classes of every application it might connect to. The problem is even worse for clients that are not written in the Java language. Then there may not be any way to tell the client what a MemoryUsage looks like.

This is where MXBeans differ from Standard MBeans. Although we define the management interface in almost exactly the same way, the MXBean framework converts model-specific classes into standard classes from the Java platform. Using arrays and the {@link javax.management.openmbean.CompositeData CompositeData} and {@link javax.management.openmbean.TabularData TabularData} classes from the standard {@link javax.management.openmbean} package, it is possible to build data structures of arbitrary complexity using only standard classes.

This becomes clearer if we compare what the clients of the two models might look like:

String name = (String)
mbeanServer.{@link MBeanServer#getAttribute
getAttribute}(objectName, "Name");
MemoryUsage usage = (MemoryUsage)
mbeanServer.getAttribute(objectName, "Usage");
long used = usage.getUsed();

String name = (String)
mbeanServer.{@link MBeanServer#getAttribute
getAttribute}(objectName, "Name");
{@link CompositeData} usage = (CompositeData)
mbeanServer.getAttribute(objectName, "Usage");
long used = (Long) usage.{@link CompositeData#get get}("used");

For attributes with simple types like String, the code is the same. But for attributes with complex types, the Standard MBean code requires the client to know the model-specific class MemoryUsage, while the MXBean code requires no non-standard classes.

The client code shown here is slightly more complicated for the MXBean client. But, if the client does in fact know the model, here the interface MemoryPoolMXBean and the class MemoryUsage, then it can construct a proxy. This is the recommended way to interact with managed objects when you know the model beforehand, regardless of whether you are using Standard MBeans or MXBeans:

MemoryPoolMBean proxy =
JMX.{@link JMX#newMBeanProxy(MBeanServerConnection, ObjectName,
Class) newMBeanProxy}(
mbeanServer,
objectName,
MemoryPoolMBean.class);
String name = proxy.getName();
MemoryUsage usage = proxy.getUsage();
long used = usage.getUsed();

MemoryPoolMXBean proxy =
JMX.{@link JMX#newMXBeanProxy(MBeanServerConnection, ObjectName,
Class) newMXBeanProxy}(
mbeanServer,
objectName,
MemoryPoolMXBean.class);
String name = proxy.getName();
MemoryUsage usage = proxy.getUsage();
long used = usage.getUsed();

Implementing the MemoryPool object works similarly for both Standard MBeans and MXBeans.

public class MemoryPool
implements MemoryPoolMBean {
public String getName() {...}
public MemoryUsage getUsage() {...}
// ...
}

public class MemoryPool
implements MemoryPoolMXBean {
public String getName() {...}
public MemoryUsage getUsage() {...}
// ...
}

Registering the MBean in the MBean Server works in the same way in both cases:

{
MemoryPoolMBean pool = new MemoryPool();
mbeanServer.{@link MBeanServer#registerMBean
registerMBean}(pool, objectName);
}

{
MemoryPoolMXBean pool = new MemoryPool();
mbeanServer.{@link MBeanServer#registerMBean
registerMBean}(pool, objectName);
}

Definition of an MXBean

An MXBean is a kind of MBean. An MXBean object can be registered directly in the MBean Server, or it can be used as an argument to {@link StandardMBean} and the resultant MBean registered in the MBean Server.

When an object is registered in the MBean Server using the {@code registerMBean} or {@code createMBean} methods of the {@link MBeanServer} interface, the object's class is examined to determine what type of MBean it is:

• If the class implements the interface {@link DynamicMBean} then the MBean is a Dynamic MBean. Note that the class {@code StandardMBean} implements this interface, so this case applies to a Standard MBean or MXBean created using the class {@code StandardMBean}.
• Otherwise, if the class matches the Standard MBean naming conventions, then the MBean is a Standard MBean.
• Otherwise, it may be an MXBean. The set of interfaces implemented by the object is examined for interfaces that:
  • have a class name SMXBean where S is any non-empty string, and do not have an annotation {@code @MXBean(false)}; and/or
  • have an annotation {@code @MXBean(true)} or just {@code @MXBean}.
  If there is exactly one such interface, or if there is one such interface that is a subinterface of all the others, then the object is an MXBean. The interface in question is the MXBean interface. In the example above, the MXBean interface is {@code MemoryPoolMXBean}.
• If none of these conditions is met, the MBean is invalid and the attempt to register it will generate {@link NotCompliantMBeanException}.

Every Java type that appears as the parameter or return type of a method in an MXBean interface must be convertible using the rules below. Additionally, parameters must be reconstructible as defined below.

An attempt to construct an MXBean that does not conform to the above rules will produce an exception.

Naming conventions

The same naming conventions are applied to the methods in an MXBean as in a Standard MBean:

1. A method T getN(), where T is a Java type (not void) and N is a non-empty string, specifies that there is a readable attribute called N. The Java type and Open type of the attribute are determined by the mapping rules below. The method {@code final Class getClass()} inherited from {@code Object} is ignored when looking for getters.
2. A method boolean isN() specifies that there is a readable attribute called N with Java type boolean and Open type SimpleType.Boolean.
3. A method void setN(T x) specifies that there is a writeable attribute called N. The Java type and Open type of the attribute are determined by the mapping rules below. (Of course, the name x of the parameter is irrelevant.)
4. Every other method specifies that there is an operation with the same name as the method. The Java type and Open type of the return value and of each parameter are determined by the mapping rules below.

The rules for getN and isN collectively define the notion of a getter. The rule for setN defines the notion of a setter.

It is an error for there to be two getters with the same name, or two setters with the same name. If there is a getter and a setter for the same name, then the type T in both must be the same. In this case the attribute is read/write. If there is only a getter or only a setter, the attribute is read-only or write-only respectively.

Type mapping rules

An MXBean is a kind of Open MBean, as defined by the {@link javax.management.openmbean} package. This means that the types of attributes, operation parameters, and operation return values must all be describable using Open Types, that is the four standard subclasses of {@link javax.management.openmbean.OpenType}. MXBeans achieve this by mapping Java types into Open Types.

For every Java type J, the MXBean mapping is described by the following information:

• The corresponding Open Type, opentype(J). This is an instance of a subclass of {@link javax.management.openmbean.OpenType}.
• The mapped Java type, opendata(J), which is always the same for any given opentype(J). This is a Java class.
• How a value is converted from type J to type opendata(J).
• How a value is converted from type opendata(J) to type J, if it can be.

For example, for the Java type {@code List}:

• The Open Type, opentype({@code List}), is {@link ArrayType}(1, {@link SimpleType#STRING}), representing a 1-dimensional array of Strings.
• The mapped Java type, opendata({@code List}), is {@code String[]}.
• A {@code List} can be converted to a {@code String[]} using {@link List#toArray(Object[]) List.toArray(new String[0])}.
• A {@code String[]} can be converted to a {@code List} using {@link Arrays#asList Arrays.asList}.

If no mapping rules exist to derive opentype(J) from J, then J cannot be the type of a method parameter or return value in an MXBean interface.

If there is a way to convert opendata(J) back to J then we say that J is reconstructible. All method parameters in an MXBean interface must be reconstructible, because when the MXBean framework is invoking a method it will need to convert those parameters from opendata(J) to J. In a proxy generated by {@link JMX#newMXBeanProxy(MBeanServerConnection, ObjectName, Class) JMX.newMXBeanProxy}, it is the return values of the methods in the MXBean interface that must be reconstructible.

Null values are allowed for all Java types and Open Types, except primitive Java types where they are not possible. When converting from type J to type opendata(J) or from type opendata(J) to type J, a null value is mapped to a null value.

The following table summarizes the type mapping rules.

The following sections give further details of these rules.

Mappings for primitive types

The 8 primitive Java types ({@code boolean}, {@code byte}, {@code short}, {@code int}, {@code long}, {@code float}, {@code double}, {@code char}) are mapped to the corresponding boxed types from {@code java.lang}, namely {@code Boolean}, {@code Byte}, etc. The Open Type is the corresponding {@code SimpleType}. Thus, opentype({@code long}) is {@code SimpleType.LONG}, and opendata({@code long}) is {@code java.lang.Long}.

An array of primitive type such as {@code long[]} can be represented directly as an Open Type. Thus, openType({@code long[]}) is {@code ArrayType.getPrimitiveArrayType(long[].class)}, and opendata({@code long[]}) is {@code long[]}.

In practice, the difference between a plain {@code int} and {@code Integer}, etc, does not show up because operations in the JMX API are always on Java objects, not primitives. However, the difference does show up with arrays.

Mappings for collections ({@code List<}E{@code >} etc)

A {@code List<}E{@code >} or {@code Set<}E{@code >}, such as {@code List} or {@code Set}, is mapped in the same way as an array of the same element type, such as {@code String[]} or {@code ObjectName[]}.

A {@code SortedSet<}E{@code >} is also mapped in the same way as an E{@code []}, but it is only convertible if E is a class or interface that implements {@link java.lang.Comparable}. Thus, a {@code SortedSet} or {@code SortedSet} is convertible, but a {@code SortedSet} or {@code SortedSet>} is not. The conversion of a {@code SortedSet} instance will fail with an {@code IllegalArgumentException} if it has a non-null {@link java.util.SortedSet#comparator() comparator()}.

A {@code List<}E{@code >} is reconstructed as a {@code java.util.ArrayList<}E{@code >}; a {@code Set<}E{@code >} as a {@code java.util.HashSet<}E{@code >}; a {@code SortedSet<}E{@code >} as a {@code java.util.TreeSet<}E{@code >}.

Mappings for maps ({@code Map<}K,V{@code >} etc)

A {@code Map<}K,V{@code >} or {@code SortedMap<}K,V{@code >}, for example {@code Map}, has Open Type {@link TabularType} and is mapped to a {@link TabularData}. The {@code TabularType} has two items called {@code key} and {@code value}. The Open Type of {@code key} is opentype(K), and the Open Type of {@code value} is opentype(V). The index of the {@code TabularType} is the single item {@code key}.

For example, the {@code TabularType} for a {@code Map} might be constructed with code like this:

String typeName =
"java.util.Map<java.lang.String, javax.management.ObjectName>";
String[] keyValue =
new String[] {"key", "value"};
OpenType[] openTypes =
new OpenType[] {SimpleType.STRING, SimpleType.OBJECTNAME};
CompositeType rowType =
new CompositeType(typeName, typeName, keyValue, keyValue, openTypes);
TabularType tabularType =
new TabularType(typeName, typeName, rowType, new String[] {"key"});

The {@code typeName} here is determined by the type name rules detailed below.

A {@code SortedMap<}K,V{@code >} is mapped in the same way, but it is only convertible if K is a class or interface that implements {@link java.lang.Comparable}. Thus, a {@code SortedMap} is convertible, but a {@code SortedMap} is not. The conversion of a {@code SortedMap} instance will fail with an {@code IllegalArgumentException} if it has a non-null {@link java.util.SortedMap#comparator() comparator()}.

A {@code Map<}K,V{@code >} is reconstructed as a {@code java.util.HashMap<}K,V{@code >}; a {@code SortedMap<}K,V{@code >} as a {@code java.util.TreeMap<}K,V{@code >}.

{@code TabularData} is an interface. The concrete class that is used to represent a {@code Map<}K,V{@code >} as Open Data is {@link TabularDataSupport}, or another class implementing {@code TabularData} that serializes as {@code TabularDataSupport}.

Mappings for MXBean interfaces

An MXBean interface, or a type referenced within an MXBean interface, can reference another MXBean interface, J. Then opentype(J) is {@code SimpleType.OBJECTNAME} and opendata(J) is {@code ObjectName}.

For example, suppose you have two MXBean interfaces like this:

public interface ProductMXBean {
public ModuleMXBean[] getModules();
}

public interface ModuleMXBean {
public ProductMXBean getProduct();
}

The object implementing the {@code ModuleMXBean} interface returns from its {@code getProduct} method an object implementing the {@code ProductMXBean} interface. The {@code ModuleMXBean} object and the returned {@code ProductMXBean} objects must both be registered as MXBeans in the same MBean Server.

The method {@code ModuleMXBean.getProduct()} defines an attribute called {@code Product}. The Open Type for this attribute is {@code SimpleType.OBJECTNAME}, and the corresponding {@code ObjectName} value will be the name under which the referenced {@code ProductMXBean} is registered in the MBean Server.

If you make an MXBean proxy for a {@code ModuleMXBean} and call its {@code getProduct()} method, the proxy will map the {@code ObjectName} back into a {@code ProductMXBean} by making another MXBean proxy. More formally, when a proxy made with {@link JMX#newMXBeanProxy(MBeanServerConnection, ObjectName, Class) JMX.newMXBeanProxy(mbeanServerConnection, objectNameX, interfaceX)} needs to map {@code objectNameY} back into {@code interfaceY}, another MXBean interface, it does so with {@code JMX.newMXBeanProxy(mbeanServerConnection, objectNameY, interfaceY)}. The implementation may return a proxy that was previously created by a call to {@code JMX.newMXBeanProxy} with the same parameters, or it may create a new proxy.

The reverse mapping is illustrated by the following change to the {@code ModuleMXBean} interface:

public interface ModuleMXBean {
public ProductMXBean getProduct();
public void setProduct(ProductMXBean c);
}

The presence of the {@code setProduct} method now means that the {@code Product} attribute is read/write. As before, the value of this attribute is an {@code ObjectName}. When the attribute is set, the {@code ObjectName} must be converted into the {@code ProductMXBean} object that the {@code setProduct} method expects. This object will be an MXBean proxy for the given {@code ObjectName} in the same MBean Server.

If you make an MXBean proxy for a {@code ModuleMXBean} and call its {@code setProduct} method, the proxy will map its {@code ProductMXBean} argument back into an {@code ObjectName}. This will only work if the argument is in fact another proxy, for a {@code ProductMXBean} in the same {@code MBeanServerConnection}. The proxy can have been returned from another proxy (like {@code ModuleMXBean.getProduct()} which returns a proxy for a {@code ProductMXBean}); or it can have been created by {@link JMX#newMXBeanProxy(MBeanServerConnection, ObjectName, Class) JMX.newMXBeanProxy}; or it can have been created using {@link java.lang.reflect.Proxy Proxy} with an invocation handler that is {@link MBeanServerInvocationHandler} or a subclass.

If the same MXBean were registered under two different {@code ObjectName}s, a reference to that MXBean from another MXBean would be ambiguous. Therefore, if an MXBean object is already registered in an MBean Server and an attempt is made to register it in the same MBean Server under another name, the result is an {@link InstanceAlreadyExistsException}. Registering the same MBean object under more than one name is discouraged in general, notably because it does not work well for MBeans that are {@link NotificationBroadcaster}s.

Mappings for other types

Given a Java class or interface J that does not match the other rules in the table above, the MXBean framework will attempt to map it to a {@link CompositeType} as follows. The type name of this {@code CompositeType} is determined by the type name rules below.

The class is examined for getters using the conventions above. (Getters must be public instance methods.) If there are no getters, or if any getter has a type that is not convertible, then J is not convertible.

If there is at least one getter and every getter has a convertible type, then opentype(J) is a {@code CompositeType} with one item for every getter. If the getter is

T getName()

String getOwner()

boolean isName()

Notice that the first character (or code point) is converted to lower case. This follows the Java Beans convention, which for historical reasons is different from the Standard MBean convention. In a Standard MBean or MXBean interface, a method {@code getOwner} defines an attribute called {@code Owner}, while in a Java Bean or mapped {@code CompositeType}, a method {@code getOwner} defines a property or item called {@code owner}.

If two methods produce the same item name (for example, {@code getOwner} and {@code isOwner}, or {@code getOwner} and {@code getowner}) then the type is not convertible.

When the Open Type is {@code CompositeType}, the corresponding mapped Java type (opendata(J)) is {@link CompositeData}. The mapping from an instance of J to a {@code CompositeData} corresponding to the {@code CompositeType} just described is done as follows. First, if J implements the interface {@link CompositeDataView}, then that interface's {@link CompositeDataView#toCompositeData toCompositeData} method is called to do the conversion. Otherwise, the {@code CompositeData} is constructed by calling the getter for each item and converting it to the corresponding Open Data type. Thus, a getter such as

{@code List getNames()}

will have been mapped to an item with name "{@code names}" and Open Type {@code ArrayType(1, SimpleType.STRING)}. The conversion to {@code CompositeData} will call {@code getNames()} and convert the resultant {@code List} into a {@code String[]} for the item "{@code names}".

{@code CompositeData} is an interface. The concrete class that is used to represent a type as Open Data is {@link CompositeDataSupport}, or another class implementing {@code CompositeData} that serializes as {@code CompositeDataSupport}.

Reconstructing an instance of Java type J from a {@code CompositeData}

If opendata(J) is {@code CompositeData} for a Java type J, then either an instance of J can be reconstructed from a {@code CompositeData}, or J is not reconstructible. If any item in the {@code CompositeData} is not reconstructible, then J is not reconstructible either.

For any given J, the following rules are consulted to determine how to reconstruct instances of J from {@code CompositeData}. The first applicable rule in the list is the one that will be used.

1. If J has a method
   {@code public static }J {@code from(CompositeData cd)}
   then that method is called to reconstruct an instance of J.

2. Otherwise, if J has at least one public constructor with a {@link ConstructorProperties} annotation, then one of those constructors (not necessarily always the same one) will be called to reconstruct an instance of J. Every such annotation must list as many strings as the constructor has parameters; each string must name a property corresponding to a getter of J; and the type of this getter must be the same as the corresponding constructor parameter. It is not an error for there to be getters that are not mentioned in the {@code ConstructorProperties} annotation (these may correspond to information that is not needed to reconstruct the object).

   An instance of J is reconstructed by calling a constructor with the appropriate reconstructed items from the {@code CompositeData}. The constructor to be called will be determined at runtime based on the items actually present in the {@code CompositeData}, given that this {@code CompositeData} might come from an earlier version of J where not all the items were present. A constructor is applicable if all the properties named in its {@code ConstructorProperties} annotation are present as items in the {@code CompositeData}. If no constructor is applicable, then the attempt to reconstruct J fails.

   For any possible combination of properties, it must be the case that either (a) there are no applicable constructors, or (b) there is exactly one applicable constructor, or (c) one of the applicable constructors names a proper superset of the properties named by each other applicable constructor. (In other words, there should never be ambiguity over which constructor to choose.) If this condition is not true, then J is not reconstructible.

3. Otherwise, if J has a public no-arg constructor, and for every getter in J with type T and name N there is a corresponding setter with the same name and type, then an instance of J is constructed with the no-arg constructor and the setters are called with the reconstructed items from the {@code CompositeData} to restore the values. For example, if there is a method
   {@code public List getNames()}
   then there must also be a method
   {@code public void setNames(List names)}
   for this rule to apply.

   If the {@code CompositeData} came from an earlier version of J, some items might not be present. In this case, the corresponding setters will not be called.

4. Otherwise, if J is an interface that has no methods other than getters, an instance of J is constructed using a {@link java.lang.reflect.Proxy} with a {@link CompositeDataInvocationHandler} backed by the {@code CompositeData} being converted.

5. Otherwise, J is not reconstructible.

Here are examples showing different ways to code a type {@code NamedNumber} that consists of an {@code int} and a {@code String}. In each case, the {@code CompositeType} looks like this:

{@link CompositeType}(
"NamedNumber", // typeName
"NamedNumber", // description
new String[] {"number", "name"}, // itemNames
new String[] {"number", "name"}, // itemDescriptions
new OpenType[] {SimpleType.INTEGER,
SimpleType.STRING} // itemTypes
);

1. Static {@code from} method:

   public class NamedNumber {
   public int getNumber() {return number;}
   public String getName() {return name;}
   private NamedNumber(int number, String name) {
   this.number = number;
   this.name = name;
   }
   public static NamedNumber from(CompositeData cd) {
   return new NamedNumber((Integer) cd.get("number"),
   (String) cd.get("name"));
   }
   private final int number;
   private final String name;
   }
   

2. Public constructor with @ConstructorProperties annotation:

   public class NamedNumber {
   public int getNumber() {return number;}
   public String getName() {return name;}
   @ConstructorProperties({"number", "name"})
   public NamedNumber(int number, String name) {
   this.number = number;
   this.name = name;
   }
   private final int number;
   private final String name;
   }
   

3. Setter for every getter:

   public class NamedNumber {
   public int getNumber() {return number;}
   public void setNumber(int number) {this.number = number;}
   public String getName() {return name;}
   public void setName(String name) {this.name = name;}
   public NamedNumber() {}
   private int number;
   private String name;
   }
   

4. Interface with only getters:

   public interface NamedNumber {
   public int getNumber();
   public String getName();
   }
   

It is usually better for classes that simply represent a collection of data to be immutable. An instance of an immutable class cannot be changed after it has been constructed. Notice that {@code CompositeData} itself is immutable. Immutability has many advantages, notably with regard to thread-safety and security. So the approach using setters should generally be avoided if possible.

Recursive types

Recursive (self-referential) types cannot be used in MXBean interfaces. This is a consequence of the immutability of {@link CompositeType}. For example, the following type could not be the type of an attribute, because it refers to itself:

public interface Node {
public String getName();
public int getPriority();
public Node getNext();
}

It is always possible to rewrite recursive types like this so they are no longer recursive. Doing so may require introducing new types. For example:

public interface NodeList {
public List<Node> getNodes();
}

public interface Node {
public String getName();
public int getPriority();
}

MBeanInfo contents for an MXBean

An MXBean is a type of Open MBean. However, for compatibility reasons, its {@link MBeanInfo} is not an {@link OpenMBeanInfo}. In particular, when the type of an attribute, parameter, or operation return value is a primitive type such as {@code int}, or is {@code void} (for a return type), then the attribute, parameter, or operation will be represented respectively by an {@link MBeanAttributeInfo}, {@link MBeanParameterInfo}, or {@link MBeanOperationInfo} whose {@code getType()} or {@code getReturnType()} returns the primitive name ("{@code int}" etc). This is so even though the mapping rules above specify that the opendata mapping is the wrapped type ({@code Integer} etc).

The array of public constructors returned by {@link MBeanInfo#getConstructors()} for an MXBean that is directly registered in the MBean Server will contain all of the public constructors of that MXBean. If the class of the MXBean is not public then its constructors are not considered public either. The list returned for an MXBean that is constructed using the {@link StandardMBean} class is derived in the same way as for Standard MBeans. Regardless of how the MXBean was constructed, its constructor parameters are not subject to MXBean mapping rules and do not have a corresponding {@code OpenType}.

The array of notification types returned by {@link MBeanInfo#getNotifications()} for an MXBean that is directly registered in the MBean Server will be empty if the MXBean does not implement the {@link NotificationBroadcaster} interface. Otherwise, it will be the result of calling {@link NotificationBroadcaster#getNotificationInfo()} at the time the MXBean was registered. Even if the result of this method changes subsequently, the result of {@code MBeanInfo.getNotifications()} will not. The list returned for an MXBean that is constructed using the {@link StandardMBean} or {@link StandardEmitterMBean} class is derived in the same way as for Standard MBeans.

The {@link Descriptor} for all of the {@code MBeanAttributeInfo}, {@code MBeanParameterInfo}, and {@code MBeanOperationInfo} objects contained in the {@code MBeanInfo} will have a field {@code openType} whose value is the {@link OpenType} specified by the mapping rules above. So even when {@code getType()} is "{@code int}", {@code getDescriptor().getField("openType")} will be {@link SimpleType#INTEGER}.

The {@code Descriptor} for each of these objects will also have a field {@code originalType} that is a string representing the Java type that appeared in the MXBean interface. The format of this string is described in the section Type Names below.

The {@code Descriptor} for the {@code MBeanInfo} will have a field {@code mxbean} whose value is the string "{@code true}".

Type Names

Sometimes the unmapped type T of a method parameter or return value in an MXBean must be represented as a string. If T is a non-generic type, this string is the value returned by {@link Class#getName()}. Otherwise it is the value of genericstring(T), defined as follows:

• If T is a non-generic non-array type, genericstring(T) is the value returned by {@link Class#getName()}, for example {@code "int"} or {@code "java.lang.String"}.
• If T is an array E[], genericstring(T) is genericstring(E) followed by {@code "[]"}. For example, genericstring({@code int[]}) is {@code "int[]"}, and genericstring({@code List[][]}) is {@code "java.util.List[][]"}.
• Otherwise, T is a parameterized type such as {@code List} and genericstring(T) consists of the following: the fully-qualified name of the parameterized type as returned by {@code Class.getName()}; a left angle bracket ({@code "<"}); genericstring(A) where A is the first type parameter; if there is a second type parameter B then {@code ", "} (a comma and a single space) followed by genericstring(B); a right angle bracket ({@code ">"}).

Note that if a method returns {@code int[]}, this will be represented by the string {@code "[I"} returned by {@code Class.getName()}, but if a method returns {@code List}, this will be represented by the string {@code "java.util.List"}.

Exceptions

A problem with mapping from Java types to Open types is signaled with an {@link OpenDataException}. This can happen when an MXBean interface is being analyzed, for example if it references a type like {@link java.util.Random java.util.Random} that has no getters. Or it can happen when an instance is being converted (a return value from a method in an MXBean or a parameter to a method in an MXBean proxy), for example when converting from {@code SortedSet} to {@code String[]} if the {@code SortedSet} has a non-null {@code Comparator}.

A problem with mapping to Java types from Open types is signaled with an {@link InvalidObjectException}. This can happen when an MXBean interface is being analyzed, for example if it references a type that is not reconstructible according to the rules above, in a context where a reconstructible type is required. Or it can happen when an instance is being converted (a parameter to a method in an MXBean or a return value from a method in an MXBean proxy), for example from a String to an Enum if there is no Enum constant with that name.

Depending on the context, the {@code OpenDataException} or {@code InvalidObjectException} may be wrapped in another exception such as {@link RuntimeMBeanException} or {@link UndeclaredThrowableException}. For every thrown exception, the condition C will be true: "e is {@code OpenDataException} or {@code InvalidObjectException} (as appropriate), or C is true of e.{@link Throwable#getCause() getCause()}".