- GraalVM for JDK 23 (Latest)
- GraalVM for JDK 24 (Early Access)
- GraalVM for JDK 21
- GraalVM for JDK 17
- Archives
- Dev Build
- Getting Started with Native Image
- Guides
- Native Image Basics
- Build Overview
- Reachability Metadata
- Compatibility Guide
- Collect Metadata with the Tracing Agent
- Experimental Agent Options
- Optimizations and Performance
- Debugging and Diagnostics
- Dynamic Features
- Interoperability with Native Code
- LLVM Backend
- Workshops and Labs
Reachability Metadata
The dynamic language features of the JVM (for example, reflection and resource handling) compute the dynamically-accessed program elements such as fields, methods, or resource URLs at run time. On HotSpot this is possible because all class files and resources are available at run time and can be loaded by the runtime. Availability of all classes and resources, and their loading at run time, comes with an extra overhead in memory and startup time.
To make native binaries small, the native-image
builder performs static analysis at build time to determine only the necessary program elements that are needed for the correctness of the application.
Small binaries allow fast application startup and low memory footprint, however they come at a cost: determining dynamically-accessed application elements via static analysis is infeasible as reachability of those elements depends on data that is available only at run time.
To ensure inclusion of necessary dynamically-accessed elements into the native binary, the native-image
builder requires reachability metadata (hereinafter referred to as metadata).
Providing the builder with correct and exhaustive reachability metadata guarantees application correctness and ensures compatibility with third-party libraries at runtime.
Metadata can be provided to the native-image
builder in the following ways:
- By computing metadata in code when the native binary is built and storing required elements into the initial heap of the native binary.
- By providing the reachability-metadata.json file(s) stored in the META-INF/native-image/<groupId>/<artifactId>/ directory on the classpath. For more information about how to collect metadata for your application automatically, see Collecting Metadata Automatically.
- For more advanced use cases, where classpath scanning or build-time initialization is needed, by using the Native Image Feature API.
Note: Native Image is migrating to the more user-friendly implementation of reachability metadata that shows problems early on and allows easy debugging.
To enable the new user-friendly reachability-metadata mode for your application, pass the option
--exact-reachability-metadata
at build time. To enable the user-friendly mode only for concrete packages, pass--exact-reachability-metadata=<comma-separated-list-of-packages>
.To get an overview of all places in your code where missing registrations occur, without committing to the exact behavior, you can pass
-XX:MissingRegistrationReportingMode=Warn
when starting the application.To detect places where the application accidentally ignores a missing registration error (with
catch (Throwable t)
blocks), pass-XX:MissingRegistrationReportingMode=Exit
when starting the application. The application will then unconditionally print the error message with the stack trace and exit immediately. This behavior is ideal for running application tests to guarantee all metadata is included.The user-friendly implementation for reflection will become the default in future releases of GraalVM so the timely adoption is important to avoid project breakage.
Table of Contents #
- Computing Metadata in Code
- Specifying Metadata with JSON
- Metadata Types
- Reflection (Including Dynamic Proxies)
- Java Native Interface
- Resources
- Resource Bundles
- Serialization
- Sample Reachability Metadata
- Defining Classes at Run Time
Computing Metadata in Code #
Computing metadata in code can be achieved in two ways:
-
By providing constant arguments to functions that dynamically access elements of the JVM. See, for example,
Class#forName
in the following code:class ReflectiveAccess { public Class<Foo> fetchFoo() throws ClassNotFoundException { return Class.forName("Foo"); } }
Here,
Class.forName("Foo")
is evaluated into a constant at build time. When the native binary is built, this value is stored in its initial heap. If the classFoo
does not exist, the call toClass#forName
will be transformed intothrow ClassNotFoundException("Foo")
.The constant is defined as:
- A literal (for example,
"Foo"
or1
). - Access to a static field that is initialized at build time.
- Access to an effectively final variable.
- Defining an array that whose lenght is constant, and all values are constant.
- Simple computations on other constants (for example,
"F"
+"oo"
, or an indexing into an array).
When passing constant arrays, the following approaches to declare and populate an array are equivalent from the point of view of the
native-image
builder:Class<?>[] params0 = new Class<?>[]{String.class, int.class}; Integer.class.getMethod("parseInt", params0);
Class<?>[] params1 = new Class<?>[2]; params1[0] = Class.forName("java.lang.String"); params1[1] = int.class; Integer.class.getMethod("parseInt", params1);
Class<?>[] params2 = {String.class, int.class}; Integer.class.getMethod("parseInt", params2);
Note that Native Image currently aggressively computes constants, and therefore it is not possible to specify exactly what is a constant at build time.
- A literal (for example,
-
By initializing classes at build time and storing dynamically accessed elements into the initial heap of the native executable. This way of providing metadata is suited for cases when specifying metadata with constants or in JSON is not possible. This is necessary in cases when:
- The user code needs to generate new class bytecode.
- The user code needs to traverse the classpath to compute the dynamically accessed program elements necessary for the application.
In the following example:
class InitializedAtBuildTime { private static Class<?> aClass; static { try { aClass = Class.forName(readFile("class.txt")); } catch (ClassNotFoundException e) { throw RuntimeException(e); } } public Class<?> fetchFoo() { return aClass; } }
The dynamically accessed elements will be included into the native executable’s heap only if that part of the heap is reachable through an enclosing method (for example, InitializedAtBuildTime#fetchFoo
) or a static field (for example, InitializedAtBuildTime.aClass
).
Specifying Metadata with JSON #
All metadata specified in the reachability-metadata.json file that is located in any of the classpath entries at META-INF/native-image/<groupId>/<artifactId>/. The JSON schema for the reachability metadata is defined in reachability-metadata-schema-v1.0.0.json.
A sample reachability-metadata.json file can be found in the sample section. The reachability-metadata.json configuration contains a single object with one field for each type of metadata. Each field in the top-level object contains an array of metadata entries:
{
"reflection":[],
"resources":[],
"bundles":[],
"serialization":[],
"jni":[]
}
For example, Java reflection metadata is specified under reflection
, and an example entry looks like:
{
"reflection": [
{
"type": "Foo"
}
]
}
Conditional Metadata Entries #
Each entry in JSON-based metadata should be conditional to avoid unnecessary growth of the native binary size.
A conditional entry is specified by adding a condition
field to the entry in the following way:
{
"condition": {
"typeReached": "<fully-qualified-class-name>"
},
<metadata-entry>
}
A metadata entry with a typeReached
condition is considered available at run time, only when the specified fully-qualified type is reached at run time.
Before that, all dynamic accesses to the element represented with the metadata-entry
will behave as if the metadata-entry
does not exist.
This means that those dynamic accesses will throw a missing-registration error.
A type is reached at run time, right before the class-initialization routine starts for that type (class or interface), or any of the type’s subtypes are reached.
For "typeReached": "ConditionType"
that guards a metadata entry in the following example, the type is considered reached:
class SuperType {
static {
// ConditionType reached (subtype reached) => metadata entry available
}
}
class ConditionType extends SuperType {
static {
// ConditionType reached (before static initializer) => metadata entry available
}
static ConditionType singleton() {
// ConditionType reached (already initialized) => metadata entry available
}
}
public class App {
public static void main(String[] args) {
// ConditionType not reached => metadata entry not available
ConditionType.class;
// ConditionType not reached (ConditionType.class doesn't start class initialization) => metadata entry not available
ConditionType.singleton();
// ConditionType reached (already initialized) => metadata entry available
}
}
A type is also reached, if it is marked as initialize-at-build-time
or any of its subtypes are marked as initialize-at-build-time
and they exist on the classpath.
Array types are never marked as reached and as such cannot be used in conditions.
A conditional metadata entry is included into the image when the fully-qualified type is reachable at build time. This entry affects the image size, and it will be available at run time only when the condition is reached at run time.
You can find more examples of the metadata files in the GraalVM Reachability Metadata repository.
Metadata Types #
Native Image accepts the following types of reachability metadata:
- Java reflection (the
java.lang.reflect.*
API) enables Java code to examine its own classes, methods, fields, and their properties at run time. - JNI allows native code to access classes, methods, fields and their properties at run time.
- Resources allow arbitrary files present on the classpath to be dynamically accessed in the application.
- Resource Bundles Java localization support (
java.util.ResourceBundle
) that enables Java code to load L10N resources. - Serialization enables writing (and reading) Java objects to (and from) streams.
- (Experimental) Predefined Classes provide support for dynamically generated classes.
Reflection #
For all methods in this section Native Image will compute reachability at build time given that all the call arguments are constant. Providing constant arguments in code is a preferred way to provide metadata as it requires no duplication of information in external JSON files.
Reflection in Java starts with java.lang.Class
that allows fetching further reflective elements such as methods and fields.
The class can be fetched reflectively via the following static functions on java.lang.Class
:
java.lang.Class forName(java.lang.String) throws java.lang.ClassNotFoundException
java.lang.Class forName(java.lang.String, boolean, java.lang.ClassLoader) throws java.lang.ClassNotFoundException
java.lang.Class forName(java.lang.Module, java.lang.String)
java.lang.Class arrayType()
- requires metadata for the array type. The classes can be also fetched reflectively loading a class from a name withjava.lang.ClassLoader#loadClass(String)
.
To provide metadata for the calls that fetch a Class
reflectively, the following entry must be added to the reflection
array in reachability-metadata.json:
{
"type": "FullyQualifiedReflectivelyAccessedType"
}
For proxy classes, the java.lang.Class
is fetched with the following methods on java.lang.reflect.Proxy
:
java.lang.Class getProxyClass(java.lang.ClassLoader, java.lang.Class[]) throws java.lang.IllegalArgumentException
java.lang.Object newProxyInstance(java.lang.ClassLoader, java.lang.Class[], java.lang.reflect.InvocationHandler)
Metadata, for proxy classes, is in the form an ordered collection of interfaces that defines a proxy:
{
"type": {
"proxy": ["FullyQualifiedInterface1", "...", "FullyQualifiedInterfaceN"]
}
}
Invocation of methods above without the provided metadata will result in throwing MissingReflectionRegistrationError
which extends java.lang.Error
and
should not be handled. Note that even if a type does not exist on the classpath, the methods above will throw a MissingReflectionRegistrationError
.
The following methods on java.lang.Class
will throw a MissingRegistrationError
if the metadata is not provided for a given type:
Constructor getConstructor(Class[]) throws NoSuchMethodException,SecurityException
Constructor getDeclaredConstructor(Class[]) throws NoSuchMethodException,SecurityException
Constructor[] getConstructors() throws SecurityException
Constructor[] getDeclaredConstructors() throws SecurityException
Method getMethod(String,Class[]) throws NoSuchMethodException,SecurityException
Method getDeclaredMethod(String,Class[]) throws NoSuchMethodException,SecurityException
Method[] getMethods() throws SecurityException
Method[] getDeclaredMethods() throws SecurityException
Field getField(String) throws NoSuchFieldException,SecurityException
Field getDeclaredField(String) throws NoSuchFieldException,SecurityException
Field[] getFields() throws SecurityException
Field[] getDeclaredFields() throws SecurityException
RecordComponent[] getRecordComponents()
Class[] getPermittedSubclasses()
Object[] getSigners()
Class[] getNestMembers()
Class[] getClasses()
Class[] getDeclaredClasses() throws SecurityException
Furthermore, all reflective lookups via java.lang.invoke.MethodHandles.Lookup
will also require metadata for the type to be present, or they will throw MissingReflectionRegistrationError
.
Note that for lambda-proxy classes, metadata can not be provided. This is a known issue that will be addressed in the future releases of GraalVM.
Reflective Method Invocation #
To reflectively invoke methods, the method signature must be added to the type
metadata:
{
"type": "TypeWhoseMethodsAreInvoked",
"methods": [
{"name": "<methodName1>", "parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]},
{"name": "<methodName2>", "parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]}
]
}
As a convenience, one can allow method invocation for groups of methods by adding the following in reachability-metadata.json:
{
"type": "TypeWhoseMethodsAreInvoked",
"allDeclaredConstructors": true,
"allPublicConstructors": true,
"allDeclaredMethods": true,
"allPublicMethods": true
}
allDeclaredConstructors
and allDeclaredMethods
allow calls invocations of methods declared on a given type.
allPublicConstructors
and allPublicMethods
allow invocations of all public methods defined on a type and all of its supertypes.
In case the method-invocation metadata is missing, the following methods will throw a MissingReflectionRegistrationError
:
java.lang.reflect.Method#invoke(Object, Object...)
java.lang.reflect.Constructor#newInstance(Object...)
java.lang.invoke.MethodHandle#invokeExact(Object...)
java.lang.invoke.MethodHandle#invokeWithArguments
(all overloaded versions)
Reflective Field-Value Access #
To reflectively access (get or set) field values, metadata about field names must be added to the type:
{
"type": "TypeWhoseFieldValuesAreAccessed",
"fields": [{"name": "<fieldName1>"}, {"name": "<fieldNameI>"}, {"name": "<fieldNameN>"}]
}
As a convenience one can allow field-value access for all fields by adding the following in reachability-metadata.json:
{
"type": "TypeWhoseFieldValuesAreAccessed",
"allDeclaredFields": true,
"allPublicFields": true
}
allDeclaredFields
allow access to all fields declared on a given type, and allPublicFields
allows access to all public fields of the given type and all of its supertypes.
In case the field-value-access metadata is missing, the following methods will throw a MissingReflectionRegistrationError
:
java.lang.reflect.Field#get(Object)
java.lang.reflect.Field#set(Object, Object)
- All accessor methods on
java.lang.reflect.VarHandle
.
Unsafe Allocation of a Type #
For unsafe allocation of a type via sun.misc.Unsafe#allocateInstance(Class<?>)
, or from native code via AllocObject(jClass)
, we must provide the following metadata:
{
"type": "FullyQualifiedUnsafeAllocatedType",
"unsafeAllocated": true
}
Otherwise, these methods will throw a MissingReflectionRegistrationError
.
Reflection Metadata Summary #
The overall definition of a type in JSON can have the following values:
{
"condition": {
"typeReached": "<condition-class>"
},
"type": "<class>|<proxy-interface-list>",
"fields": [
{"name": "<fieldName>"}
],
"methods": [
{"name": "<methodName>", "parameterTypes": ["<param-type>"]}
],
"allDeclaredConstructors": true,
"allPublicConstructors": true,
"allDeclaredMethods": true,
"allPublicMethods": true,
"allDeclaredFields": true,
"allPublicFields": true,
"unsafeAllocated": true
}
Java Native Interface #
Java Native Interface (JNI) allows native code to access arbitrary Java types and type members. Native Image cannot predict what such native code will lookup, write to or invoke. To build a native binary for a Java application that uses JNI to access Java values, JNI metadata is required.
For example, the following C
code:
jclass clazz = FindClass(env, "jni/accessed/Type");
looks up the jni.accessed.Type
class, which can then be used to instantiate jni.accessed.Type
, invoke its methods or access its fields.
The metadata entry for the above call can only be provided via reachability-metadata.json. Specify the type
entry in the jni
field:
{
"jni":[
{
"type": "jni.accessed.Type"
}
]
}
Adding the metadata for a type does not allow to fetch all of its fields and methods with GetFieldID
, GetStaticFieldID
, GetStaticMethodID
, and GetMethodID
.
To access field values, we need to provide field names:
{
"type": "jni.accessed.Type",
"fields": [{"name": "value"}]
}
To access all fields one can use the following attributes:
{
"type": "jni.accessed.Type",
"allDeclaredFields": true,
"allPublicFields": true
}
allDeclaredFields
allow access to all fields declared on a given type, and allPublicFields
allows access to all public fields of the given type and all of its supertypes.
To call Java methods from JNI, we must provide metadata for the method signatures:
{
"type": "jni.accessed.Type",
"methods": [
{"name": "<methodName1>", "parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]},
{"name": "<methodName2>", "parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]}
]
}
As a convenience, one can allow method invocation for groups of methods by adding the following:
{
"type": "jni.accessed.Type",
"allDeclaredConstructors": true,
"allPublicConstructors": true,
"allDeclaredMethods": true,
"allPublicMethods": true
}
allDeclaredConstructors
and allDeclaredMethods
allow calls invocations of methods declared on a given type.
allPublicConstructors
and allPublicMethods
allow invocations of all public methods defined on a type and all of its supertypes.
To allocate objects of a type with AllocObject
, the metadata must be stored in the reflection
section:
{
"reflection": [
{
"type": "jni.accessed.Type",
"unsafeAllocated": true
}
]
}
Failing to provide metadata for an element that is dynamically accessed from native code will result in an exception (MissingJNIRegistrationError
).
Note that most libraries that use JNI do not handle exceptions properly, so to see which elements are missing
--exact-reachability-metadata
in combination with-XX:MissingRegistrationReportingMode=Warn
must be used.
Resources #
Java is capable of accessing any resource on the application class path, or the module path for which the requesting code has permission to access.
Resource metadata instructs the native-image
builder to include specified resources and resource bundles in the produced binary.
A consequence of this approach is that some parts of the application that use resources for configuration (such as logging) are effectively configured at build time.
Native Image will detect calls to java.lang.Class#getResource
and java.lang.Class#getResourceAsStream
in which:
- The class on which these methods are called is constant
- The first argument (
name
) is a constant and automatically register such resources.
The code below will work out of the box, because:
- It uses a class literal (
Example.class
) as the receiver - It uses a string literal as the
name
parameterclass Example { public void conquerTheWorld() { InputStream plan = Example.class.getResourceAsStream("plans/v2/conquer_the_world.txt"); } }
Resource Metadata in JSON #
Resource metadata is specified in the resources
field of the reachability-metadata.json file.
Here is the example of resource metadata:
{
"resources": [
{
"glob": "path1/level*/**"
}
]
}
The glob
field uses a subset of glob-pattern rules for specifying resources.
There are several rules to be observed when specifying a resource path:
- The
native-image
tool supports only star (*
) and globstar (**
) wildcard patterns.- Per definition, star can match any number of any characters on one level while globstar can match any number of characters at any level.
- If there is a need to treat a star literally (without special meaning), it can be escaped using
\
(for example,\*
).
- In the glob, a level represents a part of the pattern separated with
/
. - When writing glob patterns the following rules must be observed:
- Glob cannot be empty (for example,
""
) - Glob cannot end with a trailing slash (
/
) (for example,"foo/bar/"
) - Glob cannot contain more than two consecutive (non-escaped)
*
characters on one level (for example,"foo/***/"
) - Glob cannot contain empty levels (for example,
"foo//bar"
) - Glob cannot contain two consecutive globstar wildcards (example,
"foo/**/**"
) - Glob cannot have other content on the same level as globstar wildcard (for example,
"foo/**bar/x"
)
- Glob cannot be empty (for example,
Given the following project structure:
app-root
└── src
└── main
└── resources
├── Resource0.txt
└── Resource1.txt
You can:
- Include all resources with glob
**/Resource*.txt
({ "glob":}
) - Include Resource0.txt with glob
**/Resource0.txt
- Include Resource0.txt and Resource1.txt with globs
**/Resource0.txt
and**/Resource1.txt
Resources in Java Modules #
For every resource or resource bundle, it is possible to specify the module from which the resource or resource bundle should be taken.
You can specify a module name in the separate module
field in each entry.
For example:
{
"resources": [
{
"module:": "library.module",
"glob": "resource-file.txt"
}
]
}
This will cause the native-image
tool to only include resource-file.txt from the Java module library.module
.
If other modules or the classpath containing resources that match the pattern resource-file.txt, only the one in library-module is registered to be included into a native executable.
Native Image will also ensure that the modules are guaranteed to be accessible at runtime.
Take the following code pattern:
InputStream resource = ModuleLayer.boot().findModule("library.module").getResourceAsStream(resourcePath);
It will always work as expected for resources registered as described above (even if the module does not contain any code that is considered reachable by static analysis).
Embedded Resources Information #
There are two ways to see which resources were included in a native executable:
- Use the option
--emit build-report
to generate a build report for your native executable. There you can find information about all included resources under theResources
tab. - Use the option
-H:+GenerateEmbeddedResourcesFile
to generate a JSON file embedded-resources.json, listing all included resources.
For each registered resource you get:
- Module (or
unnamed
if a resource does not belong to any module) - Name (resource path)
- Origin (location of the resource on the system)
- Type (whether the resource is a file, directory, or missing)
- Size (actual resource size)
Note: The size of a resource directory represents only the size of the names of all directory entries (not a sum of the content sizes).
Resource Bundles #
Java localization support (java.util.ResourceBundle
) enables to load L10N resources and show messages localized for a specific locale.
Native Image needs knowledge of the resource bundles that your application uses so that it can include appropriate resources and program elements to the application.
A simple bundle can be specified in the bundles
section of reachability-metadata.json:
{
"bundles": [
{
"name":"your.pkg.Bundle"
}
]
}
To request a bundle from a specific module:
{
"bundles": [
{
"name":"app.module:module.pkg.Bundle"
}
]
}
By default, resource bundles are included for all locales that are included into the image. Below is the example how to include only specific locales for a bundle:
{
"bundles": [
{
"name": "specific.locales.Bundle",
"locales": ["en", "de", "sk"]
}
]
}
Locales #
It is also possible to specify which locales should be included in a native executable and which should be the default. For example, to switch the default locale to Swiss German and also include French and English, use the following options:
native-image -Duser.country=CH -Duser.language=de -H:IncludeLocales=fr,en
The locales are specified using language tags.
You can include all locales via -H:+IncludeAllLocales
, but note that it increases the size of the resulting executable.
Serialization #
Java can serialize (or deserialize) any class that implements the Serializable
interface.
Native Image supports serialization (or deserialization) with proper serialization metadata registration.
This is necessary because serialization usually requires reflective accesses to the object that is being serialized.
Serialization Metadata Registration In Code #
Native Image detects calls to ObjectInputFilter.Config#createFilter(String pattern)
and if the pattern
argument is constant, the exact classes mentioned in the pattern will be registered for serialization.
For example, the following pattern will register the class pkg.SerializableClass
for serialization:
var filter = ObjectInputFilter.Config.createFilter("pkg.SerializableClass;!*;")
objectInputStream.setObjectInputFilter(proof);
Using this pattern has a positive side effect of improving security on the JVM as only pkg.SerializableClass
can be received by the
objectInputStream
.
Wildcard patterns do the serialization registration only for lambda-proxy classes of an enclosing class. For example, to register lambda serialization in an enclosing class pkg.LambdaHolder
use:
ObjectInputFilter.Config.createFilter("pkg.LambdaHolder$$Lambda*;")
Patterns like "pkg.**"
and "pkg.Prefix*"
will not perform serialization registration as they are too general and would increase image size significantly.
For calls to the sun.reflect.ReflectionFactory#newConstructorForSerialization(java.lang.Class)
and sun.reflect.ReflectionFactory#newConstructorForSerialization(java.lang.Class, )
native image detects calls to these functions when all arguments and the receiver are constant. For example, the following call will register SerializlableClass
for serialization:
ReflectionFactory.getReflectionFactory().newConstructorForSerialization(SerializableClass.class);
To create a custom constructor for serialization use:
var constructor = SuperSuperClass.class.getDeclaredConstructor();
var newConstructor = ReflectionFactory.getReflectionFactory().newConstructorForSerialization(BaseClass.class, constructor);
Proxy classes can only be registered for serialization via the JSON files.
Serialization Metadata in JSON #
Serialization metadata is specified in the serialization
section of reachability-metadata.json.
To specify a regular serialized.Type
use
{
"serialization": [
{
"type": "serialized.Type"
}
]
}
To specify a proxy class for serialization, use the following entry:
{
"serialization": [
{
"type": {
"proxy": ["FullyQualifiedInterface1", "...", "FullyQualifiedInterfaceN"]
}
}
]
}
In rare cases an application might explicitly make calls to:
ReflectionFactory.newConstructorForSerialization(Class<?> cl, Constructor<?> constructorToCall);
In which the passed constructorToCall
differs from what would automatically be used if regular serialization of cl
.
To also support such serialization use cases, it is possible to register serialization for a class with a
custom constructorToCall
.
For example, to allow serialization of org.apache.spark.SparkContext$$anonfun$hadoopFile$1
, use the declared constructor of java.lang.Object
as a custom targetConstructor
, use:
{
"serialization": [
{
"type": "<fully-qualified-class-name>",
"customTargetConstructorClass": "<custom-target-constructor-class>"
}
]
}
Sample Reachability Metadata #
See below is a sample reachability metadata configuration that you can use in reachabilty-metadata.json:
{
"reflection": [
{
"type": "reflectively.accessed.Type",
"fields": [
{
"name": "field1"
}
],
"methods": [
{
"name": "method1",
"parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]
}
],
"allDeclaredConstructors": true,
"allPublicConstructors": true,
"allDeclaredFields": true,
"allPublicFields": true,
"allDeclaredMethods": true,
"allPublicMethods": true,
"unsafeAllocated": true
}
],
"jni": [
{
"type": "jni.accessed.Type",
"fields": [
{
"name": "field1"
}
],
"methods": [
{
"name": "method1",
"parameterTypes": ["<param-type1>", "<param-typeI>", "<param-typeN>"]
}
],
"allDeclaredConstructors": true,
"allPublicConstructors": true,
"allDeclaredFields": true,
"allPublicFields": true,
"allDeclaredMethods": true,
"allPublicMethods": true
}
],
"resources": [
{
"module": "optional.module.of.a.resource",
"glob": "path1/level*/**"
}
],
"bundles": [
{
"name": "fully.qualified.bundle.name",
"locales": ["en", "de", "other_optional_locales"]
}
],
"serialization": [
{
"type": "serialized.Type",
"customTargetConstructorClass": "optional.serialized.super.Type"
}
]
}
Defining Classes at Run Time #
Java has support for loading new classes from bytecode at run time, which is not possible in Native Image as all classes must be known at build time (the “closed-world assumption”). To overcome this issue there are the following options:
- Modify or reconfigure your application (or a third-party library) so that it does not generate classes at runtime or load them via non-built-in class loaders.
- If the classes must be generated, try to generate them at build time in a static initializer of a dedicated class.
The generated java.lang.Class objects should be stored in static fields and the dedicated class initialized by passing
--initialize-at-build-time=<class_name>
as the build argument. - If none of the above is applicable, use the Native Image Agent to run the application and collect predefined classes with
java -agentlib:native-image-agent=config-output-dir=<config-dir>,experimental-class-define-support <application-arguments>
. At runtime, if there is an attempt to load a class with the same name and bytecode as one of the classes encountered during tracing, the predefined class will be supplied to the application.
Predefined classes metadata is specified in a predefined-classes-config.json file and conform to the JSON schema defined in predefined-classes-config-schema-v1.0.0.json. The schema also includes further details and explanations how this configuration works. Here is the example of the predefined-classes-config.json:
[
{
"type": "agent-extracted",
"classes": [
{
"hash": "<class-bytecodes-hash>",
"nameInfo": "<class-name"
}
]
}
]
Note: Predefined classes metadata is not meant to be manually written. Note: Predefined classes are the best-effort approach for legacy projects, and they are not guaranteed to work.