--- a/nashorn/src/jdk.dynalink/share/classes/module-info.java Wed Jul 05 22:39:06 2017 +0200
+++ b/nashorn/src/jdk.dynalink/share/classes/module-info.java Tue Jan 03 22:14:41 2017 +0530
@@ -24,7 +24,198 @@
*/
/**
- * Dynalink
+ * <p>
+ * Dynalink is a library for dynamic linking of high-level operations on objects.
+ * These operations include "read a property",
+ * "write a property", "invoke a function" and so on. Dynalink is primarily
+ * useful for implementing programming languages where at least some expressions
+ * have dynamic types (that is, types that can not be decided statically), and
+ * the operations on dynamic types are expressed as
+ * {@link java.lang.invoke.CallSite call sites}. These call sites will be
+ * linked to appropriate target {@link java.lang.invoke.MethodHandle method handles}
+ * at run time based on actual types of the values the expressions evaluated to.
+ * These can change between invocations, necessitating relinking the call site
+ * multiple times to accommodate new types; Dynalink handles all that and more.
+ * <p>
+ * Dynalink supports implementation of programming languages with object models
+ * that differ (even radically) from the JVM's class-based model and have their
+ * custom type conversions.
+ * <p>
+ * Dynalink is closely related to, and relies on, the {@link java.lang.invoke}
+ * package.
+ * <p>
+ *
+ * While {@link java.lang.invoke} provides a low level API for dynamic linking
+ * of {@code invokedynamic} call sites, it does not provide a way to express
+ * higher level operations on objects, nor methods that implement them. These
+ * operations are the usual ones in object-oriented environments: property
+ * access, access of elements of collections, invocation of methods and
+ * constructors (potentially with multiple dispatch, e.g. link- and run-time
+ * equivalents of Java overloaded method resolution). These are all functions
+ * that are normally desired in a language on the JVM. If a language is
+ * statically typed and its type system matches that of the JVM, it can
+ * accomplish this with use of the usual invocation, field access, etc.
+ * instructions (e.g. {@code invokevirtual}, {@code getfield}). However, if the
+ * language is dynamic (hence, types of some expressions are not known until
+ * evaluated at run time), or its object model or type system don't match
+ * closely that of the JVM, then it should use {@code invokedynamic} call sites
+ * instead and let Dynalink manage them.
+ * <h2>Example</h2>
+ * Dynalink is probably best explained by an example showing its use. Let's
+ * suppose you have a program in a language where you don't have to declare the
+ * type of an object and you want to access a property on it:
+ * <pre>
+ * var color = obj.color;
+ * </pre>
+ * If you generated a Java class to represent the above one-line program, its
+ * bytecode would look something like this:
+ * <pre>
+ * aload 2 // load "obj" on stack
+ * invokedynamic "GET:PROPERTY:color"(Object)Object // invoke property getter on object of unknown type
+ * astore 3 // store the return value into local variable "color"
+ * </pre>
+ * In order to link the {@code invokedynamic} instruction, we need a bootstrap
+ * method. A minimalist bootstrap method with Dynalink could look like this:
+ * <pre>
+ * import java.lang.invoke.*;
+ * import jdk.dynalink.*;
+ * import jdk.dynalink.support.*;
+ *
+ * class MyLanguageRuntime {
+ * private static final DynamicLinker dynamicLinker = new DynamicLinkerFactory().createLinker();
+ *
+ * public static CallSite bootstrap(MethodHandles.Lookup lookup, String name, MethodType type) {
+ * return dynamicLinker.link(
+ * new SimpleRelinkableCallSite(
+ * new CallSiteDescriptor(lookup, parseOperation(name), type)));
+ * }
+ *
+ * private static Operation parseOperation(String name) {
+ * ...
+ * }
+ * }
+ * </pre>
+ * There are several objects of significance in the above code snippet:
+ * <ul>
+ * <li>{@link jdk.dynalink.DynamicLinker} is the main object in Dynalink, it
+ * coordinates the linking of call sites to method handles that implement the
+ * operations named in them. It is configured and created using a
+ * {@link jdk.dynalink.DynamicLinkerFactory}.</li>
+ * <li>When the bootstrap method is invoked, it needs to create a
+ * {@link java.lang.invoke.CallSite} object. In Dynalink, these call sites need
+ * to additionally implement the {@link jdk.dynalink.RelinkableCallSite}
+ * interface. "Relinkable" here alludes to the fact that if the call site
+ * encounters objects of different types at run time, its target will be changed
+ * to a method handle that can perform the operation on the newly encountered
+ * type. {@link jdk.dynalink.support.SimpleRelinkableCallSite} and
+ * {@link jdk.dynalink.support.ChainedCallSite} (not used in the above example)
+ * are two implementations already provided by the library.</li>
+ * <li>Dynalink uses {@link jdk.dynalink.CallSiteDescriptor} objects to
+ * preserve the parameters to the bootstrap method: the lookup and the method type,
+ * as it will need them whenever it needs to relink a call site.</li>
+ * <li>Dynalink uses {@link jdk.dynalink.Operation} objects to express
+ * dynamic operations. It does not prescribe how would you encode the operations
+ * in your call site, though. That is why in the above example the
+ * {@code parseOperation} function is left empty, and you would be expected to
+ * provide the code to parse the string {@code "GET:PROPERTY:color"}
+ * in the call site's name into a named property getter operation object as
+ * {@code StandardOperation.GET.withNamespace(StandardNamespace.PROPERTY).named("color")}.
+ * </ul>
+ * <p>What can you already do with the above setup? {@code DynamicLinkerFactory}
+ * by default creates a {@code DynamicLinker} that can link Java objects with the
+ * usual Java semantics. If you have these three simple classes:
+ * <pre>
+ * public class A {
+ * public String color;
+ * public A(String color) { this.color = color; }
+ * }
+ *
+ * public class B {
+ * private String color;
+ * public B(String color) { this.color = color; }
+ * public String getColor() { return color; }
+ * }
+ *
+ * public class C {
+ * private int color;
+ * public C(int color) { this.color = color; }
+ * public int getColor() { return color; }
+ * }
+ * </pre>
+ * and you somehow create their instances and pass them to your call site in your
+ * programming language:
+ * <pre>
+ * for each(var obj in [new A("red"), new B("green"), new C(0x0000ff)]) {
+ * print(obj.color);
+ * }
+ * </pre>
+ * then on first invocation, Dynalink will link the {@code .color} getter
+ * operation to a field getter for {@code A.color}, on second invocation it will
+ * relink it to {@code B.getColor()} returning a {@code String}, and finally on
+ * third invocation it will relink it to {@code C.getColor()} returning an {@code int}.
+ * The {@code SimpleRelinkableCallSite} we used above only remembers the linkage
+ * for the last encountered type (it implements what is known as a <i>monomorphic
+ * inline cache</i>). Another already provided implementation,
+ * {@link jdk.dynalink.support.ChainedCallSite} will remember linkages for
+ * several different types (it is a <i>polymorphic inline cache</i>) and is
+ * probably a better choice in serious applications.
+ * <h2>Dynalink and bytecode creation</h2>
+ * {@code CallSite} objects are usually created as part of bootstrapping
+ * {@code invokedynamic} instructions in bytecode. Hence, Dynalink is typically
+ * used as part of language runtimes that compile programs into Java
+ * {@code .class} bytecode format. Dynalink does not address the aspects of
+ * either creating bytecode classes or loading them into the JVM. That said,
+ * Dynalink can also be used without bytecode compilation (e.g. in language
+ * interpreters) by creating {@code CallSite} objects explicitly and associating
+ * them with representations of dynamic operations in the interpreted program
+ * (e.g. a typical representation would be some node objects in a syntax tree).
+ * <h2>Available operations</h2>
+ * Dynalink defines several standard operations in its
+ * {@link jdk.dynalink.StandardOperation} class. The linker for Java
+ * objects can link all of these operations, and you are encouraged to at
+ * minimum support and use these operations in your language too. The
+ * standard operations {@code GET} and {@code SET} need to be combined with
+ * at least one {@link jdk.dynalink.Namespace} to be useful, e.g. to express a
+ * property getter, you'd use {@code StandardOperation.GET.withNamespace(StandardNamespace.PROPERTY)}.
+ * Dynalink defines three standard namespaces in the {@link jdk.dynalink.StandardNamespace} class.
+ * To associate a fixed name with an operation, you can use
+ * {@link jdk.dynalink.NamedOperation} as in the previous example:
+ * {@code StandardOperation.GET.withNamespace(StandardNamespace.PROPERTY).named("color")}
+ * expresses a getter for the property named "color".
+ * <h2>Operations on multiple namespaces</h2>
+ * Some languages might not have separate namespaces on objects for
+ * properties, elements, and methods, and a source language construct might
+ * address several of them at once. Dynalink supports specifying multiple
+ * {@link jdk.dynalink.Namespace} objects with {@link jdk.dynalink.NamespaceOperation}.
+ * <h2>Language-specific linkers</h2>
+ * Languages that define their own object model different than the JVM
+ * class-based model and/or use their own type conversions will need to create
+ * their own language-specific linkers. See the {@link jdk.dynalink.linker}
+ * package and specifically the {@link jdk.dynalink.linker.GuardingDynamicLinker}
+ * interface to get started.
+ * <h2>Dynalink and Java objects</h2>
+ * The {@code DynamicLinker} objects created by {@code DynamicLinkerFactory} by
+ * default contain an internal instance of
+ * {@code BeansLinker}, which is a language-specific linker
+ * that implements the usual Java semantics for all of the above operations and
+ * can link any Java object that no other language-specific linker has managed
+ * to link. This way, all language runtimes have built-in interoperability with
+ * ordinary Java objects. See {@link jdk.dynalink.beans.BeansLinker} for details
+ * on how it links the various operations.
+ * <h2>Cross-language interoperability</h2>
+ * A {@code DynamicLinkerFactory} can be configured with a
+ * {@link jdk.dynalink.DynamicLinkerFactory#setClassLoader(ClassLoader) class
+ * loader}. It will try to instantiate all
+ * {@link jdk.dynalink.linker.GuardingDynamicLinkerExporter} classes visible to
+ * that class loader and compose the linkers they provide into the
+ * {@code DynamicLinker} it creates. This allows for interoperability between
+ * languages: if you have two language runtimes A and B deployed in your JVM and
+ * they export their linkers through the above mechanism, language runtime A
+ * will have a language-specific linker instance from B and vice versa inside
+ * their {@code DynamicLinker} objects. This means that if an object from
+ * language runtime B gets passed to code from language runtime A, the linker
+ * from B will get a chance to link the call site in A when it encounters the
+ * object from B.
*/
module jdk.dynalink {
requires java.logging;