Accessing Resources in a Location-Independent Manner

Motivation

Applets, Beans, Applications, and Libraries often have resource files associated with them. These Applets, Beans, and so on, want to use these files in a way that is independent of where these Applets, et al., are located.

Existing code in JDK1.0.x uses two types of mechanisms. The first mechanism is used in applets. Applet.getCodeBase() returns a URL to the base of the code for the applet. This base can be extended with a "relative path" to point to the desired resource, which can then be loaded (for example using Applet.getAudioClip(url)). The second mechanism is used by applications. Applications use "well known locations" (such as System.getProperty("user.home") or System.getProperty("java.home")). They add "/lib/" to the location and then open that file.

Until now, JDK1.0.x has not had a mechanism to locate resources that are independent of the code. That is, JDK1.0.x. has not had the means to locate resources for an applet loaded from the net using multiple http connects, or for an applet loaded using JAR files, or for a Bean loaded or a Bean installed in the CLASSPATH, or for a "library" installed in the CLASSPATH, and so on. The APIs described here provide such a mechanism.

The I18N APIs use this API as a primitive operation to locate ResourceBundles. See the latest I18N documentation for details.

Resources, names, and contexts

A resource is identified by a String. This String, while possibly empty, is a /-separated sequence of substrings, each a valid Java^TM Identifier, followed by a name of the form "<shortName>" or "<shortName>.<extension>". Both "shortName" and "extension" are composed of valid Java Letters and Numbers (section 3.8 in JLS). If the optional sequence exists, it is separated from the "shortName" by a /.

The name of a resource is independent of the Java implementation; in particular, the / is always used as a separator. However, the Java implementation controls the details of how the contents of the resource are mapped into a file, database, or other object containing the actual resource.

The interpretation of a resource name is relative to a ClassLoader instance. Methods implemented by the ClassLoader do this interpretation.

System Resources

A system resource is similar to a system class (section 20.14.5 of the JLS). A system resource is a resource that is either built-in to the system, or it is kept by the host implementation in, for example, a local file system. System resources are accessed through special methods (getSystemResource and getSystemResourceAsStream) that consult the base host implementation.

For example, in a particular implementation, locating a system resource may involve searching the entries in the CLASSPATH. Each directory, zip file, or jar file entry in the CLASSPATH is searched for the resource file, and, if found, either an InputStream, or its name, is returned. If not found, null is returned. Note that a resource may be found in a different entry in the CLASSPATH than where the class file was loaded.

Non-System Resources

The implementation of getResource on a given ClassLoader will depend on the details of the ClassLoader. For example AppletClassLoader will:

First try to locate the resource as a system resource; then, if not found,
Search through the resources in ARCHIVES (JAR files) already loaded in this CODEBASE; then, if not found,
Use CODEBASE and attempt to locate the resource (which may involve contacting a remote site).

All ClassLoaders will search for a resource first as a system resource, in a manner analogous to searcing for class files. This search rule permits overwriting locally any resource. Clients should choose a resource name that will be unique (using the company or package name as a prefix, for instance).

Resource Names

A common convention for the name of a resource used by a class is to use the fully qualified name of the package of the class, convert all "." to "/", and add a resource name of the form "<Name>.<ext>". To support this, and to simplify handling the details of system classes (for which getClassLoader returns null), the class Class provides two convenience methods that call the appropriate methods in ClassLoader.

The resource name given to a Class method may have an initial starting "/" that identifies it as an "absolute" name. Resource names that do not start with a "/" are "relative".

Absolute names are stripped of their starting "/" and are passed, without any further modification, to the appropriate ClassLoader method to locate the resource. Relative names are modified according to the convention described previously and then are passed to a ClassLoader method.

Manipulating Resources

The method getResource() returns a URL for the resource. The URL (and its representation) is implementation and JVM-instance specific (the URL obtained in one JVM instance may not work in another) and may vary depending on the implementation details (it may also change between JDK1.1 and JDK1.1.1). Its protocol is (usually) specific to the ClassLoader loading the resource. If the resource does not exist, a null will be returned; if the resource is not visible due to security considerations, a null will also be returned.

If the client code wants to read the contents of the resource as an InputStream, it can apply the openStream() method on the url. This is common enough to justify adding getResourceAsStream() to Class and ClassLoader. getResourceAsStream() is semantically identical to getResource().openStream(), except that IO exceptions are caught and returned as a null InputStream.

The client code code can also request the contents of the resource as an object by applying the getContent() method on the url. This is useful when the resource contains the data for an image, for instance. Note that in this case, the result is an awt.image.ImageProducer object, not an Image object.

API Additions to Class

Specifically, the class Class methods are of the following form:


class Class {

    /**
     * Find a resource with a given name.  Will return null if no
     * resource with this name is found.  The rules for searching a
     * resources associated with a given class are implemented by the
     * ClassLoader of the class.
     *
     * The Class methods delegate to ClassLoader methods, after applying
     * a naming convention: if the resource name starts with "/", it is used
     * as is.  Otherwise, the name of the package is prepended, after
     * converting "." to "/".
     *
     * @see java.lang.ClassLoader
     */
    public InputStream getResourceAsStream(String name) {
	name = resolveName(name);
	ClassLoader cl = getClassLoader();
	if (cl==null) {
	    // A system class.
	    return ClassLoader.getSystemResourceAsStream(name);
	}
	return cl.getResourceAsStream(name);
    }

    public java.net.URL getResource(String name) {
	name = resolveName(name);
	ClassLoader cl = getClassLoader();
	if (cl==null) {
	    // A system class.
	    return ClassLoader.getSystemResource(name);
	}
	return cl.getResource(name);
    }

    /**
     * Add a package name prefix if the name is not absolute
     * Remove leading "/" if name is absolute
     */
    private String resolveName(String name) {
	if (name == null) {
	    return name;
	}
	if (!name.startsWith("/")) {
	    Class c = this;
	    while (c.isArray()) {
		c = c.getComponentType();
	    }
	    String baseName = c.getName();
	    int index = baseName.lastIndexOf('.');
	    if (index != -1) {
		name = baseName.substring(0, index).replace('.', '/')
		    +"/"+name;
	    }
	} else {
	    name = name.substring(1);
	}
	return name;
    }

Note that it is possible, albeit somewhat uncommon, to have two classes in two diffent packages sharing the same resource.

API Additions to ClassLoader

We provide two sets of methods to access a resource. One set returns an InputStream on the resource. The other set returns a URL. The methods that return an InputStream are somewhat easier to use and will satisfy many needs, while the methods that return URLs provide access to more complex information, such as an Image and an AudioClip.

Resources are managed through ClassLoaders in a manner analogous to classes. A ClassLoader controls how to map the name of a resource to its content. ClassLoader also provides methods for accessing system resources, analogous to the system classes (and which have no ClassLoader in JDK1.0.x and JDK1.1). Class Class provides some convenience methods that delegate functionality to the appropriate ClassLoader methods.

Many Java programs will access these methods indirectly through the I18N APIs. Others will access it through methods in class Class. A few will directly invoke the ClassLoader methods.

The methods in ClassLoader use the given String as the name of the resource without applying any absolute/relative transformation (cf. the methods in Class). The name should not have a leading "/".


class ClassLoader {

    /**
     * A resource is some data (images, audio, text, etc) that wants to be
     * accessed by some class code in a way that is independent of the
     * location of the code.  Resources are found with cooperation of the
     * class loaders, since they are the only ones who know where the class
     * actually came from. 
     *
     * System resources are those that are handled by the host implemenation
     * directly.  For example, they may be located in the CLASSPATH.

     *
     * The name of a resource is a "/"-separated sequence of identifiers.
     * The class Class provides convenience methods for accessing resources;
     * the methods implement a convention where the package name is prefixed
     * to the short name of the resource.

     * 
     * Resources can be accessed as an InputStream, or as a URL.
     *
     * @see	Class
     */

    /**
     * Get an InputStream on a given resource..  Will return null if no
     * resource with this name is found. 

     *
     * The resource name may be any system resource (e.g. follows CLASSPATH order)
     * @param	name	the name of the resource, to be used as is.
     * @return		an InputStream on the resource, or null if not found.
     */

    public static final InputStream getSystemResourceAsStream(String name) {
	... this is equivalent to getSystemResource() call plus a openStream()
    }

    /**
     * Find a resource with a given name.  The return is a URL to the resource
     * Doing a getContent() on the URL may return an ImageProducer, an AudioClip, or
     * an InputStream.

     *
     * The resource name may be any system resource (e.g. follows CLASSPATH order)
     * @param	name	the name of the resource, to be used as is.
     * @return		the URL on the resource, or null if not found.
     */
    public static final java.net.URL getSystemResource(String name) {
	...
    }

    /**
     */

    /**
     * Get an InputStream on a given resource.  Will return null if no
     * resource with this name is found. 

     *
     * The class loader can choose what to do to locate the resource.
     * @param	name	the name of the resource, to be used as is.
     * @return		an InputStream on the resource, or null if not found.
     */
    public InputStream getResourceAsStream(String name) {
	return null;
    }

    /**
     * Find a resource with a given name.  The return is a URL to the resource.
     * Doing a getContent() on the URL may return an ImageProducer, an AudioClip,
     * or an InputStream.
     *
     * The class loader can choose what to do to locate the resource.
     * @param	name	the name of the resource, to be used as is.
     * @return		an InputStream on the resource, or null if not found.
     */
    public java.net.URL getResource(String name) {
	return null;
    }

    }

Client code

Below are two examples of client code. The first example uses "absolute resource" names and traditional mechanisms to get a class Class object:


package pkg;

import java.io.IOException;
import java.io.InputStream;
import java.io.PrintStream;

class Test {

    private static final String absName = "/pkg/mumble.baf";

    public static void test1() {
	Class c=null;
	try {
	    c = Class.forName("pkg.Test");
	} catch (Exception ex) {
	    // This should not happen.
	}
	InputStream s = c.getResourceAsStream(absName);
	// do something with it.
    }

    public void test2() {
	InputStream s = this.getClass().getResourceAsStream(absName);
	// do something with it.
    }

The second example uses "relative resource" names and the new mechanism, available from the compiler through the -experimental flag, to get a class Class object:


package pkg;

import java.io.IOException;
import java.io.InputStream;
import java.io.PrintStream;

class Test {

    private static final String relName = "mumble.baf";

    public static void test1() {
	InputStream s = Test.class.getResourceAsStream(relName);
	// do something with it.
    }

    public void test2() {
	InputStream s = Test.class.getResourceAsStream(relName);
	// do something with it.
    }

Security Details

Since getResource() provides access to information, it must have well defined and well founded security rules that at the same time support the intended use of this mechanism. Below we describe the exact details, as specified and implemented in releases starting at version 1.1.5 of the Java Development Kit. The semantics are described only for ClassLoader.getResource and ClassLoader.getSystemResource() and extend to the AsStream methods as defined in the previous section.

If security considerations do not allow a resource to be visible in some security context, the getResource() method will fail (will return null) as if the resource was not present at all, this addresses existence attacks.

All classloaders will not provide access to the contents of a .class file. This is for both security and performance issues. Whether it is possible to obtain a URL into a .class file depends on the specifics, as shown below.

There are no specified security issues or restrictions regarding resources that are found by a non-system class loader. AppletClassLoader provides access to information that is loaded from some source location, either individually, or in a group through a JAR file; thus AppletClassLoader should apply the same checkConnect() rules when dealing with URLs through getResource().

The system Class Loader provides access to information in the CLASSPATH. A CLASSPATH may have directories and/or JAR files. Since a JAR file is created intentionally, we ascribe a different significance to it than in a directory where things may end up in a more casual manner. In particular, we are more strict on getting information out of a directory than out from a JAR file.

If the resource is in a Directory:

getResource() invocations will use File.exists() to determine whether to make the corresponding file visible to the user. Recall that File.exists() uses the checkRead() method in the security manager.
the same applies to getResourceAsStream().

If the resource is in a JAR file:

getResource() invocations will succeed for all files, regardless of whether the invocation is done from within a System or a non-System class.
getResourceAsStream() invocations will succeed for non .class resources, and so will for getContent() on corresponding URLs.