java - Strange behavior of Class.getResource() and ClassLoader.getResource() in executable jar

Question

I understand from What is the difference between Class.getResource() and ClassLoader.getResource()? and from own code, that

getClass().getResource("/path/image.png")

is identical to

getClass().getClassLoader().getResource("path/image.png")

The posting Cannot read an image in jar file shows an issue where using

getClass().getClassLoader().getResource("path/image.png")

in an executable jar file returns null, while

getClass().getResource("/path/image.png")

returns the correct URL.

Since Class.getResource() delegates to ClassLoader.getResource() after removing the leading slash, I would expect that these calls are identical, but obviously they are not in this case. Even when a special class loader is attached to the particular class, it should still be the same one for each call, again resulting in the same behavior.

So, the question is: are there any obvious circumstances under which the following code returns null for the first call but the proper URL for the second call?

package com.example;

import java.net.URL;

public class ResourceTest {

   public void run() {
      URL iconUrl1 = getClass().getClassLoader().getResource("path/image.png");
      System.out.println("ClassLoader.getResource("path/image.png"): " + iconUrl1);

      URL iconUrl2 = getClass().getResource("/path/image.png");
      System.out.println("Class.getResource("/path/image.png"): " + iconUrl2);
   }

   public static void main(String[] args) {
      ResourceTest app = new ResourceTest();
      app.run();
   }
}

Answer 1

I thought this question was already asked and answered!

• What is the difference between Class.getResource() and ClassLoader.getResource()?

getClass().getResource() searches relative to the .class file while getClass().getClassLoader().getResource() searches relative to the classpath root.

If there's an SSCCE here, I don't understand why it doesn't

1) Show the directory organization in the .jar, and...

2) Take package into consideration

Q: What (if anything) hasn't already been answered by What is the difference between Class.getResource() and ClassLoader.getResource()? (and the links it cites)?

=========================================================================

I'm still not sure what isn't clear, but this example might help:

/*
  SAMPLE OUTPUT:
  ClassLoader.getResource(/subdir/readme.txt): NULL
  Class.getResource(/subdir/readme.txt): SUCCESS

  ClassLoader.getResource(subdir/readme.txt): SUCCESS
  Class.getResource(subdir/readme.txt): NULL
 */
package com.so.resourcetest;

import java.net.URL;

public class ResourceTest {

    public static void main(String[] args) {
        ResourceTest app = new ResourceTest ();
    }

    public ResourceTest () {
        doClassLoaderGetResource ("/subdir/readme.txt");
        doClassGetResource ("/subdir/readme.txt");
        doClassLoaderGetResource ("subdir/readme.txt");
        doClassGetResource ("subdir/readme.txt");
    }

    private void doClassLoaderGetResource (String sPath) {
        URL url  = getClass().getClassLoader().getResource(sPath);
        if (url == null)
            System.out.println("ClassLoader.getResource(" + sPath + "): NULL");
        else
            System.out.println("ClassLoader.getResource(" + sPath + "): SUCCESS");
    }

    private void doClassGetResource (String sPath) {
        URL url  = getClass().getResource(sPath);
        if (url == null)
            System.out.println("Class.getResource(" + sPath + "): NULL");
        else
            System.out.println("Class.getResource(" + sPath + "): SUCCESS");
    }
}

Here's the corresponding directory tree. It happens to be an Eclipse project, but the directories are the same regardless if it's Eclipse, Netbeans ... or a .jar file:

C:.
├───.settings
├───bin
│   ├───com
│   │   └───so
│   │       └───resourcetest
│   └───subdir
└───src
    ├───com
    │   └───so
    │       └───resourcetest
    └───subdir

The file being opened is "subdir/readme.txt"

---

ADDENDUM 11/9/12:

Hi -

I copied your code verbatim from github, re-compiled and re-ran:

ClassLoader.getResource(/subdir/readme.txt): NULL
Class.getResource(/subdir/readme.txt): SUCCESS
ClassLoader.getResource(subdir/readme.txt): SUCCESS
Class.getResource(subdir/readme.txt): NULL

If that's not the output you're getting ... I'm baffled.

For whatever it's worth, I'm running:

• Eclipse Indigo (it shouldn't matter)

• Running inside the IDE (it shouldn't matter if it's filesystem or .jar, inside or outside an IDE)

• My JRE is 1.6 (if anything, this is probably the biggie)

Sorry we haven't been able to resolve what I thought was a straightforward issue :(

---

ADDENDUM 11/21/12 (Andreas):

Since there was no recent activity on this question, I would like to summarize what we found:

• From our common understanding, the answer to the above question is: "No, it is not possible that Class.getResource("/path/image.png") returns a valid URL, while ClassLoader.getResource("path/image.png") returns null":
  • We're completely clear on the difference between ClassLoader.getResource() and Class.getResource()
  • Our sample outputs match, for both "SUCCESS" and for "null"
  • The sample outputs match what we'd expect
  • Conclusion: Either we oversaw something, or something different caused the "solution" described in the linked question to work. I think we can not currently prove one or the other.