Testlib: Add test that putIfAbsent replaces a null value

[ben-manes]

This was shown in a recent Java Collections Puzzlers. The JavaDoc for Map.putIfAbsent states the following.

If the specified key is not already associated with a value (or is mapped to {@code null}) associates it with the given value and returns {@code null}, else returns the current value.

@Test
public void putIfAbsent_nullExistingValue() {
  var map = new HashMap<Object, Object>();
  map.put("a", null);
  map.putIfAbsent("a", "b");
  assertThat(map).containsEntry("a", "b");
}

I found that multiple custom maps which permit null values do not honor this peculiarity (and similar for computeIfAbsent). For example fastutils' adopted the testlib and Object2ObjectOpenHashMap passes its suite, but it mistakenly does not replace the value as shown above.

[kevinb9n]

Thanks Ben! Looks worth adding. If you or anyone else feels like making the PR we can merge it. If not, I hope we'd still get around to it sooner or later.

[alvonellos]

@kevinb9n what would be the best way for me to get started contributing? This looks like a good first issue.

[ben-manes]

Take a look at MapPutIfAbsentTester. These cases would be to allow null values so follow the pattern elsewhere for defining test matching constraints, e.g.

guava/guava-testlib/src/com/google/common/collect/testing/testers/MapPutTester.java

Lines 180 to 184 in b337be6

	@MapFeature.Require({SUPPORTS_PUT, ALLOWS_NULL_VALUES})
	public void testPut_nullValueSupported() {
	assertNull("put(key, null) should return null", put(nullValueEntry));
	expectAdded(nullValueEntry);
	}

It would also probably be good to review the default methods promoted from ConcurrentMap (e.g. replace) and the newer ones like computes and merge to see if the ALLOWS_NULL_VALUES cases are being covered. The tests should be run by OpenJdk6MapTests which includes JDK maps that support or disallow null values, so verifying that yours works as expected means to see that your new tests intentionally fail and then go green when implemented. Otherwise, simply try to stay within the coding style of the surrounding code.

[alvonellos]

Hey ben, I made the code change but I'm running into some issues getting guava to build on my system. I get an error: package sun.misc does not exist

any tips?

[ben-manes]

I believe it should work, but I haven't run Guava's build for a long time. I know that Eclipse defaults to that error for restricted apis, if that is what you mean? If so, you can disable it under,

[ben-manes]

I gave it a build and another likely reason is if you are running on recent JDK which requires access modifiers. It worked for me on Java 11, and I believe the library targets Java 8.

[alvonellos]

Take a look at my PR: #6225

[ben-manes]

Looks like a good first iteration. Of course that doesn't cover this scenario.

I don't have commit access to the repository, so you'll need to have a champion on the Guava team marshall it through.

[alvonellos]

I'm using IntelliJ. I'm having a hard time getting it to build so I can test that it all works. I signed the CLA. This will be my first contribution to a google project

[ben-manes]

You can use command line, e.g. mvn test -Dtest=OpenJdk6MapTests -pl guava-testlib.