Add RandomAccessFile digest methods

[behrangsa]

Added two new methods:

• digest(final MessageDigest messageDigest, final RandomAccessFile data)
• updateDigest(final MessageDigest digest, final RandomAccessFile data)

For computing digest of RandomAccessFiles.

Performance-wise there's almost no improvements offered by these new NIO based methods, but being non-blocking they could potentially use fewer resources and provide better throughput in highly multi-threaded scenarios.

[coveralls]

Coverage increased (+0.01%) to 93.339% when pulling 8ee6510 on behrangsa:master into 3ab9ce4 on apache:master.

[behrangsa]

Not sure why the coveralls is reporting that the new methods are not tested. Tests have been added for both of them.

[aherbert]

Your tests do not call the new methods.

[aherbert]

This should call nonblockingDigest with getTestRandomAccessFile()

[aherbert]

This should call nonblockingDigest.

[aherbert]

I am also not sure why coveralls dropped. Looking at the Travis log for JDK 8 it runs the tests:

[INFO] Running org.apache.commons.codec.digest.MessageDigestAlgorithmsTest
[WARNING] Tests run: 88, Failures: 0, Errors: 0, Skipped: 32, Time elapsed: 4.816 s - in org.apache.commons.codec.digest.MessageDigestAlgorithmsTest

When I run locally mvn clean test -Dtest=MessageDigestAlgorithmsTest jacoco:report I get the same numbers of tests run (those skipped are for algorithms in later JDKs) and the local jacoco report has coverage of the new methods.

If you fix the new tests to make a duplicate call to the new NIO methods and push it will run again and we can see if this is some strange glitch in coveralls.

[behrangsa]

Silly me. Thanks for the review.

[behrangsa]

@aherbert I fixed those issues, but for some reason coveralls reports that they are not covered yet:

    @Test
    public void testNonBlockingDigestRandomAccessFile() throws IOException {
        Assume.assumeTrue(DigestUtils.isAvailable(messageDigestAlgorithm));

        final byte[] expected = digestTestData();

        Assert.assertArrayEquals(expected,
                DigestUtils.nonblockingDigest(
                        DigestUtils.getDigest(messageDigestAlgorithm), getTestRandomAccessFile()
                )
        );

        Assert.assertArrayEquals(expected,
                DigestUtils.nonblockingDigest(
                        DigestUtils.getDigest(messageDigestAlgorithm), getTestRandomAccessFile()
                )
        );
    }

    @Test
    public void testNonBlockingDigestFile() throws IOException {
        Assume.assumeTrue(DigestUtils.isAvailable(messageDigestAlgorithm));

        final byte[] expected = digestTestData();

        Assert.assertArrayEquals(
                expected,
                DigestUtils.nonblockingDigest(
                        DigestUtils.getDigest(messageDigestAlgorithm), getTestFile()
                )
        );

        Assert.assertArrayEquals(
                expected,
                DigestUtils.nonblockingDigest(
                        DigestUtils.getDigest(messageDigestAlgorithm), getTestFile()
                )
        );
    }

[aherbert]

The failure is due to the post travis action to build the coveralls report. This command:

mvn clean cobertura:cobertura coveralls:report -Ptravis-cobertura

Fails to run the test:

[INFO] Running org.apache.commons.codec.digest.MessageDigestAlgorithmsTest
[ERROR] Tests run: 1, Failures: 0, Errors: 1, Skipped: 0, Time elapsed: 0.001 s <<< FAILURE! - in org.apache.commons.codec.digest.MessageDigestAlgorithmsTest
[ERROR] org.apache.commons.codec.digest.MessageDigestAlgorithmsTest  Time elapsed: 0.001 s  <<< ERROR!
java.lang.ClassCastException: [I cannot be cast to java.lang.String
	at org.apache.commons.codec.digest.MessageDigestAlgorithmsTest.checkValues(MessageDigestAlgorithmsTest.java:71)

This is a problem on master that needs fixing. It is not related to your PR.

[aherbert]

I fixed master to work with cobertura. If you rebase you should get coveralls reporting.

[garydgregory]

I am fine with adding more methods to handing more inputs but I am -1 to putting implementation details like "NIO" in and "nonblocking" in method names.

[garydgregory]

Let's step back and ask: Why would we want two implementations for File objects? If NIO does the job better, then let's change the implementation to do that.

[behrangsa]

Let's step back and ask: Why would we want two implementations for File objects? If NIO does the job better, then let's change the implementation to do that.

I thought using NIO would improve performance but in my limited tests I didn't notice a meaningful difference. It would be great if some NIO experts could chime in.

However we can:

• Get rid of nonblockingDigest(final MessageDigest messageDigest, final File data)
• Rename nonblockingDigest(final MessageDigest messageDigest, final RandomAccessFile data) to digest(final MessageDigest messageDigest, final RandomAcessFile data)

as the first method calls wraps the File in a RandomAccessFile and passes it to the 2nd method anyway.

[garydgregory]

xref https://issues.apache.org/jira/browse/CODEC-272