CODEC-236: MurmurHash2 and MurmurHash3 implementations.

[melloware]

All MurmurHash implementations are public domain.

Credit was given in Javadoc where both implementations came from.

[coveralls]

Coverage increased (+0.4%) to 92.771% when pulling 67fb54d on melloware:CODEC-236 into 45a195e on apache:master.

[garydgregory]

@melloware ,
Thank you for your PR.
Code coverage has decreased, which is not what we want when adding new code. Please run mvn -V clean install -Pjacoco site to discover where you are missing code coverage in the new code.

[garydgregory]

Hm, XXHash32 and others implements java.util.zip.Checksum. Other classes in org.apache.commons.codec.digest do not implement Codec interfaces where some could, like UnixCrypt. Food for thought. But for the new code in this PR, the fit with the Codec interfaces might not be best.

The Travid build (https://travis-ci.org/apache/commons-codec/jobs/556975625) reveals that the Javadoc for the new code is incomplete:

[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:84: warning: no @param for l0
[ERROR] 	public static int hash32(long l0, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:84: warning: no @param for seed
[ERROR] 	public static int hash32(long l0, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:84: warning: no @return
[ERROR] 	public static int hash32(long l0, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:97: warning: no @param for l0
[ERROR] 	public static int hash32(long l0, long l1, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:97: warning: no @param for l1
[ERROR] 	public static int hash32(long l0, long l1, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:97: warning: no @param for seed
[ERROR] 	public static int hash32(long l0, long l1, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:97: warning: no @return
[ERROR] 	public static int hash32(long l0, long l1, int seed) {
[ERROR] 	                  ^
[ERROR] /home/travis/build/apache/commons-codec/src/main/java/org/apache/commons/codec/digest/MurmurHash3.java:275: warning: no @param for offset
[ERROR] 	public static long hash64(byte[] data, int offset, int length, int seed) {
[ERROR] 	                   ^

It would be best if you could address this as well as the decrease in code coverage.

[melloware]

@garydgregory Thanks for all your feedback. I went back and Javadoc'ed every method and param. I also was able to get Coveralls unit test coverage up 0.4%! When I grabbed the unit tests from Hive it looks like it had incomplete tests for 128 hashing so I created those tests.

[garydgregory]

One more comment: Did you consider providing a MurmurHash1 implementation?

[melloware]

I did consider MurmurHash1 but from what i can tell it was more of a prototype that led to the creation of Murmur2. From Wikipedia about Murmur1:

"The original MurmurHash was created as an attempt to make a faster function than Lookup3. Although successful, it hadn't been tested thoroughly and wasn't capable of providing 64-bit hashes as in Lookup3. It had a rather elegant design, that would be later built upon in MurmurHash2, combining a multiplicative hash (similar to Fowler–Noll–Vo hash function) with a Xorshift."