Create GeoWave Subproject

[jamesmcclain]

These changes add a GeoWave subproject, which includes a GeowaveAttributeStore, GeowaveLayerReader, and GeowaveLayerWriter.

• Missing metadata exception
• Reduction or elimination of use of asInstanceOf
• Fix geotools Tests
• Find work-arounds for bugs
• Unit Tests
• GridCoverage2D Kryo Serializer
• Address Feedback
• Hillshade Demo

Tests can be run locally and they pass, but there appears to be insufficient memory to bring up GeoWave on Travis:

OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x0000000768680000, 51904512, 0) failed; error='Cannot allocate memory' (errno=12)
#
# There is insufficient memory for the Java Runtime Environment to continue.
# Native memory allocation (malloc) failed to allocate 51904512 bytes for committing reserved memory.
# An error report file with more information is saved as:
# /home/travis/build/geotrellis/geotrellis/hs_err_pid4764.log
+exit 1
The command ".travis/build-and-test.sh" exited with 1.
cache.2
store build cache
0.00s
2.98snothing changed, not updating cache
Done. Your build exited with 1.

[lossyrob]

In GeoWave, is it always 1 index == 1 tile? If so, then you're not really ignoring filter index only, it's just by default doing either true or false behavior. filterIndexOnly is for when the index has more than one tile stored in the values, so that when we are updating tile entries in a backend we can update the tile information while not losing the surrounding tile info.

[jamesmcclain]

If I understand your usage of the word "index" above, then I believe there is one index (one address) for one tile and one tile for one address. Thanks for the clarification on flterIndexOnly.

[lossyrob]

A more safe way to write this would be

  val scanner = connector.createScanner(attributeTable, new Authorizations())
  try {
   scanner.iterator 
        .map { kv =>
          val List(name, zoomStr) = kv.getKey.getRow.toString.split(SEP).toList
          LayerId(name, zoomStr.toInt)
        }
        .toList
        .distinct
  } finally {
    scanner.close()
  }

The rule generally followed with parans or no parans on a method that takes no arguments is that, if the method has side effects, include the parans so that it would never be mistaken for a val

[pomadchin]

@lossyrob I can takesimilar changes into account for hbase backend too

[jamesmcclain]

Well taken, it looks like Eugene addressed this one.

[lossyrob]

Merging!

[jamesmcclain]

😄