An Apache Pig UDF to transform milliseconds to a Parquet INTERVAL data type (NanoTime).

An example pig script that convert Avro to Parquet:

-- parameters required:
-- IN  = Avro input glob to be read from HDFS
-- OUT = Target Parquet file

SET job.name 'avro_to_parquet';

REGISTER /srv/magnetic_hadoop-new/lib/external/avro-1.7.4.jar
REGISTER /srv/magnetic_hadoop-new/lib/external/json-simple-1.1.jar
REGISTER /srv/magnetic_hadoop-new/lib/external/jackson-core-asl-1.8.8.jar
REGISTER /srv/magnetic_hadoop-new/lib/external/jackson-mapper-asl-1.8.8.jar
REGISTER /srv/magnetic_hadoop-new/lib/external/magnetic_avrostorage-0.0.3-SNAPSHOT.jar
REGISTER /srv/stanislav/magnetic_hadoop.git/lib/external/nano-time-udf-0.0.1-SNAPSHOT.jar

DEFINE toNanoTimeBinary com.magnetic.nano_time_udf.NanoTimeBinary;

in = LOAD '$IN'
	USING com.magnetic.pig.storage.avro.UnorderedSparseSchemaAvroStorage();

in_nanotime = FOREACH in GENERATE *, toNanoTimeBinary(log_timestamp) as log_nanotime;

RMF $OUT;

out = STORE in_nanotime INTO '$OUT'
	USING parquet.pig.ParquetStorer();