One nice way to do this stuff is using a special SERDE, possible like the JsonSerde:
A simpler scenario, where you have to load a multidelimiter CSV file, can be addressed using the RegexSerde : which maps columns to each group matching. In your case, the JsonSerde could be used essentially to match your fields and map them to columns. For a simple example of how to use Regex Serdes to do the same for CSV files: check out https://github.com/jayunit100/bigpetstore/blob/master/src/main/java/org/bigtop/bigpetstore/etl/HiveETL.java I *think* JsonSerde comes with HCatalog now, so if you dont have it on your classpath, you can easily add that jar into hadoop/lib and then all your mappers will be able to see your json data as tabular hive data.
