I am not sure about Pig, but its easily achievable in MapReduce. We had a
similar requirement, we had to convert logs from RFC syslog format (5424)
into JSON. We have a MR job which does this for us. The reason why we chose
MR was mainly for Error Handling - like missing fields in some records,
rem
i am not sure if you need a custom loader .. you could read this as a comma
separated string into individual fields ..
convert the last field into a map data type (using a UDF may be.. ) ..
but if you still want to persue custom loader you can probably take hint
from
https://svn.apache.org/repos/a
