How to FILTER Cassandra TimeUUID/UUID in Pig

问题

Here is my Cassandra schema, using Datastax Enterprise

CREATE KEYSPACE applications
  WITH replication = {'class': 'SimpleStrategy', 'replication_factor' : 1};

USE applications;

CREATE TABLE events(
  bucket text, 
  id timeuuid,
  app_id uuid,  
  event text, 
  PRIMARY KEY(bucket, id)
);

I want to FILTER in PIG by app_id (TimeUUID) and id (UUID), here is my Pig script.

events = LOAD 'cql://applications/events'
  USING CqlStorage()
  AS (bucket: chararray, id: chararray, app_id: chararray, event: chararray);

result = FOREACH events GENERATE bucket, id, app_id;
DESCRIBE result;
DUMP result;

Here is the result

result: {bucket: chararray,id: chararray,app_id: chararray}
(2014-02-28-04,?O]??4??p??M?,;??F? (|?Mb) \n
(2014-02-28-04,?O??4??p??M?,?h^@?E????)
(2014-02-28-04,?V???4??p??M?,;??F? (|?Mb)
(2014-02-28-04,?W?0?4??p??M?,?h^@?E????)
(2014-02-28-04,?X^p?4??p??M?,?h^@?E????)

Notice, the app_id, and id fields are binary and I need to filter by some UUID, any suggestions?

回答1:

You need use a UDF to convert the binary bytes of UUID/TimeUUID to chararray. Don't try to define it as chararray directly like AS (bucket: chararray, id: chararray, app_id: chararray, event: chararray);

Or you can use https://github.com/cevaris/pig-dse/blob/master/src/main/java/com/dse/pig/udfs/AbstractCassandraStorage.java which convert UUID/TimeUUID to String

File a Cassandra ticket if you think it's good to convert UUID to string as default.

回答2:

Was able to solve using a custom version of CQLStorage UDF Pig Loader. The issue was that Cassandras CQLStorage/CassandraStorage Pig loaders do not know how to handle UUID/TimeUUID types. I assume this happens for most non-standard data types. Anyways, here is the link to the coded solution on github.

https://github.com/cevaris/pig-dse

来源：https://stackoverflow.com/questions/22159722/how-to-filter-cassandra-timeuuid-uuid-in-pig

标签

cassandra

apache-pig

datastax-enterprise