We have observed that when we store the data in Avro format, it converts byte stream to binary, due to which all the comments gets converted to “from deserializer”.
We found a jira bug for this issue as well, few confirms, this issue has been addressed with 0.13 version. We are using hive 1.1 (Cloudera). But we are still facing the issue.
Jira :- https://issues.apache.org/jira/browse/HIVE-6681
But when we change the input and output format to normal text (specified explicitly), column description can be retained, however, it seems it is losing on its actual avro functionality in such a case. So the below code cannot be used.
-- Below is input and output format using text
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe'
STORED AS INPUTFORMAT 'org.apache.hadoop.mapred.TextInputFormat'
OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'