Is it possible to get the complete json from es using the following table structure. I am able to get the data when I define the schema for json data in hive table. But when I use a single field to get the json string I am getting NULL as output. I am able to use the same structure to write json data to ES.
It is a valid json. I am not sure if it is related to having a keyword like name , start as keys in the json string. Also I have fields with Capital letters in the field name like -> englishName. I am able to pull some of the fields from the beginning of the json when I give all the fields in my hive table.
If you are trying to return the docs from ES in JSON format, that is not supported. It shouldn't be hard to use though.
Your table configuration is confusing though - you define both an input (es.input.json) and a query (basically reading from it).
It's recommended to split the two - you'll end up with two different table that point out to the same index sure, but a table with an associated query means the data is filtered.
P.S. Defining a query that returns everything it's not just redundant, it's useless.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.