Search code examples
hadooptwitterhivehql

Where is the syntax error on this simple Hive query with STRUCT?

Let's import a simple table in Hive:

hive> CREATE EXTERNAL TABLE tweets (id BIGINT, id_str STRING, user STRUCT<id:BIGINT, screen_name:STRING>)
ROW FORMAT SERDE 'org.apache.hadoop.hive.contrib.serde2.JsonSerde'
LOCATION '/projets/tweets';

OK
Time taken: 2.253 seconds

hive> describe tweets.user;

OK
id                      bigint                  from deserializer
screen_name             string                  from deserializer
Time taken: 1.151 seconds, Fetched: 2 row(s)

I cannot figure out where is the syntax error here:

hive> select user.id from tweets limit 5;
OK
Failed with exception java.io.IOException:org.apache.hadoop.hive.ql.metadata.HiveException: Error evaluating user.id
Time taken: 0.699 seconds

I am using the version 1.2.1 of Hive.

Solution

  • I finally found the answer. It seems it is a problem with the JAR used to serialize/deserialize the JSON. The default one (Apache) is not able to perform a good job on the data I have.

    I tried all these typical JAR (in parenthesis, the class for 'ROW FORMAT SERDE'):

    • hive-json-serde-0.2.jar (org.apache.hadoop.hive.contrib.serde2.JsonSerde)
    • hive-serdes-1.0-SNAPSHOT.jar (com.cloudera.hive.serde.JSONSerDe)
    • hive-serde-1.2.1.jar (org.apache.hadoop.hive.serde2.DelimitedJSONSerDe)
    • hive-serde-1.2.1.jar (org.apache.hadoop.hive.serde2.avro.AvroSerDe)

    All of them gave me different kinds of errors. I list them there so the next guy can Google them:

    • Failed with exception java.io.IOException:org.apache.hadoop.hive.ql.metadata.HiveException: Error evaluating user.id
    • java.lang.ClassCastException: org.json.JSONObject cannot be cast to [Ljava.lang.Object;
    • Failed with exception java.io.IOException:org.apache.hadoop.hive.ql.metadata.HiveException: java.lang.ClassCastException: java.lang.Integer cannot be cast to java.lang.Long Failed with exception
    • java.io.IOException:org.apache.hadoop.hive.serde2.SerDeException: DelimitedJSONSerDe cannot deserialize.
    • Failed with exception java.io.IOException:org.apache.hadoop.hive.serde2.avro.AvroSerdeException: Expecting a AvroGenericRecordWritable

    Finally, the working JAR is json-serde-1.3-jar-with-dependencies.jar which can be found here. This one is working with 'STRUCT' and can even ignore some malformed JSON. I have also to use for the creation of the table this class:

     ROW FORMAT SERDE 'org.openx.data.jsonserde.JsonSerDe'
 WITH SERDEPROPERTIES ("ignore.malformed.json" = "true")
 LOCATION ...

    If needed, it is possible to recompile it from here or here. I tried the first repository and it is compiling fine for me, after adding the necessary libs. The repository has also been updated recently.