[HIVE-9962] JsonSerDe does not support reader schema different from data schema

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Component/s: HCatalog, Serializers/Deserializers
• Assignee: Naveen Gangam
• Reporter: Johndee Burks
• Created: 13/Mar/15 19:10
• Updated: 18/Mar/15 14:45
• Resolved: 18/Mar/15 14:45

Description

To reproduce the limitation do the following.

Create a two tables the first with full schema and the second with partial schema.

add jar /opt/cloudera/parcels/CDH/lib/hive-hcatalog/share/hcatalog/hive-hcatalog-core.jar;

CREATE TABLE json_full
(autopolicy struct<is_active:boolean, policy_holder_name:string, policy_num:string, vehicle:struct<brand:struct<model:string, year:int>, price:double, vin:string>>)
ROW FORMAT SERDE 'org.apache.hcatalog.data.JsonSerDe';

CREATE TABLE json_part 
(autopolicy struct<is_active:boolean, policy_holder_name:string, policy_num:string, vehicle:struct<brand:struct<model:string, year:int>, price:double>>)
ROW FORMAT SERDE 'org.apache.hcatalog.data.JsonSerDe';

The data for the table is below:

{"autopolicy": {"policy_holder_name": "someone", "policy_num": "20141012", "is_active": true, "vehicle": {"brand": {"model": "Lexus", "year": 2012}, "vin": "RANDOM123", "price": 23450.50}}}

I put that data into a file and load it into the tables like this:

load data local inpath 'data.json' into table json_full;
load data local inpath 'data.json' into table json_part;

Then do a select against each table:

select * from json_full;
select * from json_part;

The second select should fail with an error simlar to that below:

15/03/12 23:19:30 [main]: ERROR CliDriver: Failed with exception java.io.IOException:java.lang.NullPointerException

The code that throws this error is below:

172 private void populateRecord(List<Object> r, JsonToken token, JsonParser p, HCatSchema s) throws IOException { 
173 if (token != JsonToken.FIELD_NAME) { 
174 throw new IOException("Field name expected"); 
175 } 
176 String fieldName = p.getText(); 
177 int fpos; 
178 try { 
179 fpos = s.getPosition(fieldName); 
180 } catch (NullPointerException npe) { 
181 fpos = getPositionFromHiveInternalColumnName(fieldName); 
182 LOG.debug("NPE finding position for field [{}] in schema [{}]", fieldName, s); 
183 if (!fieldName.equalsIgnoreCase(getHiveInternalColumnName(fpos))) { 
184 LOG.error("Hive internal column name {} and position " 
185 + "encoding {} for the column name are at odds", fieldName, fpos); 
186 throw npe; 
187 } 
188 if (fpos == -1) { 
189 return; // unknown field, we return. 
190 } 

Issue Links

• duplicates: HIVE-6166 JsonSerDe is too strict about table schema