[DRILL-4594] Incorrect results when querying AVRO data with record and array in schema - ASF JIRA

XML

Word

Printable

JSON

When querying avro data, if the schema has a record before the array, then the results are incorrect.

Schema:
{ "c_record":

{"nested_c_string": "str0"}

, "d_array": [

{"nested_d_string": "str0"}

]}

Query:
select t.c_record, t.d_array[0].nested_d_string from dfs_test.`file.avro` t;

Expected Result:

{"nested_c_string":"str0"}

str0

Actual Result:
{"nested_c_string":"str0","d_array":[

{"nested_d_string":"str0"}

]} null

As you can see, the result is incorrect. This issue seems to happen only when a record is included before an array in the avro schema.

However, if the array is included before the record, then the query returns the correct result. Refer to the following schema:
{ "d_array": [

{"nested_d_string": "str0"}

], "c_record": {"nested_c_string": "str0"}}