[DRILL-1818] Parquet files generated by Drill ignore field names when nested elements are queried

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Fix Version/s: 0.7.0
• Component/s: Storage - Writer
• Assignee: Steven Phillips
• Reporter: Neeraja
• Created: 07/Dec/14 06:35
• Updated: 08/May/15 21:22
• Resolved: 07/Dec/14 23:08

Description

I observed this with this parquet file and a more comprehensive testing might be needed here. The issue is that Drill seem to simply ignore field names at the leaf level and accessing data in a positional fashion.

Below is the repro.
1. Generate a parquet file using Drill. Input is the JSON doc below

create table dfs.tmp.sampleparquet as (select trans_id, cast(`date` as date) transdate,cast(`time` as time) transtime, cast(amount as double) amount,`user_info`,`marketing_info`, `trans_info` from dfs.`/Users/nrentachintala/Downloads/sample.json` )

2. Now do queries.
Note in query below, there is no field name called 'keywords' in trans_info, but data is just positionally returned (the data returned from prod_id column).
0: jdbc:drill:zk=local> select t.`trans_info`.keywords from dfs.tmp.sampleparquet t where t.`trans_info`.keywords is not null;
------------

EXPR$0

------------

[16]

[]

[293,90]

[173,18,121,84,115,226,464,525,35,11,94,45]

[311,29,5,41]

0: jdbc:drill:zk=local> select t.`marketing_info`.keywords from dfs.tmp.sampleparquet t;

Note in the query below, it is trying to return the first element in marketing_Info which is camp_id which is of int type for keywords columns. But keywords schema is string, so it throws error with type mismatch.

Query failed: Query failed: Failure while running fragment., You tried to write a VarChar type when you are using a ValueWriter of type NullableBigIntWriterImpl. [ c3761403-b8c5-43c1-8e90-2c4918d1f85c on 10.0.0.20:31010 ]
[ c3761403-b8c5-43c1-8e90-2c4918d1f85c on 10.0.0.20:31010 ]

Error: exception while executing query: Failure while executing query. (state=,code=0)

0: jdbc:drill:zk=local> select t.`marketing_info`.`camp_id`,t.`marketing_info`.keywords from dfs.tmp.sampleparquet t;
----------------------+

EXPR$0

EXPR$1

----------------------+

4	["go","to","thing","watch","made","laughing","might","pay","in","your","hold"]
6	["pronounce","tree","instead","games","sigh"]
17	[]
17	["it's"]
8	["fallout"]

----------------------+

Sample.json is below
{"trans_id":0,"date":"2013-07-26","time":"04:56:59","amount":80.5,"user_info":

{"cust_id":28,"device":"IOS5","state":"mt"}

,"marketing_info":

{"camp_id":4,"keywords":["go","to","thing","watch","made","laughing","might","pay","in","your","hold"]}

,"trans_info":{"prod_id":[16],"purch_flag":"false"}}

{"trans_id":1,"date":"2013-05-16","time":"07:31:54","amount":100.40,
"user_info":

{"cust_id":86623,"device":"AOS4.2","state":"mi"}

,"marketing_info":

{"camp_id":6,"keywords":["pronounce","tree","instead","games","sigh"]}

,"trans_info":{"prod_id":[],"purch_flag":"false"}}

{"trans_id":2,"date":"2013-06-09","time":"15:31:45","amount":20.25,
"user_info":

{"cust_id":11,"device":"IOS5","state":"la"}

,"marketing_info":

{"camp_id":17,"keywords":[]}

,"trans_info":{"prod_id":[293,90],"purch_flag":"true"}}

{"trans_id":3,"date":"2013-07-19","time":"11:24:22","amount":500.75,
"user_info":

{"cust_id":666,"device":"IOS5","state":"nj"}

,"marketing_info":

{"camp_id":17,"keywords":["it's"]}

,"trans_info":{"prod_id":[173,18,121,84,115,226,464,525,35,11,94,45],"purch_flag":"false"}}

{"trans_id":4,"date":"2013-07-21","time":"08:01:13","amount":34.20,"user_info":

{"cust_id":999,"device":"IOS7","state":"ct"}

,"marketing_info":

{"camp_id":8,"keywords":["fallout"]}

,"trans_info":{"prod_id":[311,29,5,41],"purch_flag":"false"}}

Issue Links

• is related to: DRILL-1830 Fail to read map types in parquet reader