Details
-
Bug
-
Status: Resolved
-
Blocker
-
Resolution: Fixed
-
0.7.0
-
None
Description
On TPCDS parquet data while using the default reader drill is case sensitive for column names.
0: jdbc:drill:> select c_customer_sk from `0_0_0.parquet` limit 1;
---------------
| c_customer_sk |
---------------
| 1 |
---------------
1 row selected (0.15 seconds)
0: jdbc:drill:> select c_customer_SK from `0_0_0.parquet` limit 1;
---------------
| c_customer_sk |
---------------
| 1 |
---------------
On using the new reader though
0: jdbc:drill:> alter session set `store.parquet.use_new_reader`=true;
0: jdbc:drill:> select c_customer_SK from `0_0_0.parquet` limit 1;
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| c_customer_sk | c_customer_id | c_current_cdemo_sk | c_current_hdemo_sk | c_current_addr_sk | c_first_shipto_date_sk | c_first_sales_date_sk | c_salutation | c_first_name | c_last_name | c_preferred_cust_flag | c_birth_day | c_birth_month | c_birth_year | c_birth_country | c_login | c_email_address | c_last_review_date |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| 1 | AAAAAAAABAAAAAAA | 980124 | 7135 | 32946 | 2452238 | 2452208 | Mr. | Javier | Lewis | Y | 9 | 12 | 1936 | CHILE | null | Javier.Lewis@VFAxlnZEvOx.org | 2452508 |
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row selected (0.368 seconds)
Will file a separate bug for the issue that when the new parquet reader cannot find a column it does a * query instead and returns all columns.