You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Aman Sinha (JIRA)" <ji...@apache.org> on 2015/04/13 19:05:12 UTC

[jira] [Updated] (DRILL-2332) Drill should be consistent with Implicit casting rules across data formats

     [ https://issues.apache.org/jira/browse/DRILL-2332?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Aman Sinha updated DRILL-2332:
------------------------------
    Fix Version/s:     (was: 0.9.0)
                   1.0.0

> Drill should be consistent with Implicit casting rules across data formats
> --------------------------------------------------------------------------
>
>                 Key: DRILL-2332
>                 URL: https://issues.apache.org/jira/browse/DRILL-2332
>             Project: Apache Drill
>          Issue Type: Improvement
>          Components: Query Planning & Optimization
>            Reporter: Abhishek Girish
>            Assignee: Sean Hsuan-Yi Chu
>             Fix For: 1.0.0
>
>
> Currently, the outcome of a query with a filter on a column comparing it with a literal, depends on the underlying data format. 
> *Parquet*
> {code:sql}
> select * from date_dim where d_month_seq ='1193' limit 1;
> [Succeeds]
> select * from date_dim where d_date in ('1999-06-30') limit 1;
> [Succeeds]
> {code}
> *View on top of text:*
> {code:sql}
> select * from date_dim where d_date in ('1999-06-30') limit 1;
> Query failed: SqlValidatorException: Values passed to IN operator must have compatible types
> Error: exception while executing query: Failure while executing query. (state=,code=0)
> select * from date_dim where d_month_seq ='1193' limit 1;
> Query failed: SqlValidatorException: Cannot apply '=' to arguments of type '<INTEGER> = <CHAR(4)>'. Supported form(s): '<COMPARABLE_TYPE> = <COMPARABLE_TYPE>'
> Error: exception while executing query: Failure while executing query. (state=,code=0)
> {code}
> I understand that in the case of View on Text, SQL validation fails at the Optiq layer. 
> But from the perspective of an end-user, Drill's behavior must be consistent across data formats. Also having a view by definition should abstract out this information.
> Here, both the view and parquet were created with type information. 
> *Parquet-meta*
> {code}
> parquet-schema /mapr/abhi311/data/parquet/tpcds/scale1/date_dim/0_0_0.parquet 
> message root {
>   optional int32 d_date_sk;
>   optional binary d_date_id (UTF8);
>   optional binary d_date (UTF8);
>   optional int32 d_month_seq;
>   optional int32 d_week_seq;
>   optional int32 d_quarter_seq;
>   optional int32 d_year;
>   optional int32 d_dow;
>   optional int32 d_moy;
>   optional int32 d_dom;
>   optional int32 d_qoy;
>   optional int32 d_fy_year;
>   optional int32 d_fy_quarter_seq;
>   optional int32 s_fy_week_seq;
>   optional binary d_day_name (UTF8);
>   optional binary d_quarter_name (UTF8);
>   optional binary d_holiday (UTF8);
>   optional binary d_weekend (UTF8);
>   optional binary d_following_holiday (UTF8);
>   optional int32 d_first_dom;
>   optional int32 d_last_dom;
>   optional int32 d_same_day_ly;
>   optional int32 d_same_day_lq;
>   optional binary d_current_day (UTF8);
>   optional binary d_current_week (UTF8);
>   optional binary d_current_month (UTF8);
>   optional binary d_current_quarter (UTF8);
>   optional binary d_current_year (UTF8);
> }
> {code}
> *Describe View*
> {code:sql}
> > describe date_dim;
> +-------------+------------+-------------+
> | COLUMN_NAME | DATA_TYPE  | IS_NULLABLE |
> +-------------+------------+-------------+
> | d_date_sk   | INTEGER    | NO          |
> | d_date_id   | VARCHAR    | NO          |
> | d_date      | DATE       | NO          |
> | d_month_seq | INTEGER    | NO          |
> | d_week_seq  | INTEGER    | NO          |
> | d_quarter_seq | INTEGER    | NO          |
> | d_year      | INTEGER    | NO          |
> | d_dow       | INTEGER    | NO          |
> | d_moy       | INTEGER    | NO          |
> | d_dom       | INTEGER    | NO          |
> | d_qoy       | INTEGER    | NO          |
> | d_fy_year   | INTEGER    | NO          |
> | d_fy_quarter_seq | INTEGER    | NO          |
> | s_fy_week_seq | INTEGER    | NO          |
> | d_day_name  | VARCHAR    | NO          |
> | d_quarter_name | VARCHAR    | NO          |
> | d_holiday   | VARCHAR    | NO          |
> | d_weekend   | VARCHAR    | NO          |
> | d_following_holiday | VARCHAR    | NO          |
> | d_first_dom | INTEGER    | NO          |
> | d_last_dom  | INTEGER    | NO          |
> | d_same_day_ly | INTEGER    | NO          |
> | d_same_day_lq | INTEGER    | NO          |
> | d_current_day | VARCHAR    | NO          |
> | d_current_week | VARCHAR    | NO          |
> | d_current_month | VARCHAR    | NO          |
> | d_current_quarter | VARCHAR    | NO          |
> | d_current_year | VARCHAR    | NO          |
> +-------------+------------+-------------+
> 28 rows selected (0.137 seconds)
> {code}
> For an end



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)