You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kylin.apache.org by "hongbin ma (JIRA)" <ji...@apache.org> on 2017/05/10 05:49:04 UTC

[jira] [Comment Edited] (KYLIN-2599) select * in subquery fail due to bug in hackSelectStar

    [ https://issues.apache.org/jira/browse/KYLIN-2599?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16004116#comment-16004116 ] 

hongbin ma edited comment on KYLIN-2599 at 5/10/17 5:49 AM:
------------------------------------------------------------

similar exception is thrown when I test query like:

{code:sql}
 select lstg_format_name from test_kylin_fact order by case when 1=1  then
                                              
                                                             cal_dt
                                              
                                                          ELSE
                                              
                                                             seller_id
                                              
                                                         end 
{code}

To fix the issue more completely, we'll:

1. check if rootProj has any field starting with "_KY_", if none is found, then current logical plan does not require hacking, org.apache.calcite.sql2rel.SqlToRelConverter#hackSelectStar should abort by returning root.
2. the field list of rootProj may be longer than that of root (in case of kylin-it/src/test/resources/query/sql_verifyCount/query10.sql), so when constructing validatedRowType we'll skip the longer tail on rootProj
3.  sort rel's RelCollation (if any) may become stale after removing the "_KY_" fields, need to fix its fieldIndex 


was (Author: mahongbin):
similar exception is thrown when I test query like:

{code:sql}
 select lstg_format_name from test_kylin_fact order by case
                                              
                                                           when 1=1  then
                                              
                                                             cal_dt
                                              
                                                          ELSE
                                              
                                                             seller_id
                                              
                                                         end 
{code}

To fix the issue more completely, we'll:

1. check if rootProj has any field starting with "_KY_", if none is found, then current logical plan does not require hacking, org.apache.calcite.sql2rel.SqlToRelConverter#hackSelectStar should abort by returning root.
2. the field list of rootProj may be longer than that of root (in case of kylin-it/src/test/resources/query/sql_verifyCount/query10.sql), so when constructing validatedRowType we'll skip the longer tail on rootProj
3.  sort rel's RelCollation (if any) may become stale after removing the "_KY_" fields, need to fix its fieldIndex 

> select * in subquery fail due to bug in hackSelectStar 
> -------------------------------------------------------
>
>                 Key: KYLIN-2599
>                 URL: https://issues.apache.org/jira/browse/KYLIN-2599
>             Project: Kylin
>          Issue Type: Improvement
>            Reporter: hongbin ma
>
> {code:sql}
> select fact.lstg_format_name from 
>  
>  (select * from test_kylin_fact where cal_dt > date'2010-01-01' ) as fact
>  
>  group by fact.lstg_format_name 
>  
>  order by CASE WHEN fact.lstg_format_name IS NULL THEN 'sdf' ELSE fact.lstg_format_name END 
>  
> {code}
> will generate logical plan like:
> {code}
> LogicalSort(sort0=[$1], dir0=[ASC])
>   LogicalProject(LSTG_FORMAT_NAME=[$0], EXPR$1=[CASE(IS NULL($0), 'sdf', $0)])
>     LogicalAggregate(group=[{0}])
>       LogicalProject(LSTG_FORMAT_NAME=[$3])
>         LogicalProject(TRANS_ID=[$0], ORDER_ID=[$1], CAL_DT=[$2], LSTG_FORMAT_NAME=[$3], LEAF_CATEG_ID=[$4], LSTG_SITE_ID=[$5], SLR_SEGMENT_CD=[$6], SELLER_ID=[$7], PRICE=[$8], ITEM_COUNT=[$9], TEST_COUNT_DISTINCT_BITMAP=[$10], DEAL_AMOUNT=[$11], DEAL_YEAR=[$12], _KY_COUNT__=[$13], _KY_MIN_TEST_KYLIN_FACT_PRICE_=[$14], _KY_MAX_TEST_KYLIN_FACT_PRICE_=[$15], _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_SELLER_ID_=[$16], _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_LSTG_FORMAT_NAME_TEST_KYLIN_FACT_SELLER_ID_=[$17], _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_TEST_COUNT_DISTINCT_BITMAP_=[$18], _KY_PERCENTILE_TEST_KYLIN_FACT_PRICE_=[$19])
>           LogicalFilter(condition=[>($2, 2010-01-01)])
>             OLAPTableScan(table=[[DEFAULT, TEST_KYLIN_FACT]], fields=[[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19]])
> {code}
> org.apache.calcite.sql2rel.SqlToRelConverter#hackSelectStar will by mistake treat it like a normal case and lead to throwing exception



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)