[ 
https://issues.apache.org/jira/browse/HIVE-10122?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14384874#comment-14384874
 ] 

Mostafa Mokhtar commented on HIVE-10122:
----------------------------------------

[~sershe] [~thejas] [~hagleitn]

I ran explain for this query 
{code}
select 
    ss_item_sk rowcount
from
    store_sales
where
    ss_sold_date_sk between 2450816 and 2450817
        and ss_ticket_number > 100000000
        and ss_item_sk > 50;
{code}

And the query that gets issues to MySQL looks correct to me as only the 
qualified partitions are queried.
What am I missing?

{code}
        select 
    COLUMN_NAME,
    COLUMN_TYPE,
    min(LONG_LOW_VALUE),
    max(LONG_HIGH_VALUE),
    min(DOUBLE_LOW_VALUE),
    max(DOUBLE_HIGH_VALUE),
    min(BIG_DECIMAL_LOW_VALUE),
    max(BIG_DECIMAL_HIGH_VALUE),
    sum(NUM_NULLS),
    max(NUM_DISTINCTS),
    max(AVG_COL_LEN),
    max(MAX_COL_LEN),
    sum(NUM_TRUES),
    sum(NUM_FALSES)
from
    PART_COL_STATS
where
    DB_NAME = 'tpcds_bin_partitioned_orc_30000'
        and TABLE_NAME = 'store_sales'
        and COLUMN_NAME in ('ss_item_sk' , 'ss_ticket_number')
        and PARTITION_NAME in ('ss_sold_date_sk=2450816' , 
'ss_sold_date_sk=2450817')
group by COLUMN_NAME , COLUMN_TYPE
{code}

> Hive metastore filter-by-expression is broken for non-partition expressions
> ---------------------------------------------------------------------------
>
>                 Key: HIVE-10122
>                 URL: https://issues.apache.org/jira/browse/HIVE-10122
>             Project: Hive
>          Issue Type: Bug
>          Components: Metastore
>    Affects Versions: 0.14.0, 1.0.0, 1.1.0
>            Reporter: Sergey Shelukhin
>
> See 
> https://issues.apache.org/jira/browse/HIVE-10091?focusedCommentId=14382413&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14382413
> These two lines of code
> {noformat}
>     // Replace virtual columns with nulls. See javadoc for details.
>     prunerExpr = removeNonPartCols(prunerExpr, extractPartColNames(tab), 
> partColsUsedInFilter);
>     // Remove all parts that are not partition columns. See javadoc for 
> details.
>     ExprNodeDesc compactExpr = compactExpr(prunerExpr.clone());
> {noformat}
> are supposed to take care of this; I see there were bunch of changes to this 
> code over some time, and now it appears to be broken.
> Thanks to [~thejas] for info.



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

Reply via email to