[
https://issues.apache.org/jira/browse/DRILL-3196?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14575481#comment-14575481
]
Aman Sinha commented on DRILL-3196:
-----------------------------------
I ran some queries with slice_target = 1 to force exchanges and examined the
plans.. they look correct. The hash distribution is done for each different
partition-by column and the sort is done on the combination of partition-by and
order-by columns. Here's one: (note: here I also disabled the mux exchange):
{code}
| 00-00 Screen
00-01 ProjectAllowDup(x=[$0], y=[$1])
00-02 UnionExchange
01-01 Project(w0$o0=[$5], w1$o0=[$6])
01-02 Window(window#0=[window(partition {4} order by [3] range
between UNBOUNDED PRECEDING and CURRENT ROW aggs [MAX($1)])])
01-03 SelectionVectorRemover
01-04 Sort(sort0=[$4], sort1=[$3], dir0=[ASC], dir1=[ASC])
01-05 HashToRandomExchange(dist0=[[$4]])
02-01 Window(window#0=[window(partition {2} order by [3]
range between UNBOUNDED PRECEDING and CURRENT ROW aggs [MIN($1)])])
02-02 SelectionVectorRemover
02-03 Sort(sort0=[$2], sort1=[$3], dir0=[ASC],
dir1=[ASC])
02-04 HashToRandomExchange(dist0=[[$2]])
03-01 Project(T9¦¦*=[$0], l_extendedprice=[$1],
l_partkey=[$2], l_suppkey=[$3], l_orderkey=[$4])
03-02 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=classpath:/tpch/lineitem.parquet]],
selectionRoot=/tpch/lineitem.parquet, numFiles=1, columns=[`*`]]])
{code}
Given this, I think Abhishek's point is valid - we should revisit whether it
still makes sense to disable this piece of functionality. Sometimes, disabling
a specific functionality and subsequent testing the disabled functionality is
harder than testing the enabled one.
> Disable multiple partition by clauses in the same sql query
> -----------------------------------------------------------
>
> Key: DRILL-3196
> URL: https://issues.apache.org/jira/browse/DRILL-3196
> Project: Apache Drill
> Issue Type: Bug
> Components: Query Planning & Optimization
> Affects Versions: 1.0.0
> Reporter: Victoria Markman
> Assignee: Sean Hsuan-Yi Chu
> Priority: Critical
> Labels: window_function
>
> Currently these queries parse and execute, but plan does not look correct.
> {code}
> 0: jdbc:drill:schema=dfs> explain plan for select sum(a2) over(partition by
> a2 order by a2), count(*) over(partition by a2,b2,c2) from t2 order by 1,2;
> +------+------+
> | text | json |
> +------+------+
> | 00-00 Screen
> 00-01 Project(EXPR$0=[$0], EXPR$1=[$1])
> 00-02 SelectionVectorRemover
> 00-03 Sort(sort0=[$0], sort1=[$1], dir0=[ASC], dir1=[ASC])
> 00-04 Project(EXPR$0=[CASE(>($3, 0), CAST($4):ANY, null)],
> EXPR$1=[$5])
> 00-05 Window(window#0=[window(partition {0, 1, 2} order by []
> range between UNBOUNDED PRECEDING and UNBOUNDED FOLLOWING aggs [COUNT()])])
> 00-06 SelectionVectorRemover
> 00-07 Sort(sort0=[$0], sort1=[$1], sort2=[$2], dir0=[ASC],
> dir1=[ASC], dir2=[ASC])
> 00-08 Window(window#0=[window(partition {0} order by [0]
> range between UNBOUNDED PRECEDING and CURRENT ROW aggs [COUNT($0),
> $SUM0($0)])])
> 00-09 SelectionVectorRemover
> 00-10 Sort(sort0=[$0], sort1=[$0], dir0=[ASC],
> dir1=[ASC])
> 00-11 Project(a2=[$1], b2=[$0], c2=[$2])
> 00-12 Scan(groupscan=[ParquetGroupScan
> [entries=[ReadEntryWithPath [path=maprfs:///drill/testdata/aggregation/t2]],
> selectionRoot=/drill/testdata/aggregation/t2, numFiles=1, columns=[`a2`,
> `b2`, `c2`]]])
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
