[
https://issues.apache.org/jira/browse/DRILL-4924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15542907#comment-15542907
]
Khurram Faraaz commented on DRILL-4924:
---------------------------------------
Another observation, when we do a select * from inside the in clause, we see a
NestedLoops join and later a HashJoin, where are when I replace the select *
from by (values(1)) we see an error, "query cannot be planned possibly due to
either a cartesian join or an inequality join".
Which of the two behaviors is expected ?
{noformat}
0: jdbc:drill:schema=dfs.tmp> explain plan for select * from `emp_tbl` where id
not IN ( SELECT CASE WHEN id=1 THEN 30 else id end from (select * from
`emp_tbl`));
+------+------+
| text | json |
+------+------+
| 00-00 Screen
00-01 Project(*=[$0])
00-02 Project(T11¦¦*=[$0])
00-03 SelectionVectorRemover
00-04 Filter(condition=[CASE(=($2, 0), CAST(true):BOOLEAN, IS NOT
NULL($6), CAST(false):BOOLEAN, IS NULL($4), null, <($3, $2), null,
CAST(true):BOOLEAN)])
00-05 HashJoin(condition=[=($4, $5)], joinType=[left])
00-07 Project(T11¦¦*=[$0], id=[$1], $f0=[$2], $f1=[$3], id0=[$1])
00-09 NestedLoopJoin(condition=[true], joinType=[inner])
00-12 Project(T11¦¦*=[$0], id=[$1])
00-15 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tmp/emp_tbl]],
selectionRoot=maprfs:/tmp/emp_tbl, numFiles=1, usedMetadataFile=false,
columns=[`*`]]])
00-11 StreamAgg(group=[{}], agg#0=[COUNT()],
agg#1=[COUNT($0)])
00-14 Project($f0=[CASE(=(ITEM($0, 'id'), 1), 30, ITEM($0,
'id'))], $f1=[true])
00-17 Project(T12¦¦*=[$0], id=[$1])
00-18 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tmp/emp_tbl]],
selectionRoot=maprfs:/tmp/emp_tbl, numFiles=1, usedMetadataFile=false,
columns=[`*`]]])
00-06 Project($f00=[$0], $f10=[$1])
00-08 HashAgg(group=[{0}], agg#0=[MIN($1)])
00-10 Project($f0=[CASE(=(ITEM($0, 'id'), 1), 30, ITEM($0,
'id'))], $f1=[true])
00-13 Project(T13¦¦*=[$0], id=[$1])
00-16 Scan(groupscan=[ParquetGroupScan
[entries=[ReadEntryWithPath [path=maprfs:///tmp/emp_tbl]],
selectionRoot=maprfs:/tmp/emp_tbl, numFiles=1, usedMetadataFile=false,
columns=[`*`]]])
{noformat}
> Can not use case expression within an IN predicate.
> ---------------------------------------------------
>
> Key: DRILL-4924
> URL: https://issues.apache.org/jira/browse/DRILL-4924
> Project: Apache Drill
> Issue Type: Bug
> Components: Query Planning & Optimization
> Affects Versions: 1.9.0
> Reporter: Khurram Faraaz
>
> Can not use case expression within an IN predicate.
> Drill 1.9.0 git commit ID: f3c26e34
> Similar query works on Postgres 9.3
> {noformat}
> postgres=# select * from t1 where c1 in ( select case when c1=2 then 30 else
> c1 end);
> c1 | c2
> ----+-----
> 1 | 0
> 3 | 19
> -1 | 11
> 5 | 13
> 10 | 17
> 11 | -1
> 13 | 1
> 17 | 20
> 0 | 9
> 19 | 100
> (10 rows)
> {noformat}
> Drill 1.9.0 returns an error
> {noformat}
> 0: jdbc:drill:schema=dfs.tmp> select * from `emp_tbl` where id IN ( SELECT
> CASE WHEN id=2 THEN 30 else id end );
> Error: PARSE ERROR: Encountered ")" at line 1, column 81.
> Was expecting one of:
> "FROM" ...
> "," ...
> "AS" ...
> <IDENTIFIER> ...
> <QUOTED_IDENTIFIER> ...
> <BACK_QUOTED_IDENTIFIER> ...
> <BRACKET_QUOTED_IDENTIFIER> ...
> <UNICODE_QUOTED_IDENTIFIER> ...
> "NOT" ...
> "IN" ...
> "BETWEEN" ...
> "LIKE" ...
> "SIMILAR" ...
> "=" ...
> ">" ...
> "<" ...
> "<=" ...
> ">=" ...
> "<>" ...
> "+" ...
> "-" ...
> "*" ...
> "/" ...
> "||" ...
> "AND" ...
> "OR" ...
> "IS" ...
> "MEMBER" ...
> "SUBMULTISET" ...
> "MULTISET" ...
> "[" ...
> SQL Query select * from `emp_tbl` where id IN ( SELECT CASE WHEN id=2 THEN 30
> else id end )
>
> ^
> [Error Id: e6c3f120-8776-476e-8df7-7ef30f6b7307 on centos-01.qa.lab:31010]
> (state=,code=0)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
