[jira] [Commented] (ASTERIXDB-1322) Compiler exception for queries with complex UDFs

["ASF subversion and git services (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ASTERIXDB-1322?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15169471#comment-15169471 ] 

ASF subversion and git services commented on ASTERIXDB-1322:
------------------------------------------------------------

Commit 756ba73d1d977f3cba0e86e7c96efdb2ae88deca in incubator-asterixdb-hyracks's branch refs/heads/master from [~buyingyi]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-asterixdb-hyracks.git;h=756ba73 ]

ASTERIXDB-1322: fix variable visitors for SubplanOperator.

- Simplified the implementation of OperatorPropertiesUtil;
- Fixed PushProjectDownRule to not introduce project operator
  into a subplan to project outer variables.

Change-Id: I53ee1e27b41c9c80d51a7e1baf058d97338c18a9
Reviewed-on: https://asterix-gerrit.ics.uci.edu/662
Tested-by: Jenkins <je...@fulliautomatix.ics.uci.edu>
Reviewed-by: Till Westmann <ti...@apache.org>


> Compiler exception for queries with complex UDFs
> ------------------------------------------------
>
>                 Key: ASTERIXDB-1322
>                 URL: https://issues.apache.org/jira/browse/ASTERIXDB-1322
>             Project: Apache AsterixDB
>          Issue Type: Bug
>          Components: Optimizer
>            Reporter: Yingyi Bu
>            Assignee: Yingyi Bu
>
> The push project down is not done right regarding to subplans.
> The following query will cause an compiler exception because of undefined variables (which is projected away in the query plan subtree):
> {noformat}
> create type HRMType as closed {
>   row_id: int32,
>   sid: int32,
>   date: date,
>   day: int32,
>   time: time,
>   bpm: int32,
>   RR: float
> };
> create dataset HRM(HRMType) primary key row_id;
> declare function median($x)
> {
>      let $c := count($x)
>      return avg(
>                 for $i at $p in (for $j in $x order by $j return $j)
>                 where $p >= floor(($c-1)/2.0) and $p <= $c/2
>                 return $i
>                )
> }
> declare function MAD($x)
> {
>      let $m := median($x)
>      return median(for $j in $x order by abs($j-$m)) return abs($j-$m))
> }
> for $i in dataset HRM
> group by $sid := $i.sid, $gdate := $i.date, $gday := $i.day,
> $timebin := interval-bin($i.time, time("00:00:00"), day-time-duration("PT1M")) with $i
> return
> { 
>   "sid": $sid,
>   "gdate": $gdate,
>   "gday": $gday,
>   "timebin": $timebin,
>   "stdv": (avg(for $ii in $i return $ii.RR * $ii.RR) - 
>           avg(for $ii in $i return $ii.RR) * avg(for $ii in $i return $ii.RR))^(0.5),
>   "MAD": MAD(for $ii in $i return $ii.RR)
> }
> {noformat}
> A modified version of MAD can work:
> {noformat}
> declare function MAD($x)
> {
>      median(for $j in $x order by abs($j-median($x)) return abs($j-median($x)))
> }
> {noformat}



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