[
https://issues.apache.org/jira/browse/ASTERIXDB-1623?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Xikui Wang updated ASTERIXDB-1623:
----------------------------------
Description:
I met this bug when I was adding 'upsert-option' for the feed part. The easiest
way to reproduce this bug is to checkout my change at gerrit:
https://asterix-gerrit.ics.uci.edu/#/c/1068/ and execute the added test case.
The logical plan generated by subscribe statement is:
from:
org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:81
{quote}
sink -- |UNPARTITIONED|
upsert into experiments:TwitterUsers from record: %0->$$2 partitioned by
[%0->$$3] out: ([record-before-upsert:null]) -- |UNPARTITIONED|
assign [$$3] <- [function-call: asterix:field-access-by-name,
Args:[%0->$$2, AString: {screen-name}]] -- |UNPARTITIONED|
project ([$$2]) -- |UNPARTITIONED|
assign [$$2] <- [function-call: asterix:collection-to-sequence,
Args:[%0->$$0]] -- |UNPARTITIONED|
unnest $$0 <- function-call: asterix:feed-collect, Args:[AString:
{experiments}, AString: {UserFeed}, AString: {UserFeed}, AString: {INTAKE},
AString: {TwitterUsers}, AString: {TwitterUserType}] -- |UNPARTITIONED|
empty-tuple-source -- |UNPARTITIONED|
{quote}
I managed to trace down that after runOptimizationSets
(org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer#runOptimizationSets)
worked on ruleList "ruleList.second.get(0) instanceof PullSelectOutOfEqJoin",
the plan.toString will causes NPE.
In normal scenario, this exception will not appear due to the log level
setting. However, during the integration test, this bug will be triggered by
org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:85
was:
I met this bug when I was adding 'upsert-option' for the feed part. The easiest
way to reproduce this bug is to checkout my change at gerrit:
https://asterix-gerrit.ics.uci.edu/#/c/1068/ and execute the added test case.
The logical plan generated by subscribe statement is:
from:
org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:81
{quote}
sink -- |UNPARTITIONED|
upsert into experiments:TwitterUsers from record: %0->$$2 partitioned by
[%0->$$3] out: ([record-before-upsert:null]) -- |UNPARTITIONED|
assign [$$3] <- [function-call: asterix:field-access-by-name,
Args:[%0->$$2, AString: {screen-name}]] -- |UNPARTITIONED|
project ([$$2]) -- |UNPARTITIONED|
assign [$$2] <- [function-call: asterix:collection-to-sequence,
Args:[%0->$$0]] -- |UNPARTITIONED|
unnest $$0 <- function-call: asterix:feed-collect, Args:[AString:
{experiments}, AString: {UserFeed}, AString: {UserFeed}, AString: {INTAKE},
AString: {TwitterUsers}, AString: {TwitterUserType}] -- |UNPARTITIONED|
empty-tuple-source -- |UNPARTITIONED|
{quote}
I managed to trace down after runOptimizationSets
(org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer#runOptimizationSets)
worked on ruleList "ruleList.second.get(0) instanceof PullSelectOutOfEqJoin".
The plan.toString will causes NPE.
In normal scenario, this exception will not appear due to the log level
setting. However, during the integration test, this bug will be triggered by
org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:85
> Certain Rewritten Rule causes NPE in LogicalPlan.toString()
> -----------------------------------------------------------
>
> Key: ASTERIXDB-1623
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-1623
> Project: Apache AsterixDB
> Issue Type: Bug
> Components: Optimizer
> Reporter: Xikui Wang
>
> I met this bug when I was adding 'upsert-option' for the feed part. The
> easiest way to reproduce this bug is to checkout my change at gerrit:
> https://asterix-gerrit.ics.uci.edu/#/c/1068/ and execute the added test case.
> The logical plan generated by subscribe statement is:
> from:
> org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:81
> {quote}
> sink -- |UNPARTITIONED|
> upsert into experiments:TwitterUsers from record: %0->$$2 partitioned by
> [%0->$$3] out: ([record-before-upsert:null]) -- |UNPARTITIONED|
> assign [$$3] <- [function-call: asterix:field-access-by-name,
> Args:[%0->$$2, AString: {screen-name}]] -- |UNPARTITIONED|
> project ([$$2]) -- |UNPARTITIONED|
> assign [$$2] <- [function-call: asterix:collection-to-sequence,
> Args:[%0->$$0]] -- |UNPARTITIONED|
> unnest $$0 <- function-call: asterix:feed-collect, Args:[AString:
> {experiments}, AString: {UserFeed}, AString: {UserFeed}, AString: {INTAKE},
> AString: {TwitterUsers}, AString: {TwitterUserType}] -- |UNPARTITIONED|
> empty-tuple-source -- |UNPARTITIONED|
> {quote}
> I managed to trace down that after runOptimizationSets
> (org.apache.hyracks.algebricks.core.rewriter.base.HeuristicOptimizer#runOptimizationSets)
> worked on ruleList "ruleList.second.get(0) instanceof
> PullSelectOutOfEqJoin", the plan.toString will causes NPE.
> In normal scenario, this exception will not appear due to the log level
> setting. However, during the integration test, this bug will be triggered by
> org/apache/hyracks/algebricks/core/rewriter/base/HeuristicOptimizer.java:85
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
