[
https://issues.apache.org/jira/browse/PHOENIX-6683?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17518006#comment-17518006
]
ASF GitHub Bot commented on PHOENIX-6683:
-----------------------------------------
chrajeshbabu commented on PR #73:
URL:
https://github.com/apache/phoenix-connectors/pull/73#issuecomment-1090085221
> I haven't looked into the logic closely. Do we process the parentheses
correctly ?
>
> i.e. Make sure that we do not turn
>
> (a and b or c) into (a and (b or c))
Yes this is not happening and AND is getting precedence. Just added another
test case to ensure that..
45336 [main] DEBUG org.apache.phoenix.execute.BaseQueryPlan - Iterator for
table N000001 ready: resultIterators=ResultIterators
[name=PARALLEL,id=5337088f-cf25-3568-ad24-630b5fd8c07f,scans=[[{"loadColumnFamiliesOnDemand":true,"filter":"**((\"ENTITY_INSTANCE_ID\"
= 40 AND \"CASE_MATCH_TYPE\" = 'M') OR NOT (\"CANCELLATION_FLAG\" =
'Y'))"**,"startRow":"","stopRow":"","batch":-1,"cacheBlocks":true,"totalColumns":3,"maxResultSize":-1,"families":{"0":["\\x00\\x00\\x00\\x00","\\x80\\x0B","\\x80\\x0C"]},"caching":2147483647,"maxVersions":1,"timeRange":[0,9223372036854775807]}]]],
index=0]
> Surround the OR filters with parentheses while converting spark filters to
> phoenix expessions
> ---------------------------------------------------------------------------------------------
>
> Key: PHOENIX-6683
> URL: https://issues.apache.org/jira/browse/PHOENIX-6683
> Project: Phoenix
> Issue Type: Bug
> Components: spark-connector
> Reporter: Rajeshbabu Chintaguntla
> Assignee: Rajeshbabu Chintaguntla
> Priority: Major
>
> If there are mix of *AND* and *OR* filters in the query while converting the
> to phoenix filter from spark filters no parentheses got added to *OR* filters
> which is leading to wrong filter and finally getting wrong data.
> For example if the filter is used is below
> {code}
> where ENTITY_INSTANCE_ID = 40 and (CANCELLATION_FLAG <> 'Y' OR
> CANCELLATION_FLAG IS NULL ) and CASE_MATCH_TYPE='M'
> {code}
> which is getting converted to an expression and leading to wrong result
> because result is always giving most of the records matching
> CANCELLATION_FLAG is not 'Y'
> {code}
> NOT "CANCELLATION_FLAG" = 'Y' OR "CANCELLATION_FLAG" IS NULL AND
> "CASE_MATCH_TYPE" IS NOT NULL AND "ENTITY_INSTANCE_ID" = 40 AND
> "ENTITY_INSTANCE_ID" IS NOT NULL AND "CASE_MATCH_TYPE" = 'M'
> {code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
