[
https://issues.apache.org/jira/browse/PHOENIX-6098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17183656#comment-17183656
]
Toshihiro Suzuki commented on PHOENIX-6098:
-------------------------------------------
{quote}
Do you have an example SQL query which would could be validated by hand? I can
see how the code is broken given the context you've shared, but i'm not sure
how I would write such a query to demonstrate that Phoenix-Hive does the
correct thing after your change :)
{quote}
Thank you [~elserj]. Sorry, I should have shared that first.
The steps to reproduce this issue is as follows:
1. Create a table in Phoenix and upsert a row:
{code}
CREATE TABLE TEST (ID VARCHAR PRIMARY KEY, COL1 VARCHAR, COL2 VARCHAR, COL3
VARCHAR);
UPSERT INTO TEST VALUES ('id', 'aaa', 'bbb', 'ccc');
{code}
2. Create an external table for the Phoenix table in Hive:
{code}
create external table test (
id string,
col1 string,
col2 string,
col3 string
)
STORED BY 'org.apache.phoenix.hive.PhoenixStorageHandler'
TBLPROPERTIES (
"phoenix.table.name" = "TEST",
"phoenix.zookeeper.quorum" = "<zookeeper quorum>",
"phoenix.zookeeper.znode.parent" = "<zookeeper znode parent>",
"phoenix.zookeeper.client.port" = "2181",
"phoenix.rowkeys" = "ID",
"phoenix.column.mapping" = "id:ID, col1:COL1, col2:COL2, col3:COL3"
);
{code}
3. Run a problematic query in Hive. It shouldn't return any results, but we can
see it returns a row:
{code}
hive> select * from test where col1='aaa' and col2='bbb' and col3 like 'b%';
+----------+------------+------------+------------+
| test.id | test.col1 | test.col2 | test.col3 |
+----------+------------+------------+------------+
| id | aaa | bbb | ccc |
+----------+------------+------------+------------+
{code}
As [[email protected]] mentioned in the PR, it looks like the issue is
introduced after HIVE-11398:
https://github.com/apache/phoenix-connectors/pull/27#issuecomment-679291484
And I see HDP-3.1.5 has the same issue. Thanks.
> IndexPredicateAnalyzer wrongly handles pushdown predicates and residual
> predicates
> ----------------------------------------------------------------------------------
>
> Key: PHOENIX-6098
> URL: https://issues.apache.org/jira/browse/PHOENIX-6098
> Project: Phoenix
> Issue Type: Bug
> Components: hive-connector
> Reporter: Toshihiro Suzuki
> Assignee: Toshihiro Suzuki
> Priority: Major
>
> Currently, the following code of IndexPredicateAnalyzer is assuming that
> GenericUDFOPAnd always has 2 children nodes. I think this is wrong and it
> leads wrong results:
> https://github.com/apache/phoenix-connectors/blob/5bd23ae2a0f70c3b3edf92a53780dafa643faf26/phoenix-hive/src/main/java/org/apache/phoenix/hive/ql/index/IndexPredicateAnalyzer.java#L346-L359
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
