[
https://issues.apache.org/jira/browse/HIVE-28297?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Work on HIVE-28297 started by Wechar.
-------------------------------------
> Unify table pattern definition in Hive MetaStore
> ------------------------------------------------
>
> Key: HIVE-28297
> URL: https://issues.apache.org/jira/browse/HIVE-28297
> Project: Hive
> Issue Type: Improvement
> Components: Hive
> Reporter: Wechar
> Assignee: Wechar
> Priority: Major
>
> *Current Situation*
> There are two cases for table pattern in Hive:
> - Case 1: Wildcards ('*' and '|)
> It's replaced by SQL type like expressions with '%' and '_' since Hive 4.0.0,
> details in HIVE-23359.
> DDL documents:
> https://cwiki.apache.org/confluence/display/Hive/LanguageManual+DDL#LanguageManualDDL-Show
> - Case 2: Java regular expression
> It's declared in hive code:
> [https://github.com/apache/hive/blob/45867be6cb5308566e4cf16c7b4cf8081085b58c/ql/src/java/org/apache/hadoop/hive/ql/metadata/Hive.java#L1882-L1894],
> and also match the jdo {{matches(String pattern)}} method:
> [https://www.datanucleus.org/products/accessplatform/jdo/query.html]
> *Issues*
> The different definitions cause the misuse of table pattern, in some codes it
> use * to represent match everything, while it use .* in other codes.
> The misuse would cause the wrong result because both client and server side
> of HMS need a trick handle for '*' with code like:
> {code:java}
> patterns.split("\\|")
> pattern.replaceAll("\\*", ".*"))
> {code}
> *Action*
> We should honor the definition in HMS of Case 2, because Case 1 is just the
> frontend of SQL, and it has been replaced.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
