[HIVE-27755] Quote identifiers in SQL emitted by SchemaTool for MySQL - ASF JIRA

XML

Word

Printable

JSON

Details

Type: Improvement
Status: Closed
Priority: Major
Resolution: Fixed
Affects Version/s: 4.0.0-beta-1
Fix Version/s: 4.0.0
Component/s: Standalone Metastore
Labels:
- pull-request-available

Description

Various SchemaTool options/tasks (e.g., "validate") generate and run SQL statements on the underlying database. Depending on the database identifiers in the SQL statements may be quoted (see https://github.com/apache/hive/blob/2dbfbeefc1a73d6a50f1c829658846fc827fc780/standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/tools/schematool/HiveSchemaHelper.java#L173).

Currently, all identifiers are quoted when the database is Postgres and this tickets aims to do the same for MySQL/MariaDB.

The main motivation behind this change is to avoid unexpected surprises and query failures when/if the database decides to turn some of the tables/columns we are using internally to reserved keywords.

As a concrete example, the Percona fork of MySQL recently turned SEQUENCE_TABLE into a reserved keyword (https://docs.percona.com/percona-server/8.0/flexibility/sequence_table.html) and this comes in conflict with our internal metastore table.

The installation scripts do not fail since in that case SEQUENCE_TABLE is quoted (https://github.com/apache/hive/blob/2dbfbeefc1a73d6a50f1c829658846fc827fc780/standalone-metastore/metastore-server/src/main/sql/mysql/hive-schema-4.0.0-beta-2.mysql.sql#L447) but validation queries emitted by the SchemaTool will fail (https://github.com/apache/hive/blob/2dbfbeefc1a73d6a50f1c829658846fc827fc780/standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/tools/schematool/SchemaToolTaskValidate.java#L117) if we don't use quoted identifiers.