Hibernate：模式验证：缺少表 - 区分大小写 Mysql 模式标识符变为大写

Question

(context - migrating working Java server application from Spring/Hibernate 3 to 5)

Upon server start, the Hibernate validation fails with “ Schema-validation: missing table ”. The tables exist (the app is working with Hibernate 3)

The reason is the query [1] which fetches all tables from the DB informationSchema. The database name is 'myApp', but “WHERE (TABLE_SCHEMA = 'MYAPP')“ is queried. This leads to empty result set and “missing tables”.

The reason for that is the initial value of “unquotedCaseStrategy” [2] (IdentifierCaseStrategy.UPPER) which is left unchanged within “applyIdentifierCasing” method where all metaData.storesLowerCaseIdentifiers(),metaData.storesUpperCaseIdentifiers(),metaData.storesMixedCaseIdentifiers() return FALSE, because select @@lower_case_table_names called in MariaDbDatabaseMetaData#connection#getLowercaseTableNames() [4] method returns '0'. This is the default value for Unix systems and as the JavaDoc [5] states, all table names and aliases and database names are compared in a case-sensitive manner.

• Hibernate = 5.6.7
• Java = jdk1.8.0_241
• AppServer = wildfly-18.0.1.Final
• DB = MariaDb 10.1
• DB driver = mariadb-java-client-2.2.6.jar

[1]

SELECT TABLE_SCHEMA TABLE_CAT, NULL TABLE_SCHEM, TABLE_NAME, IF(TABLE_TYPE='BASE TABLE', 'TABLE', TABLE_TYPE) as TABLE_TYPE, TABLE_COMMENT REMARKS, NULL TYPE_CAT, NULL TYPE_SCHEM, NULL TYPE_NAME, NULL SELF_REFERENCING_COL_NAME, NULL REF_GENERATION FROM INFORMATION_SCHEMA.TABLES WHERE (TABLE_SCHEMA = 'MYAPP') AND (TABLE_NAME LIKE '%') AND TABLE_TYPE IN ('BASE TABLE','VIEW') ORDER BY TABLE_TYPE, TABLE_SCHEMA, TABLE_NAME;

[2] org.hibernate.engine.jdbc.env.spi.IdentifierHelperBuilder#unquotedCaseStrategy

[3] org.hibernate.engine.jdbc.env.spi.IdentifierHelperBuilder#applyIdentifierCasing

[4] org.mariadb.jdbc.MariaDbConnection#getLowercaseTableNames

[5] getLowercaseTableNames Javadoc: - Are table case sensitive or not. Default Value: 0 (Unix), 1 (Windows), 2 (Mac OS X). If set to 0 (the default on Unix-based systems), table names and aliases and database names are compared in a case-sensitive manner. If set to 1 (the default on Windows), names are stored in lowercase and not compared in a case-sensitive manner. If set to 2 (the default on Mac OS X), names are stored as declared, but compared in lowercase.

Answer 1

It seems to me there is an error in the applyIdentifierCasing method logic and the default (initial value) of the unquotedCaseStrategy strategy [2] should be IdentifierCaseStrategy.MIXED.

WORKAROUND: extend one of the Dialects in org.hibernate.dialect and @Override buildIdentifierHelper() method directly returning the IdentifierHelper instance (calling builder.build(); instead of super.buildIdentifierHelper(…) which would overwrite the config).

public class MySQL5InnoDBDialectEx extends MySQL55Dialect {

    public MySQL5InnoDBDialectEx() {
        super();
        registerColumnType(Types.BIT, "tinyint(1)");
    }

    @Override
    public IdentifierHelper buildIdentifierHelper(IdentifierHelperBuilder builder, DatabaseMetaData dbMetaData) throws SQLException {

        // Copied from org.hibernate.dialect.Dialect#buildIdentifierHelper..
        builder.applyReservedWords(dbMetaData);
        builder.applyReservedWords(AnsiSqlKeywords.INSTANCE.sql2003());

        builder.setUnquotedCaseStrategy(IdentifierCaseStrategy.MIXED);
        builder.setQuotedCaseStrategy(IdentifierCaseStrategy.MIXED);

        // DO NOT USE   >>return super.buildIdentifierHelper( builder, dbMetaData );<<
        // as is normally done in the superClasses, because it overwrites the Quoted/UnquotedCaseStrategy  (that is a bug..)
        return builder.build();
    }
}