org.eclipse.daanse.jdbc.db

JDBC database utilities: a neutral schema/type model, database introspection (reading), and — temporarily co-hosted — the SQL dialect framework that is scheduled to move to the org.eclipse.daanse.sql repository.

Module map

Module	Package root	Role
api	org.eclipse.daanse.jdbc.db.api	Neutral model (api.type, api.sql, api.schema, api.meta), reading SPI (DatabaseService, SnapshotBuilder, MetaDataQueries, MetadataProvider, MetadataProviderFactory). Stays here; the sql repo depends on it.
record	org.eclipse.daanse.jdbc.db.record	Record implementations of the neutral model.
metadata	org.eclipse.daanse.jdbc.db.metadata	Engine-specific MetadataProvider implementations (information_schema/system-catalog readers for H2, MySQL, MariaDB, PostgreSQL, SQL Server, Oracle) + OSGi MetadataProviderFactory services + the static MetadataProviders resolver. Reading only — never depends on a dialect.
impl	org.eclipse.daanse.jdbc.db.impl	DatabaseServiceImpl (falls back to plain DatabaseMetaData when no provider is given).
importer/csv	org.eclipse.daanse.jdbc.db.importer.csv	CSV import. Depends on org.eclipse.daanse.sql.dialect.api (the accepted future jdbc.db→sql edge).
dialect/	org.eclipse.daanse.sql.dialect[.api|.db.<x>]	Copy candidate. SQL dialect framework, already renamed in place to its future sql-repo coordinates. Self-contained subtree (own parent + revision).

Dialect ⇄ reading: the cut

Dialect no longer extends MetadataProvider. Dialects are SQL spelling (quoting, generators, capabilities, TypeMapper); database reading lives in metadata. The one JDBC probe point kept on the dialect side is DialectInitData.fromConnection(Connection).

Former callers that passed a dialect as the metadata provider migrate with a one-liner (plus a dependency on org.eclipse.daanse.jdbc.db.metadata):

// before:  service.createMetaInfo(connection, dialect);
service.createMetaInfo(connection,
        MetadataProviders.forConnection(connection)
                .orElse(new MetadataProvider() { /* plain-DatabaseMetaData fallback */ }));
// or explicitly: service.createMetaInfo(connection, new PostgreSqlMetadataProvider());

All ~37 reading methods keep their signatures; nothing was removed, only moved. Note Oracle: the plain DatabaseMetaData fallback can trip ORA-17027 (LONG-column streaming) — use OracleMetadataProvider there.

Copy runbook: moving dialect/ to org.eclipse.daanse.sql

1. Copy only the active parts into the sql repo: dialect/pom.xml, dialect/api/, dialect/db/pom.xml, dialect/db/{common,test-support,clickhouse,derby,duckdb,h2,mariadb,mssqlserver,mysql,oracle,postgresql,sqlite}/ → org.eclipse.daanse.sql/dialect/…. Do not copy the attic directories (see below).
2. sql root pom: add <module>dialect</module>.
3. Flip the sql-repo dependencies (statement/api+impl, guard, deparser) from org.eclipse.daanse.jdbc.db.dialect.* to org.eclipse.daanse.sql.dialect.* (${revision}, reactor-internal). Java imports already match.
4. In jdbc.db afterwards: delete the dialect/ tree and its <module> entry. importer/csv and the impl tests then resolve org.eclipse.daanse.sql.dialect.* from the snapshot repository. Build order: jdbc.db (api/record/metadata/impl) → sql (dialect, statement) → jdbc.db (importer).
5. Downstream repos (rolap, legacy.xmla, olap, etl, server bndruns) need a mechanical import/BSN sed org.eclipse.daanse.jdbc.db.dialect → org.eclipse.daanse.sql.dialect — separate effort.

The subtree's cross-repo dependencies (jdbc.db.api, record, impl, metadata — the latter three test-scope only) are pinned to the literal 0.0.1-SNAPSHOT so they survive leaving the reactor.

Attic

The 27 commented-out dialect modules under dialect/db/ (access … vertica) keep their old org.eclipse.daanse.jdbc.db.dialect.* package names, are not part of the reactor, and are not part of the copy. Known defect kept as-is: the redshift/snowflake sources are cross-wired.