Summary
Introduces catalogs.yml v2 support, a `skip_optimize` config for opting out of post-materialization OPTIMIZE, and Rust kernel backend for SQL warehouses. Fixes numerous incremental model bugs around constraints and tags, but now requires `--full-refresh` to apply changes to primary/foreign key expressions.
Summary generated by brickster.ai. For the full changelog and any code/binary attachments, follow the GitHub link above.
More from databricks/dbt-databricks
Metric views now replace existing tables or views using a backup-and-create strategy rather than direct create-or-replace syntax. Lazily formatted Databricks SQL log records are also now properly interpolated when mirrored into dbt logs.
v1.12.4 fixes three bugs: case-insensitive SQL credential redaction, composite unique key deletion on DBR <17.1, and quote escaping in relation comments. databricks-sql-connector is updated to 4.4.0 to fix three CVEs including TLS hostname bypass and DoS vulnerabilities.
Adds catalog_database in v2 catalogs.yml to route Unity catalog models to independent physical catalogs (requires dbt-core 1.12+). Resolves bugs affecting Iceberg Python models, dbt clone operations, non-ASCII column incremental strategies, streaming table tags, and governed tag column drops.
1.12.1
dbt-databricks 1.12.1 exposes job run IDs in adapter responses for correlating dbt runs with Databricks workflow executions and adds support for SPOG account-level vanity URLs. The release fixes critical issues including Iceberg incremental models losing clustering, column constraint enforcement now gated to require explicit contract.enforced: true, and various SQL warehouse compatibility problems.
v1.12.0 adds metric views, row filters, and Python UDFs as new materializations, along with SCHEDULE EVERY and TRIGGER ON UPDATE refresh modes for materialized views and streaming tables. databricks_tags now merge additively across hierarchy levels instead of child replacing parent (breaking change), and the release includes fixes for metric view validation, pydantic v1 compatibility, and streaming table refresh scheduling.
This release adds invocation_id to query comments and enforces a 255-character identifier length limit for Databricks relations. It fixes multiple issues including suppression of spurious warnings that appeared on every run and improves handling of materialized views, Iceberg, and Hudi metadata.