v0.7.8
What's Changed
-
Fix CloudFiles .schema() placement — source.schema now correctly applies .schema() on the DataStreamReader chain before .load(), instead of the invalid df.schema() post-load property access. Also fixes the description fallback bug ( → actual format string) and removes the dead batch branch from the template.
-
New catalog/schema namespace format replacing the legacy database: "catalog.schema" field across load sources and write targets. The old format is auto-converted with a deprecation warning via the new namespace_normalizer service.
-
Pipeline configuration made mandatory — deprecating the automatic catalog/schema lookup from pipeline defaults.
-
Deterministic job orchestration output — fixed non-deterministic depends_on ordering in generated job YAML files.
-
E2E coverage for the new format — added namespace_validation pipeline with delta load, SQL load, streaming table, and materialized view actions natively using catalog/schema.
-
Supply chain security hardening — pinned all GitHub Actions to SHA hashes, added dependency vulnerability scanning (pip-audit), static security analysis (bandit), secret scanning (gitleaks), license compliance (liccheck), OpenSSF Scorecard, SLSA provenance, Dependabot config, branch protection, and CODEOWNERS.
Full Changelog: v0.7.7...v0.7.8