fal 0.3.6

[new changes]

Bug Fixes

• Before generating sql models delete old ones (#432)
• Ignore user warnings caused by multiprocessing (#446)
• Add lock around dbt cache manipulation (#429)
• Use correct timeout configuration based on bigquery adapter version (#443)
• Use a general capture for multiprocessing / generator truthness (#456)
• Delete old generated models before fal flow run (#450)
• Remove cached_property not present in Python 3.7 (#458)
• Reference {{ this }} for ephemeral models to get correct schema and table every time (#442)

Documentation

• Patch typo (#434)

Features

• Implement a DBT runner based on the DBT's Python module (#420)
• Allow selection of non-existent nodes (#433)
• Add support for selections/exclusions in the planner (#423)
• Implement error recovery for the parallel runner (#426)
• Add support for scripts in parallel runner (#448)
• Enable more than once write_to_model per Python Model (#455)

Refactor

• Scheduler node statuses, only show failed/skipped nodes (#445)
• Use target.identifier instead of manually getting table alias (#454)
• Add some type-checks and asserts without changing logic (#457)

Testing

• Implement a partial DAG verifier for model orders (#431)
• Create a project that can be easily parallelized. (#430)
• Test several dbt versions on integration tests (#440)
• Run specific adapter test when mentioned in PR title (#444)
• Refactor profile test pull_request check to handle casing (#447)
• Choose adapter to test when dispatching integration tests (#452)
• Add a CI run for fal flow run with experimental threads (#439)
• Reduce dbt versions to test on schedule and fix matrix option (#451)
• Centralize integration test running in a single file (#453)
• Test with any Python version mentioned in PR title or branch (#459)