- Refactor cli/build/graph.py to use cached metadata and dependency indices
for faster graph generation and cleaner separation of concerns
- Refactor cli/build/tree.py to delegate per-role processing to process_role()
and support parallel execution via ProcessPoolExecutor
- Add unit tests for graph helper functions and build_mappings()
under tests/unit/cli/build/test_graph.py
- Add unit tests for find_roles() and process_role() behaviour
under tests/unit/cli/build/test_tree.py
- Remove the old include_role dependency integration test which relied on the
previous tree.json dependencies bucket
For details see ChatGPT conversation: https://chatgpt.com/share/6926b805-28a0-800f-a075-e5250aab5c4a
- Added sorting by application key and user key before YAML output.
- Ensures stable and reproducible file generation across runs.
- Added comprehensive unit tests verifying key order and output stability.
See: https://chatgpt.com/share/68ef4778-a848-800f-a50b-a46a3b878797
- Added logic to scan each role’s tasks/*.yml files for include_role usage
- Supports:
* loop/with_items with literal strings → adds each role
* patterns with variables inside literals (e.g. svc-db-{{database_type}}) → expanded to glob and matched
* pure variable-only names ({{var}}) → ignored
* pure literal names → added directly
- Merges discovered dependencies under graphs["dependencies"]["include_role"]
- Added dedicated unit test covering looped includes, glob patterns, pure literals, and ignoring pure variables
See ChatGPT conversation (https://chatgpt.com/share/68a4ace0-7268-800f-bd32-b475c5c9ba1d) for context.
