Improve YAML metric parsing, joins, and adapter loading #53
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0fbebaccb4
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
sidemantic/core/semantic_graph.py
Outdated
| join_key_base = f"{base_pk}={junction_self_fk}" | ||
| add_edge(model_name, junction_model, join_key_base, "one_to_many") | ||
| add_edge(junction_model, model_name, join_key_base, "many_to_one") | ||
|
|
||
| join_key_related = f"{junction_related_fk}={related_pk}" |
There was a problem hiding this comment.
In SemanticGraph.build_adjacency, the many-to-many edges are encoded as raw join-key strings like base_pk=junction_fk and junction_fk=related_pk (lines 223-227). find_relationship_path later determines which side owns each key by checking only primary_key, dimensions, or relationship.sql_expr. If the junction model doesn’t declare its FK columns as dimensions/relationships (common for pure bridge tables), model_has_key will not recognize junction_fk, so the join path can flip to the junction’s primary key (e.g., order_items.id = products.product_id), yielding incorrect joins. This makes many-to-many joins silently wrong unless users explicitly model the FK columns. Consider storing explicit join direction in the edge or using the relationship metadata directly when building the JoinPath so junction FKs are used even when not modeled as dimensions.
Useful? React with 👍 / 👎.
Summary:\n- parse time_comparison/conversion fields from native YAML and export relationship junction metadata\n- add many_to_many join inference via junction tables and include junction keys in CTEs\n- parse DuckDB URL query params, auto-detect Omni/Rill/Superset formats, and include MCP join conditions\n\nTests:\n- uv run pytest tests/db/test_duckdb_adapter.py tests/adapters/sidemantic_adapter/test_parsing.py tests/joins/test_many_to_many_joins.py