malloy-scope

Inspect the package with get_context to list sources and fields, then run execute_query for row counts and data-quality checks. Classify tables by role, identify 2-3 analytical domains, and present lettered scope options with one clear recommendation for the user to select.

Facts are events or transactions you measure, such as orders or payments. Dimensions are entities you slice by, such as customers or products. Bridge tables resolve many-to-many links, and operational tables like audit logs or staging tables should be excluded from modeling.

Skip operational and ETL tables such as audit logs, staging tables where a production equivalent exists, and pre-aggregated summary tables that should be computed fresh in Malloy. Record each skipped table with its reason in modeling-notes.md.

Scope proposal happens after inspecting the model and underlying data but before writing Malloy sources, dimensions, and measures. It ensures the user confirms which tables and analytical questions the model targets before modeling work begins.

Entity or snapshot datasets with cumulative counters cannot support period-over-period analysis, unlike event-based fact models with trends over time. The scope proposal should state the model kind and name concrete consequences, such as missing date fields ruling out calendar analysis.