semantic-view-modeling-patterns

Use the time intelligence pattern: create a role-playing alias of the fact table, define a computed fact that shifts the date with DATEADD, and join it to the calendar dimension via that computed key. This aligns prior-year rows into current period buckets without window functions.

Use the range join pattern with unique_keys on (key, valid_from, valid_to), a constraints distinct_range block, and a relationship column with type: range plus right_range. If the dimension only has a start date with no end date, use the ASOF join pattern instead.

Inflated numbers usually come from a wrong primary_key declaration (wrong cardinality) or summing a snapshot fact across time. Verify the declared PK is actually unique, and add non_additive_dimensions to snapshot metrics like balances or headcount.

This error occurs when a fact table has two foreign keys to the same dimension and a metric is grouped by that dimension without disambiguation. Fix it by adding using_relationships to each metric, naming which relationship path to follow.

Yes. Define a fact with a scalar row-level expression (for example CONCAT of year and quarter), then reference that fact's logical name as left_column in the relationship. Aggregation expressions are not valid as join keys.

Run CALL SYSTEM$CREATE_SEMANTIC_VIEW_FROM_YAML with the third argument set to TRUE, which validates the YAML structure without creating the view. This catches deploy-time errors like duplicate names and reversed relationship direction early.