databricks-metric-views

Use CREATE OR REPLACE VIEW with WITH METRICS LANGUAGE YAML, defining a source table, at least one dimension, and at least one measure with an aggregate expression. You can also call the manage_metric_views MCP tool with action create.

Wrap every measure in the MEASURE() function, such as MEASURE(`Total Revenue`), and group by dimensions. SELECT * is not supported on metric views; you must explicitly list dimensions and measures.

YAML version 1.1 requires Databricks Runtime 17.2 or later, while version 0.1 works on Runtime 16.4 through 17.1. Window measures currently use version 0.1 as an experimental feature.

Metric views do not support SELECT * because measures must be explicitly wrapped in the MEASURE() function and dimensions listed by name. Names containing spaces also require backtick quoting.

Yes, declare joins in the YAML definition with a name, source table, and an on or using clause. Nested joins create snowflake schemas and require Databricks Runtime 17.1 or later.

Use metric views when you need flexible aggregation at query time, safe re-aggregation of ratios, declarative joins, built-in materialization, or native AI/BI Genie integration. Standard views lock aggregation at creation time.