org.sparklinedata.druid.metadata
Order the dimensions based on the number of relations they have.
Order the dimensions based on the number of relations they have. The ones with the most relations are the base of Dimensional components. For e.g. in a Star Schema with dimensions (Product, Customer) if p_name is the key for Product, it will be related to all other product attributes. A query about p_name and other Product fields will be bounded by the cardinality of p_name for the Product dimension. So for a query that has a Group By on p_name, p_brand, p_mfr the cardinality estimate of the result set is the cardinality of p_name.
Todo: for a query on p_brand, p_mfr bound the cardinality by the cardinality of p_name. I thik this requires the user to specify the key column of a dimension.