Best Practice for Defining Composite Metrics Without Excessive Schema Overhead

[w1992wishes]

Problem

Hi,

I’m looking for advice on the best way to define composite metrics based on existing metrics. For example, I want to define a new metric C as the sum of metrics A and B, and then be able to query metric C via a REST API.

Currently, I’m considering redefining a new metric C in the schema. However, this approach leads to an excessive number of schemas, which I’ve observed can decrease API query performance. Additionally, each new schema requires a recompilation process, which becomes more time-consuming as the number of schemas increases.

Is there a recommended approach to handle such composite metrics without the drawbacks of having too many schemas? How can I efficiently define and query composite metrics without negatively impacting performance and compile times?

Thank you for your insights!

Related Cube.js schema

cube(`users`, {
  sql_table: `users`,
 
  measures: {
    m_a: {
      type: `sum`,
      sql: `m_a`,
    },
    m_b: {
      type: `sum`,
      sql: `m_c`,
    },
    m_c: {
      type: `sum`,
      sql: `${m_a} + ${m_b}`,
    },
  },
 
  dimensions: {
    organization_id: {
      sql: `organization_id`,
      type: `number`,
      primary_key: true,
    },
 
    created_at: {
      sql: `created_at`,
      type: `time`,
    },
 
    country: {
      sql: `country`,
      type: `string`,
    },
  },
});