Feature request - Specify join relationship such as many to one, one to many, one to one, and many to many joins

Description

Specify join relationship such as many to one, one to many, one to one, and many to many joins.

Here’s the docs for this in looker

What problem does this solve?

This lets the “explore” house lots of information. Empowering users to do aggregations but also dig into specific rows. With looker the relationship description ensures that upon aggregations the measure are summed, counted, etc correctly. Otherwise with we will have incorrect counting due to multiple one to many joins.

This functionality is a big part of looker’s secret sauce and one of the reasons why looker is so useful.

Example (using redshift)

generate example tables

CREATE TABLE sale AS
(select 1 as sale_id);

CREATE TABLE gift_card AS
(SELECT 1 AS gift_card_id, 1 as sale_id, 20 as cash_value
UNION ALL SELECT 2 AS gift_card_id, 1 as sale_id, 20 as cash_value);

CREATE TABLE cash AS
(SELECT 1 AS TRANSACTION_id, 1 as sale_id, 10 as cash_value
UNION ALL SELECT 2 AS TRANSACTION_id, 1 as sale_id, 10 as cash_value);

Files

v_cash.sql

select * from cash

v_gift_card.sql

select * from gift_card

v_sale.sql

select * from sale

v_sale.yml

version: 2
models:
  - name: v_sale
    meta:
      joins:
        - join: v_cash
          sql_on: ${v_sale.sale_id} = ${v_cash.sale_id}
        - join: v_gift_card
          sql_on: ${v_sale.sale_id} = ${v_gift_card.sale_id}
    columns:
      - name: sale_id
        meta:
          dimension:
            number
  - name: v_cash
    columns:
      - name: transaction_id
        meta:
          dimension:
            number
      - name: sale_id
        meta:
          dimension:
            number
      - name: cash_value
        meta:
          metrics:
            total_cash_value:
              type: sum
  - name: v_gift_card
    columns:
      - name: gift_card_id
        meta:
          dimension:
            number
      - name: sale_id
        meta:
          dimension:
            number
      - name: cash_value
        meta:
          metrics:
            total_gift_cash_value:
              type: sum

Here’s what the output looks like in lightdash The answer should be 20 and 40 not 40 and 80 (double counts due to join mechanism)

Closing thoughts

Maybe this is a design pattern / feature where lightdash and looker bifurcate? Maybe this is bad design pattern and we should handle this in the dbt joins via groups bys in the tables to be joined to. Would be curious to hear thoughts!

[TuringLovesDeathMetal]

ah yes, symmetric aggregates: Looker’s secret sauce.

We have an open issue for dealing with symmetric aggregates here: https://github.com/lightdash/lightdash/issues/961

Defining the join relationships will definitely help with this too! (if you know that you’re defining a many-to-many or many-to-one join, then we could warn you that your metrics could be wrong). And, agreed, it gives you even more power to your tables 💪

[olly-writes-code]

not stale!