Say you want to set up a Liquibase changelog to insert some data, but you don't know the IDs of some entity, and you can express it as a query. With Liquibase's computed column attribute, you can embed subqueries directly into seed data:

(select id from table2 where some_column = 'foo');BAR
(select id from table2 where some_column = 'foo');BAZ
The parentheses are important!

Mind the parentheses! Liquibase converts a loadData change into an INSERT INTO ... VALUES (...) statement, so your subqueries need to be tucked in and cozy.

Your changelog should look something like this:

  - changeSet:
      id: 1
      author: xxx
      objectQuotingStrategy: QUOTE_ONLY_RESERVED_WORDS
        - onFail: MARK_RAN
        - sqlCheck:
            expectedResult: 0
            sql: select count(*) from my_table
        - loadData:
            tableName: my_table
            file: db/seeds/my_table.csv
            separator: ;
            encoding: UTF-8
              - column:
                  header: column1
                  name: column1
                  computed: true
              - column:
                  header: column2
                  name: column2
                  computed: false
It's usually a good idea to use preconditions when you're modifying data in a table.

And that's it! Let me know if this is useful to you! Liquibase's documentation is pretty rough in parts, and it took me a little bit to figure out how to assemble these pieces into a solution to my problem.