您可以将regexp_matches和regexp_split_to_tablepattern matching functions组合在一起，创建一个包含每条消息的字符串部分的表--每行包含一个文本部分和一个${…}替换部分，最后一行不包含替换变量。然后string_agg regate将所有部分重新组合到结果中。对表中的每一行使用子查询执行此操作：

SELECT (
  SELECT string_agg(concat(literal, data->>variable[1]), '')
  FROM (
    SELECT
      regexp_split_to_table(message, '\$\{([^}]*)\}') AS literal,
      regexp_matches(message, '\$\{([^}]*)\}', 'g') AS variable
  ) AS parts
) AS result
FROM example

（online demo）

试试这个：

with tmp_table(json_row, message) as (
    values 
    ('{"key1": 1, "key2":"first"}'::jsonb, 'this ${key2} is a test of ${key1}'),
    ('{"key1": 2, "key2":"second"}'::jsonb, 'this ${key2} is a test'),
    ('{"key1": 3, "key2":"third"}'::jsonb, 'this ${key1} is a test of ${key2}')
)
select json_row,replace(replace(message,'${key1}', json_row->>'key1'),'${key2}', json_row->>'key2') as message from tmp_table

db<>fiddle
输出