如何在PostgreSQL中聚合具有可变数量的键的JSON数组？

Question

我在PostgreSQL表中有一系列的行，如下所示：

-[ RECORD 1 ]---------------------------------------------------------------------
student     | e04c0ae4709340cb8e03c52f444e723f
group       | 1
subgroup    | 1
variable    | VAR1
status      | { "track_A" : "Done", "track_B" : "Done", "track_C" : "To Do" }
-[ RECORD 2 ]---------------------------------------------------------------------
student     | e04c0ae4709340cb8e03c52f444e723f
group       | 1
subgroup    | 1
variable    | VAR2
status      | { "track_A" : "To Do", "track_B" : "Done", "track_C" : "To Do" }
-[ RECORD 3 ]---------------------------------------------------------------------
student     | 849d1e6a0c2b4530a2b550829df94556
group       | 0
subgroup    | 1
variable    | VAR3
status      | { "track_A" : "Done", "track_B" : "To Do", "track_C" : "To Do" }

我想按学生，组和子组对它们进行分组，并获取每个曲目的计数状态。 就像是：

-[ RECORD 1 ]---------------------------------------------------------------------
student     | e04c0ae4709340cb8e03c52f444e723f
group       | 1
subgroup    | 1
totals      | { "track_A" : {"done": 1, "to_do": 1}, {"track_B" : {"done": 0, "to_do": 2}, "track_C" : {"done": 0, "to_do": 2} }

问题在于轨道的数量可能会有所不同。 我确实知道它们的名称，但是它们不是静态的，因此无法进行简单的汇总。 有什么建议可以在PostgreSQL（9.5）中编写吗？ 我不想遍历所有轨道并进行汇总，因为该操作将花费一些时间。

Answer 1

你可以使用json_each_text为“unes​​t”的价值观和json_object_agg一遍结合起来。

数据：

DROP TABLE IF EXISTS tab;
CREATE TABLE tab(student VARCHAR(36), "group" INT, subgroup INT,
                 variable VARCHAR(20), status JSON);        

INSERT INTO tab(student, "group", subgroup, variable, status)
VALUES
('e04c0ae4709340cb8e03c52f444e723f',1,1,'VAR1'
,'{ "track_A" : "Done", "track_B" : "Done", "track_C" : "To Do" }'),
('e04c0ae4709340cb8e03c52f444e723f',1,1,'VAR2'
, '{ "track_A" : "To Do", "track_B" : "Done", "track_C" : "To Do" }')
,('849d1e6a0c2b4530a2b550829df94556',0,1,'VAR3'
,'{ "track_A" : "Done", "track_B" : "To Do", "track_C" : "To Do" }');

查询：

WITH cte AS
(
   SELECT student, "group", subgroup, k
     ,COUNT(CASE WHEN v='Done'  THEN 1 END) AS Done
     ,COUNT(CASE WHEN v='To Do' THEN 1 END) AS To_do
   FROM tab
   ,LATERAL json_each_text(status) s(k,v)
   GROUP BY student, "group", subgroup, k  
), cte2 AS
(
  SELECT student, "group", subgroup, k, json_object_agg(s.status, s.cnt) AS j
  FROM cte
  ,LATERAL (VALUES('Done', Done),('To Do', To_Do)) AS s(status, cnt)
  GROUP BY student, "group", subgroup, k   
)
SELECT student, "group", subgroup
      ,json_object_agg(k, j) AS totals
FROM cte2
GROUP BY student, "group", subgroup;

输出：