时间戳

首页 > 解决方案 > 如何在 PostgreSQL 中基于 order by 字段获得第二高到最低值的记录

sql - 如何在 PostgreSQL 中基于 order by 字段获得第二高到最低值的记录

问题描述

在这里，我有一个 JSON 字段，我想在数据所在的每个 id 上选择第二高到最低的记录

JSON看起来像这样

{
    "user": [
        {
            "user_name": "Devang",
            "user_weight": 0.7676846955248864
        },
        {
            "user_name": "Meet",
            "user_weight": 1.1021
        },
        {
            "user_name": "Devang",
            "user_weight": 0.16163873153859706
        },
        {
            "user_name": "Rajan",
            "user_weight": 0.22163873153859706
        }
    ],
    "address": [
        {
            "address_name": "India"
        }
    ]
}

我执行的查询是

WITH cte AS (
SELECT id, ('{user,'||index-1||'}')::text[] as json_path, value->'user_weight'
FROM user_table, jsonb_array_elements(json_field->'user')
WITH ordinality arr(value, index) WHERE arr.value->>'user_name' IN ('Devang', 'Meet') order by id, value->'user_weight' DESC
) select * from cte;

得到这样的记录

id  json_path   ?column?
1   {user,1}    1.1021
1   {user,0}    0.7676846955248864
1   {user,2}    0.16163873153859706
2   {user,0}    0.7676846955248864
2   {user,1}    0.07447325861051013

我想要这样的第二高到最低的记录

id  json_path   ?column?
1   {user,1}    1.1021 # this is the highest one in 1 id so I do not need this
1   {user,0}    0.7676846955248864
1   {user,2}    0.16163873153859706
2   {user,0}    0.7676846955248864   # this is the highest one in 2 ids so. I do not want
2   {user,1}    0.07447325861051013

输出将是

id  json_path   ?column?
1   {user,0}    0.7676846955248864
1   {user,2}    0.16163873153859706
2   {user,1}    0.07447325861051013

看这里演示

任何答案将不胜感激

标签： sqlpostgresqlcommon-table-expressionjsonb

解决方案

有很多方法可以做到这一点，但如果您想使用窗口函数，您可以使用 idRANK()或DENSE_RANK()权重，并消除外部查询中的第一行：

WITH cte AS (
  SELECT id, ('{user,'||index-1||'}')::text[] as json_path, (value->'user_weight')::text::numeric AS weight
  FROM user_table, jsonb_array_elements(json_field->'user')
  WITH ordinality arr(value, index) 
  WHERE arr.value->>'user_name' IN ('Devang', 'Meet') 
  ORDER BY id, value->'user_weight' DESC
) 
SELECT * FROM (
  SELECT cte.*, 
    RANK() OVER (PARTITION BY id ORDER BY id,weight DESC
                 RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS r
  FROM cte) j
WHERE r > 1;

演示：db<>fiddle

推荐阅读