从 jsonb 列中获取特定值

Question

我有一张桌子Customer

• customerId(int)
• 客户名称（字符串）
• customerOrders(jsonB)

customerOrders 具有以下结构：

{ 
 "nodeValue":[
   { 
      "key": "key1",
      "value": "value1"
   },
   { 
      "key": "key2",
      "value": "value2"
   },
   { 
      "key": "key3",
      "value": "value3"
   },
   { 
      "key": "key4",
      "value": "value4"
   },
   { 
      "key": "key5",
      "value": "value5"
   }
 ]
}

我正在尝试获取key = 'key3' AND key = 'key4' 行中nodeValue 的值。

例如：返回'key3'和'key4'的值，WHERE key = 'key3' AND key = 'key4'。

我正在尝试做类似的事情：

SELECT value, value 
from public.customers 
where nodeValue.key3 = 'key3' 
  AND nodeValue.key4 = 'key4'

Answer 1

with orders as 
(
   select jsonb_array_elements(customerOrders->'nodeValue') as orders
   from customers
) 
select orders->'value' as val 
from orders 
where 
    orders->>'key' = 'key3' 
or  orders->>'key' = 'key4';

结果：

值1

值2

请注意使用操作符 ->> 来获取文本的值

db-fiddle

Answer 2

获取key = 'key3' AND key = 'key4'所在行的nodeValue的值。

最好使用jsonb "contains" operator @>：

SELECT customerid, customerOrders->'nodeValue'
FROM   customer
WHERE  customerOrders->'nodeValue' @> '[{"key": "key3"}]'
AND    customerOrders->'nodeValue' @> '[{"key": "key4"}]';

为了使其 快速 为大表提供索引支持。理想情况下是jsonb_path_ops 索引：

CREATE INDEX customer_nodeValue_idx ON customer
USING gin ((customerOrders->'nodeValue') jsonb_path_ops);  -- parentheses required

见：

• What's the proper index for querying structures in arrays in Postgres jsonb?

你的问题的更高版本有点棘手：

返回'key3'和'key4'的值

SELECT c.customerid, o.values
FROM   customer c
CROSS  JOIN LATERAL (
   SELECT ARRAY(
      SELECT o.ord->>'value'
      FROM   jsonb_array_elements(c.customerOrders->'nodeValue') o(ord)
      WHERE (o.ord->>'key' = 'key3' OR
             o.ord->>'key' = 'key4')
      )
   ) o(values)
WHERE  c.customerOrders->'nodeValue' @> '[{"key": "key3"}]'
AND    c.customerOrders->'nodeValue' @> '[{"key": "key4"}]';

首先过滤符合条件的行，如上（使用快速索引查找）。然后取消嵌套 JSON 文档并在 LATERAL 子查询中构建您的答案。应该是最简单、最干净和最快的。有关该技术，请参阅：

• How to turn JSON array into Postgres array?

我在两个查询的结果中添加了customerid 以识别行。这是可选的。

dbfiddle here（同时演示）

设计

您可以将臃肿的 JSON 布局简化为：

'{
    "key1": "value1",
    "key2": "value2",
    "key3": "value3",
    "key4": "value4",
    "key5": "value5"
}'

或者至少：

'[
    {
        "key1": "value1"
    },
    {
        "key2": "value2"
    },
    {
        "key3": "value3"
    },
    {
        "key4": "value4"
    },
    {
        "key5": "value5"
    }
]'

将使一切变得更简单、更小、更快。

Answer 3

我怀疑你真正想要的是

SELECT
  (SELECT orders->>'value'
    FROM json_array_elements(customerOrders->'nodeValue') AS orders
    WHERE orders->>'key' = 'key3'
  ) AS value_key3,
  (SELECT orders->>'value'
    FROM json_array_elements(customerOrders->'nodeValue') AS orders
    WHERE orders->>'key' = 'key4'
  ) AS value_key4
FROM public.customers;

但实际上您的数据结构不适合此目的。使用具有键和值作为属性的对象，而不是具有键值对的数组。这样，您可以轻松访问结果

SELECT
  customerOrders->'nodeValue'->>'key3' AS value_key3,
  customerOrders->'nodeValue'->>'key4' AS value_key4
FROM public.customers;

Answer 4

首先你需要创建一个类型（或表）

CREATE TYPE key_value_type AS (
    key text,
    value text
);

然后就可以使用jsonb_populate_recordset函数了

SELECT
    r_key1.value AS key1,
    r_key2.value AS key2
FROM
    customers c
    JOIN LATERAL jsonb_populate_recordset(NULL::key_value_type, customerOrders -> 'nodeValue') r_key1 
      ON r_key1.key = 'key1'
    JOIN LATERAL jsonb_populate_recordset(NULL::key_value_type, customerOrders -> 'nodeValue') r_key2 
      ON r_key2.key = 'key2'

或者使用jsonb_to_recordset 函数（不创建类型）

SELECT
    r_key1.value AS key1,
    r_key2.value AS key2
FROM
    customers c
    JOIN LATERAL jsonb_to_recordset(customerOrders -> 'nodeValue') AS r_key1 (key text,
        value text) ON r_key1.key = 'key1'
    JOIN LATERAL jsonb_to_recordset(customerOrders -> 'nodeValue') AS r_key2 (key text,
        value text) ON r_key2.key = 'key2'

这是两种选择的结果；

| key1   | key2   |
| ------ | ------ |
| value1 | value2 |