PostgreSql：使用横向连接的 Json 数组到行

Question

我的表的详细信息字段中有两个以下 JSON 数组，需要评估查询，因为我在另一个关系表中使用。

{
    "city": "London",
    "name": "Sainburry",
    "quantities": [112, 145, 222, 122, 124],
    "prices": [4, 4, 4, 0, 3],
    "dates": ["13.05.2020", "14.05.2020", "15.05.2020", "16.05.2020", "17.05.2020"]
}

我想为此 JSON 数组评估以下查询：

select quantities,
       prices,
       AVG(quantities/prices::float) as ratio
from my_table
where city = 'London'
group by quantities, prices;

我使用了以下查询和许多类似的查询，包括横向连接：

select q.*
from my_table mt
  cross join json_array_elements_text(details -> 'quantities') as q

但是，当通过交叉连接将其他字段（价格和日期）添加到查询中时，行数会成倍增加。所以，我正在寻找一个新功能Lateral Join 来使用，但无法正确应用。如何在PostgreSQL中使用Lateral Join获得我之前查询的结果？任何帮助将不胜感激。

更新：

这里是fiddle。如果我成功地将 json 数组值转换为行而不进行相乘（应返回 5 条记录），我可以评估所需的结果。只需帮助我使用 lateral 连接和 json_array_elements_text 将 json 数组转换为行。

Answer 1

这是你想要的吗？

    -- just simulate table:
with my_table(details) as(
values
('{
"city": "London",
"name": "Sainburry",
"quantities": [112, 145, 222, 122, 124],
"prices": [4, 4, 4, 0, 3],
"dates": ["13.05.2020", "14.05.2020", "15.05.2020", "16.05.2020", "17.05.2020"]
}'::json)
)


-- here is query:
select  
my_table.details->>'city',  u.quantities, u.prices  
from my_table
JOIN LATERAL UNNEST( 
    ARRAY(SELECT json_array_elements_text(details->'quantities')) ,
    ARRAY(SELECT json_array_elements_text(details->'prices')) 
) u(quantities, prices) ON TRUE
WHERE
my_table.details->>'city' = 'London'

见demo

Answer 2

由于数组中的顺序，您似乎需要WITH ORDINALITY 和LEFT JOIN LATERALs 来分别匹配数组的相应元素：

SELECT q.elm AS quantities, p.elm AS prices, 
       AVG(p.elm::float/q.elm::float) AS ratio
  FROM my_table t0
  LEFT JOIN LATERAL jsonb_array_elements(details -> 'quantities') 
    WITH ORDINALITY AS q(elm, i) ON TRUE
  LEFT JOIN LATERAL jsonb_array_elements(details -> 'prices') 
    WITH ORDINALITY AS p(elm, i) ON q.i = p.i
  LEFT JOIN LATERAL jsonb_array_elements(details -> 'dates') 
    WITH ORDINALITY AS d(elm, i) ON d.i = q.i
 WHERE t0.details ->> 'city' = 'London'   
 GROUP BY q.elm, p.elm;

Demo

Answer 3

select
  quantity, price,
  avg(quantity/price) as ratio
from my_table cross join lateral (
  select
    json_array_elements_text(details->'dates') as dates,
    (json_array_elements_text(details->'quantities'))::numeric as quantity,
    (json_array_elements_text(details->'prices'))::numeric as price) as data
where details->>'city' = 'London'
group by quantity, price;

Answer 4

基本上，what Abelisto said。只是一些改进和一些解释：

SELECT to_date(d, 'DD.MM.YYYY') AS date -- ①
     , quantity, price
     , round(price / quantity, 4) AS ratio -- ③, ④
FROM   my_table
CROSS  JOIN LATERAL ( -- ②
   SELECT json_array_elements_text(details->'dates'     )          AS d -- ①
        , json_array_elements_text(details->'quantities')::int     AS quantity -- ③
        , json_array_elements_text(details->'prices'    )::numeric AS price -- ③
   ) AS data
WHERE  details->>'city' = 'London';

db小提琴here

① 默认情况下，日期字符串根据区域设置和会话变量进行解释。使用to_date() 以干净的方式进行操作。

② SELECT 列表中的多个 set-returning 函数在 Postgres 10 之前的行为方式令人惊讶，如果结果行数并不完全相同。 （考虑升级。无论如何。）请参阅：

• What is the expected behaviour for multiple set-returning functions in SELECT clause?

③ 在您的原始查询中，AVG(quantities/prices::float) 与group by quantities, prices 结合使用毫无意义。 quantities/prices 本身也没有。我按照我认为合适的方式进行了修复，并输入了round() 来格式化输出。

④如果数量可以为0，用NULLIF防被0除：

     , round(price / NULLIF(quantity, 0), 4) AS ratio