具有别名的同一列上的多个连接

Question

我正在尝试将多个表连接在一起。 需要在同一列上连接 2 个表。 但是，我需要使用指定的列别名，因为这些值仅在替换文本后才会匹配。 所以我有;

table1:
| date_time | product_quote | category_id

table2: 
| product_id | category_id

table3: 
| date | product_id | sales

table4: 
| category_id | category_name

'product_quote' 的值是一个包含 product_id 的字符串。 我可以通过使用 REPLACE 来解决这个问题。

期望的结果：

| year | month | week | day | product_id | quotes | sales | category_name 

我正在尝试使用以下查询，但它抱怨“product_id”不明确。 当我在 table2.product_id 上进行第三次连接时，它会执行查询，但我得到了错误的销售数字。 也不确定如何（重新）在 join 语句中使用日期别名（年、月等）。

SELECT  
    date_part('year', table1.date_time) as "year", 
    date_part('month', table1.date_time) as "month", 
    date_part('week', table1.date_time) as "week", 
    date_part('day', table1.date_time) as "day",
    REPLACE(REPLACE(table1.product_quote,'/quotes/',''),'.pdf','') as "product_id", 
    count(table1.product_quote) as quotes,
    table3.sales,
    table4.category_name
from table1
left join table2 ON "product_id" = table2.product_id
left join table4 ON table2.category_id = table4.category_id
left join table3 ON "product_id" = table3.product_id AND 
                        date_part('year', table3.date) = date_part('year', table1.date_time) AND 
                        date_part('month', table3.date) = date_part('month', table1.date_time) AND 
                        date_part('day', table3.date) = date_part('day', table1.date_time)  
where date_part('year', table1.date_time) = '2020'
group by year, month, week, day, table2.product_id, table1.product_quote, table4.category_name, table3.sales

这是我应该作为 1 个带有 3 个连接的查询做的事情，还是我需要使用子查询？ （如果是的话，怎么做？）

Answer 1

我认为您假设您是基于SELECT部分中的“product_id”投影（或您称之为“别名”）加入的。 但事实并非如此。 在 SQL 中， JOIN总是首先发生，因此JOIN部分不“知道” SELECT -Part。

要解决您的问题，您需要

• 在JOIN中使用两次REPLACE表达式
• 始终在表前加上 product_id

---

SELECT  
    date_part('year', table1.date_time) as "year", 
    date_part('month', table1.date_time) as "month", 
    date_part('week', table1.date_time) as "week", 
    date_part('day', table1.date_time) as "day",
    REPLACE(REPLACE(table1.product_quote,'/quotes/',''),'.pdf','') as "product_id", 
    count(table1.product_quote) as quotes,
    table3.sales,
    table4.category_name
from table1
left join table2 ON REPLACE(REPLACE(table1.product_quote,'/quotes/',''),'.pdf','') = table2.product_id
left join table4 ON table2.category_id = table4.category_id
left join table3 ON REPLACE(REPLACE(table1.product_quote,'/quotes/',''),'.pdf','') = table3.product_id AND 
                        date_part('year', table3.date) = date_part('year', table1.date_time) AND 
                        date_part('month', table3.date) = date_part('month', table1.date_time) AND 
                        date_part('day', table3.date) = date_part('day', table1.date_time)  
where date_part('year', table1.date_time) = '2020'
group by year, month, week, day, table2.product_id, table1.product_quote, table4.category_name, table3.sales

至少以上是基本解决方案。 如果您在join之前使用子选择对table1进行投影，您也可以使用别名：

SELECT  
    t1.year,
    t1.month,
    t1.week,
    t1.day,
    t1.product_id,
    t1.quotes,
    table3.sales,
    table4.category_name
FROM
(
    SELECT 
    date_part('year', table1.date_time) as "year", 
    date_part('month', table1.date_time) as "month", 
    date_part('week', table1.date_time) as "week", 
    date_part('day', table1.date_time) as "day",
    REPLACE(REPLACE(table1.product_quote,'/quotes/',''),'.pdf','') as "product_id", 
    count(table1.product_quote) as quotes,
    from table1
) as t1
left join table2 ON t1.product_id = table2.product_id
left join table4 ON table2.category_id = table4.category_id
left join table3 ON t1.product_id = table3.product_id AND 
                        date_part('year', table3.date) = t1.year AND 
                        date_part('month', table3.date) = t1.month AND
                        date_part('day', table3.date) = t1.day
where t1.year = '2020'
group by t1.year, t1.month, t1.week, t1.day, table2.product_id, t1.product_quote, table4.category_name, table3.sales