SQL 主表多个左连接与键值对查找

Question

Application - Legacy Enterprise System designed 20 yrs back

I have a Product table with 50+ Rows where each column is an attribute, like

ID, Name, Brand, Category, Category, Gender, Item Code, Color Code......

Now most(around 43/50) of the columns have a key(or id) of the value (eg For Brand, key b02 is stored, which translates to Brown Bag Brand, Key-Value is stored in different table)

This design is not exactly EAV, but different one done in a legacy application by a 3rd part application.

We are trying to get the data out of the DB into more Report/Analytics Friendly Format where-by I join each key based column in Master Table with Key-Value (using left join) and arrive at final table and store in different DB for Reporting.

This is working fine for smaller set, but if the data is large (>20K Rows), its getting slow as there are too many left joins

I see no option other than this approach, but if any one can give me more ideas to make it more efficient and fast, it will be helpful.

SELECT Product.id, Product.name, s1.value as 'Brand', s2.value as 'Category'
  , s3.value as 'Gender'
FROM Product
LEFT JOIN Attribute s1 ON s1.id = Product.brand AND s1.attribute = 'brand'
LEFT JOIN Attribute s2 ON s2.id = Product.category AND s2.attribute = 'category'
LEFT JOIN Attribute s3 ON s3.id = Product.gender AND s3.attribute = 'gender'

... 40 more

Answer 1

One way that will probably improve this query's performance is to use conditional aggregation instead of multiple joins:

SELECT   P.id
        ,P.name
        ,MAX(CASE WHEN A.Id = P.brand AND A.attribute = 'brand' THEN A.value END) as 'Brand'
        ,MAX(CASE WHEN A.Id = P.category AND A.attribute = 'category' THEN A.value END) as 'Category'
        ,MAX(CASE WHEN A.Id = P.gender AND A.attribute = 'gender' THEN A.value END) as 'Gender'
        -- 40 more
FROM Product As P
LEFT JOIN Attribute As A
    ON A.id IN(P.brand, P.category, P.gender /* 40 more */) 
    AND A.attribute IN('brand', 'category', 'gender' /* 40 more */)    
-- The group by clause must contain all the columns that are not translated using the attribute table
GROUP BY P.id, P.name