涉及集合比较的SQL查询

Question

Background

Products can be sold as bundles. Following tables are present: products , bundles , bundles_products , orders , orders_products .

An order would be said to "contain" a bundle if it contains all the bundle's products.

Problem

How would one go about counting orders for bundles?

Example

products table

id  name
1   broom
2   mug
3   spoon
4   candle

bundles table

id  name
1   dining
2   witchcraft

bundles_products table

bundle_id product_id
1         2
1         3
2         1
2         4

orders_products table

order_id  product_id
1000      1
1000      3
1001      1
1001      2
1001      3

The query would return the following table:

bundle     orders
dining     1
witchcraft 0

Notes

The example intentionally misses the orders table as it is not relevant what it contains.

Of course, this could be approached imperatively, by writing some code and gathering the data, but I was hoping there is a declarative, SQL way of querying for this kind of things?

One idea I had was to use a GROUP_CONCAT to concatenate all the products of a bundle and somehow compare that with products of each order. Still, a long way from clear.

Answer 1

One way is to use two Derived Tables (subqueries). In first subquery, we will fetch the total number of unique products for every bundle. In the second subquery, we will fetch the total products in an order, for a combination of order and bundle.

We will LEFT JOIN them on bundle_id as well as matching the total count of products per bundle in them. Eventually, we will do a grouping on bundle, and count the number of orders matching successfully.

SELECT dt1.id              AS bundle_id,
       dt1.name            AS bundle,
       Count(dt2.order_id) AS orders
FROM   (SELECT b.id,
               b.name,
               Count(DISTINCT bp.product_id) AS total_bundle_products
        FROM   bundles AS b
               JOIN bundles_products AS bp
                 ON bp.bundle_id = b.id
        GROUP  BY b.id,
                  b.name) AS dt1
       LEFT JOIN (SELECT op.order_id,
                         bp.bundle_id,
                         Count(DISTINCT op.product_id) AS order_bundle_products
                  FROM   orders_products AS op
                         JOIN bundles_products AS bp
                           ON bp.product_id = op.product_id
                  GROUP  BY bp.bundle_id,
                            op.order_id) AS dt2
              ON dt2.bundle_id = dt1.id
                 AND dt2.order_bundle_products = dt1.total_bundle_products
GROUP  BY dt1.id,
          dt1.name 

SQL Fiddle DEMO

Answer 2

Here's the brief example, which lacks some parts, I omitted because I don't know precise database structure. Logic is such:

1. Temp table is generated, which consists of 3 rows - order, count of products related to bundle, count of products in bundle
2. Then we select only orders from this table in which we have those last two variables equal

select count(order_id) from orders 
    left join(
        select count(*) from bundles_products as bundle_amount, 
                  sum(case when orders_products in (
                  select names from bundles_products where bundle_id='1') then 1 else 0) as order_total, 
                  orders.order_id 
            left join product on bundle_products.product_id = products.product_id
            left join orders on products.product_id = orders_products.product_id
        where bundle_products.bundle_id ='1'
        ) as my_table 
        on orders.order_name = my_table.orders 
        where my_table.bundle_amount = my_table.order_total

Edit: I posted this as a response to previous version of the question, without detailed explanation.

Edit2: fixed query a bit. It can be starting point. Logic is still the same, you can get amount of orders for each bundle_id using it