Oracle SQL连接三个表并按最大值排序

Question

我有以下三个表：

CREATE TABLE CUSTOMERS
(customerID     INT     PRIMARY KEY,
customerZip     VARCHAR(15) NOT NULL);

CREATE TABLE VEHICLES
(vehicleVIN     VARCHAR(25) PRIMARY KEY,
vehicleModel        VARCHAR(15) NOT NULL);

CREATE TABLE SALES
(saleID         INT     PRIMARY KEY,
customerID      INT,
vehicleVIN      VARCHAR(25),
CONSTRAINT SALES_FK1 FOREIGN KEY (customerID) REFERENCES CUSTOMERS(customerID),
CONSTRAINT SALES_FK2 FOREIGN KEY (vehicleVIN) REFERENCES VEHICLES(vehicleVIN));

我正在尝试开发一个查询来连接这三个表并生成一个输出，该输出显示按型号的销售总数和按邮政编码的销售总数，首先按最高值排序。输出应该类似于：

vehicleModel    Sales_By_Model    customerZip   Sales_By_ZIP
        S-10                12          18956              3
   Silverado                10          22789              2
       F-150                 9          12345              2

我已经尝试了以下代码，但我不认为这是正确对待这头公牛的方法，因为它没有将车辆模型合并为 Sales_By_Model 的总数：

SELECT DISTINCT v.vehicleModel, COUNT(*) OVER (PARTITION BY s.vehicleVIN) "SALES_BY_MODEL", c.customerZip, COUNT(*) OVER (PARTITION BY c.customerZip )"SALES_BY_ZIP" 
FROM SALES s, VEHICLES v, CUSTOMERS c
WHERE s. vehicleVIN = v. vehicleVIN
and  c. customerID = s. customerID    
ORDER BY 2 DESC , 4 DESC;

这是我从上面的查询中得到的输出：

VEHICLEMODEL    SALES_BY_MODEL CUSTOMERZIP     SALES_BY_ZIP
accord                       1 89523                      6
altima                       1 89523                      6
escalade                     1 89523                      6
f-150                        1 89523                      6
impala                       1 89523                      6

尽管查询确实提取了必要的数据并连接了表，但它并没有组合车辆模型来增加 SALES_BY_MODEL 的计数。 customerZIP 也是如此，尽管它似乎至少正确计算了 zip 的数量。

我的问题是：有没有办法查询这些表以获取结合了 vehicleModel 和 customerZip 计数的结果？如果是这样，我该怎么做？通过显式调用 JOIN，我知道我上面使用的查询不是完全正确的 SQL 语法。我的目标是在 ONE SELECT Query 中完成此操作。

提前感谢您的指导/帮助！

编辑：下面是几行示例输入数据：

INSERT INTO CUSTOMERS
(customerID,customerFirName,customerLasName,customerMiName,customerStreet,customerState,customerCity,customerZip)
VALUES
(1,'Steven','Christman','J','11111 Address Way','Maryland','Hollywood','20636');
INSERT INTO CUSTOMERS
(customerID,customerFirName,customerLasName,customerMiName,customerStreet,customerState,customerCity,customerZip)
VALUES
(2,'Bob','Seagram','A','22222 Seagram Lane','Texas','Houston','77001');
INSERT INTO CUSTOMERS
(customerID,customerFirName,customerLasName,customerMiName,customerStreet,customerState,customerCity,customerZip)
VALUES
(3,'Sally','Anderson','P','33333 Pheonix Drive','Arizona','Pheonix','85001');

INSERT INTO VEHICLES
(vehicleVIN,vehicleType,vehicleMake,vehicleModel,vehicleWhereFrom,vehicleWholesaleCost,vehicleTradeID)
VALUES
('147258HHE91K3RT','compact','chevrolet','spark','Maryland',20583.00,NULL);
INSERT INTO VEHICLES
(vehicleVIN,vehicleType,vehicleMake,vehicleModel,vehicleWhereFrom,vehicleWholesaleCost,vehicleTradeID)
VALUES
('789456ERT0923RFB6','Midsize','ford','Taurus','washington, d.c.',25897.22,1);
INSERT INTO VEHICLES
(vehicleVIN,vehicleType,vehicleMake,vehicleModel,vehicleWhereFrom,vehicleWholesaleCost,vehicleTradeID)
VALUES
('1234567890QWERTYUIOP','fullsize','Lincoln','towncar','Virginia',44222.10,NULL);

INSERT INTO SALES
(saleID,grossSalePrice,vehicleStatus,saleDate,saleMileage,customerID,salespersonID,vehicleVIN)
VALUES
(1,25987.28,'sold',date '2012-10-15',10,1,1,'147258HHE91K3RT');
INSERT INTO SALES
(saleID,grossSalePrice,vehicleStatus,saleDate,saleMileage,customerID,salespersonID,vehicleVIN)
VALUES
(2,29999.99,'sold',date '2012-10-17',50087,2,2,'789456ERT0923RFB6');
INSERT INTO SALES
(saleID,grossSalePrice,vehicleStatus,saleDate,saleMileage,customerID,salespersonID,vehicleVIN)
VALUES
(3,47490.88,'sold',date '2012-11-05',30,3,3,'1234567890QWERTYUIOP');

您可能会注意到，上面的输入数据反映了对我在原始帖子中未列出的其他一些属性的输入（我在此处显示了表格创建），因为这些属性中的大多数与此处的目标无关。

Answer 1

实现这一点的一个示例方法表明，您实际上是在运行两个独立的查询，并且只是假装它们是相关的。

这样做会增加执行成本，降低测试和/或维护的能力，阻止代码重用，并且通常会违反我能想到的所有相关软件工程原则。

那么，这就是如何做一件非常糟糕的事情......

WITH
  ranked_model_sales AS
(
  SELECT
    vehicles.vehicleModel,
    COUNT(*)                                     AS total_model_sales,
    ROW_NUMBER() OVER (ORDER BY COUNT(*) DESC)   AS rank_id
  FROM
    sales
  INNER JOIN
    vehicles
      ON sales.vehicleVIN = vehicles.vehicleVIN
  GROUP BY
    vehicles.vehicleModel
),
  ranked_zip_sales AS
(
  SELECT
    customers.customerZip,
    COUNT(*)                                     AS total_zip_sales,
    ROW_NUMBER() OVER (ORDER BY COUNT(*) DESC)   AS rank_id
  FROM
    sales
  INNER JOIN
    customers
      ON sales.customerID= customers.customerID
  GROUP BY
    customers.customerZip
)
SELECT
  m.vehicleModel,
  m.total_model_sales,
  z.customerZip,
  z.total_zip_sales
FROM
  ranked_model_sales    m
FULL OUTER JOIN
  ranked_zip_sales      z
    ON m.rank_id = z.rank_id
ORDER BY
  COALESCE(m.rank_id, z.rank_id)