Question

下表仅供练习。我将在更大的表上使用代码。

SELECT *
FROM price_practice;

给出

 id   company       dt       price
'16', 'Amex', '2015-07-01', '5.00'
'17', 'Amex', '2015-07-02', '5.10'
'18', 'Amex', '2015-07-03', '5.00'
'19', 'Amex', '2015-07-06', '5.88'
'20', 'Amex', '2015-07-07', '4.21'
'21', 'Citi', '2015-07-01', '1.00'
'22', 'Citi', '2015-07-02', '1.10'
'23', 'Citi', '2015-07-03', '1.00'
'24', 'Citi', '2015-07-06', '0.88'
'25', 'Citi', '2015-07-07', '1.01'
'26', 'Amex', '2015-07-08', '5.23'
'27', 'Amex', '2015-07-09', '5.35'
'28', 'Amex', '2015-07-10', '5.55'
'29', 'Amex', '2015-07-13', '5.88'
'30', 'Amex', '2015-07-14', '6.01'
'31', 'Citi', '2015-07-08', '0.95'
'32', 'Citi', '2015-07-09', '0.83'
'33', 'Citi', '2015-07-10', '0.79'
'34', 'Citi', '2015-07-13', '0.72'
'35', 'Citi', '2015-07-14', '0.59'

以下代码段计算从一个日期到下一个日期的价格变化百分比。

SELECT x.id, x.company, x.dt, x.price, (x.price - y.price)/y.price AS 'Change'
FROM 
(
    SELECT a.id AS aid, MAX(b.id) AS aPrevid
    FROM price_practice a
    INNER JOIN price_practice b
    WHERE a.id > b.id
    AND a.company = b.company
    GROUP BY a.id
) Sub1
INNER JOIN price_practice x ON Sub1.aid = x.id
INNER JOIN price_practice y ON Sub1.aPrevid = y.id
ORDER BY x.id DESC

按照预期，它返回

 id   company       dt       price      Change
'35', 'Citi', '2015-07-14', '0.59', '-0.180556'
'34', 'Citi', '2015-07-13', '0.72', '-0.088608'
'33', 'Citi', '2015-07-10', '0.79', '-0.048193'
'32', 'Citi', '2015-07-09', '0.83', '-0.126316'
'31', 'Citi', '2015-07-08', '0.95', '-0.059406'
'30', 'Amex', '2015-07-14', '6.01', '0.022109'
'29', 'Amex', '2015-07-13', '5.88', '0.059459'
'28', 'Amex', '2015-07-10', '5.55', '0.037383'
'27', 'Amex', '2015-07-09', '5.35', '0.022945'
'26', 'Amex', '2015-07-08', '5.23', '0.242280'
'25', 'Citi', '2015-07-07', '1.01', '0.147727'
'24', 'Citi', '2015-07-06', '0.88', '-0.120000'
'23', 'Citi', '2015-07-03', '1.00', '-0.090909'
'22', 'Citi', '2015-07-02', '1.10', '0.100000'
'20', 'Amex', '2015-07-07', '4.21', '-0.284014'
'19', 'Amex', '2015-07-06', '5.88', '0.176000'
'18', 'Amex', '2015-07-03', '5.00', '-0.019608'
'17', 'Amex', '2015-07-02', '5.10', '0.020000'

以下代码段完全不同：它按单个价格对每个公司的价格进行排名。

SELECT    ( 
            CASE company 
            WHEN @curType 
            THEN @curRow := @curRow + 1 
            ELSE @curRow := 1 AND @curType := company END
          ) + 1 AS rank,
          id,
          company,
          dt,
          price
FROM      price_practice,
          (SELECT @curRow := 0, @curType := '') r
ORDER BY  company DESC, price DESC;

按照预期，它返回

rank  id   company       dt       price
'1', '22', 'Citi', '2015-07-02', '1.10'
'2', '25', 'Citi', '2015-07-07', '1.01'
'3', '23', 'Citi', '2015-07-03', '1.00'
'4', '21', 'Citi', '2015-07-01', '1.00'
'5', '31', 'Citi', '2015-07-08', '0.95'
'6', '24', 'Citi', '2015-07-06', '0.88'
'7', '32', 'Citi', '2015-07-09', '0.83'
'8', '33', 'Citi', '2015-07-10', '0.79'
'9', '34', 'Citi', '2015-07-13', '0.72'
'10', '35', 'Citi', '2015-07-14', '0.59'
'1', '30', 'Amex', '2015-07-14', '6.01'
'2', '19', 'Amex', '2015-07-06', '5.88'
'3', '29', 'Amex', '2015-07-13', '5.88'
'4', '28', 'Amex', '2015-07-10', '5.55'
'5', '27', 'Amex', '2015-07-09', '5.35'
'6', '26', 'Amex', '2015-07-08', '5.23'
'7', '17', 'Amex', '2015-07-02', '5.10'
'8', '18', 'Amex', '2015-07-03', '5.00'
'9', '16', 'Amex', '2015-07-01', '5.00'
'10', '20', 'Amex', '2015-07-07', '4.21'

问题是：

如何按百分比变化对观察进行排名？

我想你可以将百分比变化数据保存在新列中，然后对其进行排名，但我怀疑这不是最好的方法。我将进行许多类似的计算（例如每周变化百分比，方差等），我有大约3,000,000个观察值，因此表格会快速增长。如果这是唯一的方法，我会，但我认为结合上面的两个片段来计算百分比变化和一次性排名会更好。或者你怎么看？

我确信你可以从我的问题中看出来，我是MySQL的初学者。任何有关如何进行的建议表示赞赏！

在MySQL中按计算变量排名

0 个答案: