如果它发生在另一个时间戳之后，则选择时间戳的最小差异答案

【问题标题】：Select min difference of timestamp if it occurred after another timestamp如果它发生在另一个时间戳之后，则选择时间戳的最小差异
【发布时间】：2014-10-08 12:47:33
【问题描述】：

我有一个表格，其中显示了customer_id、product_id、browse_date、purchase_date，以及浏览日期和购买日期之间的差异。它看起来像这样。

id  pID  b_Date     p_Date       Diff
1   001  7/20/2014  7/20/2014    0
1   001  7/20/2014  7/20/2014    0
1   002  7/20/2014  7/20/2014    0
2   001  7/20/2014  7/20/2014    0
2   001  7/20/2014  8/01/2014    -12
2   002  7/25/2014  8/01/2014    -8
2   002  7/26/2014  8/01/2014    -7
2   002  7/28/2014  8/01/2014    -5
2   002  7/28/2014  8/01/2014    -5

我试图找出客户在购买前提前多少天开始浏览特定产品。

但是，对于客户 2，他进行了两次购买。现在，他在 2014 年 7 月 20 日进行的浏览不应算作他在 2014 年 8 月 1 日进行的购买浏览，因为这与 7/20 的购买有关。

我想知道如何获得在最后一次购买后发生浏览的每个客户的 MIN Diff。另外，我不能在查询中使用 pID（产品 ID）。

我使用的是 teradata 13.1

【问题讨论】：

标签： sql teradata

【解决方案1】：

您的问题对我来说有点含糊，但我认为您是说您希望客户id 的给定购买的最低diff 与pID 无关。这是完成该任务的检查和后续查询。

以下是根据您提供的数据集得出的结果：

id  Max_b_Date  p_Date      Min_Diff
1   2014-07-20  2014-07-20      0
2   2014-07-20  2014-07-20      0
2   2014-07-28  2014-08-01      -5

以及我用来在 Teradata 13.1 中重新创建您的问题的代码

CREATE MULTISET TABLE danf.table1
    (
    id INTEGER NOT NULL,
    pID VARCHAR(50) CHARACTER SET LATIN NOT CASESPECIFIC,
    b_Date DATE FORMAT 'MM/DD/YYYY' ,
    p_Date DATE FORMAT 'MM/DD/YYYY' ,
    Diff INTEGER 
    )
PRIMARY INDEX ( id );

INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 1, '001', CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), 0 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 1, '001', CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), 0 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 1, '002', CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), 0 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '001', CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), 0 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '001', CAST('07/20/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('08/01/2014' AS DATE FORMAT 'MM/DD/YYYY'), -12 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '002', CAST('07/25/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('08/01/2014' AS DATE FORMAT 'MM/DD/YYYY'), -8 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '002', CAST('07/26/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('08/01/2014' AS DATE FORMAT 'MM/DD/YYYY'), -7 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '002', CAST('07/28/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('08/01/2014' AS DATE FORMAT 'MM/DD/YYYY'), -5 );
INSERT INTO danf.table1 (id, pID, b_Date, p_Date, Diff) VALUES ( 2, '002', CAST('07/28/2014' AS DATE FORMAT 'MM/DD/YYYY'), CAST('08/01/2014' AS DATE FORMAT 'MM/DD/YYYY'), -5 );

SELECT 
    t1.id
    ,MAX(b_Date) AS Max_b_Date
    ,t1.p_Date
    ,MAX(t1.Diff) AS Min_Diff
FROM danf.table1 t1
GROUP BY 
    t1.id, 
    t1.p_Date;

【讨论】：