Question

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

id  pID  b_Date     p_Date   
1   001  7/20/2014  7/20/2014
1   001  7/20/2014  7/20/2014
1   002  7/20/2014  7/20/2014
2   001  7/20/2014  7/20/2014
2   001  7/20/2014  8/01/2014
2   002  7/25/2014  8/01/2014
2   002  7/26/2014  8/01/2014
2   002  7/28/2014  8/01/2014
2   002  7/28/2014  8/01/2014

在最新购买之前，为每位客户追加最近购买日期的最有效方法是什么。所以，结果看起来像这样。

id  pID  b_Date     p_Date     latest_purchase_date
1   001  7/20/2014  7/20/2014  'N/A'
1   001  7/20/2014  7/20/2014  'N/A'
1   002  7/20/2014  7/20/2014  'N/A'
2   001  7/20/2014  7/20/2014  'N/A'
2   001  7/20/2014  8/01/2014  7/20/2014
2   002  7/25/2014  8/01/2014  7/20/2014
2   002  7/26/2014  8/01/2014  7/20/2014
2   002  7/28/2014  8/01/2014  7/20/2014
2   002  7/28/2014  8/01/2014  7/20/2014

我正在使用Teradata 13.1

Answer 1

Teradata中没有LAG，但很容易重写。

由于有多个行具有相同的p_date，因此无论何时更改都需要跟踪。

SELECT id, pid, b_date, p_date
   ,MAX(last_dt) -- fill the NULLs with the last date
    OVER (PARTITION BY id ORDER BY p_date, last_dt DESC
          ROWS UNBOUNDED PRECEDING)
FROM
 (
   SELECT id, pid, b_date, p_date,
      NULLIF(MIN(p_date) -- return the date only when there's a change, otherwise NULL
             OVER (PARTITION BY id ORDER BY p_date
                   ROWS BETWEEN 1 PRECEDING AND 1 PRECEDING)
            , p_date) AS last_dt
   FROM vt
 ) AS dt

附加上一个事务日期SQL

1 个答案: