遇到這樣一個需求，有一張表，要給這張表新增一個欄位delta，delta的值等於每行的c1列的值減去上一行c1列的值。
我的解決方案，可以通過python的pandas的diff來實現，也可以通過sql來實現，如下

import pandas as pd

srcTable = pd.read_csv('pos1.csv')
print(srcTable)
destTable = srcTable.loc[srcTable.tid == 1, ['ts1', 'ts2']].sort_values(by='ts1')
destTable.columns = ['deltaTs1', 'deltaTs2'
 
]
destTable = destTable.diff()
destTable = destTable.fillna(0)
destTable['delay'] = destTable['deltaTs2'] - destTable['deltaTs1']
print(destTable)

出來的效果如下：

   tid            ts1            ts2
0    1  1500443161000  1500443161240
1    1  1500443162000  1500443162994
2    1  1500443163000  1500443163067
3    1  1500443164000  1500443164993
 

   deltaTs1  deltaTs2  delay
0       0.0       0.0    0.0
1    1000.0    1754.0  754.0
2    1000.0      73.0 -927.0
3    1000.0    1926.0  926.0

若是用sql語句，我用的是mysql，自己構造行號rn

mysql> select main.t_id,
main.ts1,
ifnull(main.ts1-sub.ts1,0) deltaTs1,
main.ts2,
ifnull(main.ts2-sub.ts2,0) deltaTs2 from
(SELECT t_id,ts1,ts2,(@r1 :
 
[email protected] + 1) rn FROM pos1,(SELECT @r1 := 0) r where t_id=1 ORDER BY ts1) main
left join 
(SELECT t_id,ts1,ts2,(@r2 :[email protected] + 1) rn FROM pos1,(SELECT @r2 := 0) r where t_id=1 ORDER BY ts1) sub 
on main.rn-1=sub.rn;
+------+---------------+----------+---------------+----------+
| t_id | ts1           | deltaTs1 | ts2           | deltaTs2 |
+------+---------------+----------+---------------+----------+
|    1 | 1500443161000 |        0 | 1500443161240 |        0 |
|    1 | 1500443162000 |     1000 | 1500443162994 |     1754 |
|    1 | 1500443163000 |     1000 | 1500443163067 |       73 |
|    1 | 1500443164000 |     1000 | 1500443164993 |     1926 |
+------+---------------+----------+---------------+----------+

測試資料如下
pos1.csv

1,1500443161000,1500443161240
1,1500443162000,1500443162994
1,1500443163000,1500443163067
1,1500443164000,1500443164993

CREATE TABLE `pos1` (
  `t_id` int(11) DEFAULT NULL,
  `ts1` bigint(22) DEFAULT NULL,
  `ts2` bigint(22) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
INSERT INTO pos1 VALUES (1, 1500443161000, 1500443161240);
INSERT INTO pos1 VALUES (1, 1500443162000, 1500443162994);
INSERT INTO pos1 VALUES (1, 1500443163000, 1500443163067);
INSERT INTO pos1 VALUES (1, 1500443164000, 1500443164993);

貌似有些資料庫有這種當前行減去上一行資料的函式，具體我沒有研究過。有知道的朋友可以告訴我一下，我印象中像Sqlserver好像有。