当前位置:首页 >> 脚本专栏

pandas or sql计算前后两行数据间的增值方法

遇到这样一个需求,有一张表,要给这张表新增一个字段delta,delta的值等于每行的c1列的值减去上一行c1列的值。

我的解决方案,可以通过python的pandas的diff来实现,也可以通过sql来实现,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)

出来的效果如下:

 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语句,我用的是mysql,自己构造行号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 :=@r1 + 1) rn FROM pos1,(SELECT @r1 := 0) r where t_id=1 ORDER BY ts1) main
left join 
(SELECT t_id,ts1,ts2,(@r2 :=@r2 + 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 |
+------+---------------+----------+---------------+----------+

测试数据如下

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);

貌似有些数据库有这种当前行减去上一行数据的函数,具体我没有研究过。有知道的朋友可以告诉我一下,我印象中像Sqlserver好像有。