如何在 BigQuery 中执行 Pandas 列转换？

Question

假设我们有一个包含以下数据的数据集：

timestamp,col1

1533286270,1
1533286271,2
1533286272,3
1533286273,4
1533286274,5

我想获得之前的col1值col1_prev以便能够比较它们。结果应该是一样的pandas.shift(-1)。

如何使用纯 SQL 查询实现此功能？

查询结果应如下所示：

timestamp,col1,col1_prev

1533286270,1,NULL
1533286271,2,1
1533286272,3,2
1533286273,4,3
1533286274,5,4

Answer 1

使用该lag()函数是一种方式：

WITH
  input AS (
  SELECT
    1533286270 AS timestamp,
    1 AS col1
  UNION ALL
  SELECT
    1533286271 AS timestamp,
    2 AS col1
  UNION ALL
  SELECT
    1533286272 AS timestamp,
    3 AS col1
  UNION ALL
  SELECT
    1533286273 AS timestamp,
    4 AS col1
  UNION ALL
  SELECT
    1533286274 AS timestamp,
    5 AS col1 )
SELECT
  timestamp,
  col1,
  LAG(col1) OVER(ORDER BY col1) AS col1_prev
FROM
  input