基于几何和最近的时间戳 BigQuery SQl 连接表答案

【问题标题】：Join tables based on geometry and nearest timestamp BigQuery SQl基于几何和最近的时间戳 BigQuery SQl 连接表
【发布时间】：2020-06-23 14:06:29
【问题描述】：

我有两张桌子； df1 包含 Date1（时间戳）和 PolygonWKT（几何），df2 包含 Date2（时间戳）和 PointWKT（几何）。我根据几何加入了df1和df2，所以每个PointWKT都属于对应的PolygonWKT。问题是，Date1 和 Date2e 列搞砸了，我还需要匹配 Date1 和 Date2。我想加入基于几何的表格以及Date1和Date2之间最接近的时间戳匹配。

 df2
 | PointWKT  |      Date2            |     
 --------------------------------------
 |    b    | 2020-05-05 12:00:00 UTC |
 |    b    | 2020-05-05 12:00:10 UTC |
 |    b    | 2020-05-05 12:00:20 UTC |
 |    b    | 2020-05-05 12:17:00 UTC |
 |    c    | 2020-05-06 18:00:00 UTC |


df1
 | PolygonWKT  |      Date1          |
--------------------------------------
 |    A    | 2020-05-03 9:00:00 UTC  |
 |    A    | 2020-05-03 9:30:10 UTC  |
 |    B    | 2020-05-05 12:05:00 UTC |
 |    B    | 2020-05-05 12:25:00 UTC |
 |    C    | 2020-05-06 18:05:00 UTC |

代码的第一部分是正确的，但第二部分没有返回我想要的：

    SELECT *
    FROM `xxx.yyy.df1` as df1 ,
    `xxx.yyy.df2` as df2
    WHERE ST_Contains (df1.PolygonWKT, df2.PointWKT) 


    AND ( 
    df2.Date2 BETWEEN df1.Date1 AND TIMESTAMP_ADD(df1.Date1, INTERVAL 10 MINUTE)



    
desired df

| PointWKT  |      Date2            || PolygonWKT  |      Date1          |     
----------------------------------------------------------------------------
 |    b    | 2020-05-05 12:00:00 UTC | |    B    | 2020-05-05 12:05:00 UTC |
 |    b    | 2020-05-05 12:00:10 UTC | |    B    | 2020-05-05 12:05:00 UTC |
 |    b    | 2020-05-05 12:00:20 UTC | |    B    | 2020-05-05 12:05:00 UTC |
 |    b    | 2020-05-05 12:17:00 UTC | |    B    | 2020-05-05 12:25:00 UTC |
 |    c    | 2020-05-06 18:00:00 UTC | |    C    | 2020-05-06 18:05:00 UTC |

这样做的正确方法是什么？

【问题讨论】：

标签： sql google-bigquery

【解决方案1】：

我想基于几何以及Date1 和Date2 之间最接近的时间戳匹配加入表。

以下是 BigQuery 标准 SQL

SELECT  
  ARRAY_AGG(STRUCT(df2.PointWKT, df2.Date2, df1.PolygonWKT, df1.Date1) 
    ORDER BY ABS(TIMESTAMP_DIFF(df2.Date2, df1.Date1, SECOND)) 
    LIMIT 1)[OFFSET(0)].*
FROM `xxx.yyy.df1` AS df1 ,
  `xxx.yyy.df2` AS df2
WHERE ST_CONTAINS(df1.PolygonWKT, df2.PointWKT) 
GROUP BY TO_JSON_STRING(STRUCT(df2.PointWKT, df2.Date2))

如果适用于与您的示例中类似的示例数据 -

WITH `xxx.yyy.df1` AS (
  SELECT ST_GEOGPOINT(1,2) PolygonWKT, TIMESTAMP '2020-05-03 9:00:00 UTC' Date1 UNION ALL
  SELECT ST_GEOGPOINT(1,2), '2020-05-03 9:30:10 UTC' UNION ALL
  SELECT ST_GEOGPOINT(1,3), '2020-05-05 12:05:00 UTC' UNION ALL
  SELECT ST_GEOGPOINT(1,3), '2020-05-05 12:25:00 UTC' UNION ALL
  SELECT ST_GEOGPOINT(1,4), '2020-05-06 18:05:00 UTC' 
), `xxx.yyy.df2` AS (
  SELECT ST_GEOGPOINT(1,3) PointWKT, TIMESTAMP '2020-05-05 12:00:00 UTC' Date2 UNION ALL
  SELECT ST_GEOGPOINT(1,3), '2020-05-05 12:00:10 UTC' UNION ALL
  SELECT ST_GEOGPOINT(1,3), '2020-05-05 12:00:20 UTC' UNION ALL
  SELECT ST_GEOGPOINT(1,3), '2020-05-05 12:17:00 UTC' UNION ALL /* this value adjusted based on exapected result sample - as it looks as a typo */
  SELECT ST_GEOGPOINT(1,4), '2020-05-06 18:00:00 UTC' 
)

输出是

Row PointWKT    Date2                   PolygonWKT  Date1    
1   POINT(1 3)  2020-05-05 12:00:00 UTC POINT(1 3)  2020-05-05 12:05:00 UTC  
2   POINT(1 3)  2020-05-05 12:00:10 UTC POINT(1 3)  2020-05-05 12:05:00 UTC  
3   POINT(1 3)  2020-05-05 12:00:20 UTC POINT(1 3)  2020-05-05 12:05:00 UTC  
4   POINT(1 3)  2020-05-05 12:17:00 UTC POINT(1 3)  2020-05-05 12:25:00 UTC  
5   POINT(1 4)  2020-05-06 18:00:00 UTC POINT(1 4)  2020-05-06 18:05:00 UTC

【讨论】：

非常感谢您！我在 GROUP BY 行中收到错误消息：[6:10] 不允许按 GEOGRAPHY 类型的表达式分组。有什么建议如何处理吗？
查看对地理类型进行更正处理的更新。原来的例子是字符串 - 所以我错过了这部分。现在修复了
太棒了！完美运行，非常感谢！

【解决方案2】：

根据您的示例数据，您以错误的顺序提取日期。这符合您的要求吗？

df2.Date1 BETWEEN df2.Date1 AND TIMESTAMP_ADD(df2.Date1, INTERVAL 10 MINUTE)

【讨论】：

我认为这一行的整个概念对于我的目的来说是错误的。我认为米哈伊尔的回答会有所帮助。