CQL (cassandra) - 仅选择其中一列中具有最大值的行

Question

我需要找到具有给定 stationid 且 time1 大于指定时间和最大 time2 的行。

表是这样创建的：

CREATE TABLE forec (
    stationid int,
    time1 timestamp,
    time2 timestamp,
    value double,
    PRIMARY KEY ((stationid), time1, time2)
) WITH CLUSTERING ORDER BY (time1 DESC)

假设表中的数据是这样的：

    +------------+-----------------------+----------------------+--------+
    | stationid  | time1                 |  time2               |  value |
    +------------+-----------------------+----------------------+--------+
    | 1          | 2020-10-21 06:00:00   | 2020-10-21 05:00:00  | 1      |                                  
    | 1          | 2020-10-21 06:00:00   | 2020-10-21 04:00:00  | 2      |                                   
    | 1          | 2020-10-21 06:00:00   | 2020-10-21 03:00:00  | 3      |                                   
    | 1          | 2020-10-21 05:00:00   | 2020-10-21 04:00:00  | 4      |
    | 1          | 2020-10-21 05:00:00   | 2020-10-21 03:00:00  | 5      |
    | 1          | 2020-10-21 04:00:00   | 2020-10-21 02:00:00  | 6      |
    +------------+-----------------------+----------------------+--------+

我想查询： 给我所有 stationid = 1 和 time1 >= 2020-10-21 05:00:00 并且 time2 具有最大值的行。查询应返回以下行：

    +------------+-----------------------+----------------------+--------+
    | stationid  | time1                 |  time2               |  value |
    +------------+-----------------------+----------------------+--------+
    | 1          | 2020-10-21 06:00:00   | 2020-10-21 05:00:00  | 1      |        
    | 1          | 2020-10-21 05:00:00   | 2020-10-21 04:00:00  | 4      | 
    +------------+-----------------------+----------------------+--------+

我知道我可以这样查询：

SELECT * FROM forec WHERE stationid = 1 AND time1 >= '2020-10-21 05:00:00';

然后在客户端过滤结果（并仅保留具有最大时间的行2），但是我想知道这是否可以更有效地完成（在 Cassandra 端过滤结果）。

或者我应该改变表格模型？

Answer 1

使用 UDA/UDF 的解决方案：

状态函数：

CREATE OR REPLACE FUNCTION curValState ( state tuple<timestamp,double>, time timestamp, value double ) CALLED ON NULL INPUT RETURNS tuple<timestamp, double> LANGUAGE java AS 'if (time != null && value != null) { if(state == null) {com.datastax.driver.core.TupleType tupleType = com.datastax.driver.core.TupleType.of(com.datastax.driver.core.ProtocolVersion.NEWEST_SUPPORTED, com.datastax.driver.core.CodecRegistry.DEFAULT_INSTANCE, com.datastax.driver.core.DataType.timestamp(), com.datastax.driver.core.DataType.cdouble()); state = tupleType.newValue(time, value);} else {if(state.getTimestamp(0).compareTo(time)<0){ state.setTimestamp(0, time); state.setDouble(1, value);}}} return state;';

最终功能：

CREATE OR REPLACE FUNCTION finalVal ( state tuple<timestamp, double> ) CALLED ON NULL INPUT RETURNS double LANGUAGE java AS 'return state.getDouble(1);';

聚合函数：

CREATE OR REPLACE AGGREGATE valueatlatesttime (timestamp, double) SFUNC curValState STYPE tuple<timestamp, double> FINALFUNC finalVal INITCOND null;

查询：

SELECT
  stationid,
  time1,
  max(time2) AS max_time2,
  valueatlatesttime(time2, value) AS value_at_max_time2
FROM
  forec
WHERE
  stationid = 1
AND
  time1 >= '2020-10-21 05:00:00'
GROUP BY time1;

Answer 2

编辑：根据Cassandra document，“如果在没有聚合函数的情况下选择了列，则在带有 GROUP BY 的语句中，将返回每个组中遇到的第一个值。”因此，下面的示例仅在 time2 以 DESC 顺序存储时有效。

如果您使用的是最新版本的 Cassandra（如 3.11.x），那么您可以使用 GROUP BY 来执行类似的操作

SELECT
  stationid,
  time1,
  max(time2) AS max_time2,
  value
FROM
  forec
WHERE
  stationid = 1
AND
  time1 >= '2020-10-21 05:00:00'
GROUP BY time1;

你得到

cqlsh:test> SELECT stationid, time1, max(time2) as max_time2, value FROM forec WHERE stationid = 1 AND time1 >= '2020-10-21 05:00:00' GROUP BY  time1;

 stationid | time1                           | max_time2                       | value
-----------+---------------------------------+---------------------------------+-------
         1 | 2020-10-21 06:00:00.000000+0000 | 2020-10-21 05:00:00.000000+0000 |     1
         1 | 2020-10-21 05:00:00.000000+0000 | 2020-10-21 04:00:00.000000+0000 |     4

(2 rows)

请注意，这会扫描您的分区，因此请注意分区大小，尤其是当您在集群列中使用时间戳时。