Cassandra CQL选择查询不返回时间戳为clusterkey的记录

Question

Cassandra CQL：使用复合键和集群键创建的表。当我尝试从分区键执行 select * 时，我能够检索所有数据并且它也适用于关系运算符（）。 但是，当我使用具有正确值的等于（=）运算符查询特定集群键时，它返回 0 行。

表格：

CREATE TABLE entity_data (
received_date timestamp,
entity text,
received_time timestamp,
node int,
primary key ((received_date ,entity),received_time));

数据（从实体中选择 *）：

received_date              | entity | received_time            | node_id
2014-09-24 00:00:00+0400   |     NA | 2014-09-24 18:56:55+0400 |       0  | 

使用条件查询： -- 这里不起作用

select * from entity_data 
where received_date = '2014-09-24 00:00:00+0400' and entity = 'NA' 
and received_time='2014-09-24 18:56:55+0400';
(0 rows)

-- 返回 0 行。

Answer 1

我明白发生了什么。您正在使用 now() 生成时间 UUID。但是，当您使用dateOf() 将其转换为timestamp 时，您正在截断它的毫秒数。因此，查询等于 2014-09-24 18:56:55+0400 的 received_time 将不会产生任何结果，因为 timestamp 类型仍以毫秒为单位存储（由于您的 @987654327，您只是看不到它@)。

解决此问题的最佳方法是将您的时间存储为timeuuids（注意：我将received_date 保留为时间戳，仅用于示例目的）。然后在 SELECT 时使用 dateOf，在 WHERE 子句中使用 minTimeuuid() 函数：

CREATE TABLE entity_data2 (
    received_date timestamp,
    entity text,
    received_time timeuuid,
    node int,
PRIMARY KEY ((received_date, entity), received_time));

INSERT INTO entity_data2 (received_date, entity, received_time , node) 
VALUES ('2014-09-24 00:00:00+0400','NA',now(),0);

aploetz@cqlsh:stackoverflow> SELECT * FROM entity_data2 
    WHERE received_date = '2014-09-24 00:00:00+0400' AND entity = 'NA'  
    AND received_time>minTimeuuid('2014-10-08 08:13:53-0500') 
    AND received_time<minTimeuuid('2014-10-08 08:13:54-0500');

 received_date            | entity | received_time                        | node
--------------------------+--------+--------------------------------------+------
 2014-09-23 15:00:00-0500 |     NA | f3b548b0-4eec-11e4-9d05-7991a041665c |    0

(1 rows)

aploetz@cqlsh:stackoverflow> SELECT received_date, entity, dateof(received_time), node 
    FROM entity_data2 WHERE received_date = '2014-09-24 00:00:00+0400' AND entity = 'NA'
    AND received_time>minTimeuuid('2014-10-08 08:13:53-0500') 
    AND received_time<minTimeuuid('2014-10-08 08:13:54-0500');

 received_date            | entity | dateof(received_time)    | node
--------------------------+--------+--------------------------+------
 2014-09-23 15:00:00-0500 |     NA | 2014-10-08 08:13:53-0500 |    0

(1 rows)

基本上，dateOf() 函数旨在用于查询数据，而不是存储数据。这是一篇博客文章，描述了（更详细地）如何进行这项工作：

Time series based queries in Cassandra 1.2+ and CQL3