总结
通过对两个函数的测试,发现有如下特点:
- 针对的字段类型不同:trunc是针对date类型处理,date_trunc是针对time_stamp类型处理;如果输入的时候不是对应的类型,会自动用cast函数处理;
- 参数的位置:trunc的参数在后面,date_trunc的参数在前面;
- 仅从个人练习的情况看,trunc的用法比date_trunc少些,因为trunc到date级别,date_trunc到time_stamp级别,所以以后可以多用date_trunc和sub_str基本能够解决绝大部分需求;
trunc
测试结果:在spark_sql中只有到日期的位置可以用trunc函数,即输入可以是date、time_stamp类型,但是输出只能是date类型;
select trunc('2021-12-12 09:12:12', 'yyyy')
输出:2021-01-01
select trunc(cast('2021-12-12' as date), 'yy')
输出:2021-01-01
select trunc(cast('2021-12-12' as date), 'mm')
输出:2021-12-01
select trunc('2021-12-12 09:12:12', 'week')
输出:2021-12-06
date_trunc
用法:date_trunc(格式, 日期字段)
说明:返回该时间/日期对应的取断时间/日期,格式输入支持:'year','week','day','hour'等
示例:date_trunc('week', [下单时间]),返回该行"下单时间"字段对应的当周第一天
测试结果:在spark_sql中用date_trunc函数可以截取到任意位置(目前仅测试到秒,未测试微秒),然后让后面位置归为0等,且输入可以是date、time_stamp类型,但是输出time_stamp是型;
select date_trunc('year', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-01-01 00:00:00
select date_trunc('mm', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-12-01 00:00:00
select date_trunc('dd', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-12-12 00:00:00
select date_trunc('week', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-12-06 00:00:00
select date_trunc('hour', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-12-12 09:00:00
select date_trunc('minute', cast('2021-12-12 09:32:05' as timestamp));
输出:2015-03-05 09:32:00
select date_trunc('second', cast('2021-12-12 09:32:05' as timestamp));
输出:2021-12-12 09:32:05
select date_trunc('minute','2021-12-12');
输出:2021-12-12 00:00:00
标签:00,12,区别,09,2021,date,trunc
From: https://www.cnblogs.com/javaxubo/p/17498295.html