您可以通过使用这些内置函数来实现这一点。

• unix_timestamp->获取epoch时间戳值
• date_format->获取毫秒
• concat(".",<int>)->将毫秒转换为十进制
• round->舍入毫秒值
  Example:

spark.sql("set spark.sql.legacy.timeParserPolicy=LEGACY")
spark.sql("""select *,to_timestamp(unix_timestamp(ts) + round(concat(".",date_format(ts, "S")))) round_ts from (
  select timestamp('2023-06-16 00:00:19.645') as ts)a""").\
    show(10,False)

#+-----------------------+-------------------+
#|ts                     |round_ts           |
#+-----------------------+-------------------+
#|2023-06-16 00:00:19.645|2023-06-16 00:00:20|
#+-----------------------+-------------------+

字符串

您可以将时间戳增加1秒，然后使用DATE_TRUNC删除毫秒

q = """
SELECT
  ts,
  DATE_TRUNC('SECOND', ts + INTERVAL 1 SECOND) AS round_ts
FROM (
  SELECT TIMESTAMP('2023-06-16 00:00:19.645') AS ts
)
"""
spark.sql(q).show(10, False)

+-----------------------+-------------------+
|ts                     |round_ts           |
+-----------------------+-------------------+
|2023-06-16 00:00:19.645|2023-06-16 00:00:20|
+-----------------------+-------------------+

字符串

请也试试这个：

from_unixtime(round(unix_millis(timestamp('2023-06-16T23:59:59.999'))/1000) ,"yyyy-MM-dd HH:mm:ss")

字符串