将时间四舍五入到特定时间，即使是前一天答案

【问题标题】：Rounding time to specific hour even if it is the day before将时间四舍五入到特定时间，即使是前一天
【发布时间】：2021-07-09 23:38:23
【问题描述】：

我有时间数据，我想从暮光中减去，我还有其他代码。为此，我首先需要根据最近的暮光来标准化时间。我遇到的麻烦是在午夜后四舍五入，回到前一天的黄昏（参见：Df$time[2]）

我已经在同一天成功地使用了带有 period() 的 lubridate 函数 round_date，但是对于清晨的时间，它只是恢复到当天的午夜。帮助菜单中的“change_on_boundary”参数似乎不再起作用。有人有其他方法吗？

*注意：我知道 tz 改变了，我并不担心。这是这个问题的虚拟数据集。

Df<-data.frame(time = as.POSIXct(c("2020-12-29 21:02:23 UTC", "2020-12-15 00:48:21 UTC", "2020-12-09 21:55:52 UTC" ,"2020-12-09 18:40:06 UTC",
 "2020-12-08 18:25:27 UTC" ,"2020-12-14 20:08:51 UTC", "2020-12-14 17:55:21 UTC","2020-12-05 23:53:36 UTC",
 "2020-12-05 21:21:21 UTC", "2020-12-06 19:11:11 UTC", "2020-12-05 21:09:31 UTC", "2020-12-05 18:13:53 UTC",
 "2020-12-15 20:36:56 UTC", "2020-12-26 06:00:00 UTC", "2020-12-16 00:40:46 UTC" ,"2020-12-26 05:34:42 UTC")))

Df$time
 [1] "2020-12-29 21:02:23 EST" "2020-12-15 00:48:21 EST" "2020-12-09 21:55:52 EST" "2020-12-09 18:40:06 EST" "2020-12-08 18:25:27 EST"
 [6] "2020-12-14 20:08:51 EST" "2020-12-14 17:55:21 EST" "2020-12-05 23:53:36 EST" "2020-12-05 21:21:21 EST" "2020-12-06 19:11:11 EST"
[11] "2020-12-05 21:09:31 EST" "2020-12-05 18:13:53 EST" "2020-12-15 20:36:56 EST" "2020-12-26 06:00:00 EST" "2020-12-16 00:40:46 EST"
[16] "2020-12-26 05:34:42 EST"

round_date(Df$time, unit = period(16, units = "hour"))
 [1] "2020-12-29 16:00:00 EST" "2020-12-15 00:00:00 EST" "2020-12-09 16:00:00 EST" "2020-12-09 16:00:00 EST" "2020-12-08 16:00:00 EST"
 [6] "2020-12-14 16:00:00 EST" "2020-12-14 16:00:00 EST" "2020-12-05 16:00:00 EST" "2020-12-05 16:00:00 EST" "2020-12-06 16:00:00 EST"
[11] "2020-12-05 16:00:00 EST" "2020-12-05 16:00:00 EST" "2020-12-15 16:00:00 EST" "2020-12-26 00:00:00 EST" "2020-12-16 00:00:00 EST"
[16] "2020-12-26 00:00:00 EST"

【问题讨论】：

标签： r time lubridate

【解决方案1】：

如果我对问题的理解正确，您希望每次四舍五入到最接近的 16:00（下午 4 点）。正如您所发现的，lubridate::round_date() 在时间倒流时似乎会在午夜停止。

对于您的示例，您可以找到哪些四舍五入的日期时间已达到午夜的最低点，并减去 8 小时以到达前一天的 16:00。

library(tidyverse)
library(lubridate)

Df %>% 
  tibble %>% 
  force_tz('UTC') %>% 
  mutate(
    time_round = round_date(time, unit = period(16, units = 'hour')),
    time_round_final = as_datetime(ifelse(hour(time_round) == 0,       # if the hour is 0 (midnight), then...
                                     time_round - hours(24 - 16),      # we subtract 8 (24 - 16) hours, otherwise...
                                     time_round))                      # we keep the original rounded time
  )

# A tibble: 16 x 3
#    time                time_round          time_round_final   
#    <dttm>              <dttm>              <dttm>             
#  1 2020-12-29 21:02:23 2020-12-29 16:00:00 2020-12-29 16:00:00
#  2 2020-12-15 00:48:21 2020-12-15 00:00:00 2020-12-14 16:00:00
#  3 2020-12-09 21:55:52 2020-12-09 16:00:00 2020-12-09 16:00:00
#  4 2020-12-09 18:40:06 2020-12-09 16:00:00 2020-12-09 16:00:00
#  5 2020-12-08 18:25:27 2020-12-08 16:00:00 2020-12-08 16:00:00
#  6 2020-12-14 20:08:51 2020-12-14 16:00:00 2020-12-14 16:00:00
#  7 2020-12-14 17:55:21 2020-12-14 16:00:00 2020-12-14 16:00:00
#  8 2020-12-05 23:53:36 2020-12-05 16:00:00 2020-12-05 16:00:00
#  9 2020-12-05 21:21:21 2020-12-05 16:00:00 2020-12-05 16:00:00
# 10 2020-12-06 19:11:11 2020-12-06 16:00:00 2020-12-06 16:00:00
# 11 2020-12-05 21:09:31 2020-12-05 16:00:00 2020-12-05 16:00:00
# 12 2020-12-05 18:13:53 2020-12-05 16:00:00 2020-12-05 16:00:00
# 13 2020-12-15 20:36:56 2020-12-15 16:00:00 2020-12-15 16:00:00
# 14 2020-12-26 06:00:00 2020-12-26 00:00:00 2020-12-25 16:00:00
# 15 2020-12-16 00:40:46 2020-12-16 00:00:00 2020-12-15 16:00:00
# 16 2020-12-26 05:34:42 2020-12-26 00:00:00 2020-12-25 16:00:00

我使用lubridate::force_tz() 将时区应用于时间（为空），否则as_datetime() 似乎默认为 UTC（它必须假定无时区时间是您的本地时区？）并返回意外时间，通过静默更改（应用）时区。这可能是导致错误的常见原因。

我们可以在不强制时区时观察到这一点（我在 GMT+11）：

Df %>% 
  tibble %>% 
  # force_tz('UTC') %>%
  mutate(
    time_round = round_date(time, unit = period(16, units = 'hour')),
    time_round_final = as_datetime(ifelse(hour(time_round) == 0,       # if the hour is 0 (midnight), then...
                                          time_round - hours(24 - 16),      # we subtract 8 (24 - 16) hours, otherwise...
                                          time_round)),                      # we keep the original rounded time
    tz_1 = tz(time), 
    tz_2 = tz(time_round),
    tz_3 = tz(time_round_final),
  )

# # A tibble: 16 x 6
#    time                time_round          time_round_final    tz_1  tz_2  tz_3 
#    <dttm>              <dttm>              <dttm>              <chr> <chr> <chr>
#  1 2020-12-29 21:02:23 2020-12-29 16:00:00 2020-12-29 05:00:00 ""    ""    UTC  
#  2 2020-12-15 00:48:21 2020-12-15 00:00:00 2020-12-14 05:00:00 ""    ""    UTC  
#  3 2020-12-09 21:55:52 2020-12-09 16:00:00 2020-12-09 05:00:00 ""    ""    UTC  
#  4 2020-12-09 18:40:06 2020-12-09 16:00:00 2020-12-09 05:00:00 ""    ""    UTC  
#  5 2020-12-08 18:25:27 2020-12-08 16:00:00 2020-12-08 05:00:00 ""    ""    UTC  
#  6 2020-12-14 20:08:51 2020-12-14 16:00:00 2020-12-14 05:00:00 ""    ""    UTC  
#  7 2020-12-14 17:55:21 2020-12-14 16:00:00 2020-12-14 05:00:00 ""    ""    UTC  
#  8 2020-12-05 23:53:36 2020-12-05 16:00:00 2020-12-05 05:00:00 ""    ""    UTC  
#  9 2020-12-05 21:21:21 2020-12-05 16:00:00 2020-12-05 05:00:00 ""    ""    UTC  
# 10 2020-12-06 19:11:11 2020-12-06 16:00:00 2020-12-06 05:00:00 ""    ""    UTC  
# 11 2020-12-05 21:09:31 2020-12-05 16:00:00 2020-12-05 05:00:00 ""    ""    UTC  
# 12 2020-12-05 18:13:53 2020-12-05 16:00:00 2020-12-05 05:00:00 ""    ""    UTC  
# 13 2020-12-15 20:36:56 2020-12-15 16:00:00 2020-12-15 05:00:00 ""    ""    UTC  
# 14 2020-12-26 06:00:00 2020-12-26 00:00:00 2020-12-25 05:00:00 ""    ""    UTC  
# 15 2020-12-16 00:40:46 2020-12-16 00:00:00 2020-12-15 05:00:00 ""    ""    UTC  
# 16 2020-12-26 05:34:42 2020-12-26 00:00:00 2020-12-25 05:00:00 ""    ""    UTC

【讨论】：