将日期分成几个块，以YYYY-12-31结尾

Posted 2021-04-02

df <- data.frame(group = c("a", "a", "b", "b"),
                 start = c("2017-05-01", "2019-04-03", "2011-03-03", "2014-05-07"),
                 end = c("2018-09-01", "2020-04-03", "2012-05-03", "2016-04-02"))  

假设我有以下df：

  group      start        end
1     a 2017-05-01 2018-09-01
2     a 2019-04-03 2020-04-03
3     b 2011-03-03 2012-05-03
4     b 2014-05-07 2016-04-02

我想把它变成这种格式，每条记录分为开始日期和后续年份的31/12：

  group      start        end
1     a 2017-05-01 2017-12-31
2     a 2018-01-01 2018-09-01
3     a 2019-04-03 2019-12-31
4     a 2020-01-01 2020-04-03
5     b 2011-03-03 2011-12-31
6     b 2012-01-01 2012-05-03
7     b 2014-05-07 2014-12-31
8     b 2015-01-01 2015-12-31
9     b 2016-01-01 2016-04-02

关于如何解决这个问题的任何想法？

编辑：

我主要关心的不是同一年内的日期范围。然而，正如chinsoon12指出的那样，如果方法也可以处理它们确实会有所帮助，例如在这个数据集中：

df <- data.frame(group = c("a", "a", "b", "b", "c"),
                 start = c("2017-05-01", "2019-04-03", "2011-03-03", "2014-05-07", "2017-02-01"),
                 end = c("2018-09-01", "2020-04-03", "2012-05-03", "2016-04-02", "2017-04-05")) 

最终结果将保留最后一行：

   group      start        end
1      a 2017-05-01 2017-12-31
2      a 2018-01-01 2018-09-01
3      a 2019-04-03 2019-12-31
4      a 2020-01-01 2020-04-03
5      b 2011-03-03 2011-12-31
6      b 2012-01-01 2012-05-03
7      b 2014-05-07 2014-12-31
8      b 2015-01-01 2015-12-31
9      b 2016-01-01 2016-04-02
10     c 2017-02-01 2017-04-05  

答案

data.table可能的解决方案：

library(data.table)
setDT(df)

df[df[, rep(.I, 1 + year(end) - year(start))]
   ][, `:=` (start = pmax(start[1], as.Date(paste0(year(start[1]) + 0:(.N-1), '-01-01'))),
             end = pmin(end[.N], as.Date(paste0(year(end[.N]) - (.N-1):0, '-12-31'))))
     , by = .(group, rleid(start))][]

这使：

    group      start        end
 1:     a 2017-05-01 2017-12-31
 2:     a 2018-01-01 2018-09-01
 3:     a 2019-04-03 2019-12-31
 4:     a 2020-01-01 2020-04-03
 5:     b 2011-03-03 2011-12-31
 6:     b 2012-01-01 2012-05-03
 7:     b 2014-05-07 2014-12-31
 8:     b 2015-01-01 2015-12-31
 9:     b 2016-01-01 2016-04-02
10:     c 2017-02-01 2017-04-05

使用data.table的两种替代解决方案：

# alternative 1:
df[, ri := rowid(group)
   ][df[, rep(.I, 1 + year(end) - year(start))]
     ][, `:=` (start = if (.N == 1) start else c(start[1], as.Date(paste0(year(start[1]) + 1:(.N-1), '-01-01') )),
               end = if (.N == 1) end else c(as.Date(paste0(year(end[.N]) - (.N-1):1, '-12-31') ), end[.N]))
       , by = .(group, ri)][, ri := NULL][]

# alternative 2:
df[, ri := rowid(group)
   ][df[, rep(.I, 1 + year(end) - year(start))]
     ][, `:=` (start = pmax(start[1], as.Date(paste0(year(start[1]) + 0:(.N-1), '-01-01'))),
               end = pmin(end[.N], as.Date(paste0(year(end[.N]) - (.N-1):0, '-12-31'))))
       , by = .(group, ri)][, ri := NULL][]

---

使用数据：

df <- data.frame(group = c("a", "a", "b", "b", "c"),
                 start = c("2017-05-01", "2019-04-03", "2011-03-03", "2014-05-07", "2017-02-01"),
                 end = c("2018-09-01", "2020-04-03", "2012-05-03", "2016-04-02", "2017-04-05")) 
df[2:3] <- lapply(df[2:3], as.Date)

另一答案

这是一个no-tidyverse / no-data.table版本：

df <- data.frame(group = c("a", "a", "b", "b"),
                 start = c("2017-05-01", "2019-04-03", "2011-03-03", "2014-05-07"),
                 end = c("2018-09-01", "2020-04-03", "2012-05-03", "2016-04-02"), stringsAsFactors=FALSE)
# added stringsAsFactors =FALSE to your df for sanity
# reformatting start and end as Date  
df$start <- as.Date(df$start)
df$end <- as.Date(df$end)

dfs <- split(df, rownames(df))
# split the data frame by rows

res <- do.call(rbind, lapply(dfs, function(.){
  s <- seq(from=.$start, to=.$end, by="day")
  # sequence form df$start to df$end, by days
  y <- format(s, "%Y")
  # years of that sequence
  s2 <- as.character(s)
  # formatting s as character -- otherwise sapply will get rid of the 
  # Date class and the result will look as numeric
  ys <- split(s2,y)
  # split the sequence by years
  data.frame(group=.$group, start=sapply(ys, head,1), end = sapply(ys, tail, 1), stringsAsFactors=FALSE)
  # take the first and last element from each "sub-vector" of the split sequence
}))
rownames(res) <- NULL # kill the nasty rownames
res

  group      start        end
1     a 2017-05-01 2017-12-31
2     a 2018-01-01 2018-09-01
3     a 2019-04-03 2019-12-31
4     a 2020-01-01 2020-04-03
5     b 2011-03-03 2011-12-31
6     b 2012-01-01 2012-05-03
7     b 2014-05-07 2014-12-31
8     b 2015-01-01 2015-12-31
9     b 2016-01-01 2016-04-02

请注意，结果将start和end列作为character，就像它在原始数据框中一样。

我很抱歉基地R对待Date（和POSIXct）对象的方式 - 你永远不知道他们什么时候会失去他们的课程并成为简单的数字。在这里，我通过将日期视为字符来避免这个“特征”，除非需要日期操作，例如创建日期序列。

另一答案

library(tidyverse)
library(lubridate)
df%>%
  mutate(end=as.Date(end),
     start=as.Date(start),
     diff=Map(":",0,1+year(end)-year(start)-1))%>%
  unnest()%>%
  mutate(end=pmin(end,as.Date(paste0(year(start)+diff,"-12-31"))),
         start=pmax(start,as.Date(paste0(year(start)+diff,"-1-1"))),
         diff=NULL)

A tibble: 9 x 3
  group start      end       
  <fct> <date>     <date>    
1 a     2017-05-02 2017-12-31
2 a     2018-01-01 2018-09-01
3 a     2019-04-03 2019-12-31
4 a     2020-01-01 2020-04-03
5 b     2020-01-01 2011-12-31
6 b     2012-01-01 2012-05-03
7 b     2014-05-07 2014-12-31
8 b     2015-01-01 2015-12-31
9 b     2015-01-01 2016-04-02

使用更新的数据只需运行您将获得的确切功能：

 group      start        end
1      a 2017-05-01 2017-12-31
2      a 2018-01-01 2018-09-01
3      a 2019-04-03 2019-12-31
4      a 2020-01-01 2020-04-03
5      b 2011-03-03 2011-12-31
6      b 2012-01-01 2012-05-03
7      b 2014-05-07 2014-12-31
8      b 2015-01-01 2015-12-31
9      b 2016-01-01 2016-04-02
10     c 2017-02-01 2017-04-05