我有一个包含超过10000个项目的Excel电子表格，每个项目的开始日期和结束日期都不同。我需要在他们活跃的几年中分散项目，以便按年计算确切的数量。从这个例子开始我该怎么做？
start date end date entityNo amount
4/1/2001 8/31/2012 1 500
1/1/2005 12/31/2007 2 100
我最初想到的解决方案是在Excel中获取开始日期和结束日期（结束日期 - 开始日期+ 1）之间的差异，这样我就可以获得每个项目的持续时间。在获得天数（例如4100天）之后，我将持续时间除以一年中的总天数（365.25天），这将给出我的年份范围但是当我将每个项目分配到正确的年份。
超过10,000个项目的数据集的预期输出：
start date end date no of years entityNo amount
4/1/2001 8/31/2012 11 1 500
2002
2003
2004
2005
2006
2007 upto 8/31/2012
后跟另一个具有不同实体编号的项目
start date end date entity no amount
1/1/2005 12/31/2007 2 100
输出
years entity no amount
1/1/2005 2 100
2006 2 100
2/31/2007 2 100
答案 0 :(得分：0)
此解决方案需要r包lubridate。
library(tidyverse)
library(lubridate)
# Create example data frame
dat <- tribble(
~`start date`, ~`end date`, ~`entityNo`, ~`amount`,
"4/1/2001", "8/31/2012", 1, 500,
"1/1/2005", "12/31/2007", 2, 100
)
dat %>%
mutate(`start date` = mdy(`start date`), `end date` = mdy(`end date`)) %>%
mutate(`start year` = year(`start date`), `end year` = year(`end date`)) %>%
mutate(`no of years` = `end year` - `start year`) %>%
select(`start date`, `end date`, `no of years`, entityNo, amount)
或者你可以使用一些字符串操作方法。
dat %>%
mutate(`no of years` = as.numeric(substring(`end date`, nchar(`end date`) - 3)) -
as.numeric(substring(`start date`, nchar(`start date`) - 3))) %>%
select(`start date`, `end date`, `no of years`, entityNo, amount)
答案 1 :(得分：0)
根据最新的编辑，OP希望按日历年划分每个项目的持续时间。这可以通过使用foverlaps()包的data.table函数来完成。
library(data.table)
projects <- fread(
"start_date end_date entityNo amount
4/1/2001 8/31/2012 1 500
1/1/2005 12/31/2007 2 100")
fread()可用于从磁盘快速读取csv个文件。这里使用了一个便利功能，它允许从字符变量中读取数据。
library(lubridate)
# convert dates from character to Date class
date_cols <- c("start_date", "end_date")
projects[, (date_cols) := lapply(.SD, mdy), .SDcols = date_cols]
# compute duration of project = number of years in which project was active
projects[, years_active := year(end_date) - year(start_date) + 1]
请注意，years_active与OP提供的no of years不同。 years_active是传播数据所需的行数。
date_range <- projects[, .(year = seq(year(min(start_date)),
year(max(end_date))))]
date_range[, start_in_year := ymd(paste0(year, "-01-01"))]
date_range[, end_in_year := ymd(paste0(year, "-12-31"))]
setkey(date_range, start_in_year, end_in_year)
date_range
# year start_in_year end_in_year
# 1: 2001 2001-01-01 2001-12-31
# 2: 2002 2002-01-01 2002-12-31
# 3: 2003 2003-01-01 2003-12-31
# ...
#10: 2010 2010-01-01 2010-12-31
#11: 2011 2011-01-01 2011-12-31
#12: 2012 2012-01-01 2012-12-31
请注意，此方法可以扩展为按季度，月份，ISO周或天数创建持续时间细分。
projects_by_year <- foverlaps(projects, date_range, by.x = date_cols)
# adjust start_in_year to coincide with project start date
projects_by_year[, start_in_year := pmax(start_in_year, start_date)]
# adjust end_in_year to coincide with project end date
projects_by_year[, end_in_year := pmin(end_in_year, end_date)]
projects_by_year
# year start_in_year end_in_year start_date end_date entityNo amount years_active
# 1: 2001 2001-04-01 2001-12-31 2001-04-01 2012-08-31 1 500 12
# 2: 2002 2002-01-01 2002-12-31 2001-04-01 2012-08-31 1 500 12
# 3: 2003 2003-01-01 2003-12-31 2001-04-01 2012-08-31 1 500 12
# ...
#10: 2010 2010-01-01 2010-12-31 2001-04-01 2012-08-31 1 500 12
#11: 2011 2011-01-01 2011-12-31 2001-04-01 2012-08-31 1 500 12
#12: 2012 2012-01-01 2012-08-31 2001-04-01 2012-08-31 1 500 12
#13: 2005 2005-01-01 2005-12-31 2005-01-01 2007-12-31 2 100 3
#14: 2006 2006-01-01 2006-12-31 2005-01-01 2007-12-31 2 100 3
#15: 2007 2007-01-01 2007-12-31 2005-01-01 2007-12-31 2 100 3
项目1分布在12年/每年，预计2年超过3年。调整start_in_year和end_in_year以匹配每个项目的respctive开始和结束年中的正确开始和结束日期。
希望这是预期的结果。
长格式非常适合每年计算聚合。例如，每年的项目数量：
projects_by_year[, .N, by = year]
# year N
# 1: 2001 1
# 2: 2002 1
# 3: 2003 1
# 4: 2004 1
# 5: 2005 2
# 6: 2006 2
# 7: 2007 2
# 8: 2008 1
# 9: 2009 1
#10: 2010 1
#11: 2011 1
#12: 2012 1
或每年的总金额：
projects_by_year[, sum(amount), by = year]
# year V1
# 1: 2001 500
# 2: 2002 500
# 3: 2003 500
# 4: 2004 500
# 5: 2005 600
# 6: 2006 600
# 7: 2007 600
# 8: 2008 500
# 9: 2009 500
#10: 2010 500
#11: 2011 500
#12: 2012 500