按年和周分割日期，并获得每年正确的周数

Question

With the following code i split dates into years and this years into weeks :

library(lubridate)

start = as.Date('2002-01-01')
end = as.Date('2017-01-01')

dates = sample(seq(as.Date('2002-01-01 00:00:00'), as.Date('2017-04-01 00:00:00'), by="day"), end-start,replace = FALSE)

splitByYears = split(dates, year(dates))
splitYearsByWeeks = lapply(splitByYears, function(x) split(x, isoweek(x)))

Based on this output i have done several calculations. Only when i was plotting some data i noticed that this procedure does not work perfectly:

>splitYearsByWeeks
...

$`2011`$`52`
[1] "2011-01-01" "2011-01-02" "2011-12-26"


$`2012`
$`2012`$`1`
[1] "2012-12-31" "2012-01-02" "2012-01-06" "2012-01-08"

...

Here 2011-01-01 and 2011-01-02 a part of the 52th week of 2010, but because of splitting first by year the dates are assigned to 52th week of 2011. Same problem appears in 2012-12-31, this date is part of week one of 2013, but is assigned to first week of 2012 becaus i apply the function on each year seperate.

Splitting by year and than splitting every year into weeks give me the format i need, but the week-year relation cannot be correct. To get the correct week number i can split first by week and than by year:

splitByWeek = split(dates, isoweek(dates))
splitWeeksByYear = lapply(splitByWeek, function(x) split(x, year(x)))

But the format is not that what i need:

>splitWeeksByYear
...
$`53`
$`53`$`2004`
[1] "2004-12-31" "2004-12-29" "2004-12-28"

$`53`$`2005`
[1] "2005-01-01"

$`53`$`2009`
[1] "2009-12-28"

$`53`$`2015`
[1] "2015-12-30"

$`53`$`2016`
[1] "2016-01-03"

What is the best way to get the correct weeks in the format i need : list of $year $weekNum? (maybe transform the second result or do it in an complete other way?)

Answer 1

The week-numbering according to ISO 8601 has the benefit that ISO weeks always consist of 7 days without overlap or gap (as opposed to the US and UK week-numbering conventions).

However, it may happen that a few days around New Year belong to an ISO week of a different ISO week-year other than the calendar date year.

This is why lubridate has an isoyear() and an isoweek() function and format() recognizes the format specifiers %G , %g (ISO week-based year), and %V (ISO week).

So, with a slight modification OP's code works as expected:

library(lubridate)
splitByYears = split(dates, isoyear(dates))
splitYearsByWeeks = lapply(splitByYears, function(x) split(x, isoweek(x)))
splitYearsByWeeks$`2011`$`52`

 [1] "2011-12-28" "2011-12-27" "2011-12-29" "2011-12-31" "2012-01-01" "2011-12-30" [7] "2011-12-26" 

splitYearsByWeeks$`2012`$`1`

 [1] "2012-01-03" "2012-01-07" "2012-01-06" "2012-01-04" "2012-01-08" "2012-01-05" [7] "2012-01-02" 

However, splitting dates by the ISO week-based year and ISO week can be achieved also in one go in three slightly different ways:

splitted <- split(dates, format(dates, "%G-W%V"))
splitted$`2011-W52`

 [1] "2011-12-28" "2011-12-27" "2011-12-29" "2011-12-31" "2012-01-01" "2011-12-30" [7] "2011-12-26" 

splitted$`2012-W01`

 [1] "2012-01-03" "2012-01-07" "2012-01-06" "2012-01-04" "2012-01-08" "2012-01-05" [7] "2012-01-02" 

Alternatively, you may use the ISOweek package of which I am the author:

splitted <- split(dates, ISOweek::ISOweek(dates))

The split() function also accepts a list of factors in which case their interaction is used for the grouping:

library(lubridate)
splitted <- split(dates, list(isoyear(dates), isoweek(dates)))
splitted$`2011.52`

 [1] "2011-12-28" "2011-12-27" "2011-12-29" "2011-12-31" "2012-01-01" "2011-12-30" [7] "2011-12-26" 

splitted$`2012.1`

 [1] "2012-01-03" "2012-01-07" "2012-01-06" "2012-01-04" "2012-01-08" "2012-01-05" [7] "2012-01-02"