从每日时间序列转换为每周时间序列时，R apply.weekly（）返回不正确的时间段

Question

I am working with time-series data and experiencing a problem with apply.weekly() . It would appear that after a certain date, the weeks to not aggregate correctly.

library(xts)

value <-  c(46.40269, 47.27100 ,47.73311, 46.12858, 44.54989 ,42.79287, 41.70017 ,41.22373, 40.16180, 38.48705 ,37.02111 ,35.95312, 37.47187, 42.59649 ,49.22880, 53.96820, 57.97346, 61.22755,61.79824, 65.05720, 65.30233 ,61.86191,58.03687, 55.17815, 52.88933, 51.47876, 50.31402, 48.91674, 47.47042)
DATE <- as.Date(c("2038-01-03", "2038-01-04", "2038-01-05", "2038-01-06", "2038-01-07" ,"2038-01-08", "2038-01-09", "2038-01-10", "2038-01-11", "2038-01-12", "2038-01-13" ,"2038-01-14", "2038-01-15" ,"2038-01-16" ,"2038-01-17", "2038-01-18", "2038-01-19", "2038-01-20", "2038-01-21", "2038-01-22", "2038-01-23", "2038-01-24" ,"2038-01-25", "2038-01-26", "2038-01-27", "2038-01-28", "2038-01-29", "2038-01-30", "2038-01-31"))

DF <- data.frame(DATE, value)
DF_daily <- xts(DF$value, order.by = DF$DATE) 
DF_weekly <- apply.weekly(DF_daily, FUN=sum)

print(DF_weekly)

This generates the following output:

                [,1]
2038-01-03  46.40269
2038-01-10 311.39935
2038-01-16 231.69144
2038-01-31 840.70198

Notice how the final period is 15 days long. Now, if I instead use dates from 2010, I get exactly what you'd expect. That is, using

DATE <- as.Date(c("2010-01-03", "2010-01-04", "2010-01-05", "2010-01-06", "2010-01-07" ,"2010-01-08" ,"2010-01-09" ,"2010-01-10", "2010-01-11", "2010-01-12" ,"2010-01-13" ,"2010-01-14" ,"2010-01-15" ,"2010-01-16", "2010-01-17", "2010-01-18", "2010-01-19" ,"2010-01-20" ,"2010-01-21" ,"2010-01-22", "2010-01-23", "2010-01-24", "2010-01-25" ,"2010-01-26","2010-01-27" ,"2010-01-28" ,"2010-01-29" ,"2010-01-30", "2010-01-31"))

in the above code generates the output:

                [,1]
2010-01-03  46.40269
2010-01-10 311.39935
2010-01-17 280.92024
2010-01-24 427.18889
2010-01-31 364.28429

Is there something weird about the year 2038 I don't know about?

I am running this code on 64-bit Windows 7 Enterprise, sessionInfo() returns the following output

R version 3.2.3 (2015-12-10)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                           LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] xts_0.9-7  zoo_1.7-12

loaded via a namespace (and not attached):
[1] tools_3.2.3     grid_3.2.3      lattice_0.20-33

Answer 1

January 19, 2038 is a special date: At 3:14:08 AM, a 32-bit unix epoch (counting the number of seconds since midnight, January 1, 1970) will overflow. It possible that there is a bug in handling timestamps that causes a counter break at this date. Many numbers are stored as signed 32-bit integers, which have a maximum value of 2,147,483,647.

This is called the "Year 2038 Problem" , similar to the Y2K Problem.

However, the R Date type is in the number of days , instead of the number of seconds , since the Unix Epoch. To me, this suggests that there is an issue with the xts package.

You are not alone in this problem ( here is a 2012 discussion on a mailing list ), and it appears the bug is coming from a bad handoff between system date handling and R date handling.