Y2K problem

[Agin.Patrick@hydro.qc.ca]

Hi,

Under Windows NT on Splus 4.5, I try to read a text file which contains in column 1 dates and in column 2 numbers. Here are the first few lines of the file:

12-Dec-85                                         10.05
04-Jan-86                                         10.55
11-Feb-86                                         10.26
13-Mar-86                                          9.79
15-Apr-86                                          9.83

First, I read the data in vec:

> vec <- scan(file = "D:\\timeseries.txt", what = list("character", numeric()))

After, I construct an univariate time series with arbitrary sampling time intervals in ts0:

> ts0 _ its(vec[[2]],times=dates(vec[[1]], format = "day-mon-yr"))
Error in its(vec[[2]], times = dates(vec[[1]], format = "day-mon-y..: times not in ascending order
Dumped

As you see, the error message seems to indicate that dates in my file are not in ascending order. I found after many tries that the problem concerns year 2000 dates. The command

       

> ts0 _ its(vec[[2]][1:2334],times=dates(vec[[1]][1:2334], format = "day-mon-yr"))

returns no error message and
       

> ts0 _ its(vec[[2]][1:2335],times=dates(vec[[1]][1:2335], format = "day-mon-yr"))

returns the same error message as above. Here is the vector of dates around 2335:

> vec[[1]][2333:2336]
[1] "30-Dec-99" "31-Dec-99" "04-Jan-00" "05-Jan-00"
> dates(vec[[1]][2333:2336],format="day-mon-year")
[1] 12/30/99 12/31/99 01/04/0  01/05/0

The 2335th element is the first occurence in my file of year 2000. Now, if you sort the vector, the result is:

> sort(dates(vec[[1]][2333:2336],format="day-mon-yr"))
[1] 01/04/0  01/05/0  12/30/99 12/31/99

S-Plus seems to consider 01/04/0 as 01/04/1900 (which obviously is less than 12/30/1999). Is there a setting that you can set to tell S-Plus to consider 99 as 1999 and 00 as 2000?

Thank you very much,
Patrick