[R] Problem with very old dates

[Gabor Grothendieck]

On 10/29/06, Marc Schwartz <MSchwartz at mn.rr.com> wrote:
> On Sun, 2006-10-29 at 10:31 -0600, tom soyer wrote:
> > Hi,
> >
> > I noticed that as.Date() could not convert date string to date type if the
> > dates are very old. For example, if the date string is "1-Mar-50", then
> > as.Date() would convert this to "2050-03-01", NOT "1950-03-01". This seems
> > to be the behavior of as.Date() for dates older than 1969-1-1, and it is not
> > documented in the R as.Date() documentation. It seems very strange that R
> > would fail to convert old dates correctly. Does anyone know if this is the
> > correct behavior? If so, then which method should one use to convert old
> > dates?
> >
> > Thanks,
> >
> > Tom
> >
> > P.S., I am using R 2.4.0 for Windows.
>
> This is covered in ?strftime, which is also noted in the "See Also"
> for ?as.Date, where it says:
>
> "Your system's help pages on strftime and strptime to see how to specify
> their formats."
>
> In this case, the former help page in R indicates:
>
> %y
>        Year without century (00–99). If you use this on input, which
>        century you get is system-specific. So don't! Often values up to
>        69 (or 68) are prefixed by 20 and 70(or 69) to 99 by 19.
>
>
> Thus on FC5 Linux, I get:
>
> > as.Date("1-Mar-50", format = "%d-%b-%y")
> [1] "2050-03-01"
>
>
> Ideally, you should change the representation of the Year component of
> the dates you are working with to show a full four digit year and then
> use (note %Y (capital 'Y') instead of %y):
>
> > as.Date("1-Mar-1950", format = "%d-%b-%Y")
> [1] "1950-03-01"
>
> If this data was exported from another data source (ie. Excel) change
> the format in that program prior to exporting.
>
> Otherwise, you could do something like this in R using sub():
>
> > sub("-([0-9]+)$", "-19\\1", "1-Mar-50")
> [1] "1-Mar-1950"
>
> Which will change the two digit year ('50') to a four digit year
> ('1950').  See ?sub and ?regexp for more information.
>
> HTH,
>
> Marc Schwartz

As mentioned in the Help Desk article of Rnews 4-1, chron uses
the chron.year.expand option with a default of year.expand to do
the conversion from 2 digit to 4 digit.    year.expand has a
default cutoff of 30, i.e. years after 30 are regarded to be 19xx
and ones before 30 are 20xx.  Thus if that cutoff is ok for you:

library(chron)
as.Date(chron("1-Mar-50", format = "day-month-year"))

If the cutoff of 30 is ok then we have a solution.  Its possible to change
that in chron although as discussed in the article but as mentioned
it is not recommended that you change the chron.* options since it
might interfere with making your software interoperable with other software.

You could also do the 2 to 4 digit conversion yourself as suggested by Marc
or using gsubfn like this (where this example uses a cutoff of 10):

library(gsubfn)
gsubfn("..$", ~ as.numeric(x) + 100*(as.numeric(x) < 10) + 1900, "1-Mar-50")

This matches the last two digits and then adds 100+1900 if year <10
or adds 1900 if year is greater replacing those digits with the new 4 digit
number.  Then we can convert the output of gsubfn using as.Date
unambiguously with the appropriate format argument.