I was trying to figure out the best way to deal with Postgresql's ability to represent Infinity and -Infinity in their timestamps when using RPostgreSQL to bring data over into R. Along the way I found some strange behavior when trying to represent infinite dates in R.
I can attempt to create a date at negative and positive infinity in the following manner:
❥ as.Date(-1/0, origin="1970-01-01")
[1] NA
❥ as.Date(1/0, origin="1970-01-01")
[1] NA
They both appear to be NAs. However when comparing them, there seems to be an understanding that one is less than the other.
❥ as.Date(-1/0, origin="1970-01-01") < as.Date(1/0, origin="1970-01-01")
[1] TRUE
❥ as.Date(-1/0, origin="1970-01-01") > as.Date(1/0, origin="1970-01-01")
[1] FALSE
❥ as.Date(1/0, origin="1970-01-01") > as.Date("1970-01-01")
[1] TRUE
❥ as.Date(1/0, origin="1970-01-01") < as.Date("1970-01-01")
[1] FALSE
How does R know the difference, if they both convert to NA?
They don't convert to NA, that's just how they're printed.
R> d <- as.Date(-Inf, origin="1970-01-01")
R> is.na(d)
# [1] FALSE
R> is.infinite(d)
# [1] TRUE
If you want them to print differently, you can override the print.Date method and add special cases for +/- infinity.