endpoints fails on sparse monthly data

[joshuaulrich]

Since endpoints() only uses POSIXlt$mon to determine endpoints, it misses the case when the data is sparse and there is a year between observations. For example:

(x <- xts(1:6, as.Date(sprintf("199%d-06-%02d", c(2, 2, 2, 3:5), 21:26))))
#            [,1]
# 1992-06-21    1
# 1992-06-22    2
# 1992-06-23    3
# 1993-06-24    4
# 1994-06-25    5
# 1995-06-26    6
endpoints(x, "months", 1)
# [1] 0 6

When the correct result is:

endpointsMonthHack <- function(x, on = "months", k = 1) {
  # yearmon index
  ymIndex <- as.yearmon(index(x))
  # month change locations
  locations <- which(diff(ymIndex) != 0)
  # endpoints
  ep <- c(0, locations, nrow(x))
  unique(ep)
}
period.apply(x, endpointsMonthHack(x), mean)
#            [,1]
# 1992-06-23    2
# 1993-06-24    4
# 1994-06-25    5
# 1995-06-26    6

Thanks to Scott Waichler for the helpful report.