Find whether a particular date is an Option Expiration Friday - problem with timeDate package

I am trying to write a simple function that (should) return true if the parameter date(s) is an Op-Ex Friday.

require(timeDate)
require(quantmod)

getSymbols("^GSPC", adjust=TRUE, from="1960-01-01")
assign("SPX", GSPC, envir=.GlobalEnv)
names(SPX) <- c("SPX.Open", "SPX.High", "SPX.Low", "SPX.Close", 
                "SPX.Volume", "SPX.Adjusted")

dates <- last(index(SPX), n=10)
from <- as.numeric(format(as.Date(min(dates)), "%Y"))
to <- as.numeric(format(as.Date(max(dates)), "%Y"))

isOpExFriday <- ifelse(
  isBizday(
    timeDate(as.Date(dates)), 
    holidayNYSE(from:to)) & (as.Date(dates) == as.Date(
      format(timeNthNdayInMonth(timeFirstDayInMonth(dates), nday=5, nth=3)))
  ), TRUE, FALSE)

Now, the result should be [1] "2011-09-16". But instead I get [1] "2011-09-15":

dates[isOpExFriday]
[1] "2011-09-15"

Am I doing something wrong, expecting something 开发者_JAVA百科that timeDate package is not doing by design or is there a bug in timeDate?

I am guessing it's a timezone problem. What happens if you use this:

format(dates[isOpExFriday], tz="UTC")

On second look, you probably need to put the 'tz=' argument inside the format call inside the as.Date(format(...)) call. The format function "freezes" that dates value as text.

EDIT: On testing however I think you are right about it being a bug. (And I sent a bug report to the maintainer with this response.) Even after trying to insert various timezone specs and setting myFinCenter in RmetricsOptions, I still get the which stems from this error deep inside your choice of functions:

timeNthNdayInMonth(as.Date("2011-09-01"), nday=5, nth=3)
America/New_York
[1] [2011-09-15]

I suspect it is because of this code since as I understand it Julian dates are not adjusted for timezones or daylight savings times:

ct = 24 * 3600 * (as.integer(julian.POSIXt(lt)) + 
          (nth - 1) * 7 + (nday - lt1$wday)%%7)
class(ct) = "POSIXct"

The ct value in seconds is then coverted to POSIXct from second since "origin" simply by coercion of class. If I change the code to:

ct=as.POSIXct(ct, origin="1970-01-01")  # correct results come back

My quantmod and timeDate versions are both current per CRAN. Running Mac with R 2.13.1 in 64 bit mode with a US locale. I have not yet tried to reproduce with a minimal session so there could still be some collision or hijacking with other packages:

> sessionInfo()
R version 2.13.1 RC (2011-07-03 r56263)
Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)

locale:
[1] en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
 [1] quantmod_0.3-17   TTR_0.20-3        xts_0.8-2        
 [4] Defaults_1.1-1    timeDate_2130.93  zoo_1.7-4        
 [7] gplots_2.10.1     KernSmooth_2.23-6 caTools_1.12     
[10] bitops_1.0-4.1    gdata_2.8.1       gtools_2.6.2     
[13] wordnet_0.1-8     ggplot2_0.8.9     proto_0.3-9.2    
[16] reshape_0.8.4     plyr_1.6          rattle_2.6.10    
[19] RGtk2_2.20.17     rms_3.3-1         Hmisc_3.8-3      
[22] survival_2.36-9   sos_1.3-0         brew_1.0-6       
[25] lattice_0.19-30