CODEWITHSUNDEEP
rxtszoo
h.l.m
h.l.m

Reputation: 13475

na.approx changing timezone display

I recently found the zoo function na.approx but was wondering if there was a way to be able to use it without it altering the way that timezones are displayed, as shown if running my reproducible example...

set.seed(111)
x <- xts(matrix(cumprod(rnorm(100,0,0.001)+1)*100,
                ncol=1,dimnames=list(rep("",100),c("JJ"))),
         Sys.time()-c(100:1), tzone="America/Chicago")

x[30:50,] <- NA

tzone(x)
[1] "America/Chicago"
.index(head(x,1))
[1] 1377270598
attr(,"tzone")
[1] "America/Chicago"
attr(,"tclass")
[1] "POSIXct" "POSIXt" 
head(x)
                           JJ
2013-08-23 10:09:57 100.02352
2013-08-23 10:09:58  99.99044
2013-08-23 10:09:59  99.95928
2013-08-23 10:10:00  99.72914
2013-08-23 10:10:01  99.71210
2013-08-23 10:10:02  99.72609
Warning message:
timezone of object (America/Chicago) is different than current timezone (). 



y <- na.approx(x)

tzone(y)
[1] "America/Chicago"
.index(head(y,1))
[1] 1377270598
head(y)
                           JJ
2013-08-23 16:09:57 100.02352
2013-08-23 16:09:58  99.99044
2013-08-23 16:09:59  99.95928
2013-08-23 16:10:00  99.72914
2013-08-23 16:10:01  99.71210
2013-08-23 16:10:02  99.72609
Warning message:
timezone of object (America/Chicago) is different than current timezone ().

N.B. I am not based in Chicago....and my Sys.getenv('TZ') is not set to "America/Chicago"

Upvotes: 3

Views: 118

Answers (1)

Roland
Roland

Reputation: 132706

Edit: I could reproduce the issue with xts 0.9-3. It has been fixed in xts 0.9-5. You need to update xts.

I believe this could be considered a bug.

head(x)
#                           JJ
#2013-08-23 14:12:19 100.02352
#2013-08-23 14:12:20  99.99044
#2013-08-23 14:12:21  99.95928
#2013-08-23 14:12:22  99.72914
#2013-08-23 14:12:23  99.71210
#2013-08-23 14:12:24  99.72609

methods(na.approx)
#[1] na.approx.default na.approx.ts*     na.approx.zoo*    na.approx.zooreg*

So, there is no xts method. Since 

class(x)
#[1] "xts" "zoo"

na.approx.zoo is dispatched. Inside that function essentially this happens:

index(x) <- index(x)
head(x)
#                           JJ
#2013-08-23 21:12:19 100.02352
#2013-08-23 21:12:20  99.99044
#2013-08-23 21:12:21  99.95928
#2013-08-23 21:12:22  99.72914
#2013-08-23 21:12:23  99.71210
#2013-08-23 21:12:24  99.72609

Note that index<-.xts exists and is used here. Unfortunately, contrary to index<-.zoo, it doesn't preserve the time zone (of the index).

So, you could use y <- na.approx(as.zoo(x)) to avoid the issue.

Upvotes: 2

Related Questions