Home/ Questions/Q 6773063
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-26T15:37:20+00:00

Given a vector of datetime values, I needed to create a data.frame containing datetimes

  • 0

Given a vector of datetime values, I needed to create a data.frame containing datetimes at 0:10 hours past each original datetime – the first column 0 hours past, the second column 1 hour past, etc.

I had some trouble finding a way to do this easily using lubridate stuff. I thought this should work:

rt <- ymd_hms(c("2011-11-03 19:24:12", "2011-10-28 20:48:21",
  "2011-11-04 10:06:14", "2011-10-31 17:10:05", "2011-10-28 06:35:59"))
result <- outer(rt, hours(0:10), "+")

But various parts in that pipeline break down. Ultimately I get this error:

Error in FUN(X[[1L]], ...) : invalid 'times' argument

which seems to come from rep.POSIXct()‘s or rep.period()‘s inability to handle a non-unit-length times argument. Or something.

And it probably wouldn’t have worked anyway, because outer() returns a matrix, and date objects, even POSIXct dates (which are internally just integers) it seems can’t be elements in a matrix.

What I figured out that worked (just to get the times, not to put them in a data frame), after about 10 other guesses, was this:

with_tz(do.call(c, lapply(rt, function(x) x+hours(0:3))), tz(rt[1]))

The with_tz() addition is necessary because c() loses the timezone attribute. I also have to do do.call(c, lapply(...)) rather than just sapply(...) because sapply() loses the fact that it’s a date.

Maybe another alternative would be to create a data frame by doing do.call(cbind, ...) or something.

In general, it would be great if, whenever we find R date/time tasks that seem conceptually easy but require a lot of gymnastics before finding a solution, we could remove the roadblocks by making changes to lubridate, or whatever. I’m thinking this might be one of those times. =)

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    This doesn’t use outer(), but I think it gets you where you want. It does use plyr.

    library("lubridate")
library("plyr")

rt <- ymd_hms(c("2011-11-03 19:24:12", "2011-10-28 20:48:21",
  "2011-11-04 10:06:14", "2011-10-31 17:10:05", "2011-10-28 06:35:59"))

offsets = 0:10
names(offsets) <- offsets

dat <- data.frame(llply(offsets, function(offset){rt+hours(offset)}))

    Giving names to the offsets variable just makes the column names of the data.frame nicer.

    > str(dat)
'data.frame':   5 obs. of  11 variables:
 $ X0 : POSIXct, format: "2011-11-03 19:24:12" "2011-10-28 20:48:21" ...
 $ X1 : POSIXct, format: "2011-11-03 20:24:12" "2011-10-28 21:48:21" ...
 $ X2 : POSIXct, format: "2011-11-03 21:24:12" "2011-10-28 22:48:21" ...
 $ X3 : POSIXct, format: "2011-11-03 22:24:12" "2011-10-28 23:48:21" ...
 $ X4 : POSIXct, format: "2011-11-03 23:24:12" "2011-10-29 00:48:21" ...
 $ X5 : POSIXct, format: "2011-11-04 00:24:12" "2011-10-29 01:48:21" ...
 $ X6 : POSIXct, format: "2011-11-04 01:24:12" "2011-10-29 02:48:21" ...
 $ X7 : POSIXct, format: "2011-11-04 02:24:12" "2011-10-29 03:48:21" ...
 $ X8 : POSIXct, format: "2011-11-04 03:24:12" "2011-10-29 04:48:21" ...
 $ X9 : POSIXct, format: "2011-11-04 04:24:12" "2011-10-29 05:48:21" ...
 $ X10: POSIXct, format: "2011-11-04 05:24:12" "2011-10-29 06:48:21" ...

    UPDATE:

    Ken’s comment about ldply() versus data.frame(llply()) made me realize there is another way to approach this.

    dat <- ldply(rt, `+`, hours(0:10))

    which gives

    > str(dat)
'data.frame':   5 obs. of  11 variables:
 $ V1 : POSIXct, format: "2011-11-03 12:24:12" "2011-10-28 13:48:21" ...
 $ V2 : POSIXct, format: "2011-11-03 13:24:12" "2011-10-28 14:48:21" ...
 $ V3 : POSIXct, format: "2011-11-03 14:24:12" "2011-10-28 15:48:21" ...
 $ V4 : POSIXct, format: "2011-11-03 15:24:12" "2011-10-28 16:48:21" ...
 $ V5 : POSIXct, format: "2011-11-03 16:24:12" "2011-10-28 17:48:21" ...
 $ V6 : POSIXct, format: "2011-11-03 17:24:12" "2011-10-28 18:48:21" ...
 $ V7 : POSIXct, format: "2011-11-03 18:24:12" "2011-10-28 19:48:21" ...
 $ V8 : POSIXct, format: "2011-11-03 19:24:12" "2011-10-28 20:48:21" ...
 $ V9 : POSIXct, format: "2011-11-03 20:24:12" "2011-10-28 21:48:21" ...
 $ V10: POSIXct, format: "2011-11-03 21:24:12" "2011-10-28 22:48:21" ...
 $ V11: POSIXct, format: "2011-11-03 22:24:12" "2011-10-28 23:48:21" ...

    Note that, in addition to different column names (V1-V11 rather than X0-X10), these dates have been converted to local time (PDT, in my case):

    > dat$V1
[1] "2011-11-03 12:24:12 PDT" "2011-10-28 13:48:21 PDT"
[3] "2011-11-04 03:06:14 PDT" "2011-10-31 10:10:05 PDT"
[5] "2011-10-27 23:35:59 PDT"
    • 0