[R] c() and dates
I'm a bit puzzled by a certain behavior with dates. (R version 3.1.1) temp1 - as.Date(1:2, origin=2000/5/3) temp1 [1] 2000-05-04 2000-05-05 temp2 - as.POSIXct(temp1) temp2 [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT So far so good. On 5/4, midnight in Greenwich it was 19:00 on 5/3 in my time zone. The manual page has a clear explanation of what goes on. c(temp1, temp2) [1] 2000-05-042000-05-052623237-10-15 2623474-05-06 class(c(temp1, temp2)) [1] Date c(temp2, temp1) [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT [3] 1969-12-31 21:04:41 CST 1969-12-31 21:04:42 CST class(c(temp2, temp1)) [1] POSIXct POSIXt I would have expected c() to determine a common class, somehow, then do the conversion and concatonate. That is obviously not what happens. I've read the manual page but I must be missing something. I make no claim that R is broken, mistaken, or otherwise deficient, only that my understanding is so. Could someone illuminate? Terry T. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] c() and dates
On Oct 3, 2014, at 7:19 AM, Therneau, Terry M., Ph.D. wrote: I'm a bit puzzled by a certain behavior with dates. (R version 3.1.1) temp1 - as.Date(1:2, origin=2000/5/3) temp1 [1] 2000-05-04 2000-05-05 temp2 - as.POSIXct(temp1) temp2 [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT So far so good. On 5/4, midnight in Greenwich it was 19:00 on 5/3 in my time zone. The manual page has a clear explanation of what goes on. c(temp1, temp2) [1] 2000-05-042000-05-052623237-10-15 2623474-05-06 class(c(temp1, temp2)) [1] Date c(temp2, temp1) [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT [3] 1969-12-31 21:04:41 CST 1969-12-31 21:04:42 CST class(c(temp2, temp1)) [1] POSIXct POSIXt I would have expected c() to determine a common class, somehow, then do the conversion and concatonate. That is obviously not what happens. I've read the manual page but I must be missing something. I make no claim that R is broken, mistaken, or otherwise deficient, only that my understanding is so. It doesn't appear that any check is made on the commonality of class by either c.Date (which one would expect to be called when the Date object is the first argument) . or with c.POSIXctt: c.Date function (..., recursive = FALSE) structure(c(unlist(lapply(list(...), unclass))), class = Date) bytecode: 0x10bc1c0e0 environment: namespace:base c.POSIXct function (..., recursive = FALSE) .POSIXct(c(unlist(lapply(list(...), unclass bytecode: 0x10bc0d470 environment: namespace:base I don't find any description of the behavior of `c.Date` when I go to a help page with ?c.Date. The only description of the action of `c.POSIXct` that I can find in the page to which we are sent with ?c.POSIXct says: Using c on POSIXlt objects converts them to the current time zone, and on POSIXct objects drops any tzone attributes (even if they are all marked with the same time zone). Could someone illuminate? Agree it's unexpected behavior and worthy of a feature request to R Core, but until that obvious undesireable behavior is corrected, I guess the user is left with the responsibility of converting all objects to a common class. -- David Winsemius Alameda, CA, USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] c() and dates
On Oct 3, 2014, at 7:19 AM, Therneau, Terry M., Ph.D. wrote: I'm a bit puzzled by a certain behavior with dates. (R version 3.1.1) temp1 - as.Date(1:2, origin=2000/5/3) temp1 [1] 2000-05-04 2000-05-05 temp2 - as.POSIXct(temp1) temp2 [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT So far so good. On 5/4, midnight in Greenwich it was 19:00 on 5/3 in my time zone. The manual page has a clear explanation of what goes on. c(temp1, temp2) [1] 2000-05-042000-05-052623237-10-15 2623474-05-06 class(c(temp1, temp2)) [1] Date c(temp2, temp1) [1] 2000-05-03 19:00:00 CDT 2000-05-04 19:00:00 CDT [3] 1969-12-31 21:04:41 CST 1969-12-31 21:04:42 CST class(c(temp2, temp1)) [1] POSIXct POSIXt I would have expected c() to determine a common class, somehow, then do the conversion and concatonate. That is obviously not what happens. I've read the manual page but I must be missing something. I make no claim that R is broken, mistaken, or otherwise deficient, only that my understanding is so. Could someone illuminate? Followup to my earlier post: It's pretty easy to redefine c.Date and c.POSIXct to behave in hte manner is expected: c.Date - function (..., recursive = FALSE) structure(c(unlist(lapply(list(...), as.Date))), class = Date) temp1 - as.Date(1:2, origin=2000/5/3) temp1 #[1] 2000-05-04 2000-05-05 temp2 - as.POSIXct(temp1) c(temp1, temp2) #[1] 2000-05-04 2000-05-05 2000-05-04 2000-05-05 # class(c(temp1, temp2)) #[1] Date # c(temp2, temp1) [1] 2000-05-03 17:00:00 PDT 2000-05-04 17:00:00 PDT 1969-12-31 19:04:41 PST [4] 1969-12-31 19:04:42 PST c.POSIXct - function (..., recursive = FALSE) .POSIXct(c(unlist(lapply(list(...), as.POSIXct c(temp1, temp2) [1] 2000-05-04 2000-05-05 2000-05-04 2000-05-05 class(c(temp1, temp2)) [1] Date c(temp2, temp1) [1] 2000-05-03 17:00:00 PDT 2000-05-04 17:00:00 PDT 2000-05-03 17:00:00 PDT [4] 2000-05-04 17:00:00 PDT -- David Winsemius Alameda, CA, USA __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] c() and dates
Well duh -- type c.Date at the command prompt to see what is going on. I suspected I was being dense. Now that the behaior is clear can I follow up on David W's comment that redfining the c.Date function as structure(c(unlist(lapply(list(...), as.Date))), class = Date) allows for a more intellegent response, since it allows all of the as.Date machinery to be brought into play. It seems like a good idea in general. Would it be a good exchange between the current nonsense result, no warning and the new error messages that would arise, e.g., from c(as.Date(2000/10/1), factor('b')). Terry T. On 10/03/2014 09:52 AM, peter dalgaard wrote: S3 only has single dispatch, so in one case it dispatches to c.Date and in the other to c.POSIXct, both of those return an object of the corresponding class. In both cases, the arguments pass through c(unlist(lapply(list(...), unclass))) which doesn't look at the class at all. Since Date objects unclass to days and POSIXct to seconds, something is bound to go wrong. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.