R devel - Dec 2002 - difftime arithmetic (PR#2345)

Full_Name: Barry Rowlingson
Version: 1.6.0
OS: RH8 i386
Submission from: (NULL) (148.88.136.205)


Strange things happen if I premultiply a difftime() object with a number.

Example:
> d1 <- difftime(Sys.time(),Sys.time())
> d2 <- 1 * difftime(Sys.time(),Sys.time())
> d3 <- difftime(Sys.time(),Sys.time()) * 1
> d1
Time difference of 0 secs
 
  - thats fine
> d2
[1] 0
attr(,"units")
[1] "secs"
attr(,"class")
[1] "difftime"

 - d2 prints out with print.default, although it is of class(difftime)!
print.difftime isnt being called, even though:
> class(d2)
[1] "difftime"
> d3
Time difference of 0 secs

 - d3 prints out fine.

But if I dput those three objects, the files look identical, and when dgetted
(dgot?) they all act the same.

But if instead I save() them and then load() them then d2 acts weird again.

 More weirdness:
> unclass(d1)
[1] 0
attr(,"units")
[1] "secs"
 
 - as expected, but:
> unclass(d2)
[1] 0
attr(,"units")
[1] "secs"
attr(,"class")
[1] "difftime"

 similarly for unclass(unclass(unclass(d2))) !

 Any ideas?

Barry Rowlingson
Lancaster University
Lancaster, UK

>>>>> "BaRow" == B Rowlingson
<B.Rowlingson@lancaster.ac.uk>
>>>>>     on Wed, 4 Dec 2002 17:58:51 +0100 (MET) writes:
    BaRow> Full_Name: Barry Rowlingson Version: 1.6.0 OS: RH8
    BaRow> i386 Submission from: (NULL) (148.88.136.205)

    BaRow> Strange things happen if I premultiply a difftime()
    BaRow> object with a number.

Thank you for the bug report!
Well, it's ``just'' the internal  object bit that got lost ->
is.object()

The bug is still there in R-devel,
and a shorter example is
> now <- Sys.time(); d1 <- difftime(now,now); d2 <- 1 * d1
> d1
Time difference of 0 secs
> d2
[1] 0
attr(,"units")
[1] "secs"
attr(,"class")
[1] "difftime"
> is.object(d1)
[1] TRUE
> is.object(d2)
[1] FALSE

----

We will have to fix that in the C code, too.
Martin

>>>>> "MM" == Martin Maechler
<maechler@stat.math.ethz.ch>
>>>>>     on Wed, 4 Dec 2002 18:09:08 +0100 (MET) writes:
    MM> We will have to fix that in the C code, too.  Martin

well, probably, not.

I found that we are simply missing an  Ops.difftime()
{method implementation for `arithmetic' generics}.

Martin