[R] Strange zoo behaviour, possible bug?

[Achim Zeileis]

On Wed, 28 Apr 2010, Giuseppe Milicia wrote:

> Hi there,
>
> thanks for the reply. I agree that it's essentially a user error.
>
> I was really wondering whether the behaviour has changed from past 
> version of zoo and whether it'd be worth to handle this case in a ad-hoc 
> manner. I'm using old code and I was a bit surprised I didn't bump on 
> this earlier...

I'm fairly sure that zoo always behaved like this but I didn't go back to 
check.

In any case: I now added explicit handling of this case

   rbind(zoo_vector, zoo_one_column_matrix)
   rbind(zoo_one_column_matrix, zoo_vector)

where the former returns a vector and the latter a matrix.

You can access it by checking out the SVN from R-Forge or wait a couple of 
days for the package builds on R-Forge.

Best,
Z


> I have a function which can take as input zoo object whose coredata is either a matrix or a vector, to do that I'll have to handle the two cases slightly different. Not a bit deal, but kind of inelegant.
>
> Cheers,
>
> //Giuseppe
>
> ________________________________________
> From: Achim Zeileis [Achim.Zeileis at uibk.ac.at]
> Sent: 28 April 2010 16:25
> To: Giuseppe Milicia
> Cc: r-help at r-project.org
> Subject: Re: [R] Strange zoo behaviour, possible bug?
>
> On Wed, 28 Apr 2010, Giuseppe Milicia wrote:
>
>> Hi all,
>>
>> I bumped into this awkward zoo behaviour. I'd be half tempted to call it
>> a bug, what do you think?
>
> The situation could probably be improved on the zoo side but the source of
> the problem is clearly user error.
>
>> It's annoying to work around it :(
>
> You would just have to use it appropriately :-)
> In addition to that, we can maybe improve warnings/errors.
>
>>> t1 = zoo(-100,   as.POSIXct("2009-12-31")+(2:10)*60*60*24)
>>> t2 = zoo(matrix(0), index(t1)[1]-1)
>
> Just to be clear: t1 is a vector, t2 is a one-column matrix.
>
>>> colnames(t1)="test"
>
> This doesn't work and it tells you that it doesn't work.
>
> Error in `colnames<-`(`*tmp*`, value = "test") :
>   attempt to set colnames on object with less than two dimensions
>
> If you want it to work, then the data in the time series should be a
> matrix (as you showed at the bottom of your post), e.g.:
>
> t1 <- zoo(as.matrix(coredata(t1)), time(t1))
> colnames(t1) <- "test"
>
> If you do that, everything else will work as expected.
>
>>> rbind(t1,t2)
>>
>> 2010-01-01 23:59:59    0
>> 2010-01-02 00:00:00 -100
>> 2010-01-03 00:00:00    0
>> 2010-01-04 00:00:00 -100
>> 2010-01-05 00:00:00    0
>> 2010-01-06 00:00:00 -100
>> 2010-01-07 00:00:00    0
>> 2010-01-08 00:00:00 -100
>> 2010-01-09 00:00:00    0
>> 2010-01-10 00:00:00 -100
>> Warning message:
>> In rbind(c(-100, -100, -100, -100, -100, -100, -100, -100, -100),  :
>>  number of columns of result is not a multiple of vector length (arg 1)
>
> Yes, this is not nice, but it does warn you that something went wrong!
>
>>> rbind(as.xts(t1),as.xts(t2))
>
> "xts" stores everything internally as a matrix, hence there cannot be a
> confusion between vectors and matrices.
>
> I'll have a look whether we can throw a better warning or maybe handle the
> border case of rbinding a vector with a one-column matrix.
>
> Best,
> Z
>
>
> ---- MAKO ----
> This email and any files transmitted with it are confi...{{dropped:4}}