[R] 1.8.1 behavior change?

[Prof Brian Ripley]

On Sat, 22 Nov 2003, Al Piszcz wrote:

> In <R 1.8.1 the following fragment worked properly, now (1.8.1)

I take it you mean R < 1.8.1?  In trying to reproduce something
like this

stt <- data.frame()
stt[1] <- as.matrix(data.frame(foo=1:10, bar=letters[1:10]))

in 1.8.0 I got

> stt
[1] V1
<0 rows> (or 0-length row.names)

so is that what `worked properly' means?  (It is also what S+6.1 does, but
I am pretty sure it is not what anyone wanted.  Note the number of 
columns is wrong, too.)

In 1.7.1 it gave a similar error to 1.8.1.

So as far as I can see (given you did not supply a test example) 1.8.1 is
just more helpful by pointing out your user error.  The rule is that the
existing columns should be replicated an exact number of times when
extending a data frame: and that includes the row names so applies even if
there are no other columns.

I hope each of your input files except the last have just one column (in 
which case your code is very clunky) as otherwise you are asking for 
columns to be overwritten.  I would advise you to use cbind for this, 
or dimension the data frame correctly in the first place and replace 
sections at a time.

> it creates the following warning/error:
> 
> Any advice appreciated.

Supply reproducible examples, as the FAQ asks.  I have no idea why you
have given the same code twice here, BTW, so I remove the first copy.


> >   stt <- data.frame()
> >   # load all datasets into a dataframe
> >   for (ds in 1:n) {
> +     stt[ds] <- as.matrix(read.table(fileList[ds]))
> +   }
> Error in "[<-.data.frame"(`*tmp*`, ds, value =
> as.matrix(read.table(fileList[ds]))) :
>         replacement has 358 rows, data has 0

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595