R devel - Sep 2008 - indexing by character with identical names

Hi all,

I've been using R for a while, but was puzzled when I just barely
noticed this behavior:
> x <- 1:10
> names(x) <- c("a","b","a",letters[1:7])
> x
 a  b  a  a  b  c  d  e  f  g 
 1  2  3  4  5  6  7  8  9 10 
> x["a"]
a 
1 

... that is, if a vector has identical (and therefore ambiguous) names,
it only returns the first match when the vector is indexed by a
non-unique name.  From my perspective, a more sensible behavior would be
to return *all* elements which carry the non-unique name (e.g.
x[names(x)=="a"] ), since it's ambiguous as to which element the
user
actually wanted to extract.  It just seems safer than quietly returning
only the first matched element with no warning or other indication as to
what's actually happening.

I couldn't find that this behavior is explicitly documented in the help
file for '[' or 'names'.

I also couldn't find any prior discussion of this behavior.  Is this
something that could be addressed in an upcoming release of R? 

Thanks,

jake

R version 2.7.2 (2008-08-25) 
x86_64-unknown-linux-gnu

2008/9/19 Jake Michaelson <jake.michaelson at gmail.com>
> ... that is, if a vector has identical (and therefore ambiguous) names,
> it only returns the first match when the vector is indexed by a
> non-unique name.  From my perspective, a more sensible behavior would be
> to return *all* elements which carry the non-unique name (e.g.
> x[names(x)=="a"] ), since it's ambiguous as to which element
the user
> actually wanted to extract.  It just seems safer than quietly returning
> only the first matched element with no warning or other indication as to
> what's actually happening.
 Some people may think it sensible that indexing a vector with an
index of length N returns a vector of length N:

 > x[c("a","b","a")]
 a b a
 1 2 1
> I also couldn't find any prior discussion of this behavior.  Is this
> something that could be addressed in an upcoming release of R?
 Changing this would probably (no, definitely!) break stuff. Breaking
stuff is a bad thing. Perhaps the upcoming R version Aleph-null
release would be the right place for it...

 R is not like other programming languages, so your perspective of
sensibility of minimal use here. You will get bitten by these gotchas.
I bet everyone on this mailing list has done a X[1,] and got bitten by
not having drop=FALSE....

 Ouch. Writing tests on edge cases is a good thing, since it catches
these problems (if you write your tests well!).

Barry