R help - May 2004 - Surprise when indexing with a factor.

Hello,

I had been thinking for years, without having ever checked (shame on 
me), that indexing a named vector by a factor 'f' produced the same 
results as indexing it by 'as.character(f)'. I was wrong, as the 
following example shows:

  (m <- c(a=1,b=2))
  (f <- factor(c(1,2),labels=c('b','a')))

  m[f]
  m[as.character(f)]
  m[as.numeric(f)]

When the labels of a factor are not sorted alphabetically, m[f] and 
m[as.character(f)] yield different results. m[f] is actually equivalent 
to m[as.numeric(f)].

I'm not the first to have been beaten by this (documented) feature, as I 
discovered in a recent thread ("[R] Indexing by factor misfeature", 
<http://tolstoy.newcastle.edu.au/R/help/04/04/0937.html> see 
http://tolstoy.newcastle.edu.au/R/help/04/04/0938.html).

If I am breaking the posting rules by writing again about a previous 
subject, it is only
to suggest to add a short notice about that behavior in the section of 
'R-intro.pdf' about 'index vectors'. Currently, it does not
mention the
use of factors as indexes (It is true that factors are only defined 
later).  At least, this manual could tell the reader to read the help 
about Extract, or type '?"["' (which did not occur to me
before writing
this message...).

Christophe Pallier

It may be educational to read ?factor before you use a factor for some
operation (such as subscripting), I guess.  In part, it says:

Value:

     'factor' returns an object of class '"factor"'
which has a set of
     numeric codes the length of 'x' with a '"levels"'
attribute of
     mode 'character'.  If 'ordered' is true (or
'ordered' is used) the
     result has class 'c("ordered", "factor")'.

In other words, a factor is a numeric vector with a "levels"
attribute.
What do you expect to happen when you use a numeric vector as subscript?

Andy
> From: pallier
> 
> Hello,
> 
> I had been thinking for years, without having ever checked (shame on 
> me), that indexing a named vector by a factor 'f' produced the same
> results as indexing it by 'as.character(f)'. I was wrong, as the 
> following example shows:
> 
>   (m <- c(a=1,b=2))
>   (f <- factor(c(1,2),labels=c('b','a')))
> 
>   m[f]
>   m[as.character(f)]
>   m[as.numeric(f)]
> 
> When the labels of a factor are not sorted alphabetically, m[f] and 
> m[as.character(f)] yield different results. m[f] is actually 
> equivalent 
> to m[as.numeric(f)].
> 
> I'm not the first to have been beaten by this (documented) 
> feature, as I 
> discovered in a recent thread ("[R] Indexing by factor
misfeature",
> <http://tolstoy.newcastle.edu.au/R/help/04/04/0937.html> see 
> http://tolstoy.newcastle.edu.au/R/help/04/04/0938.html).
> 
> If I am breaking the posting rules by writing again about a previous 
> subject, it is only
> to suggest to add a short notice about that behavior in the 
> section of 
> 'R-intro.pdf' about 'index vectors'. Currently, it does not
> mention the 
> use of factors as indexes (It is true that factors are only defined 
> later).  At least, this manual could tell the reader to read the help 
> about Extract, or type '?"["' (which did not occur to me 
> before writing 
> this message...).
> 
> Christophe Pallier
> 

------------------------------------------------------------------------------
Notice:  This e-mail message, together with any attachments,...{{dropped}}