R help - Apr 2004 - Indexing by factor misfeature

Yesterday I was biten by a feature, which I find too dangerous.

I wanted to use a factor `Subject?? as index into a data frame, whose row
names were the levels of this factor. So there a 2 different possible
interpretations of this: Either Subject is coerced to numeric or to
character. The intended interpretation was, of course,
`as.character(Subject)'.
R did `as.numeric(Subject)??. This will not cause a semantic error, when
the levels of Subject happen to be in numeric order; so it will be
unnoticed quite often.

I propose to issue a warning whenever a factor is implicitly coerced,
at least in an ambiguous context.

This *is* a documented feature: see ?Extract.

I am afraid your `of course' is completely non-obvious to me.  I would 
object strongly to code written to use a documented feature issuing a 
warning.

On Tue, 20 Apr 2004, Wilhelm B. Kloke wrote:
> Yesterday I was biten by a feature, which I find too dangerous.
> 
> I wanted to use a factor `Subject?? as index into a data frame, whose row
> names were the levels of this factor. So there a 2 different possible
> interpretations of this: Either Subject is coerced to numeric or to
> character. The intended interpretation was, of course,
`as.character(Subject)'.
> R did `as.numeric(Subject)??. This will not cause a semantic error, when
> the levels of Subject happen to be in numeric order; so it will be
> unnoticed quite often.
> 
> I propose to issue a warning whenever a factor is implicitly coerced,
> at least in an ambiguous context.
-- 
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

> Yesterday I was biten by a feature, which I find too dangerous.
>From your description, it sounds like you confused row lables and vector
values -- two very different things.  It also sounds like you mixed up
factor values and levels.  Let me know if I'm wrong.
> I wanted to use a factor `Subject?? as index into a data frame, whose row
> names were the levels of this factor.
> So there a 2 different possible
> interpretations of this: Either Subject is coerced to numeric or to
> character. The intended interpretation was, of course,
> `as.character(Subject)'.
> R did `as.numeric(Subject)??.
What did you do to make R do this coersion?  I can't tell from here.

Try this, with the same "Subject" vector:

foo <- factor(Subject)
foo
levels(foo) <- names(Subject)
foo

Does that help?

Cheers

Jason