[R] Referring to an object by a variable containing its name: 6 failures

Joshua Wiley jwiley.psych at gmail.com
Sat Sep 17 20:15:06 CEST 2011

Hi Andrew,

On Sat, Sep 17, 2011 at 12:04 AM, andrewH <ahoerner at rprogress.org> wrote:
> Dear Folks --
> The anonymous poster ("rmailbox") is perfectly correct.  I had forgotten you
> could use names in this way.  When referring to rows or columns by name
> rather than by number, I usually use either attach() or the $ operator,
> neither of which works here. If anyone understands why data.df[,colName]
> works in this setting but datadf$colName and the use of as.symbol(colName)
> after attach(data.df) do not work, i would love an explanation, because I
> sure don't.

Perhaps this will be a roughly correct rule of thumb to help: `$`
assumes it is being passed a variable name, `[` (and ilk) assume they
are passed a quoted name or a variable containing a variable name.
Six example:

d <- list(a = 1:10)
myname <- "a"


note that each does not work in one case but which case is different.

> Thanks, Timothy, for helping to clarify what I was trying to do. You are
> exactly right, and your analogy to the $$ command in PHP – a command that
> works -- was thereby more perfect than my analogy to things in R which do
> not work.
> Elk's suggestion to use the get() function was very welcome, as I had never
> really understood what get() was for, and this is a great use that often
> arises. However, for this purpose, get() is somewhat capricious in its
> effectiveness. “get(colName)” works as the operand of class(), length(),
> mode(), and summary(), but it does not work for typeof(), where it returns
> this error:
> "Error in eval(substitute(expr), data, enclos = parent.frame()) :  numeric
> 'envir' arg not of length one"
> And it does not work for str(), where it treats the variable name as a
> character string rather than a symbol.

I do not understand your question here---get() seems to work as I
would expect.  That said I would caution you (nay, beat you over the
head with countless logs until my limbs lacked strength to continue)
to avoid using attach() for things like datasets, at least until you
are comfortable enough with R and its lexical scoping that you no
longer see any use to using it and wonder why you ever did.  attach()
makes copies of data and adds it to the search path, and I would not
be surprised if some of your surprises were related to confusions
about why things worked or did not work.  I.e., typing:


should not work in my example above, but had I done:




would work, but


would not be referring to the same thing as:


attach() also sets users up to think that they can go around typing
the name of variables contained in lists, data.frames or some other
structure and it will just work.  To complicate matters, you are
trying to use all these things in a function, which has its own
environment.  Anyway, here are some examples with get() (although I
find myself using get() about as often as I find myself using attach):

## look at the search path

## rather than attach() I will use with:

with(d, get(a))
with(d, get("a"))
with(d, get(myname))

## before you mention all the retyping:
with(d, {
  x <- a + 5
  x2 <- rnorm(10) ^ x
  mean(data.frame(a, x, x2))

The only reason I had to retype with() was that I was expecting some
of the calls to produce errors and that halts execution.  In the other
with() (which I argue is an appropriate alternative to attach() to
save typing), note that x and x2 are not assigned to the global
environment, when the function is finished executing and it is
discarded, they are discarded with it----good for temporary

> Again, I do nut understand what distinguishes the functions for which Elk's
> solution works from those for which it does not. Does anybody know? Ideas
> welcome.

Happy to give more ideas if you give the examples that work/do not work.

> --and thanks again for all the help.



> andrewH
> --
> View this message in context: http://r.789695.n4.nabble.com/Referring-to-an-object-by-a-variable-containing-its-name-6-failures-tp3817129p3819813.html
> Sent from the R help mailing list archive at Nabble.com.
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Joshua Wiley
Ph.D. Student, Health Psychology
Programmer Analyst II, ATS Statistical Consulting Group
University of California, Los Angeles

More information about the R-help mailing list