I'd like to check if some variable is defined in R - without getting an error. How can I do this?
My attempts (not successful):
> is.na(ooxx)
Error: object 'ooxx' not found
> is.finite(ooxx)
Error: object 'ooxx' not found
Thanks!
Add a comment
|
7 Answers
You want exists():
R> exists("somethingUnknown")
[1] FALSE
R> somethingUnknown <- 42
R> exists("somethingUnknown")
[1] TRUE
R>
answered Feb 20, 2012 at 21:51
Dirk is no longer here
370k60 gold badges668 silver badges742 bronze badges
4 Comments
CousinCocaine
@tim if you are inside a function, missing() is what you want.
Tomas
Might get a bit trickier if checking for list elements: stackoverflow.com/q/7719741
tim
what about for what the op wanted - using the variable name, not in quotes?
Julien
@tim
exists(quote(ooxx))See ?exists, for some definition of "...is defined". E.g.
> exists("foo")
[1] FALSE
> foo <- 1:10
> exists("foo")
[1] TRUE
Comments
if you are inside a function, missing() is what you want.
exchequer = function(x) {
if(missing(x)){
message("x is missing… :-(")
}
}
exchequer()
x is missing… :-(
As others have pointed out, you're looking for exists. Keep in mind that using exists with names used by R's base packages would return true regardless of whether you defined the variable:
> exists("data")
[1] TRUE
To get around this (as pointed out by Bazz; see ?exists), use the inherits argument:
> exists("data", inherits = FALSE)
[1] FALSE
foo <- TRUE
> exists("foo", inherits = FALSE)
[1] TRUE
Of course, if you wanted to search the name spaces of attached packages, this would also fall short:
> exists("data.table")
[1] FALSE
require(data.table)
> exists("data.table", inherits = FALSE)
[1] FALSE
> exists("data.table")
[1] TRUE
The only thing I can think of to get around this -- to search in attached packages but not in base packages -- is the following:
any(sapply(1:(which(search() == "tools:rstudio") - 1L),
function(pp) exists(_object_name_, where = pp, inherits = FALSE)))
Compare replacing _object_name_ with "data.table" (TRUE) vs. "var" (FALSE)
(of course, if you're not on RStudio, I think the first automatically attached environment is "package:stats")
3 Comments
CJB
Playing around, using argument
inherits = FALSE seems to isolate things in the global environment. Does that sound right?
MichaelChirico
@Bazz you're correct; I've edited this into the answer.
mzm
This comment should be higher up, since I use variable name "data", just using exist gave me some trouble initially.
If you don't want to use quotes, you can use deparse(substitute()) trick which I found in the example section of ?substitute:
is.defined <- function(sym) {
sym <- deparse(substitute(sym))
env <- parent.frame()
exists(sym, env)
}
is.defined(a)
# FALSE
a <- 10
is.defined(a)
# TRUE
3 Comments
chinsoon12
you can also
force or evaluate it in the function like this: is.defined <- function(sym) class(try(sym, TRUE))!='try-error'
Gregg H
this answer may be a tad more complicated...but this is the ideal answer if you don't want to contend with characters vs var-names
stucash
I have bumped into an issue that needs exactly your solution. When you wanted to use something that's not
NULL for quosure (enquo(x) if x is not null), you need a good way to check if the expression passed in is not null. Your solution does exactly that. thanksIf you don't mind using quotes, you can use:
exists("x")
If you don't want to use quotes you can use:
exists(deparse(substitute(x)))
1 Comment
Nirmal
This answer combines previous answers. Not sure if this is adding anything to the conversation so far.
There may be situations in which you do not exactly know the name of the variable you are looking for, like when an array of results have been created by a queuing system. These can possibly be addressed with "ls" and its argument "pattern" that expects a regular expression.
The "exists" function could be reimplemented that way as
exists <-function(variablename) {
#print(ls(env=globalenv()))
return(1==length(ls(pattern=paste("^",variablename,"$",sep=""),env=globalenv())))
}
While preparing this answer, I was a bit surprised about the need for the need of the specification of the environment when invoking ls() from within a function. So, thank you for that, stackoverflow! There is also an "all.names" attribute that I should have set to true but have omitted.
