How can I load an object into a variable name that I specify from an R data file?

Question

When you save a variable in an R data file using save, it is saved under whatever name it had in the session that saved it. When I later go to load it from another session, it is loaded with the same name, which the loading script cannot possibly know. This name could overwrite an existing variable of the same name in the loading session. Is there a way to safely load an object from a data file into a specified variable name without risk of clobbering existing variables?

Example:

Saving session:

x = 5
save(x, file="x.Rda")

Loading session:

x = 7
load("x.Rda")
print(x) # This will print 5. Oops.

How I want it to work:

x = 7
y = load_object_from_file("x.Rda")
print(x) # should print 7
print(y) # should print 5

Solution 1

If you're just saving a single object, don't use an .Rdata file, use an .RDS file:

x <- 5
saveRDS(x, "x.rds")
y <- readRDS("x.rds")
all.equal(x, y)

Solution 2

I use the following:

loadRData <- function(fileName){
#loads an RData file, and returns it
    load(fileName)
    get(ls()[ls() != "fileName"])
}
d <- loadRData("~/blah/ricardo.RData")

Solution 3

You can create a new environment, load the .rda file into that environment, and retrieve the object from there. However, this does impose some restrictions: either you know what the original name for your object is, or there is only one object saved in the file.

This function returns an object loaded from a supplied .rda file. If there is more than one object in the file, an arbitrary one is returned.

load_obj <- function(f)
{
    env <- new.env()
    nm <- load(f, env)[1]
    env[[nm]]
}

Solution 4

You could also try something like:

# Load the data, and store the name of the loaded object in x
x = load('data.Rsave')
# Get the object by its name
y = get(x)
# Remove the old object since you've stored it in y 
rm(x)

Solution 5

Rdata file with one object

assign('newname', get(load('~/oldname.Rdata')))

View more solutions

Comments

[Ryan C. Thompson]

Is the new.env really necessary? Doesn't the function call itself provide a temporary environment?

[Wojciech Sobala]

You can return whole environment (return(env) or return(environment()) when you modify function according to @Ryan suggestion).

[hadley]

Updated to reflect that in 2.13 these are no longer experimental.

[Ryan C. Thompson]

Does that mean that they're fully supported, just like .Rdata files?

[hadley]

If you use the function environment it will contain f and have a parent. Probably not what you want

[Ryan C. Thompson]

Since these are no longer experimental, I'm marking this as the accepted answer. This is what I've been using.

[Ryan C. Thompson]

The point is to avoid clobbering the value of x when loading.

[Aleksandr Blekh]

Do saveRDS and readRDS, correspondingly, save and restore all object's attributes, including ones created by an application (via attr)? I tried to use this approach instead of save and load, trying to find a workaround for my problem. Howver, it doesn't seem to be the case, unless I'm doing something wrong: stackoverflow.com/questions/23701195/….

[Trevor Nederlof]

This is a great little function. Was having issues with trying to load things in (didnt want to just put them in the global environment as I know that was going to problems.

[Repmat]

Something like this belongs in base R

[Sander W. van der Laan]

OMG. This is awesome. Just what I need! #loveit

[nnachefski]

Can you please expand on the logic of the GET statement? Why not match to FILENAME instead of excluding?

[Ryan C. Thompson]

Won't this load the object into the old name, and then also assign it to the new name as well? That won't help if I'm worried about the possibility of overwriting an existing variable.

[Gerhard Burger]

For an RData with multiple objects, use mget instead of get