modify variable within R function

Question

How do I modify an argument being passed to a function in R? In C++ this would be pass by reference.

g=4
abc <- function(x) {x<-5}
abc(g)

I would like g to be set to 5.

Solution 1

There are ways as @Dason showed, but really - you shouldn't!

The whole paradigm of R is to "pass by value". @Rory just posted the normal way to handle it - just return the modified value...

Environments are typically the only objects that can be passed by reference in R.

But lately new objects called reference classes have been added to R (they use environments). They can modify their values (but in a controlled way). You might want to look into using them if you really feel the need...

Solution 2

There has got to be a better way to do this but...

abc <- function(x){eval(parse(text = paste(substitute(x), "<<- 5")))}
g <- 4
abc(g)
g

gives the output

[1] 5

Solution 3

I have a solution similar to @Dason's, and I am curious if there are good reasons not to use this or if there are important pitfalls I should be aware of:

changeMe = function(x){
assign(deparse(substitute(x)), "changed", env=.GlobalEnv)
}   

Solution 4

I think that @Dason's method is the only way to do it theoretically, but practically I think R's way already does it.

For example, when you do the following:

y <- c(1,2)
x <- y

x is really just a pointer to a the value c(1,2). Similarly, when you do

abc <- function(x) {x <- 5; x}
g <- abc(g)

It is not that you are spending time copying g to the function and then copying the result back into g. I think what R does with the code

g <- abc(g)

is:

1. The right side is looked at first. An environment for the function abc is set up.
2. A pointer is created in that environment called x.
3. x points to the same value that g points to.
4. Then x points to 5
5. The function returns the pointer x
6. g now points to the same value that x pointed to at the time of return.

Thus, it is not that there is a whole bunch of unnecessary copying of large options.

I hope that someone can confirm/correct this.

View more solutions

Comments

[Paul Hiemstra]

my feeling is that the R way to do is to say g = abc() where abc returns 5. why this specific syntax?

[joran]

With a slight modification to abc so that it actually returns a value, yes.

[Alex]

yes that would work. the point is to modify the value within teh function though without throwing around all these big objects and making copies of them.

[Joshua Ulrich]

There's also the proto package.

[Alex]

interesting! i would be curious to know if this is right. in my program that has a df that is approximately a gig, it seems that df gets copied several times as my RAM usage grows pretty large

[Tommy]

@Alex: In this case, yes - but only since it is trivial. If you modify a part of a large vector it is typically (but not always) copied. x <- runif(1e6); y <- x; x[4] <- 42 makes a copy so that x and y are now different.

[Dason]

@joran nothing like getting called out ~8 years later to make one feel at home on SO. But yeah you're right. And I threw an upvote their way. But really I want to downvote myself and their answer and burn the whole question to the ground.

[Rafa]

I have arrived to this answer out of desperation, and believe there might be a justified use case for this. I am using a websocket API to receive updates on stock prices, with the websocket package. The new data is received only as 1 data point (the most recent) at a time, which I then need to collate into a data frame or similar holding all data points until current time. Every time a message is received from the websocket connection, an onMessage hook is executed. The most reasonable solution I have found is to include an assignment in the onMessage hook of the form...

[Rafa]

... that you mention to modify the value of the existing data frame to cbind(dataframe, new_data_row) . The reason it wouldn't be practical to be applying the standard "return the new value" paradigm is that the hook is internally handled by the websocket package code

[Dason]

@Rafa I don't know what you're doing but handling any input from the web using eval just seems like asking to get hacked.