If there is no variable assignment to a function, does the output get garbage collected?

General
#1

I'm having an unpredictable error relating to a core dump, one thats very hard to make an MWE. In the mean-time, I'm wondering if I can fix the error reliably by making sure I'm not storing too many things in memory

foo = function(){
    runif(10000)
}

foo()

If I run this as a script, does the 10,000 long vector get garbage collected? Or do I have to wrap calls to foo() in a function that returns nothing in order for the memory to get freed?

#2

it would be garbage collected eventually, you could call gc() to force it.

1 Like
#3

It definitely doesn't get gc()'d straight away because you can still access it with .Last.value

#8

fixest have a github which is the best place to raise errors, try to focus on making a minimal reproducible example if you can.
An example of an active github issue being raised and fixed:

#9

Thanks. Posted an issue here. Hopefully I can find a solution.

#10

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.

#4

This is likely the source of my core dump, then. thank you!

#5

Sorry, just to confirm, I have a function that looks like this

inner_fun = function(x){
    # writes to a file 
    t = runif(1000000) * x
    return(t)
}

function outer_fun(){
    inner_fun(1)
    inner_fun(2)
    inner_fun(3)
}

I'm getting a crash that I think might be related to running out of memory. My guess is that allocations caused by the inner_fun calls are not being gcd. Is this a valid hypothesis?

#6

I would seriously doubt it ...
my reason for scepticism about that is that if you try to allocate a larger than possible vector, you'll get a meaningfull error message, and not a core dump

inner_fun = function(x){
  # writes to a file 
  t = runif(10^10) * x
  return(t)
}

inner_fun(1)
#Error: cannot allocate vector of size 74.5 Gb
#7

Sorry for causing an XY problem here with my less than complete description. I'm not used to worrying about performance or memory allocations in R code.

My actual problem is with a package called fixest which I think will allocate memory using cpp, so maybe something is fishy there. I will file an issue there if I see this problem again.