Memory usage keeps accumulating when set_attribute() is used?

[SQ3 W23]

Hello,

I'm using R 4.1.2 and simmer 4.4.5, I am working on a simulation that needs to replicated many times, and I noticed that the memory usage kept increasing. It seems to be because of set_attribute(), here is an example:

simulate <- function(time) {

    env <- simmer("Hospital process")

   

    patient <- trajectory("patient")  %>%

      set_attribute("rand", function() {rnorm(1)})  

    env <- env %>%

      add_generator("patient", patient, function() {1},mon=1) %>%

      run(time) %>%

      wrap()

    env

}

mem<-c(1:10)

for (x in 1:10) {

  for (i in 1:500) {

    env <- simulate(100)

  }

  mem[x]<-memory.size()

}

Memory usage when running 10 times:

mem

[1] 535.47 534.54 539.49 537.81 542.58 552.64 546.90 548.49 608.48 614.58

Normally, memory usage fluctuates around a value and doesn't keep increasing like this. Would it help to use rm() somewhere?

Thank you.

[Iñaki Ucar]

Please try wrapping the environment into a new variable, then rm'ing everything else. Something like

out <- env %>%

  add_generator("patient", patient, function() {1},mon=1) %>%
  run(time) %>%
  wrap()

rm(env, patient)
return(out)

--

Iñaki Úcar