R: repeate replicate if output is NA

I am solving numerically stochastic differential equation with use of diffeqr in R, check if the last value is large enough and replicate the process for 5 times. The piece of code for this part is:

for(i in 1:30){
  for(j in 1:15) {
    u0<-c(1.5e+6+i*5e+4,j*2e+5)
    data<-replicate(n=5,{
      prob <- de$SDEProblem(f,g,u0,tspan,p)
      fastprob<-diffeqr::jitoptimize_sde(de,prob)
      sol <- de$solve(fastprob,saveat=0.005)
      mat <- sapply(sol$u,identity)
      udf <- as.data.frame(t(mat))
      udf[40000,1]<1e+6}
    )
    sum(data)/length(data)
    res02_4[i,j]=sum(data)/length(data)
  }
}

However, sometimes of the the results is NA for whatever reason and so sum(data)/length(data) is also NA. In this case I want to redo entire replicate piece (until it is not NA). How I can organize this? First thing comes to mind is a combination of if() {} else {} but I think it'll be quite cumbersome construction - is there more elegant way?

Solution

You could pre-define data as NA, then wrap the replicate section in a while loop that runs one or more times until sum(data) is not NA

for(i in 1:30){
  for(j in 1:15) {
    u0<-c(1.5e+6+i*5e+4,j*2e+5)
    data <- NA
    while(is.na(sum(data))) {
      data <- replicate(n = 5,{
        prob <- de$SDEProblem(f,g,u0,tspan,p)
        fastprob<-diffeqr::jitoptimize_sde(de,prob)
        sol <- de$solve(fastprob,saveat=0.005)
        mat <- sapply(sol$u,identity)
        udf <- as.data.frame(t(mat))
        udf[40000,1]<1e+6}
      )
    }
    sum(data)/length(data)
    res02_4[i,j]=sum(data)/length(data)
  }
}

Obviously, I can't demonstrate that this works because you haven't included the de object in your question.