NIMBLE crashing when running compiled code

[Bernardo S. Scaldaferri]

Hello everyone, I'm new to NIMBLE and I've been trying to run a RJMCMC for a product partition model with a stick-break parameter. The code compiles but when I try to run the sampler it causes R to crash and I'm very lost as to why since there is no error message I can try to follow for debugging.

Any help is appreciated.

Here is my attempt at a reproducible example:

[Chris Paciorek]

Hi Bernardo,

We don't support all of R's operations in the run function of a nimbleFunction because of limitations on what we can compile from R to C++.

In particular you can't use minus in indexing (I'm doing a bit of work right now to make that more clear in our documentation and error-trapping).

Also, you can't use a vector as the second argument to `rep`. 

I haven't fully checked, but I think if you rework your `getClustersN` function to do the equivalent operations in a more 'manual' way, that will hopefully resolve things. Note that since the `run` code will be compiled to C++, you can use loops there without worrying that the loop will be slow.

Feel free to iterate here if that doesn't clear things up or getting things working for you.

-chris

--
You received this message because you are subscribed to the Google Groups "nimble-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nimble-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/nimble-users/63b14068-e3bd-4e7a-b55e-8a3872fe9583n%40googlegroups.com.

[Perry de Valpine]

A clarification: it should work fine to use a vector as the second argument to `rep`, similarly to how `rep` works in R.

To view this discussion visit https://groups.google.com/d/msgid/nimble-users/CA%2B7Ts_qnXxEy3N12wpVoz7B%3DT6vYCGJzs53D%2BMBm24gGgoDZFw%40mail.gmail.com.

[Chris Paciorek]

Right, thanks, Perry - I managed to confuse myself when trying something out and botched that part of the response!

[Bernardo S. Scaldaferri]

Hello Chris,

I implemented the function as a for loop and it worked perfectly:

getclustersN <- nimbleFunction(
  run = function(U = double(1)) {
    A <- length(U)+1
    K <- A - sum(U)
    V <- c(K)
    c=K
    for(i in 1:(A-1)){
      if(U[i]==0){
        c <- c-1
      }
      V <- c(V,c)
    }
    return(V)
    returnType(double(1))
  })

Thanks a lot for the help!

-Bernardo S. S.