expose_stan_functions causes error: "$ operator is invalid for atomic vectors"

I'm working with experimental version 2.26 of RStan on Windows and I'm getting an error, "$ operator is invalid for atomic vectors", from expose_stan_functions.

I researched this a bit and did not find anything the same. Issue #899 mentions expose_stan_functions in passing but that appeared to be about other errors.

I get the error from an existing .stan file for which expose_stan_functions does not cause an error with RStan 2.26 on macOS. Instead of including the .stan file I'm working with, here is a shorter example which I nicked from another discussion. See: https://discourse.mc-stan.org/t/mixture-models/17721/9

functions {
  real repulsive_lpdf(vector mu, vector rho) {
    int K = num_elements(mu);
    matrix[K, K] S = diag_matrix(rep_vector(1, K));
    matrix[K, K] L;
    int c = 0;

    for (k1 in 1:(K - 1))
      for (k2 in (k1 + 1):K){
        c += 1;
        S[k1, k2] = log(1 - exp(- squared_distance(mu[k1], mu[k2]) / rho[c]));
        S[k2, k1] = S[k1, k2];
      }
    L = cholesky_decompose(S);

    return 2 * sum(log(diagonal(L)));
  }
}

data {
  int<lower=1> K;
  int<lower=1> N;
  real y[N];
}

parameters {
  ordered[K] mu;
  vector<lower=0>[choose(K, 2)] rho;
  real<lower=0> sigma[K];
  simplex[K] lambda;
}

model {
  // Prior model
  mu ~ normal(0, 5);
  sigma ~ std_normal();
  lambda ~ dirichlet(rep_vector(3, K));
  rho ~ gamma(1, 2);
  mu ~ repulsive(rho);

  // Observational model
  for (n in 1:N) {
    real comp_lpdf[K];
    for (k in 1:K) {
      comp_lpdf[k] = log(lambda[k]) + normal_lpdf(y[n] | mu[k], sigma[k]);
    }
    target += log_sum_exp(comp_lpdf);
  }
}

I put that into a file name example.stan and then I get, in my R session:

> expose_stan_functions ("example.stan")
Error: $ operator is invalid for atomic vectors

packageVersion ("RStan") reports 2.26.2.

R.version.string reports "R version 4.1.0 (2021-05-18)".

Operating system is MS Windows 10.

I installed RStan following these instructions. (I was trying the officially released version 2.21, however, it lacks the %/% operator which I want to use.) See: https://discourse.mc-stan.org/t/problems-installing-rstan-2-21-2-with-rtools-4/22492/5

remove.packages(c("StanHeaders", "rstan"))
install.packages("StanHeaders", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))
install.packages("rstan", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))

Start by reproducing expose_stan_functions("example.stan") with the supplied example.stan, RStan 2.26.2, R 4.1.0, and Windows 10 details. Trace the source of the "$ operator is invalid for atomic vectors" error and verify that the same call completes successfully without the error on the reported environment.