promote int[] to real[] as argument for every function

[alashworth]

Issue by bob-carpenter Wednesday Aug 06, 2014 at 20:51 GMT Originally opened as https://github.com/stan-dev/stan/issues/830

---

Make sure that all functions and distributions specifying real[] arguments also accept int[] arguments.

This should only be done where there is not a specialization for integers. mean(), for example, only has the signature (rea[]):real, and the promotion would allow signature (int[]):real --- the result is the same type because means of integers can be real. Similarly for the Dirichlet density dirichlet_log of signature (real[], real[]): real, which would accept int[] arguments (though only the size-1 int[] argument is an appropriate simplex for the first argument).

• [ ] Create a list of functions by scanning the manual index (I'm looking at stan-functions-2.13.0.txt)
  • [ ] cov_exp_quad;(real[] x, real sigma, real l);matrix;438
  • [ ] cov_exp_quad;(real[] x1, real[] x2 real sigma, real l);matrix;438
  • [ ] cumulative_sum;(real[] x);real[];437
  • [ ] integrate_ode;(function ode, real[] initial_state, real initial_time, real[] times, real[] theta, real[] x_r, int[] x_i);real[\nobreakspace ,\nobreakspace ];450
  • [ ] integrate_ode_bdf;(function ode, real[] initial_state, real initial_time, real[] times, real[] theta, real[] x_r, int[] x_i);real[];450
  • [ ] integrate_ode_bdf;(function ode, real[] initial_state, real initial_time, real[] times, real[] theta, real[] x_r, int[] x_i, real rel_tol, real abs_tol, int max_num_steps);real[];450
  • [ ] integrate_ode_rk45;(function ode, real[] initial_state, real initial_time, real[] times, real[] theta, real[] x_r, int[] x_i);real[\nobreakspace ,\nobreakspace ];450
  • [ ] integrate_ode_rk45;(function ode, real[] initial_state, real initial_time, real[] times, real[] theta, real[] x_r, int[] x_i, real rel_tol, real abs_tol, int max_num_steps);real[\nobreakspace ,\nobreakspace ];450
• [ ] doc each new signature
• [ ] test each new signature on its own and with an example model instantiating it (mostly in transformed data other than for rngs)

[alashworth]

Comment by syclik Wednesday Nov 30, 2016 at 15:46 GMT

---

I just updated the issue. We don't have too many functions to tackle.