data named “i” is overwritten by rstan

[jgabry]

An anonymous user on discourse reported an error that i helped them work around, but is a bug in RStan:

https://discourse.mc-stan.org/t/mismatch-in-dimensionality-stan-is-fixing-the-dimensionality-to-a-number-why/9158

It seems that an integer (probably other types too) named i (lower case) in the list passed to data has its value overwritten when RStan processes/checks the data. The temporary workaround is for the user to just change the name from i to something else.

[jgabry]

i’ll see if I can sort this out quickly but I may have questions since I haven’t worked much with that part of the RStan code.

[jgabry]

I don't know why it's happening but it's happening when sampling() calls parse_data():

rstan/rstan/rstan/R/stanmodel-class.R

Line 118 in ef3dace

parsed_data <- with(data, parse_data(get_cppcode(object)))

and specifically when parse_data() uses dynGet():

rstan/rstan/rstan/R/misc.R

Lines 1607 to 1608 in ef3dace

	stuff <- sapply(objects, simplify = FALSE, FUN = dynGet,
	inherits = FALSE, ifnotfound = NULL)

Here's a minimal working example to trigger it.

// test.stan
data {
  int i;
  vector[i] y;
}
parameters {
 real mu;
}
model {
  y ~ normal(mu, 1);
}

# In R
standata <- list(i = 25, y = rnorm(25))
fit <- stan("test.stan", data = standata, iter = 1, chains = 1)

I'll make a PR shortly that I think will fix this by giving parse_data() a data argument instead of using with(data, ...) and then using mget() instead of dynGet.