alashworth
commented
5 years ago 
Good point. This is very confusing. And it'll probably also crop up with x-- + x, which I'm guessing would get parsed as (x - (-+x)) == (x + x).
I see four options
- add ++x and --x to Stan's language with the usual semantics
- trap uses of ++x, --x, x++, x-- and print warnings
- trap uses and raise an error
- just get rid of (+x) as it's useless from within Stan itself (and frankly, also pretty useless within C++ itself, but we want to leave it there so all C++ programs work as expected).
Does anyone else have an opinion? And if we go with the first option, do we also add post increment,
- add x++ and x-- to the language
- Bob
On Mar 14, 2015, at 5:04 AM, Kevin S. Van Horn notifications@github.com wrote:
I had some Stan code that had the lines
idx <- 0; shparms[++idx] <- logit_lambda;
thinking that I was using the pre-increment operator on idx. Instead, it appears that "++idx" was interpreted as +(+idx), i.e., just idx. This is correct according to the published syntax reference, but unexpected. I suggest that "++" should be parsed as one token and, as long as pre-increment isn't supported, should give an error.
— Reply to this email directly or view it on GitHub.
I had some Stan code that had the lines
thinking that I was using the pre-increment operator on idx. Instead, it appears that "++idx" was interpreted as +(+idx), i.e., just idx. This is correct according to the published syntax reference, but unexpected. I suggest that "++" should be parsed as one token and, as long as pre-increment isn't supported, should give an error.