Add rng parameter to Rcpp::export to prevent inclusion of RNGScope in generated code

[jjallaire]

No description provided.

[eddelbuettel]

I like it ! What was the motivation today? Leakage? Saving a cycle here or there? Drew's post?

One suggestion: a positive term ("rng") for a negative action ("do not include") is weird. Would it make sense to call it norng instead of rng?

[kevinushey]

It seems like we default to rng = TRUE but someone could write:

// [[Rcpp::export(rng = false)]]

if so desired. Is that correct @jjallaire?

[jjallaire]

I knew this had been asked for a few times and it's hung over my head that we didn't support it. I figured as long as I am mucking around in the attributes code I might as well just do it :-)

Kevin is correct, the default is rng = true so the only meaningful value is rng = false (which doesn't have the inversion of meaning problem you described)

[jjallaire]

Do you think I should also support TRUE and FALSE (I currently only support true and false lowercase)

[kevinushey]

I would say yes (I almost reflexively wrote FALSE just because I was thinking in R mode; also R provides TRUE and FALSE defines anyway so users might be accustomed to using that already.)

[jjallaire]

Okay, I'll update to support TRUE and FALSE as well.