Revert DESCRIPTION change on Depends about rcpparmadillo HOT 8 CLOSED

Sure. That just means we nee to load Rcpp when we do unit tests. RcppArmadillo does not otherwise need to depends on Rcpp.

from rcpparmadillo.

Can we do unit tests for RcppArmadillo without Rcpp? I doubt it. Feel free to reopen if there is a better way.

from rcpparmadillo.

Done in af779d0

from rcpparmadillo.

The R code in RcppArmadillo does not call Rcpp functions, and so I think it is not right to have a Depends. We should move away from Depends as much as possible.

from rcpparmadillo.

Yes. But Imports: was just wrong.

I was just out with Anna and realized that Suggests: should be enough for unit testing, and does not conflict with LinkingTo. Will try that in a moment or two.

[ tried and didn't work. ]

You did not mention that you changed the package in the meantime. I think we get by with Suggests: alone in another package, but if the unit_test_setup() makes you happy all is well. But id does trigger this warning even under R-release:

* checking R code for possible problems ... NOTE
unit_test_setup : <anonymous>: no visible binding for global variable
   ‘pathRcppArmadilloTests’

from rcpparmadillo.

Fixed in 3a7e59f

Also, importFrom(Rcpp, foo, bar, biz) is preferable to a blanket import(Rcpp) which is not a lot better than Depends:. So if you know which functions you need, consider trying the former.

from rcpparmadillo.

Yes. I'm going to specifically import the artefacts we need (e.g. sourceCpp, Rcpp.plugin.maker, ...) instead of a big import.

from rcpparmadillo.

Good, that's much preferred.

from rcpparmadillo.