include the package.h file before Rcpp.h

[romainfrancois]

compileAttributes puts Rcpp.h before the package header file in RcppExports.cpp.

When I modernized RcppArmadillo to use attributes, I had this:

In file included from RcppExports.cpp:5:
./../inst/include/RcppArmadillo.h:27:6: error: "The file 'Rcpp.h' should not be included. Please correct to include only 'RcppArmadillo.h'."
    #error "The file 'Rcpp.h' should not be included. Please correct to include only 'RcppArmadillo.h'."
     ^

So for now, I have manually edited the [RcppExports.cpp](https://github.com/RcppCore/RcppArmadillo/blob/master/src/RcppExports.cpp#L4] file to flip the order.

@jjallaire I was wondering if we could include the package header file first. Chances are people include Rcpp.h in there anyway.

[jjallaire]

I think you could probably do this without breaking anyone but it's certainly worth checking. Is the error condition in RcppArmadillo strictly required though? (it seems like allowing duplicate inclusion should be allowed, as there are many other ways this could happen e.g. using multiple Rcpp derivative packages).

[romainfrancois]

Having the client package file included first allows te client package to add wrap and as overloads, etc ... So we can use this for Rcpparmadillo, rcppgsl, ... And any package that wants to customize wrap.

Envoyé de mon iPhone

Le 24 nov. 2013 à 15:37, JJ Allaire notifications@github.com a écrit :

I think you could probably do this without breaking anyone but it's certainly worth checking. Is the error condition in RcppArmadillo strictly required though? (it seems like allowing duplicate inclusion should be allowed, as there are many other ways this could happen e.g. using multiple Rcpp derivative packages).

—
Reply to this email directly or view it on GitHub.

[eddelbuettel]

@jjallaire Yes, it is overly strict, and yes it has saved us a few times :)

This came in direct response when people included multiple headers in random order. If you can think of a better way we'd be all ears. So far it works well enough to be worth the pain (and yes, bit me once today too after I ran compileAttributes().