The inline package is an amazing, yet simple, package for R. It allows to dynamically (within the R session) define R functions and S4 methods with inlined C/C++/Fortran code.
Together with RUnit, inline powers the entire unit test suite of Rcpp.
As agreed with Oleg Sklyar, who maintains inline, we made a few additions to inline to accomodate the needs of the Rcpp family of packages.
The main addition is cxxfunction which is very similar to cfunction, except that it only focuses on C++ code using the .Call calling convention. cxxfunction uses a plugin system allowing other packages to control the code that is generated before compilation, environment variables, etc ... For example, the next version of Rcpp defines an inline plugin that takes care of all the details (find the Rcpp include path, link against the Rcpp user library, etc ...)
Here is an example, from the cxxfunction help page using the Rcpp plugin (this will only work with the next version of Rcpp, because the current version does not know about this)
Here is an example using the plugin from the next version of RcppArmadillo
Another addition to inline concerns the package.skeleton. We've made it S4 generic in inline and defined methods for the CFunc and CFuncList classes. In short, this allows to prototype some code using inline and quickly dump the code into a proper package
For example, here we make two functions using cxxfunction and then generate a package skeleton directly from them
Furthermore, the package.skeleton methods are aware of the plugin system, which allows plugin to have some control of additional steps involved in making the package skeleton, such as Makevars files, etc ...
getDynLib has been introduced in this version of inline to grab a reference to the dynamic library associated with a package, a function (CFunc object) generated by inline, or a set of functions (CFuncList object) generated by inline