In C++ we often have containers that are not compatible with R or Rcpp
with data already in them (std::vector, std::set, etc.). One would like
to be able to operate on these containers without having to copy them into
Rcpp structures like IntegerVector. RcppHoney aims to address this
problem by providing operators and functions with R semantics that can be
used on any iterator-based container.
Introduction
RcppHoney allows any iterator-based container to be “hooked” in. Once a
container type is hooked to RcppHoney, it is granted operators (+, -, *, /, etc.)
and a host of other mathematical functions that can be run on it. It also
becomes interoperable with any other hooked data structure. This lets us
write expressions that look like std::vector + Rcpp::IntegerVector +
log(Rcpp::NumericVector) and get the expected results.
Implementation
RcppHoney has several structures that are hooked in by default. Currently
they are std::vector, std::set, and Rcpp::VectorBase. The ability to
hook in custom structures is also provided.
All operators and functions are implemented as
expression templates
to minimize memory usage and enhance performance. The goal here is to only
copy the data into an R compatible structure when we must (i.e. when we
return it to R). This is achieved through the use of the RcppHoney::operand
class. RcppHoney::operand provides an iterable interface to the result
types of operators and functions.
RcppHoney currently provides all the basic mathematical operators (+, -, *, /)
as well as some common
functions
(abs, sin, cos, exp, etc.). Eventually all of the functionality provided by
Rcpp::sugar as well as anything else we can think of will be supported.
Enough about the abstract though…let’s see it in action.
Example
The following example shows how to hook in a custom data structure
(in this case std::list) as well as the types of expressions that can be
created once a data structure is hooked in.
Conclusion
RcppHoney is a powerful tool for allowing different container types to interoperate
under Rcpp. It can save development time as well as help the user generate faster
and more readable code.
RcppHoney is available via CRAN though as it is still
in an alpha state and changing rapidly, it is recommended that you install it from
source. Source code is available at
github.com/dcdillon/RcppHoney.