Working with C/C++ in R

Duncan Temple Lang

Data Science Initiative, UC Davis

This workshop provides an overview and introduction to the ecosystem for interfacing R with compiled code, specifically C and C++. (These cover other compiled languages and also provide the lowest common denominator for all languages, including interpreted languages which can be embedded in R and vice versa)

We discuss

• the nature of compiled code,
• the "compilation" process and the importance of understanding the different steps and the make utility for managing dependencies and keeping files updated with respect to changes in others.
• calling a simple routine working on a primitive R vector
• using the .Call() interface for more complex and flexible situations, e.g. creating R objects, specifically lists in C code.
• writing an R package with compiled code
• interfacing to the 3rd party compiled libraries in C and C++
• machine-specific compilation flags using pkg-config at installation time
• configuration for machines with autoconf before compiling.
• external pointers in R for C/C++ objects (across R function calls)

Slides

Plus

• Why compiled code is faster than interpreted code
• Understanding what a compiler does and when.
• dissambler and machine code
• the make utiity

CompileSteps

• Steps in the "compilation" process
  • Pre-processor
  • compiler
  • linker
    • executable
    • shared library
  • loading

dotC

• .C() interface to simple C routine taking a numeric vector
• .Call() version

getrusage

• interface to existing system-level C routine
  • returns a structure
• .Call()
• R's C API

Rqpdf

• C & C++,
• .Call()
• pkg-config for compiler and library flags for 3rd party software,
• external pointers to C/C++ objects in R
  • finalizers and garbage collection