The IPR, short for Internal Program Representation, is an open source project originally developed as the core data structures of a framework for semantics-based analysis and transformation of C++ programs. The foundations of the IPR were laid down and implemented between 2004 and 2005, as part of The Pivot research project. An overview, the general philosophy, and the design principles behind the IPR are presented in the paper "A Principled, Complete, and Efficient Representation of C++"

That paper is a useful source of general information about the IPR and non-obvious challenges in representing C++ programs in their most general forms.

The IPR library purposefully separates the interface (a collection of abstract classes found in <ipr/interface>) from the implementation (found in <ipr/impl>) for various reasons. An interface class (say ipr::Fundecl) can admit several implementations: a class for non-defining function declarations, and another class for
function definitions. Furthermore, compilers and tools (in general) can provide their own specific optimized implementations of the interface without impacting users of the IPR as long as those users restrict themselves to the public interface. Such a separation of concerns shields users of the library from implementation vagaries. The implementation in <ipr/impl> is provided for exposition and reference.

For more information, bug reports, and suggestions, please visit

https://github.com/GabrielDosReis/ipr

Gabriel Dos Reis, Bjarne Stroustrup.