In many ways Python and C++ represent the two ends in the spectrum of programming languages. C++ has an important role in the field of computing as the language design principles promote efficiency, reliability and backward compatibility – a vital tripod for any long-lived codebase. Python has prioritized better usability and safety while making some tradeoffs on efficiency and backward compatibility. That has led developers to believe that there is a binary choice between performance and usability.

Python has become the language of data science and machine learning in particular while C++ still is the language of voice for performance-critical software. The C++ and Python ecosystems are vast and achieving seamless interoperability between them is essential to avoid risky software rewrites.

In this talk we leverage our decade-old experience in writing automatic Python to C++ bindings. We demonstrate how we could connect the Python interpreter to the new in-tree C++ interpreter called Clang-Repl. We show how we can build a uniform execution environment between both languages using the new compiler-as-a-service (CaaS) API available in Clang. The execution environment enables advanced interoperability such as the ability for Python to instantiate C++ templates on demand, inherit from C++ classes or catch std::exception. We show how CaaS can be connected to external services such as Jupyter and execute code written in both languages.