Presently, the RealTimeWeb architecture (the technology powering High Velocity data and distributed High Volume) supports generation of Racket, Python, and Java code. For better or worse, a number of institutions still use C++ for introductory courses, and several professors have personally requested C++ support. Unfortunately, supporting this language is not as clear cut as the others, because there are fewer language-level features to simplify web access and parsing.
One of the big goals of the CORGIS project is to make it as trivial as possible to use Big Data. The Racket and Python libraries are each a single file, which require no external modules or local installation - just download the .py or .rkt file and get started. The Java libraries are packaged into a single JAR file, which can then be added to the students class path (not as trivial as we'd like, but that's a different conversation). There is no such solution for C++: you are expected to either distribute a large code base, or compile for the target platform (which then raises the terrors of platform dependence).
The options as I see them:
1) Fail to support C++
Obviously, this option is least viable, but currently where we are.
2) Support C++ through a multi-file library
By leveraging the right libraries (e.g., libcurl, libjson), it's reasonable to create the necessary C++ code. However, by utilizing those libraries we make it considerably more difficult for students to get started with our libraries - they suddenly have to worry about unzipping files. Still, that may simply be a penalty for using C++.
3) Write a script to concatenate the entire library into a single C++ file
Although monstrous from a Software Engineering perspective, this approach is already used in Python and Racket libraries requiring OAuth. This is a case where proper design must take a back-seat to the user experience, but it is not something that I take lightly.
I am very interested in hearing community feedback on these options; I am even more interested in better options.