- Support copy construction, assignment, move construction, and move assignment on all analytics classes, enabling them to, among other things, be used in STL containers.
- Parameterize all algorithms by measurement type
- Change C++ library to be a header-only library
- Add documentation for all algorithms
- Mark quantile() method as const
- Add documentation to README.md
- Finish consolidation of streaming-percentiles-cpp and streaming-percentiles-js by adding support of publishing streaming-percentiles package to NPM
- Wrapped JS library into a UMD so that it works from Node, require.js, and script src=...
- Added browser-based unit tests usig nightwatch.js
- Changed node unit test framework from unit.js to tape
- Added CKMS algorithm with support for uniform, low-biased, high-biased, and targeted quantiles
- Build shared libraries on all platforms (including Windows)
- Changed ABI of all algorithms to use C++ pimpl paradigm so that the ABI should be stable from now on
- Create cross-compiled JavaScript version with emscripten (when available)
- Ensure Win32-generated packages are created as .ZIP not .MSI
- Separate out release and debug target directories when building on Linux / Mac OS X
- Build both static and shared libraries on most platforms using cpack
- Create a .zip / .tar.gz as the final step of the build process
- Fixed an off-by-one bug with Greenwald-Khanna which could cause a buffer underflow
- Make the internal state of Greenwald-Khanna private instead of public
Initial release. Supports the Greenwald-Khanna algorithm.