- Fix bug for compatibility with numba 0.12
- Add sql formats
- Add hdf5 formats
- Add support for numpy ufunc operators
- Fix bug with compatibility for numba 0.12
- Split the datashape and blz modules out.
- Add catalog and server for blaze arrays.
- Add remote arrays.
- Add csv and json persistence formats.
- Add python3 support
- Add scidb interface
- Solidifies the execution subsystem around an IR based on the pykit project, as well as a ckernel abstraction at the ABI level.
- Supports ufuncs running on ragged array data.
- Cleans out previous low level data descriptor code, the data descriptor will have a higher level focus.
- Example out of core groupby operation using BLZ.
- Brings in dynd as a required dependency for in-memory data.
- Initial preview release