Pycalphad

------------------

This is a minor bug fix release.

* Python 3.3 support has been dropped. See :issue:`46`.
* Documentation has been transitioned to a new domain, [https://pycalphad.org](https://pycalphad.org). See :issue:`47`.
* BLD: Exclude xarray 0.8 from dependencies since it has a regression. (Newer versions are fine.)
* DOC: Automated project documentation building and deployment via Travis CI.

----------------

This is a major release with bug fixes and performance improvements.

* The equilibrium solver core has been rewritten, resulting in a significant increase in robustness and accuracy,
particularly for chemical potential calculation with miscibility gaps. See :issue:`43`.
* For performance, dask-powered multiprocessing is now used to parallelize equilibrium calculations.
Because of this, dask and dill are now dependencies.
* Database and Model objects can now be pickled on all supported platforms, fixing a multiprocessing issue.

------------------

This is a minor release with bug fixes and performance improvements.

* Fix installation problem on Windows when using Anaconda.
* Add new compiled backend for phase models. This new backend provides a significant performance improvement.
* Experimental support for the numba library has been removed.

------------------

This is a minor bug fix release.

* ``tdb``: Fix TDB parsing errors on recent (>=2.1) versions of pyparsing.
* ``equilibrium``: Improve convergence and numerical stability of solver. Fix potential sign error in Hessian matrix.
Support mapping over two composition variables at once.
An error is now raised if a calculation specifies components not in the Database.

------------------

This is a minor bug fix release.

* ``Model``: Support the use of the absolute value function in the energy function.

------------------

This is a minor release with bug fixes and performance improvements.

* ``equilibrium``: Significant improvements to the speed and accuracy of the solver.
There is still some work to do for step and map calculations, planned for 0.4.
* ``Model``: Numerical accuracy improvement for the magnetic model :issue:`40`.
* ``Database``: Improvements to TDB writing, particularly for order-disorder models.
* ``Database``: Support for reading diffusion mobility databases.
Kinetic simulations are not on the roadmap, but this makes it easier to manipulate diffusion data.
Pull requests improving pycalphad's support for kinetic calculations are welcome.