Pycalphad

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

This is a minor bug fix release.

* FIX: Improved ``model_hints`` construction when reading databases with out of order type definitions, fixes detecting disordered phases with ``filter_phases`` ( :issue:`269` )
* FIX: Complex infinity in ``Model`` expressions are converted to real infinity so SymEngine can ``lambdify`` the expressions ( :issue:`267` )

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

This is a minor release with bug fixes and performance improvements. Python 2.7 support is dropped as well as Python 3.5 and below. Python 3.6-3.8 are explictly supported.

* ENH: Reading large databases via delayed parameter processing ( :issue:`266` )
* FIX: Support PhaseRecord pickling, switch SymEngine backend to LLVM ( :issue:`264` )
* DOC: Regenerate examples ( :issue:`263` )
* DOC: Update examples ( :issue:`262` )
* ENH: variables.MassFraction object implementation ( :issue:`254` )
* MAINT: Update and pin to SymPy 1.5 ( :issue:`251` )
* MAINT: Support Python 3.8, drop Python 2, <3.6 ( :issue:`257` )

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

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

* ENH: Calculation speed and accuracy improvements via exact Hessians and the SymEngine lambda backend ( :issue:`249` )
* ENH: Faster binary phase diagram mapping ( :issue:`209` )
* FIX: Calculating disordered phase only if respective ordered phase inactive. Thanks igorjrd ( :issue:`248` )
* ENH: Use better colors in phase_legend(). Thanks igorjrd ( :issue: `242` )
* FIX: Suspend a phase if only a pure-vacancy endmember would be active. Thanks igorjrd ( :issue:`239` )
* ENH: Add element reference data reading/writing to TDB parser ( :issue:`240` )
* DOC: Typo in documentation. Thanks jwsiegel2150 ( :issue:`237` )
* FIX: SymPy namespace clash with TDBs, and other deprecation fixes ( :issue:`234` )
* DOC: Update installation instructions ( :issue:`241` )
* MNT: Relax dask requirements to the minimum required for `scheduler=` syntax ( :issue:`223` )

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

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

* ENH: Major performance improvement with new Just-In-Time SymEngine/LLVM-based compiler. ( :issue:`220` )
* ENH: Support for fixing the chemical potential of an element as an equilibrium constraint. ( :issue:`200` )
* ENH: Support for shifting the reference state of an equilibrium calculation. ( :issue:`205` )
* MAINT: Internal reorganization of the phase model constructors. ( :issue:`214` :issue:`217` )
* DOC: A new example for computing properties of custom models has been added.
* MAINT: Windows Python 2.7 support has been dropped. ( :issue:`220` )

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

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

* FIX: PhaseRecord: Fix pickling, so distributed scheduling will work ( :issue:`196` )
* FIX: Max phases by Gibbs phase rule accommodated ( :issue:`184` )
* FIX: SymPy 1.2 compatibility ( :issue:`180` )
* FIX: Model: Degree of ordering property calculation when vacancy is in the system
* FIX: Species Python 2 unicode support ( :issue:`166` )
* ENH: Allow solution refinement by the Ipopt solver to be disabled. ( :issue:`187` )
* ENH: Enable custom solvers ( :issue:`177` )
* DOC: Update pycalphad logo to be smoother and have a version with text. Thanks to Joyce Yong. ( :issue:`193` )
* MNT: Refactor callables creation in equilibrium() and calculate() ( :issue:`192` )
* ENH: tdb: Move tdb grammar creation out of loop
* ENH: Add magnetic moment as default Model property BMAG
* ENH: Optimize _compute_phase_values ( :issue:`175` )

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

This is a major release with new features and performance improvements.

* ENH: Add support for calculations with species, including support for the associate, ionic liquid, and gas phase models ( :issue:`161` ).
* The compiled backed of common models has been removed. Users should expect that the first set of calculations with new phases in a Python script or session be slower as the models for each phase are compiled in real time.
* ENH: Performance of JIT compilation of phases has been improved.
* ENH: equilibrium: Performance optimizations to reduce the overhead of calling equilibrium, particularly in tight loops.