Hyperion

- Fixed compatibility with HDF5 1.10.x. [188]
- Simplify installation instructions, remove custom installer for Fortran dependencies, and remove CMake build system. [206]
- Fixed compatibility with latest version of h5py. [201]
- Fixed bug in method to truncate scattering matrix in dust properties. [199]

- Added a workaround for a bug in ifort v16 that prevented compilation. [190]
- Improved the accuracy of the Wood and Reynolds (1999) forced first scattering algorithm for very small optical depths, and implemented the Baes et al - 2016) composite biasing algorithm for forced first interaction. In addition, the `Model.set_forced_first_scattering` method has been renamed to `Model.set_forced_first_interaction`, and this method now optionally takes the algorithm to use (`wr99` or `baes16`) and in the case of the `baes16` algorithm, `baes16_xi` can be passed. [189]
- Improved the efficiency of monochromatic radiative transfer: instead of propagating photons until they are absorbed, photon packets are forced to scatter and their energy is lowered at every interaction by the albedo until the energy goes below a threshold. This threshold defaults to 1.e-10, but this can be changed using the `energy_threshold` argument to `Model.set_monochromatic`. [158]
- Remove bundled numpydoc and fix Python 3.6 compatibility. [192]

- Refactored installation instructions to provide a more streamlined experience. [161]
- Fixed a severe bug in the set-up of AMR grids that caused photons to be terminated before escaping from grids in specific cases. [167]
- Fixed bug in ordering of axes for Cartesian and AMR grids for yt 3.x. [168]
- Add installation instructions for FreeBSD (thanks to Aaron Lauterer). [173]
- Fixed a bug that caused a VoronoiGrid initialized from another VoronoiGrid to not include the sparse neighbor information. [177]
- Check for NaNs when writing out dust or model HDF5 file. [179]
- Fix the installation script for dependencies to work with GCC 5. [182]
- Remove deprecated frequencies= argument from set_monochromatic (instead, the wavelengths= argument should be used to give the wavelengths in microns). [183]
- The set_wavelength_index_range should now take zero-based indices, consistent with the Python/Numpy convention. Note that this will break compatibility with previous scripts, so you will need to be sure to update calls to this method if you were using it. [183]
- Update bundled decorator.py module. [184]
- Disabled computation of SEDs for inside observers, since the apertures are not truly circular. Users should instead set up images and then do photometry on them as appropriate. [185]
- The code base is now Python 3-compatible, and we no longer rely on 2to3. [185]
- Update installation instructions to mention conda as the preferred installation method. [186]
- Add workaround for a bug in HDF5 1.8 that caused issues with models that made use of filters. [187]

New features

- Added support for on-the-fly filter convolution. [119]
- Power-law and Ulrich envelopes can now be used on cylindrical polar grids. [136]
- Provide a way to sample random positions inside Voronoi cells. [134, 151]
- Added the ability to load AMR grids from yt datasets. [148]

Bug fixes

- Correctly raise an error if photons are emitted outside the Voronoi grid. [127]
- Avoid issues with number of photons when reading in models. [145]
- Fix a bug that caused sublimation parameters to not be correctly read in. [133]
- Fixed Fortran dependencies install script.

API changes

- `set_aperture_range` has been renamed to `set_aperture_radii` to emphasize that these are radii. [132]

Other

- Internal refactoring of how spectra are gridded for raytracing to make it easier to implement Doppler-shifted spectra in future. [126]
- Optimizations to memory and disk usage for Voronoi grids. [128, 154, 156]
- Improvements to warnings and documentation. [54, 98, 125]
- Removed deprecated mctherm2hyperion script. [144]

Bug fixes

- Fixed backward compatibility with files that don't include d_min/d_max

New features

- Added an importer function, `construct_octree`, to convert a list of SPH particles into an Octree grid. [67]
- Addded support for Voronoi grids. Voronoi grids can be set via the `~hyperion.model.Model.set_voronoi_grid` method, passing in the `x`, `y` and `z` coordinates of the sites as arguments. [92]
- Added the ability to make images that split up photons as a function of how many times they were scattered (using `set_track_origin('scatterings')`). [99]
- Added the ability to kill photons when scattered using `set_kill_on_scatter`, similarly to the existing `set_kill_on_absorb`. [101]
- Added the ability to use input files in `ModelOutput.get_quantities`. [106]
- Added the ability to control the interpretation of the specific energy passed to `add_density_grid` using the `set_specific_energy_type` argument. [117]
- Added the ability to use a grid from an HDF5 file on disk without reading it into the Python code, using `use_grid_from_file`. [116]
- Added support for cmake. [112]
- `Image` and `SED` objects now include attributes `d_min` and `d_max` that indicate the depth of the region used to contruct the image or SED. [121]
- Fixed a bug in the computation of the Rosseland mean opacity (it was in fact the Planck reciprocal mean opacity). Dust files have now been updated to version 2 to include both the Rosseland mean opacity and the Planck reciprocal mean opacity. [123]

Bug fixes

- Fixed a minor bug in the logic for killing photons that have had too many interactions. [100]
- Fixed a bug that meant that BipolarCavity instances could not be subtracted from AmbientMedium instances. [106]

Other improvements

- The `to_yt()` methods are now compatible with yt 3.x (3.0.1 and later recommended). [113]
- The `uncertainties=True` mode for `get_sed` and `get_image` has now been properly vectorized and should be faster by a factor of a few when requesting polarization results. [114]