Pyspeckit

~~~~~~~~~~~~~~~~~~~~~~~~~~~
* none

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
* Very minor fix - just change pyspeckit.__doc__

~~~~~~~~~~~~~~~~~~~~~~~~~~~~
* Virtually nothing, just compatibility stuff

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
* Compatibility 287: fix for mpl3
* Minor 283: fix incorrect use of _traceback variable in fiteach
* Enhancement 282: Allow instantiation from things with .hdus
* Enhancement 267: Improved NH3 fit defaults
* Bugfix 266: Fix an ammonia normalization issue
* Enhancement 257: Implement multi-component guessing for non-gaussian models
* Enhancement 256: New NH3 model
* Bugfix 249: Fix includemask initialization
* Bugfix 248: Prevent zero-width interactive guesses
* Bugfix 245: Corrections to powerlaw baseline fitting
* Bugfix 244: Fix typo in array arithmetic
* Bugfix 233: Fix problem retrieving multicomponent NH3 models

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

* Minor 210, 212: add cold_ammonia to the registry
* Bugfix 175: Fixed an error when indexing dxarr
* Minor 178: error message typo fixed
* Bugfix 159: Cube fitting was incorrectly ignoring some pixels. Fixed.
* Bugfix 176: Force fits to operate on float64s because float32s were giving
gnorm=0 errors.
* Minor 180: In cube fitting, default to Integral=False
* Bugfix 135: Clear fits from plots each time a new one is made to avoid a
memory leak
* Bugfix 183: Fix the usage of the reference value in doppler convention
equivalency creation in line_ids
* Bugfix 185: Prevent divide-by-zero error in renormalization
* Minor 188,189,197,200: Some py3 compatibility updates (especially for
CLASS)
* Minor 186,191,192,195: fix links, badges
* Bugfix 194: fix an infinite loop bug in cube init
* Bugfix 182: fix cube copying
* Enhancement 196: updated cube tests
* Bugfix 202: Fix filling fraction definition
* Docs 203: Add contributors section
* Bugfix 205: Make a figure if no figure exists
* Bugfix 207: clobber->overwrite to handle upstream astropy changes
* Bugfix 206: Fix units in ``hydrogen.find_lines``
* Bugfix 209: have to index with integers in numpy >= 1.12
* Bugfix 211,214: Fix rescaling for multi-component fits
* Bugfix 215: some text restructuring and fixes for default font sizes in
matplotlib 2.0. An improved error message when units are pixels.
Some bug fixes related to 'reset' in fitting. Documentation for hill5
* Bugfix 216: len(pars) division for python 3 fixed in ammonia code
* Bugfix 218,219: fix return parsing of parallelized fits
* Enhancement 217: Parallelize get_modelcube
* Bugfix 223: fix lmfit parameter issue

~~~~~~~~~~~~~~~~~~~~~~~~~~~

* Enhancement 170: Added a ``cold_ammonia`` model, which uses a specific
set of approximations to fit the rotational temperature and retrieve the
kinetic temperature
* Bugfix 166: Spectrum objects should, and now do, inherit the registry
from their parents. This is critical for fitting non-default models to
cubes
* Bugfix 163: Allow parameter cubes that have NaNs to be loaded
* Enhancement 161: Cleaned up moment-specified guesses for cubes
* Enhancement 160: Added multiply ionized RRLs to the RRL models
* Enhancement 156: NaN pixels are treated as though they were masked out
(previously, numpy masked arrays had to be used for this functionality)
* Bugfix 154: pyspeckit script upgraded to py3 compatible
* Bugfix 152: SpectroscopicAxis comparisons did not properly yield
booleans due to an inheritance error.
* Bugfix 147: Spectrum arithmetic was being allowed between spectrum
objects with different X-axes. This has been disallowed, and a test has
been added.
* Enhancement 147: ``from_spectrum1d`` implemented to instantiate a
``pyspeckit.Spectrum`` from a ``specutils.Spectrum1D`` object
* Bugfix 128: fiteach would try to measure an integral even if the
spectral fit failed
* MAJOR bugfix 136: There was an error in the equation used to create the
Ammonia model that would result in fits reporting incorrect values of the
column density. The Ammonia model has been substantially refactored,
removing the ``thin`` option in favor of putting the thin approximation
into its own function.