v2.0.0 is a (long overdue) major release with several backward-incompatible changes. There are several major features still to come in v2.1.0, which may again be backward incompatible. Though this is not ideal (ideally backwards-incompatible changes will be restricted to increase in the major version number), this has been driven by time constraints.
Known issues with this version, to be addressed by the next, are that both scripts (hmf and hmf-fit) are buggy and untested. Don’t use these until the next version unless you’re crazy.
- New methods on all frameworks to list all parameters, defaults and current values.
- New general structure for Frameworks and Components makes for simpler delineation and extensibility
- New growth_factor module which adds extensibility to the growth factor calculation
- New transfer_models module which separates the transfer models from the general framework
- New Component which can alter dn/dm in WDM via ad-hoc adjustment
- Added a Prior() abstract base class to the fitting routines
- Added a guess() method to fitting routines
- Added ll() method to Prior classes for future extendibility
- New fit from Ishiyama+2015, Manera+2010 and Pillepich+2009
- Removed nz and z2 from MassFunction. They will return in a later version but better.
- Improved structure for FittingFunction Component, with cutmask property defining valid mass range. NOTE: the default
MassFunction is no longer to mask values outside the valid range. In fact, the parameter
cut_fitno longer exists. One can achieve the effect by accessing a relevant array as dndm[MassFunction.hmf.cutmask]
- Renamed some parameters/quantities for more consistency (esp. M –> m)
- No longer dependent on cosmolopy, but rather uses Astropy (v1.0+)
- mean_dens now mean_density0, as per Astropy
- Added exception to catch when dndm has many NaN values in it.
- Many more tests
- Made the cosmo class pickleable by cutting out a method and using it as a function instead.
- Added normalise() to Transfer class.
- Updated fit.py extensively, and provided new example config files
- Send arbitrary kwargs to downhill solver
- New internal _utils module provides inheritable docstrings
- fixed problem with _gtm method returning nans.
- fixed simple bugs in BBKS and BondEfs transfer models.
- fixes in _cache module
- simple bug when updating sigma_8 fixed.
- Made the EnsembleSampler object pickleable by setting __getstate__
- Major bug fix for EH transfer function without BAO wiggles
- @parameter properties now return docstrings
February 2, 2015
- Better WDM models
- Added SharpK and SharpKEllipsoid filters and overhauled filter system.
- Separated WDM models from main class for extendibility
- Enhanced caching to deal with subclassing
- Minor bugfixes
January 28, 2015
- Added warning to docstring of _dlnsdlnm and n_eff for non-physical oscillations.
October 28, 2014
- Very much updated fitting routines, in class structure
- Made fitting_functions more flexible and model-like.
- Modified get_hmf to be more general
- Moved get_hmf and related functions to “functional.py”
September 16, 2014
- New HALOFIT option for original co-oefficients from Smith+03
- Better Singleton labelling in get_hmf
- Much cleaning of mass function integrations. New separate module for it.
- IMPORTANT: Removal of nltm routine altogether, as it is inconsistent.
- IMPORTANT: mltm now called rho_ltm, and mgtm called rho_gtm
- IMPORTANT: Definition of rho_ltm now assumes all mass is in halos.
- Behroozi-specific modifications moved to Behroozi class
- New property hmf which is the actual class for mf_fit
- Fixed bug in Behroozi fit which caused an infinite recursion
- Tests fixed for new cumulants.
September 8, 2014
- Better get_hmf function
- Fixed “transfer” property
- Updates fixed for transfer_fit
- Updates fixed for nu
- Fixed cache bug where unexecuted branches caused some properties to be misinterpreted
- Fixed bug in CAMB transfer options, where defaults would overwrite user-given values (introduced in 1.6.0)
- Fixed dependence on transfer_options
- Fixed typo in Tinker10 fit at z = 0
August 19, 2014
- New Tinker10 fit (Tinker renamed Tinker08, but Tinker still available)
- Completely re-worked caching module to be easier to code and faster.
- Better Cosmology class – more input combinations available.
- Fixed all tests.
May 08, 2014
- Introduced _cache module: Extracts all caching logic to a separate module which defines decorators – much simpler coding!
January 24, 2014
- Added get_hmf function to tools.py – easy iteration over models!
- Added hmf script which provides cmd-line access to most functionality.
- Added Behroozi alias to fits
- Changed kmax and k_per_logint back to have transfer__ prefix.
- Fixed a bug on updating delta_c
- Changed default kmax and k_per_logint values a little higher for accuracy.
January 23, 2014
- Added ability to change the default cosmology parameters
- Made updating Cosmology simpler.
- Fixed a bug in the Tinker function (log was meant to be log10): - thanks to Sebastian Bocquet for pointing this out!
- Fixed a bug in updating n and sigma_8 on their own (introduced in 1.4.0)
- Fixed a bug when using a file for the transfer function.
January 10, 2014
- Mocked imports of cosmolopy for setup
- Cleaner imports of cosmolopy
January 10, 2014
- Upgraded API once more: - Now Perturbations –> MassFunction
- Added transfer.py which handles all k-based quantities
- Upgraded docs significantly.
January 03, 2014
- A few more documentation updates (especially tools.py)
- Removed new_k_bounds function from tools.py
- Added w parameter to cosmolopy dictionary in cosmo.py
- Changed cosmography significantly to use cosmolopy in general
- Generally tidied up some of the update mechanisms.
- API CHANGE: cosmography.py no longer exists – I’ve chosen to utilise cosmolopy more heavily here.
- Added Travis CI usage
- Fixed a pretty bad bug where updating h/H0 would crash the program if only one of omegab/omegac was updated alongside it
- Fixed a compatibility issue with older versions of numpy in cumulative functions
December 5, 2013
- Addition of cosmo module, which deals with the cosmological parameters in a cleaner way
- Major documentation overhaul – most docstrings are now in Sphinx/numpydoc format
- Some tidying up of several functions.
October 29, 2013
- Better updating – checks if update value is actually different.
- Now performs a check to see if mass range is inside fit range.
- Fixed bug in mltm property
October 2, 2013
- The normalization of the power spectrum now saved as an attribute
September 19, 2013
- Updated “ST” fit to “SMT” fit to avoid confusion. “ST” is still available for now.
- Now uses trapezoid rule for integration as it is faster.
September 05, 2013
- Included an option to use delta_halo as compared to critical rather than mean density (thanks to A. Vikhlinin and anonymous referree)
- Couple of bugfixes for fitting_functions.py
- Fixed mass range of Tinker (thanks to J. Tinker and anonymous referee for this)
September 03, 2013
-Added a whole suite of tests against genmf that actually work
- Fixed bug in mgtm (thanks to J. Mirocha)
- Fixed an embarrassing error in Reed07 fitting function
- Fixed a bug in which dndlnm and its dependents (ngtm, etc..) were calculated wrong if dndlog10m was called first.
- Fixed error in which for some choices of M, the whole extension in ngtm would be NAN and give error
August 27, 2013
- Added ability to change resolution in CAMB from hmf interface (This requires a re-install of pycamb to the newest version on the fork)
July 02, 2013
- Ability to calculate fitting functions to whatever mass you want (BEWARE!!)
July 02, 2013
- Added Eisenstein-Hu fit to the transfer function
- Improved docstring for Perturbations class
- Corrections to Watson fitting function from latest update on arXiv (thanks to W. Watson)
- IMPORTANT: Fixed units for k and transfer function (Thanks to A. Knebe)
June 27, 2013
- Massive overhaul of structure: Now dependencies are tracked throughout the program, making updates even faster
June 19, 2013
- Fixed an error with an extra ln(10) in the mass function (quoted as dn/dlnM but actually outputting dn/dlog10M)
June 19, 2013
- Took out log10 from cumulative mass functions
- Better cumulative mass function logic
June 6, 2013
- Added Bhattacharya fitting function
- Fixed concatenation of list and dict issue
May 20, 2013
- Added better warnings for non-updated variables passed to update()
- Made default cosmology WMAP7
May 10, 2013
- Added warning for k*R limits
- Couple of minor bugfixes
- Important Angulo fitting function corrected (arXiv version had a typo).
April 09, 2013
- Added cascading variable changes for optimization
- Added the README
- Added update() function to simply change parameters using cascading approach
April 08, 2013
- First version in its own package
- Added pycamb integration
- Removed fitting function from being a class variable
- Removed overdensity form being a class variable
March 18, 2013
- Modified set_z() so it only does calculations necessary when z changes
- Made calculation of dlnsdlnM in init since it is same for all z
- Removed mean density redshift dependence