NEWS

Notes for 0.926 release:

The gen_test_number class has been rewritten and improved to allow for
more flexbility.

Improvements to polylogs and associated testing. 

Started process of using C++ std random number generators instead of
GSL.

Improved the verbose output in root_brent_gsl, the behavior of
tensor::rearrange(), the polynomial classes (especially for larger
floating-point types).

Improved OpenMP integration.

Created functions to compute KL divergence and a new set of classes,
matrix_invert_det to handle the inversion and determinants with Eigen
or Armadillo.

Updated to more recent versions of AME and FRDM.

Created a new vector_autocorr_fftw function for autocorrelations.

Created dockerfiles in a docker directory for common configurations.

The fnmatch() function is being replaced by regex, enabled if
O2SCL_REGEX is defined. This is not yet enabled by default, because it
has caused some problems with o2graph.

A new version of the calculator class which partially supports UTF8 is
in implemented in calc_utf8. This new UTF-8 compatible class is used
in O2scl unless O2SCL_NO_CALC_UTF8 is defined. There are still some
complications with regex and UTF8 which haven't fully been resolved.

The convert_units class has been reworked to enable UTF8 support.

Created a new set of HDF5 input functions, hdf_input_n(), to allow
input with or without a name.

More work on the python interface everywhere and the snapcraft
configuration.

Improved citations for nuclear masses, handling of electrons and
photons in eos_sn, the fermion_rel nondegenerate expansion, the
fermion_rel pair_density() function, the nstar_cold class (including a
small change to the default crust), the fermion_mag_zerot class, the
tov_love class, the eos_tov_buchdahl class, the command ser-hist-t3d
in acol, the function table3d::create_hist_set().

Updated nucmass_ldrop to avoid unphysical central densities.

Created experimental exp_max, nstar_hot, eos_quark_njl_vec, and
eos_had_virial_spin. Progress on the experimental interp_krige
classes.

Improved documentation everywhere.

Cleaned the "acol -commands" and "acol -help" output.

------------------------------------------------------------------------
Notes for 0.925 release:

Fixed interpolation bug and log grid bug in
tensor_grid::rearrange_and_copy().

Improved acol docs everywhere. Created acol table3d to-tensor-grid,
acol table3d to-hist-2d, acol table3d get-grid, acol create table-mv,
acol table3d slice-hist, acol table3d select, acol h5-copy, acol table
correl, acol constant, acol binary, acol docs, and acol wdocs. The
acol help documentation has also been reformatted and improved.

The integration routines for semi-infinite intervals have been
refactored to use \ref integ_iu(), \ref integ_il() and \ref integ_i().

Improved error handling and documentation in cloud_file.

Several updates to eos_sn, nstar_rot, eos_had_rmf, eos_had_rmf_hyp_ts,
and ex_eos_gibbs. Created new beta-equilibrium functions in some
EOS classes. Created new function nucmass_fit::fit_covar() to
fit nuclear masses with a covariance matrix.

Reworked documentation in sphinx/breathe.

Fixed spacing bug in rewrap_keep_endlines().

Updated strings_spec(), vector_spec(), mult_vector_spec() and
value_spec().

Added OpenMP support to some table functions. Updated table::swap().

Created experimental new convert_units::convert_calc() function
and created find_constants class. Updated solar system constants.

Improvements to the vec_index class.

Updated the vector_acor() function. Created vectors_equal() and
vectors_equal_tol().

Fixed hidden virtual function warnings.

Created columnify::add_spaces().

Worked on multiprecision support for several particle classes.
Cleaned up polylogs and created bessel_K_exp class.

Fixed HDF5 1.12-related deprecation warnings.

Renamed integration classes for infinite or semi-infinite limits.

Created a new experimental auto_format class.

Created new vt100 string functions and moved them into a new
terminal class.

Added quadratic and cubic discriminants.

Added experimental part_pdg. 

------------------------------------------------------------------------
Notes for 0.924 release:

Updated second derivatives of the pressure for the particle clases and
updated the eos_had_skyrme class to compute second derivatives.

Created a new system for OS detection to make compilation a bit
more user-friendly. HDF5 headers have been reworked a bit
accordingly. 

Improved integration between acol and o2sclpy. 

Improved the handling of aliases in acol and cli.

Generalized convert_units, several of the integration classes, a few
derivative and solver classes, and several of the particle classes to
templates to allow for other floating point types.

Improvements to the ex_mmin example and the example plots. Created
new example ex_fermion_summ.

Improvements to the ode_iv_solve_grid class.

Improved error handling for calculator class.

Created functions vector_refine(), file_exists(), and o2isfinite().

Created classes inte_tanh_sinh_boost, inte_sinh_sinh_boost,
inte_kronrod_boost, inte_exp_tanh_boost, nucmass_gen, and frib_mex.

Created a couple_threads option and an ac_coeffs() function for
mcmc_para.

Creating strings_spec() for string list specifications and updated
vector_spec() and mult_vector_spec().

Improved tabulate option for tov_love::calc_y().

Updated constants for CODATA 2018 release.

------------------------------------------------------------------------
Notes for 0.923 release:

A new function "rearrange_and_copy()" was added to the tensor
and tensor_grid classes for several kinds of tensor rearrangements.

A new matrix_view_transpose class and a new wordexp() function were
created.

The classes cloud_file, boson_rel, eos_had_skyrme, eos_tov_buchdahl,
fermion_deriv_rel, fermion_eff, fermion_nonrel, fermion_rel,
eos_had_rmf, hist, interpm_krige, and interp_krige_optim classes were
improved improved. The particle calibration routines were improved.

The constants were updated for the CODATA 2014 release.

Several updates were made to acol. The handling of tensor and
tensor_grid classes was improved. Vector and value specifications were
added. The help text was reworked. The autocorr command was improved.

A new class slack_messenger was created to send messages to Slack.

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

Notes for 0.922 release:

In the table class, the set() function no longer automatically adds
rows or columns. This prevents some behavior which was not intuitive for
users. 

Several tensor functions have been added, including
tensor::swap_data(), tensor_out(), tensor::min(), tensor::max(),
tensor::copy(), tensor::is_valid(). Slicing for tensor_grid objects
has been improved. Copy constructors for tensor and tensor_grid have
been created.

A few long double integration classes have been added when higher
accuracy is required.

The MCMC classes have been completely reworked, offering better
support for OpenMP and MPI parallelism.

The acol command has been expanded and improved in several ways.
Code for o2graph has been moved out and the documentation has
been improved.

The conversion functions in convert_units now has a const version

The documentation has been improved everywhere.

A new nearest neighbor interpolation type has been added.

A seed bug in rng_gsl has been fixed.

HDF5 I/0 for table3d types now allows const references.

The cloud_file class is updated and simplified.

The prob_dens_func_amr class has been updated.

Some work has been done on heat capacity functions for particle
objects.

The fermion_rel::calc_mu() and pair_mu() functions work better
now with bad initial guesses.

The tov_solve, eos_tov, eos_cs2_const, and eos_had_rmf_hyp classes
are improved.

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

Notes for 0.921 release:

New vector_autocorr() functions were created to analyze
autocorrelations.

The class serach_vec now has improved thread safety.

Kriging classes interpm_krige_nn, interpm_krige, and interp_krige were
created. These are still a bit experimental.

A new Robbins-Monro iteration root-finding class has been created. A
class smooth_func for smoothing multivariate functions by averaging in
a neighborhood has been created. These should help dealing with noisy
functions.

All 'funct11' classes have been renamed to 'funct'.

New vector bin size functions have been created. 

The 'acol' command has been reworked to handle HDF5 I/O for more
types. It can now manipulate vectors, matrices, and tensors
in several new ways. The interaction with o2graph in o2sclpy
has also been improved.

The Markov chain Monte Carlo classes have been updated, but
are still experimental. Similarly with the probability
density distribution classes where new bivariate gaussian
and AMR-like probability densities were created. 

The internal bibliography mechanics have been improved.

Error handling in nucleus_rmf has been improved.

Updates to eos_tov_vectors, eos_had_rmf, eos_tov_vectors,
eos_had_skyrme.

Created an eos_cs2_poly class and anneal_para_ts. 

Nuclear masses were updated for the 2016 AME data set.

The calculation of dndT, dndmu, and dsdT for relativistic
fermions has been improved. 

The SFHo and SFHx EOSs were updated.

Several internal changes have been made to allow Ubuntu-style
packages. The 'make check' target now works without 'make install'. A
snapcraft snap has also been created. These are still in progress.

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

Notes for 0.920 release:

A new set of Markov chain Monte Carlo classes has been created.

Experimental conditional and multidimensional probability
distributions haven't been added.

A new experimental class cloud_file has been created which
assists in using and downloading files from URLs.

The interpm_idw class has been improved, including a new function
which computes partial derivatives and uncertainties.

A new helper class vector_range has been created for vector
objects which don't already have associated range types. 

The command-line option 'acol -filelist' has been updated to be more
informative. New commands 'cat' and 'sum' perform concatenation and
addition of table objects.

The class root_cern now throws if the root is infinite.

Several classes have clear() methods, and the nomenclature has
been made more consistent.

A new function tov_solve::fixed_pr() computes a configuration with
a fixed central pressure. The surface gravity units have been fixed.

Error handling in nstar_cold has been improved.

The cubature classes have been improved. 

Much more work has been done on python integration with o2sclpy,
and this is still experimental. The o2sclpy package now has a
binary o2graph which supports plotting of O2scl data files with
matplotlib.

Several documentation improvements everywhere.

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

Notes for 0.919 release:

The old "FunctionParser" class has been replaced by a much nicer
"calculator" class adapted from Brandon Amos' code.

Travis testing has been implemented, and works for Linux and Mac OS X.

The 'table' class copy constructor was fixed. 

A new tov_love class has been created to compute tidal deformability.
The class 'nstar_rot' has been improved significantly.

Some upgrades have been implemented to work well with GSL versions 2.0
and later. A new flag to ./configure, --enable-gsl2 , enables
extensions built on new GSL code.

Improvements have been made to ode_it_solve and associated
testing.

Exception safety has been improved in several classes.

The 'contour' class internals have been reworked to make
the code easier to understand. 

A new acol command gen3-list was created. The auxillary python
code o2py.py has been updated.

Documentation has been improved everywhere.

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

Notes for 0.918 release:

The GNU-style target 'make check' now works, and is equivalent to
'make o2scl-test'

The ./configure script now properly detects the flags for c++11. 

O2scl can now be installed using homebrew, using 'brew tap
awsteiner/science'. 

O2scl is now reimplementing Cubature from Steven G. Johnson for
multi-dimensional integration. The current implementation just uses
simple wrappers to allow the use of std::function objects. Later work
will more fully implement template methods in the Cubature code.

Fixes for step size and normalization in anneal_gsl.

Added tensor clear(), min() and max() methods.

Fixed several warnings from gcc-5 and clang.

Created a new vec_index class similar to a python dictionary. 

Improved solving for eos_had_base::calc_temp_p().

Created RMF EOS example ex_eos_had_rmf.

Improvements and better documentation for eos_nse.

Added units and updated I/O in eos_sn.

Updates and new testing for dense_matter and nucmass_densmat. 

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

Notes for 0.917 release:

Documentation improvements everywhere.

Moved tensor_grid to template class.

Better exception descriptions in test_mgr.

Added UNEDF parameterizations and exceptions for negative
masses in eos_had_skyrme.

Added monotonicity preserving interpolation from GSL.

Created 'upper' and 'lower' functions to test upper and lower
triangularity for matrices.

Added functions to vector.h to operate on vectors without an
additional size_t parameter.

Improved bracketing to avoid undefined regions in root_bkt::solve().

Created a new class nstar_rot, based on RNS, which computes the
structure of rotating neutron stars, and an associated set of
EOS classes. This infrastructure is still in progress.

Created functions for computing number susceptibilities for the
hadronic EOSs.

Created new copy constructors for part, fermion, and nucleus.

Created eos_had_base::check_mu() to check chemical potentials.

Improvements to the nuclear mass example.

Added new tests for the eos_had_hlps class.

Improved accuracy and convergence of fermion_rel::pair_density() in
degenerate and nondegenerate limits.

Improved eos_crust_virial and eos_sn classes.

More work on eos_nse_full. 

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

Notes for 0.916 release:

Several updates for C++11 and function types. Created funct_gsl.

Several improvements to the particle classes. 

Some updates to EOS classes including: eos_had_base, eos_tov_buchdahl,
eos_sn, eos_had_apr, eos_tov_interp, eos_tov, eos_had_hlps, and
tov_solve.

Updated vector stats functions and added wvector_covariance().

Updates to the nuclear mass files, adding new nuclear mass tables,
reorganizing nucmass entry structures, and recast nucdist as 
vector<nucleus>. 

Documentation improvements everywhere. 

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

Notes for 0.915 release:

O2scl now assumes some C++11 support by default. This can be disabled
by ./configure --disable-cpp11. C++11 support may be required in
future versions as supporting earlier versions is becoming intrusive.

The sourceforge svn repository for the development branch is now
synchronized to github.

The EOS, particle and nuclei classes have been renamed to be more
consistent everywhere. Documentation has improved in several places.

The two-dimensional and multi-dimensional interpolation classes have
been reworked. Preliminary work on monotonicity-preserving
interpolation has been done.

Latex documentation is being phased out and may not appear in
future releases. 

Some return values were incorrect in functions in vector.h. These are
fixed. Calls to finite() have been replaced with the O2scl version,
which is a compatibility wrapper.

Linear interpolation was improved in the tensor_grid classes. 

Some copy constructors were missing and have been properly added.

Some of the example plots have been converted to python.

HDF5 I/O for uniform grid classes was implemented.

Convergence has been improved in fermion_rel and
eos_had_rmf::calc_temp_e(). A new eos_had_hlps class was added. The
TOV EOS class, eos_tov, has been updated to do interpolation a bit
better. The TOV solver now computes enclosed baryon masses for
pressures in pr_list. The eos_crust_virial and eos_sn_base classes
have been updated.

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

Notes for 0.914 release:

Several classes have been renamed to make the nomenclature
more consistent everywhere. The exception enums have also
been renamed.

Interpolation classes have been reworked in several ways to be
more consistent. Two-dimensional interpolation has been improved
in several ways, including the addition of a faster interp2_direct.

The tensor class was updated to a template and based by default
on std::vector<double>.

The last_conv member has been removed everywhere to be more
consistent. Convergence errors are reported in integer
return values instead.

The interp2_planar class has been updated to avoid failure
from colinear points in the data.

Some I/O functions in hdf_file have been renamed to make it
clear that they require a copy. 

The rmf_eos class has better verbose output. It now handles
the inc_rest_mass=false case. A bug was fixed in calc_temp_e()
and some typos were fixed in the documentation.

A bug was fixed in table::delete_rows().

The new function arb_dist::add_dist() was added.

The core-crust boundary was improved in tov_interp_eos. The tov_solve
class is now faster. Fixed radius interpolation was also implemented
in tov_solve. The documentation for the TOV solver was improved.

Several updates have been made for C++11. GSL_MIN and GSL_DBL_EPSILON
have been replaced with it's C++11 equivalent when possible.

Template types in the ODE classes have been updated to allow
more flexbility. C++11 support is also better for the ODE classes.

A function hist::normalize() has been added.

Boost-related solvers and minimizers have been added. 

Linear fitting in fit_linear has been updated.

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

Notes for 0.913 release:

All of the O2scl vector and matrix classes have been deprecated. O2scl
now requires Boost, and works with the Boost vector and matrix types. 
O2scl also has support for working directly with the Armadillo and 
Eigen linear algebra libraries. Deeper integration of these 
linear algebra libraries with O2scl is in progress.

The new 2012 AME nuclear masses are included.

The degenerate and nondegerate expansions for fermions have been
improved. 

The table class is now a template class, operating on any
suitably-defined vector type.

The contour class has been improved, and some small typos fixed. 

SV decompositions are now included, as well as linear fits based on
GSL. 

The Gogny and Virial EOSs have been added.

The ODE classes ode_it_solve and ode_bv_solve have been updated. 

Several improvements have been made in tov_eos and tov_solve. 
Some new crusts have been added. 

Several improvements have been made in the internal 'agraph'
command-line utility.

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

Notes for 0.912 release:

A new differential evolution minimizer has been added by Edwin van
Leeuwen which performs minimization of functions without the use of
derivative information.

Memory allocation in uvector/umatrix classes now uses new and delete
instead of malloc() and free().

The interp_smart classes have been deprecated.

Minor fixes have been made to the cli and ode_iv_solve classes.

The TOV solver now allows one to compute the moment of inertia after
considering rotation. The TOV documentation has been improved
considerably.

A couple of bugs have been repaired in ovector_cx_tlate,
umatrix_tlate, matrix_ev, uvector_tlate and arb_dist classes.

The gsl_fit class has been reworked to allow generic vector types and
several other improvements. For this, QRPT decomposition was also
improved. The fit_fix class has also been updated.

Some template types in vec_stats.h have been added.

Constants have been updated, and their namespaces are reorganized and
renamed. The 'convert_units' class has also been updated and a new
global unit conversion object is now present and accessible from
o2scl_settings.

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

Notes for 0.911 release:

O2scl has been updated to compile gracefully with newer versions of gcc. 

Most of the 1-dimensional function objects have been converted to
be const where appropriate. 

The interpolation classes have been reworked. Interpolation managers
have been replaced with an integer interpolation type. 

In several places functions which return 'int' now return 'void', bringing
O2scl more in line with C++ error-handling rather than GSL-like integer
return values. 

The contour and two-dimensional interpolation classes have been updated
in several places. 

The experimental expectation value classes have been improved. 

HDF I/O has been improved and extended, so that more objects are now
writable to and readable from HDF files. 

The nuclear mass classes have been updated and reworked slightly. The
associated documentation has also been improved.

Added an experimental set of classes for univariate probability density
functions and a class mag_fermion_zerot for quasi-free fermions in 
a magnetic field. Finally, an experimental Bayesian fit classes, 
bayes_fit, has been added. 

The TOV solver has been improved, and some bugs have been fixed,
including the addition of convergence information flags.

Function parsing has been moved from o2scl_ext to the main library.
As a result, o2scl_ext development has stopped temporarily while more
of its functionality is being included into o2scl. 

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

Notes for 0.910 release: 

Some of the constants have been updated to the CODATA 2010 values.

Major updates have been made on the supernova EOSs. 

The particle classes have been extensively reworked, to seprate data
classes for algorithm classes. The calibration for fermions has also
been improved.

A bug in QR_update() has been fixed.

The class gsl_mroot_hybrids now internally uses uvectors, and a
Broyden solver has been added. 

The histogram classes have been updated.

New copy constructors have been added tensor, hist, expect_val, and
related classes. 

Bugs in uvector_subvector have been fixed.

A shared_ptr class has been created.

The grid class was renamed to uniform grid. 

The hdf_nucmass routines now allow the data files to be in different
directories. 

A bug in gsl_vegas was fixed.

Epxerimental examples ex_methast, ex_apr_eos and ex_skryme_eos have 
been added.

Two-dimensional histogram I/O has been fixed.

Documentation and improvements in the rmf_nucleus class.

The Duflo-Zuker mass formula was added and updates have been made
to the nuclear mass classes. 

The hadronic_eos class has been updated to compute the 
properties of nculear saturation better. 

The classes gsl_miser and gsl_series have been improved.

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

Notes for 0.909 release:

The adaptive stepper class gsl_astep, and the ODE IV and BV solvers
have been updated considerably to correspond to the GSL 1.15 release
of ode-initval2. The classes ode_iv_table and ode_bv_mshoot are 
still experimental and under development.

The histogram classes have been updated. The hist_grid class has been
replaced with a more generic grid class.

HDF I/O has been developed for histograms and tensors.

The cblas routines have been reworked a bit to make their interface
more consistent and to improve teh documentation. 

Some experimental classes for expectation values have been added 
which inherit from expect_val. 

The interpolation classes have been reworked, to improve
acceleration and improve error handling. 

The variables tolf and tolx appearing in several classes have been
renamed to 'tol_rel' and 'tol_abs' repectively.

The examples and several associated plots have been updated.

Some bug fixes and improvements in the cold_nstar class.

A nuclear mass model based on FRDM has been added as frdm_mass.

Some new classes have been added to process supernova EOSs, based
on gen_sn_eos.

A class to compute nuclei in the Hartree approximation based 
on rmf_eos has been added, rmf_nucleus.

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

Notes for 0.908 release:

The GSL integration routines have been reworked, in large part due to
the efforts of Jerry Gagelman. The GSL integration workspace is now a
separate class and the QAWS integration now works. The documentation
has been improved considerably. Error handling and verbose reporting
has been improved. Testing is also much more complete.

A couple C++ language issues were fixed to assist compilation with
newer compilers and a few small updates to gsl_chebapp and
gsl_mmin_conf were made corresponding to the GSL-1.15 update.

The documentation of the ODE classes was improved. 

Bugs were fixed in umatrix_tlate::set_all() and hdf_nucmass_io. 

The TOV solver classes have been updated in several places. 

The command-line interface class, cli, has been updated, along with
the acol command-line utility in o2scl_ext. 

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

Notes for 0.907 release:

The most significant change for this release is that the old I/O
classes are being deprecated in favor of a new I/O system
utilizing HDF files. This means that version 0.907 is likely
the last version of O2scl which will support I/O using the old
classes. In the case of the o2scl_ext additional library, the 
command-line utility 'acol' now manipulates HDF files. 

The temporary command-line utility 'acol_old' performs input of data
tables using the old I/O format and outputs HDF files. Thus 'acol_old'
can be used to convert data files from the old to the new format. I
don't expect this to be a significant issue for a large number of
users, but please post to the o2scl-help mailing list if you have
difficulty.

The sole drawback of HDF file I/O is that users will now either
have to specify the location of the HDF library on their system
or give the configure script the option '--disable-hdf'. The new
I/O system using HDF files is much easier to use than the old I/O
classes. All of the native O2scl data files have been updated to 
the new HDF format. 

The o2scl-test makefile targets have been upgraded.

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

Notes for 0.906 release:

The table histogram functions have been updated, though these will
eventually be replaced by a new histogram data type. 

Some new statistics functions have been added to smart_interp.h,
and some of the functions in vec_stats.h have been improved.

The nuclear mass data from the Atomic Mass Evaluation has been
updated.

The fermion classes have been updated and a new calibration function
has been added to the fermion_T class. 

The IML linear solvers have been updated, and a new example
demonstrating their use, 'ex_it_iml' has been added to the O2scl_ext
package.

The source code to plot the output of the examples has been moved to
a new directory 'examples/plot'. 

A chi-squared test has been added.

A new Markov chain example has been added.

The documentation has been improved in several places and exception
messages have been modified to be more informative.

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

Notes for 0.905 release:

O2scl now uses C++ exceptions for error handling. By default O2scl and
GSL errors will now throw C++ exceptions. GSL errors are mapped to C++
exception objects. This default behavior can be easily modified.

The ODE routines have all been updated to include error information
and perform more efficiently, and a stiff integrator gsl_bsimp has
been added along with an example ex_stiff. A new example ex_ode_it
has also been added. 

Several vector functions have been added in vector.h and some of 
the currently existing functions were reworked. 

The binary search functions have been moved to vector.h and there have
been some changes to the search_vec class. Documentation in both cases
was improved significantly.

The statistics functions in vec_stats.h were updated.

Some template types have been added to the linalg functions to 
make them more generic. 

The gsl_miser and gsl_vegas classes have been updated, and there
were some small changes in the other GSL integrators. 

The annealing classes have been updated. 

A gsl_smooth class has been added which employs the GSL basis spline
functions to smooth data. 

There were a couple of fixes to table and table_units.

A few updates have been made to the nuclear mass fitting class,
mass_fit, and a class arb_dist was added for an arbitrary 
nuclear distribution.

The function tensor::matrix_slice() has been fixed.

New example ex_mandel was added.

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

Notes for 0.904 release:

Created new classes convert_units and table_units to facilitate unit
conversion. The tests have been reworked so that they no longer
require the user to have grep, awk, and tail. There were a couple
fixes to ensure compatibility with the Intel compiler. Several other
documentation updates.

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

Notes for 0.903 release:

There is no version 0.903 in order to make the version matching
with o2scl_ext more clear.

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

Notes for 0.902 release:

Reworked the vector and matrix classes to solve const-correctness
issues. Reworked gsl_anneal class to allow more flexibility in
changing the algorithm.

Added classes anneal_mt (threaded version of gsl_anneal), format_float
(which replaces the old double_to_latex() and double_to_html()
functions), gsl_rkf45, nonrel_fermion_zerot, hadronic_eos_temp (to
clearly separate the zero and finite temperature parts of
hadronic_eos), and nse_eos. Added several Skyrme parameterizations.

Fixed bugs in hh_base and qr_base found in the gsl-1.12 update. Added
gsl_mmin_simp2, corresponding to the gsl-1.12 simplex2 minimizer. 

Several documentation updates and minor bug fixes.

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

Notes for 0.901 release:

The function pointer classes didn't allow the user to specify const
member functions so that flexibility has been added with classes
named with 'cmfptr' rather than 'mfptr'. 

A ODE multiple shooting class (from Julien Garaud) and a unit
conversion class have been added. The gsl_chebapp class was updated. 
Some classes have been added associated with nuclear reactions.
A implicit ODE stepper was added based on the GSL 'bsimp' stepper.

There are also several other small fixes in nonrel_fermion, rmf_eos,
the QR routines, skyrme_eos, schematic_eos, and the ovector classes.

There are also several documentation updates.

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

Notes for 0.9 release:

This is the first beta release of O2scl. The majority of the classes
are stable enough for general use. The full library, all of the tests,
and all of the examples should compile on most systems. Generally,
the library testing should also succeed without failure.

There are still some classes and functions which are experimental. The
decision to release a beta version while some of the classes are still
in development was intentional: it allows users to use the more
experimental classes and see what may be possible with O2scl in the 
future, hopefully without causing too many difficulties in testing and
compilaion for the casual end-user. 

Since version 0.805, several changes have been made:

There have been extensive documentation updates, and all experimental
classes are explicitly indicated as such, so that users are properly
warned.

The gsl_min_brent class has been fixed. The composite integration 
classes composite_inte and comp_gen_inte have been updated. The
class search_vec was fixed. 

Most of the unnecessary void *'s have been removed.

Error handling has been reworked slightly, so that the user has more
control over when an error is called. The default error handler now
aborts execution if the error handler is called. 

Some problems with executable file name suffixes in Cygwin have 
been fixed.

Self-assignment is now properly handled in ovector and related classes.

The o2scl_arith namespace was removed, as it caused problems
with Koenig lookup.

A new table3d class was created -- a 3-dimensional version of the 
table class. 

The contour and twod_intp classes were been updated.

Interpolation managers were created, to make memory management simpler.

The linear solvers have been reworked slightly,
especially so that they will work better with the O2scl_ext package.
An error in the linalg makefile was also fixed. The basic matrix-matrix
multiply function was also fixed.

Some of the particle and EOS classes have been reworked slightly,
and the old CFL EOSs have been removed, There were small changes
in skyrme_eos, hadronic_eos, eff_boson, nonrel_femrion, and eff_fermion.

Some major work has been performed on the collection class, but this
and related classes are still experimental.

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

Notes for 0.806 release:

Version 0.806 was a quick release to fix some compilation difficulties
on Mac OS X. 

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

Notes for 0.805 release:

There are several new examples and new associated documentation. The
examples are still a work in progress, but hopefully this will make
the library more transparent.

There are several fixes to ensure classes are purely virtual when they
should be.

A few of the vector functions have been moved around slightly, 
and the documentation for vectors and matrices has been improved 
significantly. 

Finally, the linear algebra functions now have their own namespace
and offer operator[] and operator() versions. 

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

Notes for 0.804 release:

The test targets have been reimplemented so that they are automake
targets and use libtool. This should make them a little more portable
and easier to use.

A small set of linear algebra routines has been added, which provides
for LU, QR, householder, and tridiagonal solvers for generic vector
and matrix types. Also, a << operator has been defined for the
permutation class.

Several classes have been updated and or fixed, including eff_fermion,
rel_boson, eff_boson, cern_adapt, and cern_gauss56, ode_iv_solve,
tov_eos and tov_solve.  The GSL solver class gsl_mroot_hybrids has
been updated to be faster and is now properly generalized to generic
vector and matrix types.

Also, the documentation has been improved and updated in several places.

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

Notes for 0.803 release:

A permutation class, which is fully compatible with gsl_permutation
objects (ovector and omatrix are also compatible with gsl_vector and
gsl_matrix). The functions which apply a permutation to a user-specified
vector are member template functions in the permutation class. 

New classes based on the Open Optimization Library (OOL) (available
from http://ool.sourceforge.net) which perfom constrained minimization
have been added. At present, only the projected gradient and spectral
projected gradient methods are implemented, and the "gencan" method is
in progress. These minimizers work consistently with the other
multidimensional minimizers in O2scl and are now sufficiently flexible
to accept member functions as arguments, but are still essentially
identical to the original versions. 

The class gsl_miser has been updated to be compatibile with generic
vector types. 

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

Notes for 0.802 release:

The classes gsl_monte, gsl_vegas, and gsl_anneal have all been
updated. They are now compatible with any sufficiently constructed
vector type. In the case of class gsl_vegas, the interface has been
significantly updated so that the user should have a lot more freedom
(even more freedom than in the original GSL) to use or modify the
integration algorithm as desired.

There were a couple bugs in gsl_rkck, gsl_rk8pd, and the interpolation
in the new tensor class which have been fixed. The ODE classes now
properly allow one to choose the same objects for y and yout or
the same objects for dydx and dydx_out, which should be easier for 
more "casual" users. 

Statistics functions (mean, standard deviation, autocorrelation, etc)
have been added and are defined in vec_stats.h for generic double 
vector types. 

The range checking is (at present) performed with "#if
O2SCL_NO_RANGE_CHECK...#endif" blocks. This is similar to how GSL used
to do range checking in earlier versions. There is now a function in
the class lib_settings_class which checks to see if range checking was
turned on at the library compile time. 

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