Path: |
rdoc/changes.rdoc |

Last Update: |
Sun Nov 14 14:53:48 -0800 2010 |

Ruby/GSL 1.11.2 introduced enhancements and some important bug fixes since the earlier 1.10.3 release. Some of the enhancements and bug fixes involve changes that are not backwards compatible. Existing scripts that work with Ruby/GSL 1.10.3 may break under 1.11.2 (or later) if they use certain constructs that are no longer supported or have changed in an incompatible way.

The changes that could break old scripts are described below. They are followed by highlights of the other enhancements and new features.

A complete log of changes since Ruby/GSL version 1.10.3 can be obtained by using Subversion‘s log command to look at the repository…

svn log http://rb-gsl.rubyforge.org/svn/trunk/rb-gsl

- FFT interface has been extensively changed.
- Matrix#get has been modified.
- Size changing methods of GSL::Vector et al have, in general, been removed.
- The shift, unshift, push, and pop methods have been removed from
GSL::Vector et al since they result increasing the size of
`self`. This can cause new memory to be allocated and old memory to be freed, but because existing Views contain pointers to the old memory, this cannot be allowed. Also removed up and down aliases from GSL::Poly. - The concat method of GSL::Vector et al has been changed to always return a
new, longer instance rather than reallocating memory for
`self`. - Return values from delete, delete_at, and delete_if now follow same convention as similarly named methods in Ruby‘s Array class. These methods should be considered deprecated because existing Views could still refer to now unmanged (but still allocated) memory.
- GSL::Complex objects are now immutable.

- Now distributed and installable as a Ruby Gem.
- Allow nil for real and/or imag in GSL::Vector::Complex#set(i,re,im) to NOT set that component. For example, "z=GSL::Vector::Complex[1]; z[0] = [nil, 1]" will set z[0] to GSL::Complex[0,1].
- Added GSL::Vector::Complex#to_s method.
- Improved GSL::Vector::Complex#inspect.
- Added GSL::Vector::Complex#set_real and GSL::Vector::Complex#set_imag methods (and real=, re=, imag=, im= aliases). Currently only work with Numerics, but could be made to work with GSL::Vectors (and Arrays of Floats).
- Added GSL::Vector::Complex#conjugate and GSL::Vector::Complex#conjugate! methods along with corresponding "conj" and "conj!" aliases.
- Added "add!", "sub!", "mul!", and "div!" methods to GSL::Vector::Complex. Also "add_constant!", and "scale!" aliases.
- Added the following stats methods to GSL::Vector::Complex, sum, mean, tss, tss_m, variance, variance_m, variance_fm, sd, sd_m, and sd_fm. sum and mean return GSL::Complex; all others return Float.
- Added "dup" alias for "clone" method of Vector and Maxtrix.
- Added support for three parameter call to subvector to support stride.
- Added shape method with alias size for Matrix.
- Provide GSL_Object#dup method to override Object#dup. GSL_Object#dup simply raises NotImplementedError. Subclasses that want dup functionality should provide their own implementation.
- Added length alias to Vector and Vector::Complex.
- Added angle alias for arg method of Complex, Vector::Complex, and Matrix::Complex
- Added conj alias for conjugate for GSL::Complex. Added conjugate! method with conj! alias for GSL::Complex.
- Added #[]= alias for set method, conj alias for conjugate method, and conj! alias for conjugate! method to GSL::Matrix::Complex
- Support GSL::Permutation as parameter to Vector::Complex#get (and #[]).
- Disallow the stride and length parameters of the various FFT methods. The Vector being transformed could really be a Vector::View (e.g. a view of a Matrix column). In that case, we need to use the stride of the Vector::View. Users wishing to use a custom stride or length will be able to do so by transforming a subvector of the Vector.
- Added Block#length alias for Block#size.
- Fixed bug 16023.
- Calling to_v on a single column matrix view will return column vector. Related to 19757.
- Provided GSL::Matrix::Complex with reasonable to_s and inspect implementations.
- Made Vector#{add,sub,mul,div}! methods work with non-Vector parameter (i.e. constant).
- Added collect and collect! methods to Matrix::Complex.
- Added map and map! aliases for collect and collect! to Block, Vector et al, and Matrix et al.
- Vector::Complex#each, reverse_each, collect, and collect! now yield objects that are allocated on heap rather than stack.
- Made Vector#subvector work with negative offsets
- Added Vector#sgn and Matrix#sgn, both with signum alias.
- Added support for a Range parameter (with optional stride) to Vector#subvector and Vector::Complex#subvector.
- Modified Vector#get and Vector::Complex#get to return Views (via subvector) if more than one (or none) parameters are given or if a single Range parameter is given.
- Enhanced Matrix#submatrix with many new options and updated docs to reflect changes
- Modified Matrix#get to use new Matrix#submatrix functionality.
- Added Matrix#to_a and Matrix::Complex#to_a methods.
- Enhanced Vector#set functionality.
- Added more details, examples, and a warning to Vector#set documentation.
- Changed Range handling to support Floats for begin and end values where appropriate.
- Changed Vector#view_with_stride to require non-zero stride rather then positive stride.
- Enhanced Vector::Complex#set functionality.
- Enhanced Matrix#set. Changed behavior of Matrix#get when called with a single Fixnum argument; now returns individual element rather than Vector::View of row. See html/matrix.html for more details. Changed tests and examples to accomdate the change in Matrix#get.
- Match to the latest tensor extension (version 0.99)
- Modified GSL::Matrix::Complex#set to have same (i.e. newly enhanced) functionality as GSL::Matrix#set. [Possibly incomplete; please report any discrepancies.]
- Added diag alias for Matrix#diagonal.
- Added fftshift, fftshift!, ifftshift, and ifftshift! methods to GSL::Vector::Complex and documented them.