[Qtfm-announce] QTFM version 1.4

[Release a. a. u. <qtf...@li...>]

A new release of the Quaternion Toolbox for Matlab (QTFM) has
been posted on Sourceforge.

With this release there are now help files available in the
Matlab Help browser.

The release notes are given below. If updating to this release
from versions prior to 1.1, please note that you should delete
the existing file dot.m. The test suite provided in the script
test.m in the test subdirectory runs a check for the existence
of this file. With this release, it is now possible to run the
test suite code from the Matlab Start menu (paragraph 2 below).

--------------------------------------------------------------
                Quaternion toolbox for Matlab
                http://qtfm.sourceforge.net/
       Mailing list: qtf...@li...
--------------------------------------------------------------
Copyright (©) Stephen J. Sangwine        (S.S...@IE...)
2005-2008     Nicolas Le Bihan  (nic...@li...)
--------------------------------------------------------------

                     Release Notes

Version 1.4  1 November 2008

The toolbox now has help files accessible in the Matlab help browser.
Many of these lack example code - this will be added as time permits.
The authors would welcome examples from users to add to the
documentation, especially with graphical outputs.

There are now some QTFM menu items accessible from the Matlab Start
button. These include access to the help files, the ability to run
the test suite, to refresh HTML help files (useful if updates are
downloaded by CVS), and access to the QTFM website.

The show, char, and displayall functions have been modified to give
a clearer output, especially in the case where the quaternions are
complex.

The conv and conv2 functions have been rewritten to use a much
faster vectorised algorithm (due to Todd Ell, first implemented
in the LQSTFM toolbox, http://lqstfm.sourceforge.net).

The bidiagonalize function has been rewritten to operate iteratively
rather than recursively. This overcomes a limitation on the size of
matrices that could be handled by the svd function without running
out of memory or recursion depth. Further improvements to this
function are anticipated in a future release to improve the size
of matrices that can be handled by this and the svd function.

The end function has been rewritten to use the builtin Matlab
function. This corrects an error where end gave incorrect
results on row vectors.

The numel function has been modified to call the builtin Matlab
function so that subscripted indexing will work. Any existing
code that calls numel should be modified to call prod(size(q))
instead, because numel will return 1 in all cases.

Connected with the change to numel, subscripted indexing of the
form q.x now works for quaternion arrays with more than one
element. The subscripted notations available are: w, x, y, z for
the four components, scalar or s (synonyms of w); I, J, K (synonyms
of x, y, z); vector or v; and real/imag.

The reshape function is now implemented for quaternion arrays.
This was made possible by the change to numel described above.

A new private function has been added called overload. This makes
it trivial to implement overloadings of Matlab functions for which
the algorithm is to apply the Matlab function to the components of
the quaternion array and then compose a quaternion array from the
result.

The Matlab functions blkdiag and squeeze are now overloaded for
quaternions.

New functions implemented are: eyeq, onesq, zerosq, zerosv. These
are quaternion equivalents of the Matlab functions eye, ones and
zeros (zerosq returns a full quaternion zero array, whereas zerosv
returns a pure quaternion zero array).

New functions have been added for generating random quaternions.
randv and randq generate uniformly distributed unit vectors and
quaternions respectively. randf generates unit vectors concentrated
in a particular direction (in 3-space) using the Fisher distribution.
randvmf generates unit quaternions concentrated in a particular
direction (in 4-space) using the von Mises-Fisher distribution.

The Matlab function cast is now overloaded for casting a quaternion
array to have different element types. This functionality was
previously available in the function convert (which remains).