This sounds great to me, excellent work, Martin. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Chris Mattmann, Ph.D. Senior Computer Scientist NASA Jet Propulsion Laboratory Pasadena, CA 91109 USA Office: 171-266B, Mailstop: 171-246 Email: [email protected] WWW: http://sunset.usc.edu/~mattmann/ ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Adjunct Assistant Professor, Computer Science Department University of Southern California, Los Angeles, CA 90089 USA ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
-----Original Message----- From: Martin Desruisseaux <[email protected]> Organization: Geomatys Reply-To: "[email protected]" <[email protected]> Date: Wednesday, September 11, 2013 4:36 PM To: Apache SIS <[email protected]> Subject: Report on matrix work >Hello all > >The javadoc package documentation below tries to explain the rational >behind the SIS matrix package (still work in progress): > >https://builds.apache.org/job/sis-jdk7/site/apidocs/org/apache/sis/referen >cing/operation/matrix/package-summary.html > >The idea is strongly inspired by "vecmath", which was part of Java3D. >"Vecmath" is actually hard to replace by other matrix package (JBLAS, >JAMA or other) because the design goals are not the same. There is >thousands of possible operations for matrix, and the operations of >interest are not the same for resolving linear equations than for >performing coordinate transformations. Because "vecmath" was designed >for coordinate conversions in 3D spaces, it was a pretty good fit to >geospatial needs. > >There is a few methods in org.apache.sis.referencing.operation.matrix >that are more closely related to coordinate transformations rather than >general matrix operations: isAffine(), toAffineTransform(), >normalizeColumns() - more will come soon. I have not yet connected the >SIS package to JAMA. But I came to think that the connection (if any) >should not be visible in the public API. For example when inverting a >non-square matrix, JAMA uses a least squares approach. This is probably >a good choice for resolving linear equations, but sis-referencing needs >a more conservative approach. For this reason, I don't think that we >should expose the JAMA Matrix.inverse() method directly. > >I will post more when the package will be closer to completion, unless >someone would like clarifications or changes? > > Martin >
