Releases

Tasks for the Next Release 2011.8 (August 31, 2011)

All existing examples will work, along with tests [Pavel]
Sphinx documentation will be completed [Pavel]
Sphinx docs will contain a section where all default forms will be described [Pavel]
Richards done correctly via the rk_time_step() method [Pavel]
rk_time_step() will to work with all matrix solvers [Vlada Cerny]
Stable directory structure (saved in a document) [Lukas]
Clean up include hierarchy [Lukas]
Stable src file structure (saved in a document) [Lukas]
Stable namespace structure (saved in a document) [Lukas]
No method definitions in headers [Lukas]
Proper comments in headers (DOXYGEN or similar => development documentation) - no comments (only explanation in method bodies) in source files (the above 10-15 man-days) [Lukas]
Logical class structure (saved in a document) (3-6 man-days) [Lukas]
Inner functionality tests (5-10 man-days) [Lukas]
No source file inclusions (2-5 man-days) [Lukas]
(Maybe next release) unification of real & complex versions of hermes_common, hermes2d into one (10-15 man-days) (this entails another change in all examples, tutorials, benchmarks, Sphinx before release) [Lukas]

Runge-Kutta Nystrom method for second-order problems in time
rk_time_step() will be optimized for speed
rk_time_step() will be able to handle incompressible N-S equations
Redo Legendre polynomials using recurrent relations. These are stable and will allow higher polynomial degrees than Horner. We learned this in H1D.
Reuse ref. solution from previous adaptivity steps. This can be achieved by projecting the solution from the previous reference mesh on the new one via JFNK. The result will be used to solve on the new reference mesh via JFNK. None of these steps requires assembling of the stiffness or mass matrices on the new reference mesh.
Create Python wrappers including classes for weak forms.