1. 21 Dec, 2018 2 commits
  2. 17 Dec, 2018 1 commit
  3. 30 Nov, 2018 1 commit
  4. 21 Nov, 2018 2 commits
  5. 14 Nov, 2018 1 commit
  6. 04 Oct, 2018 2 commits
  7. 03 Oct, 2018 1 commit
  8. 02 Oct, 2018 1 commit
  9. 01 Oct, 2018 2 commits
  10. 30 Sep, 2018 1 commit
  11. 28 Sep, 2018 1 commit
  12. 24 Sep, 2018 3 commits
  13. 14 Sep, 2018 1 commit
  14. 13 Sep, 2018 1 commit
  15. 10 Sep, 2018 3 commits
  16. 09 Sep, 2018 1 commit
  17. 29 Aug, 2018 3 commits
  18. 27 Jul, 2018 2 commits
  19. 17 Jul, 2018 1 commit
  20. 16 Jul, 2018 1 commit
  21. 12 Jul, 2018 1 commit
  22. 10 Jul, 2018 3 commits
  23. 08 Jul, 2018 5 commits
    • Andreas Dedner's avatar
      bring snippet up to date in README.md · fd2967c0
      Andreas Dedner authored
      fd2967c0
    • Andreas Dedner's avatar
      some minor fixes: · 4daf41e2
      Andreas Dedner authored
      avoid some uint warnings and corrected shebang in rmgenerated script
      4daf41e2
    • Andreas Dedner's avatar
      41d63dc9
    • Andreas Dedner's avatar
      [!110] Extend support for polygonal grids · 37760596
      Andreas Dedner authored
      Merge branch 'feature/support-polygongrid' into 'master'
      
      True polygonal grids, like dune-polygongrid, will return none as geometry type
      for their elements. These elements do not have an associated reference element
      and some assumptions usually made in Dune are no longer valid. For example,
      elements of geometry type none may have different vertex counts. This
      assumption, however, was made in some places in dune-python. This MR removes
      some of these spots.
      
      See merge request [!110]
      
        [!110]: gitlab.dune-project.org/staging/dune-python/merge_requests/110
      37760596
    • Martin Nolte's avatar
      [bugfix] Do not rely on `operator[]` on a `JacobianInverseTransposed` · e4f51e62
      Martin Nolte authored
      Some grid implementations, like dune-spgrid, will not return a dense
      matrix for the `JacobianInverseTransposed`. While we could simply cast it
      into a `FieldMatrix`, this might reduce performance due to unnecessary
      dense matrix vector multiplications.
      
      This patch uses the `mv` method, which must be provided by any
      `JacobianInverseTransposed` instead. Apart from retaining the
      performance of the specialized implementation, this improves the overall
      clarity of the code.
      
      As a drawback, the local and global gradient vectors are copied for the
      time being. This unnessary copy can be avoided by a specialized
      `DenseVector` implementation for a single axis in a pybind11::array
      accessor.
      e4f51e62