Proposed in the July 2026 AI-assisted code review (Claude Code), from investigation of the v1.4 development branch (octave).
-oct generates a native DEFUN_DLD oct-file (compiled with plain mkoctfile, no MEX emulation layer): src/mwrap-octgen.cc + src/mwrap-oct-support.c, mirrored in python/mwrap_octgen.py + python/mwrap_oct_support.c. Uses fortran_vec() for direct data access; several correctness fixes already landed on the branch (is_complex_type, goto-over-init, real-to-complex promotion).
Benefit: first-class Octave support instead of relying on Octave's MEX compatibility layer.
Before it lands on master:
- C++/Python byte-parity coverage in
testing/test_python.sh
- an end-to-end Octave CI test (Octave is already in CI, so this is cheap)
- user-guide/README documentation
See #45 for dispatch-table compatibility and merge scope.
Proposed in the July 2026 AI-assisted code review (Claude Code), from investigation of the v1.4 development branch (
octave).-octgenerates a nativeDEFUN_DLDoct-file (compiled with plainmkoctfile, no MEX emulation layer):src/mwrap-octgen.cc+src/mwrap-oct-support.c, mirrored inpython/mwrap_octgen.py+python/mwrap_oct_support.c. Usesfortran_vec()for direct data access; several correctness fixes already landed on the branch (is_complex_type, goto-over-init, real-to-complex promotion).Benefit: first-class Octave support instead of relying on Octave's MEX compatibility layer.
Before it lands on master:
testing/test_python.shSee #45 for dispatch-table compatibility and merge scope.