Hi list, it has been unusually quiet here so it's time to speak up. How hard would it be to implement support for specifying buffer dtype (and friends) in @cython.locals() in a .pxd file? Following code (in a .pxd file)
cimport numpy @cython.locals(local1 = numpy.ndarray[double]) cdef double f() currently yields "test.pxd:4:32: Not a type" error. The problem is that one currently cannot make use of buffer access optimisation for local variables in pure python mode. I use pure mode extensively in my project [1]. [1] https://github.com/strohel/PyBayes I understand that it has no sense in .py files as type type is not usually subscriptable. On the same note, is there a plan to implement fast buffer access for class variables? ;) (this is obviously not that crucial) Regards and keep up the good work! Matěj Laitl _______________________________________________ cython-devel mailing list cython-devel@python.org http://mail.python.org/mailman/listinfo/cython-devel
