After discussion with Greg in previous thread and careful reading of
C99 std (TC3 from Nov. 2007) , I believe we should do this:
def widest_numeric_type(type1, type2):
# Given two numeric types, return the narrowest type
# encompassing both of them.
if type1.is_complex or type2.is_complex:
def real_type(ntype):
if ntype.is_complex:
return ntype.real_type
return ntype
widest_type = CComplexType(
widest_numeric_type(
real_type(type1),
real_type(type2)))
elif type1.is_enum and type2.is_enum:
widest_type = c_int_type
elif type1.rank < type2.rank:
widest_type = type2
elif type1.rank > type2.rank:
widest_type = type1
elif type1.signed < type2.signed:
widest_type = type1
else:
widest_type = type2
return widest_type
This algorithm will be wrong for binops mixing signed and unsigned
integrals, because we cannot know if if a signed type can or not
represent all the values of other unsigned type.
As part of my on-going work to cleanup PyrexTypes.py, I propose to use
Greg's algorithm, and adding a warning somewhere would be nice (can
someone can point me were should I look for adding such warning?).
Comments?
--
Lisandro Dalcin
---------------
Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
PTLC - Güemes 3450, (3000) Santa Fe, Argentina
Tel/Fax: +54-(0)342-451.1594
_______________________________________________
Cython-dev mailing list
[email protected]
http://codespeak.net/mailman/listinfo/cython-dev
