On Fri, Nov 5, 2010 at 6:12 PM, Martin Albrecht <martinralbre...@googlemail.com> wrote: > Hi there, > > in the Sage library we cannot write the following C++ code: > > Class a = Class() > > since Cython turns the statement > > cdef Class a = Class() > > into > > Class a > a = Class() > > which only works when there is a copy constructor. In the Sage library we > sometimes use the follow construct in order to work around this Cython vs. C++ > incompatibility: > > Class a > Construct<Class>(&a) > > For example check the PolyBoRi wrapper and the NTL wrapper. There are also > variants of this called Construct_p<Class>(&a, one_parameter) and so forth. > > I'm wrapping CryptoMiniSat at the moment (a general purpose (!) SAT solver > which one this year's SAT Race). When I try to use this construct in my > wrapper I get a complaint from g++ that Construct wasn't defined in the > scope; yet I cannot see what we are doing special in the PolyBoRi or NTL > wrapper which would have an affect on that. Also searching online for this > construction also returned nothing of interest as far as I can tell, mainly > because one is overwhelmed with C++ constructor introductions. Thus, my > question: does any of you C++ gurus now where this thing is defined and what I > have to do (or not to do) in order to use it?
If heap-allocated instances are fast enough, you can use the much easier C++ support (recently merged into Sage): cdef Class *a = new Class() del a What we use in Sage is probably buried in some .h file. I think the official term is "placement new" where we allocate a chunk of the Python object struct to hold the data. - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
