On Fri, Dec 10, 2010 at 10:34 AM, Jason Grout <jason-s...@creativetrax.com> wrote: > On 12/10/10 10:47 AM, Simon King wrote: >> >> Dear Jason, >> >> On 10 Dez., 14:55, Jason Grout<jason-s...@creativetrax.com> wrote: >>>> >>>> In sage/modules/vector_rational_dense.pyx there is a class definition >>>> Vector_rational_dense. It has a cdef'd method _cmp_c_impl, which is >>>> supposed to do the comparison. >>> >>> ... >>>> >>>> Hence, the comparison method is consulted *only* when calling the >>>> special method __cmp__ explicitly! So, how else is comparison >>>> achieved?? >>> >>> Perhapshttp://groups.google.com/group/sage-devel/msg/ca37624c4ad655c6 >>> could be helpful? That message deals with comparison between matrices, >>> but you might find some useful information for tracking down what is >>> happening. >> >> Thank you, the message is helpful, to some extent. I see that >> __richcmp__ and __cmp__ can not be inherited, so that the code from >> sage.structure.element should be copied. So, it confirms some of my >> suspicions. > > > I think you misunderstood the post. They *can* be inherited. The point is > that "according to the Python/C API docs [3], either all of __hash__, > __cmp__, and __richcmp__ are inherited, or none are. " > > > In our case, since none of these three is defined in vector_rational_dense, > they are all inherited from free_module_element. > > >> >> But now, my confusion has increased. I see that >> sage.modules.vector_rational_dense does *not* define its own >> __richcmp__ and __cmp__. So, it is even less clear why it works. > > > I inserted > > print "_cmp_c_impl in vector_rational_dense.pyx" > > in the vector_rational_dense.pyx _cmp_c_impl function. I inserted > > print "_cmp_c_impl in FreeModuleElement (free_module_element.pyx)" > > in the free_module_element.pyx _cmp_c_impl function, and > > print "__richcmp__ in FreeModuleElement (free_module_element.pyx)" > > in the free_module_element.pyx __richcmp__ function. Then I get the > following: > > sage: v = vector(QQ, [0,0,0,0]) > sage: w = vector(QQ, [-1,3/2,0,0]) > sage: cm = sage.structure.element.get_coercion_model(); cm > <sage.structure.coerce.CoercionModel_cache_maps object at 0x101f686b0> > sage: cm.explain(v,w,operator.eq) > Identical parents, arithmetic performed immediately. > Result lives in Vector space of dimension 4 over Rational Field > Vector space of dimension 4 over Rational Field > sage: v==w > __richcmp__ in FreeModuleElement (free_module_element.pyx) > False > sage: cmp(v,w) > __richcmp__ in FreeModuleElement (free_module_element.pyx) > __richcmp__ in FreeModuleElement (free_module_element.pyx) > __richcmp__ in FreeModuleElement (free_module_element.pyx) > 1 > > > So I see that it is the free module element class (the superclass of > vector_rational_dense) has the __richcmp__ function defined, and it calls > the special FreeModuleElement._cmp_same_ambient_c function. So we are not > using the coercion system, nor are we using the vector_rational_dense > function, in either of the above calls. > > At least, that's the way I understand it.
Yes, I think you are spot on. As for the more general question, the current state of comparison of elements in Sage is a huge mess, and not much should be read into the current structure. It was written before cpdef existed and before the interaction between __cmp__, __richcmp__ and __hash__ was fully understood, and then more hacked on rather than fixed with every iteration of the coercion model. It really just needs to be fixed for once and for all (e.g. make a system where elements define a cpdef _cmp_ and _hash_, with a default __cmp__, __righcmp__, and __hash__ provided, and anything needing to do something more fancy (e.g. symbolic elements) define their own __richcmp__ etc. doing coercion themselves (the binop function is simple enough to use here). There was an attempt made as part of the penultimate coercion re-write, but it was too invasive on top of everything else that was changing at the time. - 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
