On Sun, Oct 26, 2014 at 2:50 PM, Nils Bruin <nbr...@sfu.ca> wrote: > On Sunday, October 26, 2014 2:08:27 PM UTC-7, Volker Braun wrote: >> >> Python 3 only uses __eq__, __lt__, ..... Python also provides >> functools.partial_ordering to synthesizing the remaining methods just from >> __eq__ and __lt__. We either use that or write a similar decorator to also >> deal with coercion. >> >> The __richcmp__ method is a Cython convention and doesn't exist in plain >> Python 3 (or 2 for that matter). >> > > A little more elaborately: On C-API level PyTypeObject has a > tp_richcompare slot with the interface that cython's __richcompare__ has. > Cython puts that special method in that slot for a cdef class. Python's > "python" classes inherit a tp_richcompare that dispatches to __eq__ and > __lt__ methods, probably because the constants Py_EQ etc. were deemed > unappetizing to expose on python level. > > This also means that on cython level, we should stick with __richcompare__ > by default, because dispatching to __eq__ python functions would be slow. >
I've been toying with the idea of letting (the default) __richcmp__ delegate to (c-level) __eq__, etc. if defined in Cython to further minimize the differences between Cython and Python. For elements, it'd be nice to have a them work with coercion by default, with an _eq_ (and possibly _lt_) method that is guaranteed to get elements of the same parent. There's the wrinkle that __richcmp__ and __hash__ must be inherited together in the C API. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
