Re: [eigen] Signed or unsigned indexing

[ Thread Index | Date Index | More Archives ]


Am 20.01.2017 um 19:41 schrieb Benoit Jacob:
That leads to one of the strongest arguments in favor of unsigned: at least, that has well-defined (wrapping) overflow behavior. That might be 'broken', but it isn't nearly as dangerous as the undefined behavior of signed overflow!
But if you take this point of view the unsigned-signed promotion and comparison rules are just a as big hazard. We should never be forced to really think about them.
E.g. try to wrap your had around the following:
signed int a{-1 };
unsigned int b{1};
according to C++ language rules now a < b is false.

This is possible through consistent use of one type of indexes, sizes, slices etc. And negative slices and strides have well-defined meaning and should be allowed, so they must be signed. Deciding on a case-by-case basis forces peoples into all the nitty gritty details of integer promotion and comparision, which is exactly what we should IMHO protect users from.

So that argument works both ways.


Mail converted by MHonArc 2.6.19+