BTW, Do we want to fix "continuous" / "contiguous" mess?

Memory block is contiguous.
Mathematic function is continuous.

Related StackOverflow questions:

• https://stackoverflow.com/questions/26998223/what-is-the-difference-between-contiguous-and-non-contiguous-arrays
• https://stackoverflow.com/questions/4059363/what-is-a-contiguous-memory-block
• https://stackoverflow.com/questions/28930465/what-is-the-difference-between-flatten-and-ravel-functions-in-numpy

yes, definitely. And it's not a mess at all. On small images (like 20x20) this trick with

Size size = mat.size();
if( mat.isContinuous() ) { size.width *= size.height; size.height = 1; }

can accelerate processing by 10-20% in the case of simple functions. And it's put in many places of the library and in the user code as well. Obviously, this trick can only be done with size.width * size.height not bigger than INT_MAX. And it's also obviously that it only needs to be checked during matrix construction (just few places in the library), not when you process the matrix (many places in the library and in the user code). We attempted to implement such check in updateContinuityFlag() function, but 1) it contained a bug and 2) it was not used everywhere.

I mean this typo:

-continous
+contiguous

One more reference: https://docs.scipy.org/doc/numpy-1.10.0/reference/generated/numpy.ascontiguousarray.html

Do I understand correctly that huge arrays (matrix 1xN) are still not supported by cv::Mat? Or is it somehow solved too?