Repository navigation
Counted ranges abuse mathematical notation #2932
Description
Activity
FWIW, +1 from me as the originator of this particular abuse of interval notation. It was originally less offensive since we used it in contexts where it was plain that the second member of the tuple is not a sentinel a la
[i, 42), but we've had to avoid using it in less clear contexts ([i, n)isn't horrible, but[i, j)surely is). This context-sensitivity is a clear indicator that overloading the syntax is a bad idea.- changed the title
[-]counted ranges abuse mathematical notation[/-][+]Counted ranges abuse mathematical notation[/+]on Jun 15, 2019 I agree that we should not overload existing, established notation for counted ranges.
I don't have an "obvious" solution, but maybe we can define a slightly more elaborate device, maybe like “
p + [0, N)” or so, to denote the set of iteratorsp + 0,p + 1, ...,p + (N - 1)? Something like a pair of a starting iterator plus the set of all valid offsets?Editorial meeting: Define range of iterators. "Elements in the range" means dereferencing the iterators in the range. Use "
p+ [0,N)".
Counted ranges are presented as
[
A,N)but don't have the normal mathematical meaning; instead, they mean [
A,A+N) (where the "+" means "the value that would be obtained byNapplications of ++ toA" rather thanoperator+).We should not abuse established mathematical notation in this way, and should find a different and unambiguous way to present counted ranges.