20114, "bradcray", "Should the '#' operator on ranges take boolean values?", "2022-06-29T20:30:40Z"
In #20098, I re-implemented # on ranges to take any integral value by
changing the second argument types from int(w)/uint(w) to integral.
This had the effect of disabling the ability to apply a boolean using the
count operator, like 1..10#true, which our tests didn't catch because we
didn't have any. However, it turns out that Arkouda does rely on this
behavior, so stopped compiling.
In #20111 , I am re-enabling # on booleans for the time being, but this
issue asks whether we should continue to support it or deprecate it
(requiring the user to cast booleans to integers before passing them to #).
Arguments against:
-
#logically takes a numeric count value in order to count off that many
values
Arguments in favor:
- Since booleans coerce to integers, and in many cases, we permit this
style of coercion to get things done, maybe we should here as well?