# New Ticket Created by Zefram
# Please include the string: [perl #127019]
# in the subject line of all future correspondence about this issue.
# <URL: https://rt.perl.org/Ticket/Display.html?id=127019 >
I'm not sure what identities we can expect coercions to obey in Perl 6,
but this combination of coercions sure looks surprising:
> True.^does(Numeric)
True
> True.^isa(Int)
1
> True.Numeric
1
> True.Int
True
The two coercions are surprisingly inconsistent. In both cases the output
of the coercion isa Int, and the type to coerce to includes all of Int,
so one would think that, having found a satisfactory Int to coerce to,
the same Int would be the output of both coercions. But they differ.
Apparently True isn't Numeric enough to be the output of a coercion,
despite it being Int enough, when that's a more specific type.
Furthermore, in this case the input value is already of both of the
types to which it is requested to coerce. In this situation, it is
surprising that a coercion produces any result different from the input.
There is no work for the coercion to do.
The Numeric coercion shown above is also inconsistent with what is
obtained from a coercing type constraint:
> (sub (Numeric() $a) { $a })(True)
True
The coercion that looks inconsistent with the others and with common
sense is True.Numeric. If you're serious about Bool being a subclass
of Int, surely True.Numeric needs to return True. (Overall it would be
more sensible for Bool to have no particular relation to Int and Numeric,
though. The subclass arrangement is detrimental to type discrimination
and to correctness.)
-zefram
