From @masak

<jercos> m​: for -1..2 ->\_{(10 ** _).WHAT.say}
<camelia> rakudo-moar 3fc98b​: OUTPUT«(Num)␤(Int)␤(Int)␤(Int)␤»
<TimToady> one supposes that first one could be Rat
<jercos> I would expect negative exponents to be Rat until Rat overflows.
<TimToady> could be worth an RT
* masak submits rakudobug
<masak> m​: say (10 ** -1).^name
<camelia> rakudo-moar 3fc98b​: OUTPUT«Num␤»
<masak> m​: say (1/ 10).^name
<camelia> rakudo-moar 3fc98b​: OUTPUT«Rat␤»
<TimToady> question after that is whether it should underflow to
FatRat, but probably should just go to Num
<masak> TimToady​: yeah, I think Num is less surprising somehow.
<masak> TimToady​: otherwise people might run into unexpected slowness
for unrequited precision.
<jercos> I think Num is fairly reasonable for the under/overflow, yeah.

From @pmichaud

Having (10 ** -1) produce a Rat sounds potentially slippery-slope to me.

For example, what about (9 ** 0.5) ?

Also, while researching this I ran into the following​:

  > say (9.0 ** -1)
  Type check failed in binding nu; expected 'Int' but got 'Num'
  in block <unit> at <unknown file>​:1

Pm

The RT System itself - Status changed from 'new' to 'open'

From @usev6

As a status update​: 10 ** -1 is now a Rat​:

$ perl6-m -e 'say (10 ** -1).WHAT'
(Rat)

9.0 ** -1 is also a Rat, and there is no type check error​:

$ perl6-m -e 'say (9.0 ** -1).WHAT'
(Rat)

$ perl6-m -e 'say (9.0 ** -1)'
0.111111

9 ** 0.5 is a Num​:

$ perl6-m -e 'say (9 ** 0.5).WHAT'
(Num)

From @AlexDaniel

Yes, this was fixed in rakudo/rakudo@563abdd

Testneeded.

On 2015-11-12 12​:19​:37, bartolin@​gmx.de wrote​:

As a status update​: 10 ** -1 is now a Rat​:

$ perl6-m -e 'say (10 ** -1).WHAT'
(Rat)

9.0 ** -1 is also a Rat, and there is no type check error​:

$ perl6-m -e 'say (9.0 ** -1).WHAT'
(Rat)

$ perl6-m -e 'say (9.0 ** -1)'
0.111111

9 ** 0.5 is a Num​:

$ perl6-m -e 'say (9 ** 0.5).WHAT'
(Num)

From @zoffixznet

Tests in Raku/roast#346

@zoffixznet - Status changed from 'open' to 'resolved'