From @mscha

It's fairly rare to encounter a Rat that isn't normalized, but if you do
get one, norm() doesn't normalize the numerator/denominator.
nude() does, though, and actually changes the Rat (which is supposed to
be immutable).

Example​:

my $f = 1/13² + 1/26² + 1/39² + 1/78²
0.008218
say $f.numerator, '/', $f.denominator;
50/6084
my $g = $f.norm;
0.008218
say $g.numerator, '/', $g.denominator;
50/6084
say $f.nude.join('/');
25/3042
say $f.numerator, '/', $f.denominator;
25/3042
say $g.numerator, '/', $g.denominator;
25/3042

Even stranger, $g is now normalized as well.

This is on Rakudo Star 2017.01, 64-bit Linux.

% perl6 --version
This is Rakudo version 2017.01 built on MoarVM version 2017.01
implementing Perl 6.c.

From @mscha

Actually, it's not as rare as I thought​: the same thing happens when you do​:

my $f = 1/6 + 1/6

From @zoffixznet

The avoidance of reduction is for optimization purposes.

The .norm stuff is fixed now in rakudo/rakudo@aac9efcbda and tested in Raku/roast@7d0daf5286

However, there's a data-race in Rational.REDUCE-ME and it needs to go​: https://irclog.perlgeek.de/perl6-dev/2017-02-13#i_14093035

Renaming the ticket to that.

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

From @zoffixznet

Another issue the .REDUCE-ME thing causes​:

mscha │ m​: say (1/2+1/2, 2/2).unique;
+camelia │ rakudo-moar 9da50e​: OUTPUT​: «(1 1)␤»

infix​:<+> does not call .REDUCE-ME so the rats are different in this case

From @skids

On Mon, 06 Mar 2017 05​:48​:07 -0800, cpan@​zoffix.com wrote​:

Another issue the .REDUCE-ME thing causes​:

mscha │ m​: say (1/2+1/2, 2/2).unique;
+camelia │ rakudo-moar 9da50e​: OUTPUT​: «(1 1)␤»

infix​:<+> does not call .REDUCE-ME so the rats are different in this case

I could see where a mutable Rational could be useful. Should we parameterize that
role in-core to provide immutable (for Rat and FatRat) and mutable versions for module
authors? Or should we kick all mutable functionality out into module space and
have module authors provide their own role variant... maybe just adding a single
method to Rat/FatRat to allow introspection that they are immutable?

Naturally any mutable Rational would come with the caveat that it is not threadsafe.

Also, what behavior are we looking for Rat -- just "immutabilty" or "value-type"?

From @AlexDaniel

The issue with .unique was resolved in these commits

Rakudo​: rakudo/rakudo@8cd70d1
Roast​: Raku/roast@ad38801
Roast​: Raku/roast@032ce8d

On 2017-03-06 05​:48​:07, cpan@​zoffix.com wrote​:

Another issue the .REDUCE-ME thing causes​:

mscha │ m​: say (1/2+1/2, 2/2).unique;
+camelia │ rakudo-moar 9da50e​: OUTPUT​: «(1 1)␤»

infix​:<+> does not call .REDUCE-ME so the rats are different in this case

From @zoffixznet

Data race is now fixed (in a post release branch)​:

Rakudo fix​: rakudo/rakudo@6dd2058
Test​: Raku/roast@1d10e9dc12

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