From @masak

<masak> rakudo​: say (1..10).sort(&rand)
<jnthn> Yes.
<p6eval> rakudo e02bc0​: OUTPUT«too many arguments passed - 0 params
expected␤in sub rand [...]
<masak> shouldn't that work?
<jnthn> rakudo​: say &rand.multi
<p6eval> rakudo e02bc0​: OUTPUT«0␤»
<jnthn> rakudo​: say &rand.signature.perl
<p6eval> rakudo e02bc0​: OUTPUT«​:(Any *@​args)␤»
<jnthn> rakudo​: say &rand.arity
<masak> jnthn​: wrong, right?
<p6eval> rakudo e02bc0​: OUTPUT«0␤»
<pmichaud> Note that slurpies are supposed to be Object, not Any
<jnthn> masak​: Well the answers can't both be right. ;-)
<jnthn> Anyway, rand is 0-ary, at least as far as Rakudo knows it.
<masak> jnthn​: right, so &sort should accept it.
<jnthn> So in your sort example, Rakudo doesn't really know what to do
with an 0-ary.
<Tene> *any* 0-ary sub passed to sort is effectively a random sort.
<Tene> so we can optimize that and just return the list in the same
order it was passed. :)
<pmichaud> well, it depends on the 0-ary.
<pmichaud> a 0-ary sub that always returns zero should end up not
modifying the sort
<jnthn> Aye.
<jnthn> True.
<jnthn> It's odd anyway. ;-)
<Tene> is Perl 6's sort guaranteed to be stable?
<masak> Tene​: aye.
<jnthn> But maybe sort could be made to understnad 0-ary blocks.
<masak> jnthn​: spec says it should.
<jnthn> ah, ok
<jnthn> Then, fix it. :-)
* masak submits rakudobug
<jnthn> Or rakudobug. ;)
<pmichaud> the spec says it supports 0-ary blocks?
<masak> pmichaud​: aye.
<pmichaud> weird
<masak> aye.

From @bbkr

On Kiev build​:

$ perl6 -e 'say (1..10).sort(&rand)'
12345678910

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

From @bbkr

tests added in
pugs/t/spec/S32-list/sort.t

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

From @masak

On Mon Jul 12 08​:07​:38 2010, bbkr wrote​:

On Kiev build​:

$ perl6 -e 'say (1..10).sort(&rand)'
12345678910

I'd argue that this is the wrong behavior. I'd like for it to behave the same way this does​:

(1..10).sort( { rand } )
9 8 5 3 10 1 2 4 6 7

Oh, and I get a different result with Rakudo 2010.06-219-g7579f7a​:

(1..10).sort( &rand )
too many arguments passed - 0 params expected

From @kyleha

This is an automatically generated mail to inform you that tests are now available in t/spec/S32-list/sort.t

commit 8a55371caf69c520e9c5272fa84b24d6a4f33682
Author​: bbkr <bbkr@​c213334d-75ef-0310-aa23-eaa082d1ae64>
Date​: Mon Jul 12 15​:16​:59 2010 +0000

  [t/spec] tests for RT #​68112 sort doesnt accept &rand (or zero-arg subs in general)
 
  git-svn-id​: http://svn.pugscode.org/pugs@&#8203;31642 c213334d-75ef-0310-aa23-eaa082d1ae64

diff --git a/t/spec/S32-list/sort.t b/t/spec/S32-list/sort.t
index 2f82da3..de6f71a 100644
--- a/t/spec/S32-list/sort.t
+++ b/t/spec/S32-list/sort.t
@@ -1,6 +1,6 @@
 use v6;
 use Test;
-plan 29;
+plan 31;
 
 # L<S32::Containers/"List"/"=item sort">
 
@@ -210,4 +210,13 @@ plan 29;
         'sort by class name';
 }
 
+# RT #68112
+{
+    sub foo () { 0 }
+    lives_ok { (1..10).sort(&foo) },
+        'sort accepts 0-arity method';
+    lives_ok { (1..10).sort(&rand) },
+        'sort accepts rand method';
+}
+
 # vim: ft=perl6

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

From @bbkr

[17​:24] <bbkr> masak​: should i revert it from test suite untill expected
behavior is determined?
[17​:25] <moritz_> bbkr​: just fudge it for now

From @kyleha

This is an automatically generated mail to inform you that tests are now available in t/spec/S32-list/sort.t

commit dda2c2b53e5d047cff3110138f5982f3e8b2270d
Author​: bbkr <bbkr@​c213334d-75ef-0310-aa23-eaa082d1ae64>
Date​: Mon Jul 12 15​:35​:39 2010 +0000

  [t/spec] tests for RT #​68112 sort doesnt accept &rand (or zero-arg subs in general) - fudged
 
  git-svn-id​: http://svn.pugscode.org/pugs@&#8203;31643 c213334d-75ef-0310-aa23-eaa082d1ae64

diff --git a/t/spec/S32-list/sort.t b/t/spec/S32-list/sort.t
index de6f71a..1e0d219 100644
--- a/t/spec/S32-list/sort.t
+++ b/t/spec/S32-list/sort.t
@@ -211,6 +211,7 @@ plan 31;
 }
 
 # RT #68112
+#?rakudo skip "determine behavior of 0-arity methods passed to sort"
 {
     sub foo () { 0 }
     lives_ok { (1..10).sort(&foo) },

From @bbkr

[17​:42] <masak> bbkr​: re "determine behavior of 0-arity methods passed
to sort" -- (1) it's not a method :) (2) I think the behavior is already
clear, at least if the spec says this should act like a Schwartzian
transform, which I think it does.
[17​:43] <masak> then the behavior is "collect all (random) values, sort
on them".
[17​:46] <bbkr> masak​: thanks, I didn't know about Schwartzian transform.
adding this info to ticket.

From @masak

On Mon Jul 12 08​:46​:21 2010, bbkr wrote​:

[17​:42] <masak> bbkr​: re "determine behavior of 0-arity methods passed
to sort" -- (1) it's not a method :) (2) I think the behavior is
already
clear, at least if the spec says this should act like a Schwartzian
transform, which I think it does.
[17​:43] <masak> then the behavior is "collect all (random) values,
sort
on them".
[17​:46] <bbkr> masak​: thanks, I didn't know about Schwartzian
transform.
adding this info to ticket.

Changing my mind on this one. use .pick(*) if you want .sort(&rand)
semantics. Wrap your nullary function in a unary or binary function if
you want something more advanced.

Rejecting ticket.

@masak - Status changed from 'open' to 'rejected'