Can't store whole 1&2 junctions in Mu-keyed hash in Rakudo

[p6rt]

Migrated from rt.perl.org#125614 (status was 'open')

Searchable as RT125614$

[p6rt]

From @masak

<jnthn> my %h{Mu}; # now you can store objects as keys
<jnthn> m​: my %xy{Mu};
<camelia> rakudo-moar dd3bcc​: ( no output )
<jnthn> Note the keys there are objects, not strings.
<moritz> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl
<camelia> rakudo-moar ccb41d​: OUTPUT«Hash[Any,Mu].new(1 => 3, 2 => 3)␤»
<moritz> bug
<jnthn> hah...I wonder if the postcircumfix sub is to blame...
* masak submits rakudobug

[p6rt]

From @lizmat

On 14 Jul 2015, at 12​:29, Carl Mäsak (via RT) <perl6-bugs-followup@​perl.org> wrote​:

# New Ticket Created by "Carl Mäsak"
# Please include the string​: [perl #​125614]
# in the subject line of all future correspondence about this issue.
# <URL​: https://rt-archive.perl.org/perl6/Ticket/Display.html?id=125614 >

<jnthn> my %h{Mu}; # now you can store objects as keys
<jnthn> m​: my %xy{Mu};
<camelia> rakudo-moar dd3bcc​: ( no output )
<jnthn> Note the keys there are objects, not strings.
<moritz> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl
<camelia> rakudo-moar ccb41d​: OUTPUT«Hash[Any,Mu].new(1 => 3, 2 => 3)␤»
<moritz> bug
<jnthn> hah...I wonder if the postcircumfix sub is to blame...
* masak submits rakudobug

Some more tests​:
[22​:14​:53] <lizmat> Looking at #​125614 now​:
[22​:15​:02] <lizmat> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl # wrong
[22​:15​:02] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(1 => 3, 2 => 3)␤»
[22​:15​:25] <lizmat> m​: my %xy{Mu}; %xy.AT-KEY(1&2) = 3; say %xy.perl # is this right?
[22​:15​:25] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:15​:43] <lizmat> m​: my %xy{Mu}; %xy.ASSIGN-KEY(1&2,3); say %xy.perl # is this right?
[22​:15​:43] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:16​:20] <lizmat> m​: my %xy{Mu}; %xy.BIND-KEY(1&2,3); say %xy.perl # and this?
[22​:16​:21] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:17​:53] <ShimmerFairy> Hm. Unless there's something in the specs stating otherwise, I'd think %xy{1&2} = 3 would make 1 => 3, 2 => 3 , since it is a Junction
[22​:17​:53] <lizmat> m​: my %xy{Any}; %xy{1&2} = 3; say %xy.perl # and this?
[22​:17​:53] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Any].new(1 => 3, 2 => 3)␤»
[22​:17​:57] <ShimmerFairy> m​: sub foo($a) { say $a }; foo(1&2)
[22​:17​:58] <+camelia> rakudo-moar 8957fc​: OUTPUT«1␤2␤»
[22​:18​:18] <ShimmerFairy> ^ that's my reason for thinking {1&2} is the one doing the right thing.
[22​:18​:58] <lizmat> not if it is a typed hash with Mu​: then the autothreading should not work (is what conception of the problem is)
[22​:20​:58] <ShimmerFairy> Oh, I see. Well, I think in that case it should be any(1, 2) => 3 , since the Junction is the key (I don't think all(1 => 3, 2 => 3) is right, and I don't see how it could be immediately)

So, assuming AT-KEY is doing the right thing, and looking at the circumfix {} code​:

multi sub postcircumfix​:<{ }>( \SELF, \key ) is rw {
  SELF.AT-KEY(key);
}

Since we bind the key value to what is being passed to AT-KEY, and AT-KEY is doing the right thing, I can only assume there is something screwy going on with “is rw”. And that is going right into the bowels, and I’m out of my league :-)

Hope this helps.

Liz

[p6rt]

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

[p6rt]

From @AlexDaniel

Still reproducible (2017.11, HEAD(5929887))

On 2015-07-14 22​:32​:47, elizabeth wrote​:

On 14 Jul 2015, at 12​:29, Carl Mäsak (via RT) <perl6-bugs-
followup@​perl.org> wrote​:

# New Ticket Created by "Carl Mäsak"
# Please include the string​: [perl #​125614]
# in the subject line of all future correspondence about this issue.
# <URL​: https://rt-archive.perl.org/perl6/Ticket/Display.html?id=125614 >

<jnthn> my %h{Mu}; # now you can store objects as keys
<jnthn> m​: my %xy{Mu};
<camelia> rakudo-moar dd3bcc​: ( no output )
<jnthn> Note the keys there are objects, not strings.
<moritz> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl
<camelia> rakudo-moar ccb41d​: OUTPUT«Hash[Any,Mu].new(1 => 3, 2 =>
3)␤»
<moritz> bug
<jnthn> hah...I wonder if the postcircumfix sub is to blame...
* masak submits rakudobug

Some more tests​:
[22​:14​:53] <lizmat> Looking at #​125614 now​:
[22​:15​:02] <lizmat> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl #
wrong
[22​:15​:02] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(1
=> 3, 2 => 3)␤»
[22​:15​:25] <lizmat> m​: my %xy{Mu}; %xy.AT-KEY(1&2) = 3; say
%xy.perl # is this right?
[22​:15​:25] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:15​:43] <lizmat> m​: my %xy{Mu}; %xy.ASSIGN-KEY(1&2,3); say
%xy.perl # is this right?
[22​:15​:43] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:16​:20] <lizmat> m​: my %xy{Mu}; %xy.BIND-KEY(1&2,3); say
%xy.perl # and this?
[22​:16​:21] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:17​:53] <ShimmerFairy> Hm. Unless there's something in the
specs stating otherwise, I'd think %xy{1&2} = 3 would make 1 => 3, 2
=> 3 , since it is a Junction
[22​:17​:53] <lizmat> m​: my %xy{Any}; %xy{1&2} = 3; say %xy.perl #
and this?
[22​:17​:53] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Any].new(1
=> 3, 2 => 3)␤»
[22​:17​:57] <ShimmerFairy> m​: sub foo($a) { say $a }; foo(1&2)
[22​:17​:58] <+camelia> rakudo-moar 8957fc​: OUTPUT«1␤2␤»
[22​:18​:18] <ShimmerFairy> ^ that's my reason for thinking {1&2}
is the one doing the right thing.
[22​:18​:58] <lizmat> not if it is a typed hash with Mu​: then the
autothreading should not work (is what conception of the problem is)
[22​:20​:58] <ShimmerFairy> Oh, I see. Well, I think in that case
it should be any(1, 2) => 3 , since the Junction is the key (I don't
think all(1 => 3, 2 => 3) is right, and I don't see how it could be
immediately)

So, assuming AT-KEY is doing the right thing, and looking at the
circumfix {} code​:

multi sub postcircumfix​:<{ }>( \SELF, \key ) is rw {
SELF.AT-KEY(key);
}

Since we bind the key value to what is being passed to AT-KEY, and AT-
KEY is doing the right thing, I can only assume there is something
screwy going on with “is rw”. And that is going right into the
bowels, and I’m out of my league :-)

Hope this helps.

Liz

[p6rt]

From @zoffixznet

On Tue, 14 Jul 2015 22​:32​:47 -0700, elizabeth wrote​:

On 14 Jul 2015, at 12​:29, Carl Mäsak (via RT) <perl6-bugs-
followup@​perl.org> wrote​:

# New Ticket Created by "Carl Mäsak"
# Please include the string​: [perl #​125614]
# in the subject line of all future correspondence about this issue.
# <URL​: https://rt-archive.perl.org/perl6/Ticket/Display.html?id=125614 >

<jnthn> my %h{Mu}; # now you can store objects as keys
<jnthn> m​: my %xy{Mu};
<camelia> rakudo-moar dd3bcc​: ( no output )
<jnthn> Note the keys there are objects, not strings.
<moritz> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl
<camelia> rakudo-moar ccb41d​: OUTPUT«Hash[Any,Mu].new(1 => 3, 2 =>
3)␤»
<moritz> bug
<jnthn> hah...I wonder if the postcircumfix sub is to blame...
* masak submits rakudobug

Some more tests​:
[22​:14​:53] <lizmat> Looking at #​125614 now​:
[22​:15​:02] <lizmat> m​: my %xy{Mu}; %xy{1&2} = 3; say %xy.perl #
wrong
[22​:15​:02] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Mu].new(1
=> 3, 2 => 3)␤»
[22​:15​:25] <lizmat> m​: my %xy{Mu}; %xy.AT-KEY(1&2) = 3; say
%xy.perl # is this right?
[22​:15​:25] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:15​:43] <lizmat> m​: my %xy{Mu}; %xy.ASSIGN-KEY(1&2,3); say
%xy.perl # is this right?
[22​:15​:43] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:16​:20] <lizmat> m​: my %xy{Mu}; %xy.BIND-KEY(1&2,3); say
%xy.perl # and this?
[22​:16​:21] <+camelia> rakudo-moar 8957fc​:
OUTPUT«Hash[Any,Mu].new(all(1 => 3, 2 => 3))␤»
[22​:17​:53] <ShimmerFairy> Hm. Unless there's something in the
specs stating otherwise, I'd think %xy{1&2} = 3 would make 1 => 3, 2
=> 3 , since it is a Junction
[22​:17​:53] <lizmat> m​: my %xy{Any}; %xy{1&2} = 3; say %xy.perl #
and this?
[22​:17​:53] <+camelia> rakudo-moar 8957fc​: OUTPUT«Hash[Any,Any].new(1
=> 3, 2 => 3)␤»
[22​:17​:57] <ShimmerFairy> m​: sub foo($a) { say $a }; foo(1&2)
[22​:17​:58] <+camelia> rakudo-moar 8957fc​: OUTPUT«1␤2␤»
[22​:18​:18] <ShimmerFairy> ^ that's my reason for thinking {1&2}
is the one doing the right thing.
[22​:18​:58] <lizmat> not if it is a typed hash with Mu​: then the
autothreading should not work (is what conception of the problem is)
[22​:20​:58] <ShimmerFairy> Oh, I see. Well, I think in that case
it should be any(1, 2) => 3 , since the Junction is the key (I don't
think all(1 => 3, 2 => 3) is right, and I don't see how it could be
immediately)

So, assuming AT-KEY is doing the right thing, and looking at the
circumfix {} code​:

multi sub postcircumfix​:<{ }>( \SELF, \key ) is rw {
SELF.AT-KEY(key);
}

Since we bind the key value to what is being passed to AT-KEY, and AT-
KEY is doing the right thing, I can only assume there is something
screwy going on with “is rw”. And that is going right into the
bowels, and I’m out of my league :-)

Hope this helps.

Liz

I managed to fix[^1] this, though almost immediately using Junctions as keys caused
Map.sort to blow up. I fixed[^2] that too, but not sure what other crucial methods
explode due to Junction keys.

I aborted spectest when multiple failures occurred; not due to my fix, but rather due
to RT#​128905[^3]. I've no idea how to fix that ticket, so I've stashed my work for this
ticket into fix-Hash-Mu-keys--RT-1357201 branch[^4] for now. It still needs a full
spectest run before merge, once RT#​128905 is fixed.

[1] rakudo/rakudo@a5cf61c
[2] rakudo/rakudo@a5cf61c
[3] https://rt-archive.perl.org/perl6/Ticket/Display.html?id=128905#txn-1516048
[4] https://github.com/rakudo/rakudo/tree/fix-Hash-Mu-keys--RT-1357201