On Sun, Sep 14, 2008 at 09:13:33PM +1000, Vasily Chekalkin wrote:
> Ilya Belikin wrote:
>>
>> Hi,
>> look like bug (in r31017):
>>
>> my $k = 'bar';
>> my %s;
>> say defined $k; # 1
>> say defined %s{$k}; # 0
>> say defined $k and defined %s{$k}; # 1 (!) must be 0
>> say 1 && 0; # 0
>
> I replyed in ticket, but it doesn't hit mailinglist. This is bug in
> parsing. "say defined $k and defined %s{$k};" parsed as "(say defined
> $k) and defined %s{$k};"
As Carl correctly noted, Rakudo has the parsing for infix:<and> correct --
infix:<and> has looser precedence than listops (including 'say'), so
Rakudo is doing the right thing here.
The intended version is likely:
say defined $k && defined %s{$k};
However, as of r31017 Rakudo did not have 'defined' as a named
unary, so parsed this version as
say( defined( $k && defined( %s{$k} ) ) );
which is also not what is wanted. In r31116 I've changed 'defined'
to be a named unary, and we now get the desired output:
$ cat x
my $k = 'bar';
my %s;
say defined $k; # 1
say defined %s{$k}; # 0
say defined $k and defined %s{$k}; # 1
say defined $k && defined %s{$k}; # 0
$ ./parrot perl6.pbc x
1
0
1
0
We probably need to test that 'defined' is a named unary in t/spec.
Closing ticket, thanks.
Pm
