On Thu, 26 Sep 2002, John Williams wrote:
: I'm trying to write a revised operator precedence table for perl6,
: similar to the one in perlop.pod.
:
: This is what I have come up with based on Apocalypse 3 and Exegesis 3.
: Does anyone have comments? I'm not sure if the precedence
: for : (adverb) or 'is' and 'but' are quite right.
Me either.
: perl6 operator precedence
:
: left terms and list operators (leftward) [] {} () quotes
: left . and unary .
Unary . can't be left associative. Perhaps unary . is nonassoc like ++.
: nonassoc ++ --
: left is but
Should probably be down about where .. is, maybe same as.
: right **
: right ! \ and unary ~ + - * _
Probably unary ? goes here too.
: left =~ !~
Probably should be the same as all the other comparison ops.
: left * / % x
And & maybe.
: left + - _
And | maybe.
: left << >>
Can be argued these are really just multiplicative. Or even exponential.
: right named unary operators, -X
: left < > <= >= lt gt le ge == != <=> eq ne cmp
: left &
: left | ~
Not sure what to do with ~ if we move & and |. Probably just follows |.
: left &&
: left || ~~ //
: nonassoc .. ...
Maybe "but" here. Maybe "is" too, though compile-time declarations
don't necessarily have the same syntactic constraints as ordinary
expressions.
: right ??::
: right = := **= += -= _= *= /= %= x= &= |= ~=
: <<= >>= &&= ||= ~~= //=
: left , =>
=> is no longer a comma. I think it has to be tighter than , now,
or we can't say
a => "a", b => "b"
Perhaps it goes in as another nonassoc .. operator.
: left ;
: left :
While semicolon is definitely looser than comma, it's not clear where
colon should go yet.
: nonassoc list operators (rightward)
: right not
I've also considered unifying C<not> with the list operators, which are
actually right associative, or you couldn't say
print sort 1,2,3;
On the other hand, it would be strange for C<not> to give its right
side a list context.
: left and
: left or xor err
Those are fine. :-)
Larry
