On Tue, Jun 06, 2006 at 02:06:05AM +0100, Jonathan Worthington wrote:
: "Sage La Torra" <[EMAIL PROTECTED]> wrote:
: >I've got a design document detailing the first few translations I'll be
: >handling, and I'd greatly appreciate feedback and advice.
: >
: I may be off base here, or in the alternate pointing out a corner case
: that's of little significance, but:
:
: -Compound statements: if($foo eq "a" or $foo eq "b" or $foo eq "c") {...}
: -> if $foo eq "a"|"b"|"c" {...}
:
: Is there an issue if $foo is tied here, in that (unless the a or b
: condition match) $foo will be evaluated three times; for testing against a
: junction is it feasible that the compiler can say "aha, I'll only evaluate
: $foo once and test it three times"? Reading $foo may have some real world
: visible effects and the semantics would change. The P6 spec may nail this
: down, but I haven't time to go look (and I'll surely not have time before I
: forgot this thought completely...)
Yes, but I think the translator can assume that most scalar variables
aren't tied, especially if there isn't a tie anywhere else in the file.
Could have a user option to pessimize that, and it should probably be logged
as one of the places the translator guessed. But unlimited tying is
one of the problems with Perl 5 that Perl 6 is trying to control, so it's
sort of natural that the translator would run into this here and there.
Another minor difference is that "or" guarantees an ordered
short-circuit while "|" may short-circuit in any order the compiler
pleases. Again, unlikely to be a problem for the variable in the
absence of ties, but if some of the value calculations that can blow
up are guarded by "or" conditions, then you've got a problem.
We'll definitely need to provide some knobs for the user to turn to
control how much refactoring is attempted.
Larry
