John W. Krahn <[EMAIL PROTECTED]> wrote:
> Kevin Meltzer wrote:
[ back and forth ]
>> >>> > This is actually a bug. It just seems that nobody seems to
>> >>> > care :) It would break too many JAPHs which use this.
>> >>> >
>> >>> > So, don't depend on it, in case it is ever fixed.
>> >>>
>> >>> Can you cite a reference to this behavior described as a bug?
>> >>
>> >> Does that behavior not seem like a bug to you?
>> >
>> > Not necessarily. :-)
>>
>> I'm curious as to why. When I mentioned it on channel, a few people
>> didn't see it as a bug either, at first. Being that it is using --
>> in a way which isn't consistent with -- (it increments as opposed to
>> decrement). In fact, it isn't just with --/++ but + and - will yield
>> the same results. Anyways, just curious why you think that subtracting
>> from 0 yields a 1 doesn't seem like a bug (and when adding 1 never
>> yields a 0).
>
> Well, because Perl has lots of special cases like this. Most people
> don't ever use $| let alone the special properties of $|--. The average
> programmer just needs to know that setting $| to 1 turns on autoflush
> and setting $| to 0 turns off autoflush. What about the fact that ++
> will increment a string but -- will not decrement it?
My 2p:
$| is a boolean. It's an alias to a *bit*.
And I'm not sure what the best semantic for $bool--
would be, but two possibilities come to mind:
- make it falser
This might be kind of nice, actually.
$|-- would always turn autoflushing off.
- treat it as zero and decrement it
This is what it does now.
"But that's not what it does now!"
Sure it is:
Any non-zero number is "true", so decrementing zero
has to yield a "true" value.
$x = 0; # false
$x--; # true
And just so with $|:
$| = 0; # false
$|--; # true
If you ignore what the numeric value is afterward,
this does exactly what it should.
Obviously, the fact that the numeric value is *one*
afterward is a little jarring. But maybe it's best
if you don't think of $| as ever being zero or one.
Think of it as a "pure" boolean, which gets
temporarily coerced to zero or one when you use
it in a non-boolean context.
And anyway...
$^W behaves the same way so it can't be a bug. :)
--
Steve
perldoc -qa.j | perl -lpe '($_)=m("(.*)")'
--
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
