if (my $result = some_func()) {
    # do something with $result
}
[download]

I use this construct all the time, and am not aware of any problems with it. The conditional is testing the return value of the assignment, which is true if some_func() returns something true. If it's false, then $result gets whatever false value was returned and the block does not execute (or the compiler may optimize away the assignment completely -- I don't know how that works.)

Off hand, I can't think of any reason why this would not also be allowed in Perl 6.

friedo,
Off hand, I can't think of any reason why this would not also be allowed in Perl 6.

In perl 5, we write:

for my $name (keys %people) {
    # do something with $name
}
[download]

The $name lexical is implicitly scoped to the for block. This is not the case in perl 6. The way perl 6 solves this is by introducing parameters to the block so you achieve the same thing in a cleaner way (from a design perspective).

Cheers - L~R

Eh? Going away? What's the corresponding Perl 6 construct and how/why is it considered to be "cleaner"?

From Synopsis 4:

if    testa() -> $a { say $a }
elsif testb() -> $b { say $b }
else          -> $b { say $b }
[download]

I think you have eaten too much Christmas pudding :) You seem to be thinking the following constructs are equivalent:

• my $foo = 0
• my $foo if 0

These constructs are hardly equivalent and there is no ambiguity that needs to be cleaned up, regardless of whether they happen to appear within a conditional.

A lexical declared within an if is not prone to the my $foo if 0 persistency problem, period. Consider the construct

assuming some_func() is some sort of cyclic iterator that returns a positive number, then 0, then undef, the if branch will be taken only when $result contains a positive number. If $result contains 0 or undef, the conditional will evaluate to false, and the else branch (if any) will be taken instead.

At no point will $result retain its previous contents. Even if it did at a purely implementational level, you're initialising it with a value via the assignment of the results of some_funct(), and so you're safe.

grinder,
No, I assure you I know they are different and completely understand why postfix is wrong. Please read my response to dragonchild. My questions are specific to this construct.

The reason we don't do:

if (some_func()) {
    my $val = some_func();
}
[download]

Is because some_func may change return values in subsequent invocations. The reason why we don't do:

my $val = some_func();
if ($val) {
    # ... 
}
[download]

Is because we want $val scoped as small as possible.

This only works because of the implicit scoping in perl 5 which is going away in perl 6 (at least in for blocks). I am wondering if there is an alternative way of writing this that accomplishes both things.

Cheers - L~R

if you want to avoid:

if (my $val = some_func()) {
    # ...
}
[download]

and keep as small a scope as possible, how about:

{
    my $val = some_func();
    if ($val) {
        # ... 
    }
}
[download]



John

See Re^2: Duh. 'my' scope in if else blocks.. If you still want that, then you just put an explicit scoping block around the area where the variable is used, of course. I'd probably put the assignment outside of the conditional as well.

dragonchild,
Yes, I know they are different constructs (I said as much). Was giving background information into why I was asking the question confusing? If so, I am only interested in the construct in the thread title and the questions I asked about it. I know I must sound like a broken record by now, I have a newborn and a 2 year old at home and I haven't had much sleep lately.

Cheers - L~R

I'm confused. if BLOCK already creates a scope (in 5.8, at least)

>perl -c -Mstrict -we"if (my $r = f()) { g() }  print($r);"
Global symbol "$r" requires explicit package name at -e line 1.
-e had compilation errors.
[download]

Note that the scope is around the entire statement, so any elsif clauses, "then" blocks, elsif blocks and else blocks will have access to the variable.

>perl -c -Mstrict -we"if (my $r = f1()) { g1() } elsif (my $r = f2()) 
+{ g2() }"
"my" variable $result masks earlier declaration in same scope at -e li
+ne 1.
-e syntax OK
[download]

Update: Nevermind, I understand the question now. Like someone's already said, there's nothing stopping you from creating an explicit block.

>perl -c -Mstrict -we"{ my $r = f(); if ($r) { g() } }  print($r);"
Global symbol "$r" requires explicit package name at -e line 1.
-e had compilation errors.
[download]

Caveat: last, next and redo will ignore the implicit block but will treat the explicit bare block as a loop.

>perl -le"sub f { 1 }  for (1..2) { if (my $r = f()) { next }  print '
+!' }"

>perl -le"sub f { 1 }  for (1..2) { { my $r = f(); if ($r) { next } } 
+ print '!' }"
!
!
[download]

I would also add that the idiom could be useful even when the function is pure, but heavy on CPU. Just as a form of self-baked cache.

polettix,
Thanks! I tried hard to find this thread when I posted this because I remembered it. Unfortunately, I misremembered TheDamian being the one providing the solution and thought he was the one asking the question. Using Super Search to find posts by TheDamian asking a question (not a reply) didn't help and so I gave up.

Cheers - L~R

I've just found some code on my scratchpad which shows that apart from postfix modifiers, there's at least one other way where scoping rules get crazy. The following code behaves differently on perl 5.10.0 from 5.8.8.

perl -we '@a = qw"a b"; while (my $x = pop @a) { push @x, \$x; print \
+$x, ": ", ($x || 0), "\n"; redo if 0 != @x % 3; }; '
[download]

Note: I don't understand what happens in either case. I don't even really understand the simple my $x if 0 case.

For the quote challenged

perl -we " @a = qw!a b!;while (my $x = pop @a) { push @x, \$x; print \
+$x, q!: !, ($x || 0), qq!\n!; redo if 0 != @x % 3; };
[download]

I believe the old behaviour is a bug, because $x isn't supposed to change