... for (@_) x= 2;

[rsFalse]

Hello,

I wrote codes, which seems for me equivalent, but their behaviour differ. Any ideas, why they differ? ...I can't catch.

#!/usr/bin/perl

use warnings;
use strict;

$\ = $/;

@_ = ('a' .. 'b');
print for (@_) x= 2;

OUTPUT:

Can't modify array dereference in repeat (x) ...
Execution of arrX2.pl aborted due to compilation errors.

#!/usr/bin/perl

use warnings;
use strict;

$\ = $/;

print for (@_ = ('a' .. 'b')) x= 2;

OUTPUT:

2
2

UPD. Added:

#!/usr/bin/perl

use warnings;
use strict;

$\ = $/;

(@_ = ('a' .. 'b')) x= 2;
print for @_;

OUTPUT:

a
b

P.S. parentheses around 'a' .. 'b' are redundant.

[1nickt]

x= , the repetition assignment operator, only works on scalars.

So you can do:

my $str = 'ab';
$str x= 2;
print $str;
# 'abab'

You don't say what you want as output, but maybe you are looking for:

#!/usr/bin/perl

use warnings;
use strict;

$\ = $/;

@_ =  ( 'a' .. 'b' ) x 2;
print for @_;

Output:

a
b
a
b

edit: show example of repetition assignment on a scalar

[rsFalse]

Yes, firstly, I wanted so for an output, but later I found that behaviours are interesting with using 'x=' operator :) . I experimented if I can multiply the array in 'for' sentence with using 'x='.

[jeffa]

Just in case this is not obvious, map tends to produce a more consistent solution in these situations:

perl -le'print map $_ x=2, ("a" .. "f")'

[GrandFather]

I suspect the key is that x= only operates on scalars. So (@_ = ('a' .. 'b')) is in scalar context and becomes the scalar value 2 which x then make two of.

[LanX]

ok but ...

> which x then make two of.

.. it's not an x but a combined assignment x= and it's tricky to see what exactly the target of this assignment is.

[LanX]

I had problems to understand what your goal is, this code already looks broken.

It's true that x has a double nature (repeat string vs repeat list) like .. depending on context (flip flop vs range)

But "combined assignments" like += only allow scalars on the LHS...

See perlop

> These combined assignment operators can only operate on scalars, whereas the ordinary assignment operator can assign to arrays, hashes, lists and even references.

update

Besides you don't need any assignment

print for @_ x 2 should already work perfectly.

update

correction print for (@_) x 2 should already work perfectly.

  DB<100> @_=1..3
 => (1, 2, 3)

  DB<101> print for @_ x 2
 => ""
33
  DB<102> print for (@_) x 2
 => ""
123123

from perlop

> In list context, if the left operand is enclosed in parentheses or is a list formed by qw/STRING/, it repeats the list.

[rsFalse]

I supposed that three of my codes looks equivalently (for me), so and output should be the same - e.g. warning + compilation error.

[LanX]

the list assignment returns the length of the list in scalar context, so

   DB<114> $l = ( @a = qw/a b/ )
 => 2

So it's not the same code!

The difference between your code and my example is that there is no explicit variable (like $l) used here, but somehow it's still possible to change it.

Maybe it's a side effect of for ? I dunno.

I'd prefer a Can't modify warning here too!

But it's still a very peculiar construction, so I'm not surprised if this edge case wasn't covered.

HTH!

update

More insights:

... it's not the temporary scalar which is modified but the resulting list in brackets

  DB<121> print "$_\n" for ($a=666) x= 2
666
666
 => ""

  DB<122> $a
 => 666

According to the already cited documentation of "combined assignments" this shouldn't be possible, b/c its a list operation.

[rsFalse]

LanX cited: "> In list context, if the left operand is enclosed in parentheses or is a list formed by qw/STRING/, it repeats the list."

Now I tried to write some lines with 'qw' and found next interesting situation, which I can't understand:


#!/usr/bin/perl

use warnings;
use strict;

$\ = $/;
$, = '+';

print qw/a b/ x 2;
print scalar(qw/a b/ x 2);
print for (qw/a b/ x 2) x 2;
print for qw/a b/ x 2 x 2;

OUTPUTS:

a+b+a+b

bb

a
b
a
b
a
b
a
b

bbbb

The strange for me is the output of second 'print' (of course so does 4th too): it prints "bb" instead of supposed "b". I don't catch that, because print scalar('a','b','a','b'); would print "b".

[LanX]

the documentation says that x can only act like a list repeater in list context.

> > In list context, if the left operand is enclosed in parentheses or is a list formed by qw/STRING/, it repeats the list.

[Anonymous Monk]

simpler:

$ perl -E '$x = "foo"; say for $x x= 2'

that compiles to (only relevant part):

BINOP (0x189a938) repeat [1] 
    UNOP (0x189a9c0) null [15] 
         SVOP (0x189aa00) gvsv  GV (0x197dac8) *x 
    SVOP (0x189a980) const  IV (0x197dae0) 2

while

$ perl -E '$x = "foo"; say for ($x) x= 2'

is actually

BINOP (0xecd938) repeat [1] 
    UNOP (0xecd8f0) null [158] 
        OP (0xecd8b8) pushmark 
        UNOP (0xecd9c0) null [15] 
            SVOP (0xecda00) gvsv  GV (0xfb0ac8) *x 
    SVOP (0xecd980) const  IV (0xfb0ae0) 2 

and

$ perl -E '@x = (@_ = qw( a b )) x= 2'

is

BINOP (0x14b79b8) repeat [4] 
    UNOP (0x14b7970) null [158] 
        OP (0x14b7938) pushmark 
        BINOP (0x14b7a40) aassign [3] 
            UNOP (0x14b7b40) null [158] 
                OP (0x14b7b08) pushmark 
                SVOP (0x13ffc38) const  PV (0x159ab28) "a" 
                SVOP (0x13ffbe0) const  PV (0x159ab70) "b" 
            UNOP (0x14b7ac0) null [158] 
                OP (0x14b7a88) pushmark 
                UNOP (0x13ffc78) rv2av [2] 
                    SVOP (0x13ffcb8) gv  GV (0x13de0d0) *_ 
    SVOP (0x14b7a00) const  IV (0x159ab58) 2

So pretty much like GrandFather said. Hopefully that answers your question. I don't think it's documented anywhere; see toke.c and perly.y if you want more details. Certainly this behaviour is subject to change without notice.