And, indeed, B::Deparse shows it to be so:

Win8 Strawberry 5.8.9.5 (32)  Thu 10/14/2021 17:49:15
C:\@Work\Perl\monks
>perl -Mstrict -Mwarnings -MO=Deparse,-p

my ($u, $v) = split /,/, 'U,V';
print "'$u' '$v' \n";

(my $w, my $x) = split /,/, 'W,X';
print "'$w' '$x' \n";

^Z
use warnings;
use strict 'refs';
(my($u, $v) = split(/,/, 'U,V', 3));
print("'${u}' '${v}' \n");
(my($w, $x) = split(/,/, 'W,X', 3));
print("'${w}' '${x}' \n");
- syntax OK

Win8 Strawberry 5.8.9.5 (32)  Thu 10/14/2021 18:54:22
C:\@Work\Perl\monks
>perl -Mstrict -Mwarnings

my ($u, $v) = split /,/, 'U,V';
print "'$u' '$v' \n";

(my $w, my $x) = split /,/, 'W,X';
print "'$w' '$x' \n";

^Z
'U' 'V'
'W' 'X'
[download]

B::Deparse is occasionally wrong, it has to guess which code was optimized to what it finally sees.

But the op-tree is more reliable

D:\tmp\pm>perl -Mstrict -Mwarnings -MO=Concise
my ($u, $v) = split /,/, 'U,V';
__END__
9  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter v ->2
2     <;> nextstate(main 1 -:1) v:*,&,{,x*,x&,x$,$ ->3
8     <2> aassign[t4] vKS ->9
-        <1> ex-list lK ->7
3           <0> pushmark s ->4
6           </> split(/","/)[t3] lK/IMPLIM ->7
4              <$> const[PV "U,V"] s ->5
5              <$> const[IV 3] s ->6
-        <1> ex-list lKPRM* ->8
7           <0> padrange[$u:1,2; $v:1,2] RM/LVINTRO,range=2 ->8
-           <0> padsv[$u:1,2] sRM*/LVINTRO ->-
-           <0> padsv[$v:1,2] sRM*/LVINTRO ->-
- syntax OK

D:\tmp\pm>perl -Mstrict -Mwarnings -MO=Concise
(my $u, my $v) = split /,/, 'U,V';
__END__
9  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter v ->2
2     <;> nextstate(main 1 -:1) v:*,&,{,x*,x&,x$,$ ->3
8     <2> aassign[t4] vKS ->9
-        <1> ex-list lK ->7
3           <0> pushmark s ->4
6           </> split(/","/)[t3] lK/IMPLIM ->7
4              <$> const[PV "U,V"] s ->5
5              <$> const[IV 3] s ->6
-        <1> ex-list lKPRM* ->8
7           <0> padrange[$u:1,2; $v:1,2] RM/LVINTRO,range=2 ->8
-           <0> padsv[$u:1,2] sRM*/LVINTRO ->-
-           <0> padsv[$v:1,2] sRM*/LVINTRO ->-
- syntax OK

D:\tmp\pm>
[download]

Now, spot the difference! ;-)

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}

Now, spot the difference! ;-)

$ diff -u \
   <( perl -Mstrict -Mwarnings -MO=Concise -e'my ($u, $v)    = split /
+,/, "U,V";' 2>&1 ) \
   <( perl -Mstrict -Mwarnings -MO=Concise -e'(my $u, my $v) = split /
+,/, "U,V";' 2>&1 ) \
&& echo same
same
[download]

PS, I usually use -MO=Concise,-exec

1  <0> enter
2  <;> nextstate(main 1 -e:1) v:*,&,{,x*,x&,x$,$
3  <0> pushmark s
4  <$> const[PV "U,V"] s
5  <$> const[IV 3] s
6  </> split(/","/)[t3] lK/IMPLIM
7  <0> padrange[$u:1,2; $v:1,2] RM/LVINTRO,range=2
8  <2> aassign[t4] vKS
9  <@> leave[1 ref] vKP/REFC
-e syntax OK
[download]

it's explicitly documented

I had looked at the docs for my and didn't find it very explicit at all. It says "If more than one variable is listed, the list must be placed in parentheses" but makes no reference to having the my statement inside the parenthesis. Or am I missing something subtle in the documentation?

This declaration seems unusual and I don't recall seeing it anywhere else which is why it stood out.

this

> "If more than one variable is listed, the list must be placed in parentheses"

means:

If VARLIST is

$x,$y,$z

you must write

my ($x,$y,$z)

The alternative

my $x,$y,$z

would only declare $x because of precedence, like

(my $x),$y,$z

Hence an alternative to what we want is

my $x,my $y,my $z

But since we need a list assignment on the LHS, we still need to put it all into parens

(my $x,my $y,my $z) = split ...

because this

my $x,my $y,my $z = split ...

would only assign to $z because of precedence°, like

(my $x),(my $y),(my $z = split ...)

HTH! :)

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}

Errata

s/declare/assign to/ choroba++

UPDATES

°) Even worse, it would only be a scalar assignment to $z, i.e. the number of possible splits.