CODEWITHSUNDEEP
perlslurp
jja
jja

Reputation: 2098

Why can I slurp a file by assigning the result of angle brackets to the input record separator?

I know I can slurp a file by setting the input record separator ($/) to an undefined value, like

open my $fh, '<', $filename or die "Cannot read from $file: $!";
my $contents = do { local $/; <$fh> };

But recently I came across a very similar but different idiom:

open my $fh, '<', $filename or die "Cannot read from $file: $!";
my $contents = do { local $/ = <$fh> };

(Note the local $/ = <$fh> instead of local $/; <$fh>).

Both of these work, and there are examples on CPAN of both the variant with the assignment and the one without (although the latter is not surprisingly, much more common).

But my question is why does it work? What is the variant with the assignment doing?

PS: I also know that I should be using eg. File::Slurper to slurp files, but life is funny like that sometimes.

Upvotes: 6

Views: 132

Answers (1)

ikegami
ikegami

Reputation: 385645

This is the result of an undocumented optimization on which you shouldn't rely.

Normally LHS = RHS is evaluated as follows:

  1. RHS is evaluated.
  2. LHS is evaluated.
  3. Assignment is evaluated.

As you can see, the right-hand side of the assignment is evaluated first[1]. This allows the following to work:

my $x = 123;

{
   my $x = $x * 2;
   say $x;  # 456
}

say $x;  # 123

Obviously, something different —and undocumented— is happening in your case. That's because LHS = <$fh> is special. Rather than reading from the file then assigning the result to the left-hand side, readline (<>) writes directly to the result of the left-hand side of the assignment.[2]

  1. LHS is evaluated. (This backs up $/ and sets it to undef in your case.)
  2. $fh is evaluated.
  3. readline is evaluated, writing directly to the result of the left-hand side of the assignment.

No assignment is performed.

This optimization is undocumented, and you shouldn't rely on it.

local $/ = uc(<$fh>) wouldn't work, for example.

  1. The compiled code has the right-hand side evaluated first:

    $ perl -MO=Concise,-exec -e'$L = $R'
1  <0> enter
2  <;> nextstate(main 1 -e:1) v:{
3  <#> gvsv[*R] s                   <- $R
4  <#> gvsv[*L] s                   <- $L
5  <2> sassign vKS/2                <- =
6  <@> leave[1 ref] vKP/REFC
-e syntax OK

    The following shows the right-hand side evaluated first:

    $ perl -e'sub f :lvalue { CORE::say $_[0]; $x } f("L") = f("R")'
R
L

  2. $x = uc(<>) evaluates uc(<>) before $x, then performs an assignment:

    $ perl -MO=Concise,-exec -e'$x = uc(<>)'
1  <0> enter
2  <;> nextstate(main 1 -e:1) v:{
3  <#> gv[*ARGV] s                  \
4  <1> readline[t3] sK/1             > RHS
5  <1> uc[t4] sK/1                  /
6  <#> gvsv[*x] s                   -> LHS
7  <2> sassign vKS/2
8  <@> leave[1 ref] vKP/REFC
-e syntax OK

    $x = uc(<>) evaluates $x before <>, and it doesn't perform an assignment:

    $ perl -MO=Concise,-exec -e'$x = <>'
1  <0> enter
2  <;> nextstate(main 1 -e:1) v:{
3  <#> gvsv[*x] s                   -> LHS
4  <#> gv[*ARGV] s                  \  RHS
5  <1> readline[t3] sKS/1           /
6  <@> leave[1 ref] vKP/REFC
-e syntax OK

    Note the (uppercase) S next to readline that wasn't there before. This "special" flag is what tells readline to write to $x.

    Adding local doesn't change anything.

    $ perl -MO=Concise,-exec -e'local $x = <>'
1  <0> enter
2  <;> nextstate(main 1 -e:1) v:{
3  <#> gvsv[*x] s/LVINTRO
4  <#> gv[*ARGV] s
5  <1> readline[t3] sKS/1
6  <@> leave[1 ref] vKP/REFC
-e syntax OK

Upvotes: 4

Related Questions