Perl无障碍评估

Question

I am learning perl eval. I understand how to use eval BLOCK, but I have came across the code below. What is the code below doing?

while(<>) {
    eval; 
    warn $@ if $@;
}

Answer 1

while(<>) {

This reads input, and places it in the variable $_ . The input used by <> is first @ARGV (if you called your script with arguments), then STDIN (standard input).

Information on the diamond operator here .

eval; 

This evaluates the line that was read, since not specifying what to evaluate looks at $_ .

warn $@ if $@;

This line will display the warnings that appear in $@ , if there are any.

Answer 2

Perl's eval() builtin can take either a BLOCK or an EXPR. If it is given an EXPR that EXPR will be evaluated as a string which contains Perl code to be executed.

For example:

#!/usr/bin/env perl

use strict;
use warnings;
use feature 'say';

eval { say "Hello, Block World!"; };

eval 'say "Hello, String World!";';

This code executes both say() s as you would expect.

$ ./evals.pl
Hello, Block World!
Hello, String World!

In general the string version of eval() is considered dangerous, especially if you allow interpolation into that string based on variables that are coming from outside your control. For exmaple:

#!/usr/bin/env perl

use strict;
use warnings;
use feature 'say';

my $name = $ARGV[0] // 'World';

eval "say 'Hello, $name';";

This code is safe if called as so:

$ ./evals.pl Kaoru
Hello, Alex

$ ./evals.pl
Hello, World

But would be very dangerous if the user called it as:

$ ./evals.pl "Kaoru'; system 'rm -rf /"

On the other hand, in string eval() 's favour, it can be very useful as the opposite of Data::Dumper::Dumper() for turning dumped Perl code back into Perl-internal data structures. For example:

#!/usr/bin/env perl

use strict;
use warnings;
use feature 'say';

use Data::Dumper;

my $hashref = { a => 1, b => 2, c => 3 };

print Dumper $hashref;

my $VAR1;
my $hashref_copy = eval Dumper $hashref;

say $hashref_copy->{b};

Which, as you would expect, outputs:

$ ./evals.pl
$VAR1 = {
        'c' => 3,
        'b' => 2,
        'a' => 1
        };
2

See perldoc -f eval or http://perldoc.perl.org/functions/eval.html for more details.

As of Perl 5.16.3, there is also an evalbytes() which treats the string as a byte string rather than a character string. See perldoc -f perlunicode or http://perldoc.perl.org/perlunicode.html for more details on the difference between character strings and byte strings.

The code which you asked about explicitly:

while(<>) {
    eval; 
    warn $@ if $@;
}

Is reading in each line of either STDIN or the files specified in @ARGV, and evaluating each line of input as a line of Perl code. If that Perl code fails to compile, or throws an exception via die() , the error is warned to STDERR . perldoc -f eval has the full details of how and why eval() might set $@ .

As an example of the code being called:

$ echo 'print "foo\\n";' | ./evals.pl 
foo

$ echo 'print 1 + 1, "\\n";' | ./evals.pl
2

$ echo 'dfsdfsdaf' | ./evals.pl
Bareword "dfsdfsdaf" not allowed while "strict subs" in use at (eval 1) line 1, <> line 1.

$ echo 'die "dead";' | ./evals.pl 
dead at (eval 1) line 1, <> line 1.