Re: Escaping %params

[tangent]

As mentioned in the CB $cgi->Vars may cause problems - a safer way to build the params hash is:

my %params = map { $_ => $cgi->param($_) } $cgi->param;

And beware of params that have more than one value, e.g. checkbox groups

Update: to allow for params with multiple values:

my %params;
for my $name ($cgi->param) {
    my @values = $cgi->param($name);
    $params{$name} = @values > 1 ? \@values : $values[0];
}

When further processing %params you will need to check if $param{$name} is an array reference.

[DaisyLou]

Thank you... Interestingly, when I use this code:

use warnings;
use strict;
use CGI;

use Data::Dumper;
print "Content-type: text/html\n\n";

my $cgi = CGI->new();
my %params = map { $_ => $cgi->param($_) || '' } $cgi->param;

print Dumper \%params;

The output is when I pass test.pl?a=\'b is:

$VAR1 = { 'a' => '\\\'b' };

... so it looks like something in the stack is already taking care of the escaping for me. Am I worrying about nothing?

[tangent]

Data::Dumper is doing that - try this:

for my $param (keys %params) {
    print "$param: $params{$param}<br>"
}

[Anonymous Monk]

Data::Dumper does perl-escaping (defaults have some caveats)

Data::Dump::pp() does better perl-escaping by default

Neither ddumper does HTML-escaping

You can alway do  my $cgi = CGI->new; print $cgi->header, $cgi->Dump ; to see whats inside $query

[AnomalousMonk]

s/_/\_/g


s/_/\_/g does nothing. You probably want  s/_/\\_/g

>perl -wMstrict -le
"$_ = 'a_b__c___';
 print qq{before: '$_'};
 ;;
 s/_/\_/g;
 print qq{after:  '$_'};
"
before: 'a_b__c___'
after:  'a_b__c___'

But please allow me to add my voice to the chorus imploring you to Just Use Placeholders!.

[DaisyLou]

I've distilled all this advice this as best I can.

============
something.lib
============
sub safer {
    my $hash = shift;
    my %safer;
    while (my ($k, $v) = each %$hash) {
        s/\\//g         for $k, $v;
        s/0x00//g       for $k, $v;
        s/0x08//g       for $k, $v;
        s/0x09//g       for $k, $v;
        s/0x0a/\n/g     for $k, $v;
        s/0x0d/\r/g     for $k, $v;
        s/"/\\"/g       for $k, $v;
        s/%/\\%/g       for $k, $v;
        s/'/\\'/g       for $k, $v;
        s/_/\_/g        for $k, $v;

        $safer{$k} = $v;
    }
    return %safer;
}

================
something.cgi...
================
use warnings;
use strict;
use CGI;
use CGI::Carp;

print "Content-type: text/html\n\n";

# marker
my $cgi = CGI->new();

$cgi->param;
my %params;

for my $name ($cgi->param) {
    my @values = $cgi->param($name);
    $params{$name} = @values > 1 ? \@values : $values[0];
}

%params=safer(\%params);
# marker

for my $param (keys %params) {
    print "$param: $params{$param}<br>"
}

The stuff between the two markers will replace the existing

sub main
{
        my $cgi = CGI->new();
        my %params = $cgi->Vars();

... in the existing scripts. This is running under mod-perl (w/ regcooker). Are there any "gotchas" I should be aware of here? Thanks to all you monks for all your help!

[Anonymous Monk]

A problem is with CGI->Vars , you never want to use CGI->Vars, CGI->Vars is for perl4, Vars mangles (encodes, serializes, packs, implodes) the data, its backwards compatibility for some 1993 stuff

You want "escapeHTML" from CGI.pm