syntax for hashes with variable in name

[equick]

Hi,

I have a problem with creating a hash which has a a variable in its name. This works ok for arrays ,eg

@{"platforms_$site"} 

but not hashes

%{"platforms_$site"}

Here's my script:

#!/usr/bin/perl

sub buildconfig;

buildconfig;

foreach $key(keys %sites){
        print "$key\n";
}

sub buildconfig{
        my ($site, $platform, $webserver, $option);

        my %sites = map {$_=>1} @sites;

        # create following variables
        # %sites = BFH, LPR, UAT, TEST
        # @all_platforms and %all_platforms = TEST_MLS, TEST_RST, UAT_
+MLS, UAT_RST ...
        # %platform_TEST = TEST_MLS, TEST_RST, TEST_OPP, TEST_BO ...
        # @all_webservers and %all_webservers = mls001.test, mls002.te
+st, tomcat001.test, tomcat002.test ...
        # %webservers_TEST_MLS = mls001.test, mls002.test
        # %deploy_as = prod, standby, demo
        # %deploy_as_platforms = Prod1, Prod2, UAT, TEST

        my %sites = map {$_=>1} @sites;

        foreach $site (@sites) {

                my %all_platforms = map {$_=>1} @{"platforms_$site"};
                my %{"platforms_$site"} = map {$_=>1} @{"platforms_$si
+te"};

                foreach $platform (keys %{platforms_$site}) {

                        my %all_webservers = map {$_=>1} @{"webservers
+_$platform"};
                        my %{"webservers_$platform"} = map {$_=>1} @{"
+webservers_$platform"};

                }
        }

        my @all_platforms=sort keys(%all_platforms);
        my @all_webservers=sort keys(%all_webservers);

        my %deploy_as = map {$_=>1} @deploy_as;
        my %deploy_as_platforms = map {$_=>1} @deploy_as_platforms;

}

but when it's run I get:

Can't declare hash dereference in my at ./test.pl line 30, near "} ="
Execution of ./test.pl aborted due to compilation errors.

Could anyone tell me how to fix that please?

Thanks, Ed.

[jethro]

Your immediate problem is that perl interprets your %{ as a hash slice (see http://docstore.mik.ua/orelly/perl/learn/ch05_05.htm for a small example)

Your larger problem is that you want to use a technique that is a relict from perl4 days and only (if at all) suitable for quick hacks

Instead of using

my %{"platform_$site"};
${"platform_$site"}{"thingy"}= 5;
foreach my $x ( keys %{"platform_$site"}) {
  ...

use multidimensional hashes

my %platform;
$platform{$site}{"thingy"}= 5;
foreach my $x ( keys %{$platform{$site}}) {
  ...

You can use the same technique for your arrays as well, i.e. use a so called HashOfArrays:

push @{$platform{$site}}, 1,2;
$platform{$site}[5]= 5;
foreach my $x ( @{$platform{$site}}) {
  ..

PS: I also notice you don't use 'use warnings;' and 'use strict;' at the beginning of your script. It is highly advisable to include those two lines. Avoids many bug finding sessions later.

[ikegami]

Hash (and array) slices always start with @. No slices start with %.

my %{"$name"};

is the same as

my %{$name}

which is the same as

my %$name;  # Symbolic dereference

which no more works than

my %$ref;   # Generic deference

[ambrus]

Your immediate problem is that perl interprets your %{ as a hash slice

Is this for real?

[Tux]

is it?

Let's rule out the fact that this deals with symbolic references, strict tells us ...

use strict;
use warnings;

{   no strict "refs";
    my $site = "foo";
    my @{"platforms_$site"} = qw( foo bar );
    my %{"platforms_$site"} = map { $_ => 1 } @{"platforms_$site"};
    }

=>

Can't declare array dereference in "my" at xx.pl line 8, near "} ="
Execution of xx.pl aborted due to compilation errors.

The difference is in the assignment part. You cannot use my there. When you drop those, you get:

use strict;
use warnings;

{   no strict "refs";
    my $site = "foo";
    @{"platforms_$site"} = qw( foo bar );
    %{"platforms_$site"} = map { $_ => 1 } @{"platforms_$site"};
    }

Which is still bad coding style, but this works.

[jethro]

It was the best explanation I could find at the moment. But I didn't check with deparse or something.

[equick]

Thanks for your reply.

Yes usually I include strict, but I had just taken an extract of an old script I'm patching to work on that particular function.

I also agree multidimensional arrays are probably the way to go, but if possible I'd rather keep the current variables as they are used throughout the other script.

[Tux]

Don't go that road. Please. Done that, been there. It leads to unwanted things, in the widest sense of "unwanted". Really!

Symbolic reference (where the name of the hash/array/scalar is in a variable name) can always be made strict and safer by embedding it inside a hash:

my $hash_name = "platform_$site";
%$hash_name = ( key => $value );
:
my $value = ${$hash_name}{key};

=>

my %global_hash;
my $hash_name = "platform_$site";
$global_hash{$hash_name} = { key => $value };
:
my $value = $global_hash{$hash_name}{key};

Or - even better - go with jethro's suggestion to use multidimensional data structures.

[aquarium]

...ditto...but once you've touched some code, you own it, and anything it may or may not do with good or bad data as input. so as much as this sounds as a bit of a pain, you're much better of re-writing the code to use good/clean programming patterns. code with symbolic reference variables is harder to debug/maintain/improve. i once wrote a data reporting perl script initially (and naively) using symbolic reference variables. i immediately re-wrote it using multilevel hashes, and the spaghetti code turned magically into nicely ordered simple subroutines and loops, and halved the number of lines of code.