I'm not quite clear on what you're trying to do. It sounds as if maybe you have more than one conference at a time, and you want to store them all under the same hash key.

I suggest that you work with hash references that contain all of the conference information in one hash.

{
    time       => $time,
    conference => $value1,
    type       => $value2,
    comment    => $value3,
}
[download]

You can take one of those and stick it in a hash of arrays (of hashes) or just have an array of them.

push @AoH, { time => $t, conference => $v1, ... };
[download]

Then you can sort the array by time when you want to output them.

my @sorted_AoH = sort { $_->{time} <=> $_->{time} } @AoH;
foreach my $conf ( @sorted_AoH ) {
    my $time = $conf->{time};
    print ... # element
}
[download]

If you need random access to these records by time, then you can put them in a hash as you have them but add a layer of array to it.

push @{ $hash{$time} }, { time => $time, type => $v2, ... };
[download]

Then you need an extra loop when dealing with them.

for my $time ( sort keys %hash ) {
    for my $conf ( @{ $hash{ $time } } ) {
        print ... # $conf element
    }
}
[download]

Again, I'm not clear from your description what you're trying to do, but maybe this gives you some ideas.

In general a Hash is a String-to-Thing mapping, not a relational database.

(oh, and please use <code>...</code> tags around example code that you post here, that'll make it easier to read).

use strict;
use warnings;
my %hash = ();
my ($time, $conference, $type,$comment);
while(<FILE>)
{
   # adapt parsing values to your need here

   if (/conference=(.*)/){
       $time = $1;
   }
   if (/conference=(.*)/){
       $conference = $1;
   }
   if (/type=(.*)/){
       $type = $1;
   }
   if (/comment=(.*)/){
       $comment = $1;
   }
   if (/lastline of record/){
      $hash{$time,$conference}{time}       = $time;
      $hash{$time,$conference}{conference} = $conference;  # info is a
+lready in the key, for easy parsing in the for loop, store conference
+ twice.
      $hash{$time,$conference}{type}       = $type;
      $hash{$time,$conference}{comment}    = $comment;
   }
}

for my $timeconference (sort keys %hash)
{
   printf "time %s, conference %s, type %s, comment %s\n",
          @{hash}{qw(time conference type comment)};
}
[download]

It is certainly possible to create, say, a hash which contains an arrayref which in turn contains a list of references to records. It's also certainly possible to create more than one such structure which refers-to the same record.

In other words... “oh yes, you can do this in Perl, but...”

But before I proceeded too far in that direction, I think I'd be asking myself this hard question:   “is what I really want to use here just ‘an SQL database?’ ”

When your well-laid plans begin to “smell funny,” follow your gut.

When your well-laid plans begin to “smell funny,” follow your gut.

A detachable smelling gut? You could make a killing in selling your worms :) Now I'll make like a tree and get out of here :)

The problem that I have is to merge two files which have the same fields (time, conference, type, comment) and to sort them out. The combination of (time, conference)is unique. When I tried it, I had the problem of how to address multi value key in hash. I had a sub to read files and write contents to two hashes and return the hash reference. But that was for single value key. I don't know how to do the same thing with muti-value key and how to use key reference to access multiple values in the hash. Hope I made the problem a little bit cleare.

Perhaps a HoHoH would do it.

#!/usr/bin/perl

use warnings;
use strict;

my %conf_table = (
  1 => {
    conference1 => {
      type => q{type1},
      comment => q{comment1},
    },
    conference2 => {
      type => q{type2},
      comment => q{comment2},
    },
  },
  2 => {
    conference3 => {
      type => q{type1},
      comment => q{comment3},
    },
    conference4 => {
      type => q{type2},
      comment => q{comment4},
    },
  },
);

for my $time (keys %conf_table){
  print qq{time: $time\n};
  for my $cnf (keys %{$conf_table{$time}}){
    print qq{\tconference: $cnf\n};
    print qq{\t\ttype: $conf_table{$time}{$cnf}{type}\n};
    print qq{\t\tcomment: $conf_table{$time}{$cnf}{comment}\n};
  }
}
[download]

time: 1
  conference: conference1
    type: type1
    comment: comment1
  conference: conference2
    type: type2
    comment: comment2
time: 2
  conference: conference4
    type: type2
    comment: comment4
  conference: conference3
    type: type1
    comment: comment3
[download]

Could I ask why you want to use the conference time as the first entry? Do you often want to search the hash for conferences at an exact time? (The hash won't help you searching conferences in eg. an interval between two given times, which is what you'd more often want.) If not, why don't you just use a single array of hashes, with each entry of the array storing one conference, and the time stored in another field just like how you store the conference name and type and comment.

Update: oh, I see kyle has already suggested this.