Efficiency in this case depends on a couple of things:

• a. How many different variable names are there going to be?
• b. How many different priority levels are there going to be?
• c. Do they have (m)any "gaps" - i.e. levels are 10, 15, 20, 30 ... or 1, 2, 3, 4...
• d. How many checks on existance of a priority level are there going to be?
• e. How many checks on on existance of a variable name are there here going to be?
• f. Are there multiple uses of the same priority level or variable name?

You can assume the efficiency doesn't really matter as long as a, b, d, and e are "low" for some value of low.

If you have lots of lookups on priority, put priority first : $hash{$priority}->{$name} - you can use an array if you have a "continuous" list of priorities in this case: $array[$priority]->{$name}.

If you have lots of lookups on name you can put name first: $hash{$name}->{$priority}.

If you have lots of lookups on both name and priority and less "inserts" make 2 indexes: $names{$name}->{$priority} and $priorities{$priority}->{$name}.

If you only use 1 priority per name and vice versa, you can collapse them so it will be faster to look up:

# for named lookup
$names{$name} = 
  { 
    priority => $priority,
    value => $value
  }; 
# for priority lookup
$priorities{$priority} = 
  { 
    name => $name,
    value => $value
  };
[download]

As you can see, there are quite a lot of options.

note that, on Unix, setting ENV will only affect programs exec'd from the current program (including system(), backticks etc.), NOT the "calling" environment (shell).

Updated: slightly better list layout