Yes, in this case I can be sure that the populated hash is not empty.

I suppose that should be fast/efficient if done in boolean context.

I suppose that should be fast/efficient if done in boolean context.

Short answer: yes. Long answer: What is the most efficient way to see if a hash is empty?

Perl 6 - second systems done right

  DB<102> p defined %h
 
  DB<103> %h=(a=>1)
 
  DB<104> p defined %h
1
  DB<105> my %h=(a=>1); print defined %h
1
[download]

UPDATE: Here what the docs say:

Use of defined on aggregates (hashes and arrays) is deprecated. It used to report whether memory for that aggregate has ever been allocated. This behavior may disappear in future versions of Perl. You should instead use a simple test for size:

1. if (@an_array) { print "has array elements\n" }
2. if (%a_hash) { print "has hash members\n" }

UPDATE:

I think if you really need to distinguish between an empty hash and an non-existent hash, it can only be a package var. In this case checking the STASH should do.

UPDATE: I have to change my mind again... exists would check for a glob...

if (keys(%BigHash))
[download]

Update: moritz is right

if (%BigHash) {
[download]

if test for emptiness is not enough for you, you could flag the state of your hash by blessing it per default to a pseudo-package "EMPTY" or "Unloaded".²

  DB<123> print ref \%h
HASH
  DB<124> bless \%h, EMPTY
 
  DB<125> print ref \%h
EMPTY
  DB<126> delete ${main::}{h}
 
  DB<127> print ref \%h
HASH
[download]

but be carefull to destroy the empty hash before loading your data, unfortunately there is no "unbless" mechanism.¹

There are also tie, tied and untie, but then those pseudo-packages need to exist and to have special methods and might slow down the use of your hash.

1) of course you can also bless it to "Loaded" if it doesn't interfere with your plans...

2) or you just use one package with a method isLoaded() and store the state in a class variable.

if (\%h->isLoaded() ) {...}

From a basic REPL script:

%foo; scalar(%foo)
0
%foo = (a=>1, b=>2, c=>3); scalar(%foo)
3/8
%foo = undef; scalar(%foo);
1/8
%foo = (); scalar(%foo);
0
[download]

Does that help?

As an aside, you can start with an undefined variable and start referencing a bunch of keys (as though a nested set of “hashes containing hashrefs” with these particular keys already existed, and, miraculously, one will appear.

When you say “a public variable,” however, I immediately think that you should strictly be using a class for this.   You don’t want code that is dependent upon the arrangement and use of this variable to be scattered willy-and-yob across the system:   you want it to be in one place, and with a clearly defined abstract interface.   These methods specify what the rest of the application wants to do (or wants to know), and (only...) this one package implements how it is done.

“Efficiency?”   Schmefficiency!   Once you pass a billion clock-cycles per second, no one can hear you scream.   If you need a few more gigabytes, grab ’em at the grocery checkout stand ...   Maintainability trumps all.

My last $dayjob concerned making servers more efficient. With 200 servers in the farm and counting, adding a little RAM does become expensive, and reducing the need for more hardware by 10% becomes worth an engineer's salary.

Instead of a global hash, consider creating a singleton object having get() and set() methods.

And exists, and keys, and when I'm all done it might as well be a tied hash since that is the API I want to provide.

%h or $object -> init( \%h );
[download]

NOOO.

my %foo = (a => 1, b => 2);

for (1..10) {
    each %foo or print "re-init $_\n";
}
__END__
re-init 3
re-init 6
re-init 9
[download]

each uses an iterator tied to the variable. If you use it, you interfer with other (and even your own) usages of that iterator.

Perl 6 - second systems done right

(re the modified form) Right, using it in boolean context allows it to know I really care only if it is empty or not, and don't need to count the actual values.