I like using plain hashes unless necessary. The same goes with arrays, and subroutines. References are extremely useful tools, but using them for everything seems a bit silly.

What are my reasons? Well, one might be overhead. There is a small amount of overhead involved in dereferencing, though to be honest this doesn't concern me that much.

A much more important reason is tidiness. All those arrows do not take long to type, and they don't really make it extremely difficult to read, but they do add some noise. Frankly, I think we should take advantage of any decrease we can get, in terms of visual noise. Consider the following examples:

    @{$href}{"one", "two", "three"}; # hash slice, hashref
    @hash   {"one", "two", "three"}; # hash slice, hash

    for (0 .. $#{$aref}) {} # index iteration, arrayref
    for (0 .. $#array)   {} # index iteration, array

    # with references
    @{$aref2} = map  { $href1->{$_} + 1 }
                sort { $href2->{$a} <=> $href2->{$b} }
                grep /foo/, @{$aref1}[0 .. $#{$aref1}/2];

    # without
    @array2 = map  { $hash1{$_} + 1 }
              sort { $hash2{$a} <=> $hash2{$b} }
              grep /foo/, @array1[0 .. $#array1/2];
[download]

Sure, the versions using references aren't terribly much more difficult to follow than the ones without. It's not that big of a difference, but like I've said before, every little bit helps. This kind of thing adds up quickly to create a mess.

I find that, especially with complex data structures, the -> notation can clarify things by seperating my identifiers.

You also pick as examples of dereferencing notations that can tend to look complex to an inexperienced programmer whether they are references or not. All of the examples you cite for complexity can be rewritten using less confusing notations.

for (0 .. $#{$aref}) {}
# or
for (@$aref) {}

@{$aref2} = ...;
#or
@$aref2 = ...;
[download]

I'm not sure why you don't show that the curly braces can be left out during dereferencing.
dsb
This is my cool %SIG

I find that, especially with complex data structures, the -> notation can clarify things by seperating my identifiers

I find this view rather bizarre. I think the balanced nature of subscript delimiters clarifies things pretty well as it is. To each his own, I suppose.

All of the examples you cite for complexity can be rewritten using less confusing notations.

This statement may be true, but your for loop example is not the same as mine. Looping on array indexes and array elements is not the same thing. Sometimes you need the index as well as the value it points to.

I'm not sure why you don't show that the curly braces can be left out during dereferencing

I generally use the braces for clarity. I suppose that's a bit of a contradiction over my earlier claim of trying to avoid visual noise, but there it is. You're probably right that I should have mentioned it, though. I just didn't think to do so.

That said, this whole conversation thread is kind of strange. Am I actually defending the use of plain data structures over references? What's the point? You can use references 100% of the time if you want, but I'll stick to the regular type of variables.