Re^3: Method parameters validation and inheritance

[roman]

Thanks for your comments. I should probably give some longcalls explanation.

Code sample given is a simplified but real world example: $file encapsulates a binary file (log) produced by a phone switch. It contains cdrs (Call Data Records) describing voice calls carried out. The file is used as a source for several target applications (billing, dwh).

If a call continues longer than say half an hour it is logged by parts. If join_longcalls is set for open_for_reading only complete calls are produced by read_cdr and the total duration of the call is summed up.

So the join_longcalls flag really causes the long calls to be joined. In such case I am not sure whether to use join_long_calls or joins_long_calls?

I think open_for_reading is clearer than plain reading since I try to use imperative for my subs.

As kyle suggested, I could avoid some of the "SUPERs" and turns the code into:

my $context = $file->open_for_reading({'path' => $path});
$context->joins_long_calls(1); 
# or better like this, 
# because there is no property join_long_calls
$context->join_long_calls;

The open_for_reading, actually returns Class::Prototyped based object, so join_longcalls - whatever its implementation is - actually wraps read_cdr method.

This approach has some advantages:

• The interface of open_for_reading remains same across classes.
• It is more self checking - since the parameters were turned into method calls, which fail when called but not defined.

This approach has also some disadvatages:

• Although $lc->join_long_calls can be called anytime, it make sense only immediately after open.
• When I have target file (representation of the billing, dwh file) I could have previously used a declaratory source_params method containing the parameters to be passed to open_for_reading, now I had to redefine some open_source method.