One is tempted to ask why the
module has an interface that forces a user to create an instance, give the instance a number, and then retrieve a string from
the instance when a simple class method exported would have worked more easily. I mean, then I have to remember whether
the current number in the instance is the one I want-- which is the sort of "internal state of object" knowledge the rest of my
program shouldn't have to be thinking about.
This is exactly what I mean by "pointless OO". Even your
suggested class method is "pointless OO" to me. It doesn't
buy you anything, except additional clutter.
Abigail
| [reply] |
For a module with one essential function, parsing a number into an English string, it does seem excessive. But what if my program already has a get_string() function or an %N hash, and I add this module after the fact? The OO layer protects me from name collisions (and in my example is completely optional). While I might not worry too much about it for Lingua::EN::Numbers, I certainly would worry if CGI didn't offer me a choice-- that module has the potential to export several dozen function names, many of which (in a CGI script) might clash with my function names (poorly chosen names, perhaps).
The OO isn't really pointless. It's happening either way-- exporting symbols without being asked to do so is just a sneaky way of concealing it is all. What I've said is that it should be explicit and optional for a module where the instance is really just an instantiation of a machine, rather than an instance of a class with unique properties that would distinguish it from other instances in the same class.
Essentially the instance becomes a shorter way of doing print Lingua::EN::Numbers->get_string(1281), " fish in the seas\n";... which you can also get around by offering the ability to export get_string(). Is there an actual reason for avoiding instantiation of a machine to prevent namespace collision? Or is the objection purely aesthetic?
| [reply] [d/l] |
But what if my program already has a get_string() function or an %N hash,
and I add this module after the fact?
Well, so what? You _do_ know the working of Perl, don't you? Then
you _do_ know that if you do:
use Module ();
the import() routine isn't called, and nothing will be
exported. In which case, you can still use the function by calling
it like:
Module::get_string ()
But a name clash is an exceptional case - normally you don't have a name
clash (I've been programming Perl for over 6 years now, I cannot recall
ever having had a name clash problem), and you can just use the short
get_string. But with a class method, you have to use
Module -> sub, even if it's not necessary.
The OO isn't really pointless. It's happening either way-- exporting
symbols without being asked to do so is just a sneaky way of concealing
it is all.
Exporting symbols without being asked is only a sneaky way if there
isn't an easy way to avoid it. All you need to do is ask to _not_
import anything. That takes a whopping two characters.
A module author using Exporter gives the programmer a choice -
a module author using OO to avoid name clashes doesn't. I know who acts
Perlish, and it ain't the latter programmer.
As for the %N of Lingua::EN::Numbers::Easy, let
me quote from the manual:
By default, "Lingua::EN::Numbers::Easy" exports a hash
"%N" to the importing package. Also, by default, "Lin-
gua::EN::Numbers::Easy" uses the British mode of "Lin-
gua::EN::Numbers". Both defaults can be changed by
optional arguments to the "use Lingua::EN::Numbers::Easy;"
statement.
The first argument determines the parsing mode of "Lin-
gua::EN::Numbers". Currently, "Lingua::EN::Numbers" sup-
ports British and American. The second argument deter-
mines the name of the hash in the importing package.
use Lingua::EN::Numbers::Easy qw /American %nums/;
would use American parsing mode, and "%nums" as the tied
hash.
If %N gives you a name clash, just use any other suitable name.
OO isn't the only way of solving name clash problems!
Abigail
| [reply] [d/l] [select] |