Casey Hawthorne wrote:
I have heard, that one should always use objects when programming and
avoid the builtin types!
Who told you that? Some university teacher with no industrial
experience? ;^)
It's not good advice in Python programming, for two reasons.
- It can influence performance significantly, since Python
is so dynamic. You could change the behaviour of a class,
or change the class of an instance in runtime. This means
that the Python runtime system needs to look up a lot of
things repeatedly for class instances. For ints and floats
etc, it doesn't need to do that.
- In general, Python doesn't rely on types so much for enforcing
things. The good way to ensure correct Python code is with
automated tests, and it's considered good practice to design
your APIs so that they require as little as possible from
the parameters passed in. Stricter typing means tighter
coulpling, and thus a bigger maintenance burden.
For e.g. C++, I think it could often be a good idea to use more
specialzed classes, particularly if you want the compiler to help
you with type checking. I've rarely seen it used in practice
though. The problem with using the compiler to verify that
the code is correct is that while it can check types (and do
it better if they are more specific) it still can't verify
that the code actually does what you want it to do, just that
it does *something* which is legal C++. So, in the end you
still need automated tests to verify the program in a systematic
way, and these tests will find those type errors even if you
don't write a lot of specialized classes. I mean, if you have
a full set of tests, and all tests run OK, your program is
correct, whatever types you used in various places. It might
not be ideal, but it's correct. Your classes will mainly be a
rigid way of documenting your API.
