Re^11: eval to replace die?

[BrowserUk]

If you throw a MyError::Foo, I would have to explicitly throw that error

And if I prefix my error messages with "MyError::Foo" you'd have to explicitly throw an error that started with that string to break my error handling.

And if two modules chose bad namespaces--like MyError rather than MyModuleName::Error--then the same problem arises in code that uses both modules.

The error vectors are exactly the same! It's still just string compare.

The solution is in well chosen namespaces--whether error prefixes or execption class names.

Just moving to throwing exception objects rather than strings does not fix the underlying problem: that of badly chosen namespaces!

[phaylon]

And if I prefix my error messages with "MyError::Foo" you'd have to explicitly throw an error that started with that string to break my error handling.

Right, which can happen a lot, since I don't prefix my errors and often build error messages dynamically. So if my error is "MyError::Foo instance cannot be passed to do_something()", your error handling breaks. And it doesn't even do it consistently, it'll break depending on what I want to communicate to the user.

Yes, bad namespaces are a problem. But there's no namespaces in error messages. It's all just a message. And while you can define for yourself that the start of the message is "the namespace," I'd rather use inheritance, since that has dedicated namespace functionality.

Again, exception objects are for communicating with the rest of the program. Error messages are for communicating with the user. If you mix the two, there's worlds of pain down that road. Not to mention the fact that exception objects can carry optional stack traces and various other information, which makes them even more valuable to the user, even if it's just to communicate a better description of the error to a developer.

If you really think $foo eq "bar" and $foo->isa('Something') are exactly the same, I'm quite concerned.

[BrowserUk]

Not to mention the fact that exception objects can carry optional stack traces and various other information, which makes them even more valuable to the user, even if it's just to communicate a better description of the error to a developer.

Have you heard of Carp::Croak()?

If you really think $foo eq "bar" and $foo->isa('Something') are exactly the same, I'm quite concerned.

If you think that $foo eq 'Something' and $foo->isa( 'Something' ) are vastly different...you should be concerned.

The bottom line is, either technique requires programmer discipline. If you really used 'MyError' for your exception namespace, you are just as vulnerable to someone else choosing that badly chosen namespace for their exceptions, as I would be if I used a non-specific formating convention for my error messages.

If programmers adhere to a standard for their errors or exceptions, the problems are alleviated.

But exceptions are not a magic bullet that will fix badly architected code.

[phaylon]

Have you head of Carp::Croak()?

Have you read the word optional? Why are you ignoring all other of my points? If you don't have an actual argument but just want to be right then say so, and I'll just shut up.

If you think that $foo eq 'Something' and $foo->isa( 'Something' ) are vastly different...you should be concerned.

I disagree. Strongly.

The bottom line is, either technique requires programmer discipline. If you really used 'MyError' for your exception namespace, you are just as vulnerable to someone else choosing that badly chosen namespace for their exceptions, as I would be if I used a non-specific formating convention for my error messages.

I disagree. First off, everything in programming requires discipline. Cooking and exception handling both require discipline, that doesn't make them identical. The big difference is that class namespaces are designed to be namespaces. Strings aren't. You can assume part of a string is a namespace, but you can never be sure. A class is always a namespace. It's designed for the job you want to cram into a string.

You also say yourself why strings are less reliable for error handling: It's depending on your error message formatting, *plus* the handling of errors by a type you want to extract from a free-form value that might have come from anywhere. With objects, all you need to check is the type. It doesn't matter then how I format my errors, how you format you errors, how I format your errors and how you format my errors.

If programmers adhere to a standard for their errors or exceptions, the problems are alleviated.

With object oriented exceptions, there are the enforced standards of objects. Specifically that data and package are separate pieces of information that I can access reliably. And as I've said earlier, if you're the only programmer working the code, and never upgrade your modules, and know all formats of errors all other authors used, then yes, it might "work."

But exceptions are not a magic bullet that will fix badly architected code.

Nothing will do that, but that's not an argument on the issue. I'd argue that inspecting user error messages for error type determination is bad code.