[Tutor] when to use exceptions

[alan.gauld@bt.com]

> Where do you draw the line of when to use an exception and 
> when not to use an exception?  

By use I assume you mean writing a handler.
The answer then becomes "when you can do something useful"
If you can't do amnything useful then write a generic 
handler that prints out a friendly error message, logs the 
useful info and then exits.

In practice I usually find that means writing 2 or 3 at 
most specific handlers and one generic handler.

Often, if the try block is short then its only a single 
handler for a specific error type that is most likely 
to occur- Divide by zero, IO error etc.

> exceptions for every conceiveable problem... but you would 
> spend all of your time writing exception-catching code, right?

<Puts large scale programming hat on again!>
In commercial C++ programs its not uncommon to find 
that 30-50% of the code consists of error handlers.
Exception handling helps reduce that but old style C
error handling(still necedssary in many C++ libraries)
led to very verbose error handling code.

In my experience of Python(ie no large scale projects)
the percentage of error handling code is much reduced but
if you really wanted a bulletproof system then yes, 
you could write as much error handler code as core 
application.

Alan g.