Exception handling....dumb question?

[Felipe Almeida Lessa]

Em Dom, 2006-04-02 às 15:54 +0300, Flexx escreveu:
> Ben Finney writes:
>  > This allows all other exceptions to propogate back through the call
>  > stack.
> 
> import sys, logging
> try:
>      foo = 12/0
> except:
>      e = str(sys.exc_value)
>      print "You *knew* this was going to happen: '%s'" % (e)
>      logging.error(e)

The point (I think) Ben was trying to show is that you should not hide
exceptions from the caller's code unless you expected that exception.
For example:

def percentage(now, total):
	"""Returns the percentage of now in total."""
	return now * 100.0 / total

Very simple. What if I do "percentage(1, 0)"? The code expected me to be
clever enough to know that you can't make a percentage when the total is
zero, so it lets the exception pass by to my code (like: "it's your
problem, not mine!"):

>>> percentage(1, 0)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "<stdin>", line 3, in percentage
ZeroDivisionError: float division

But if your function is being used on a context where total can be zero
and it has a meaning, for example, returning -1, then _in_this_case_ you
should catch the exception:

def percentage(now, total):
	"""Returns the percentage of now in total.
	
	If total is zero, then return -1.
	"""
	try:
		return now * 100.0 / total
	except ZeroDivisionError:
		return -1

>>> percentage(1, 0)
-1

But this function won't catch exceptions if you don't give it numbers:

>>> percentage(None, 1)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "<stdin>", line 7, in percentage
TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'

But this is also a problem on the caller's code! You should *not* hide
it from him! But if None is valid in the context of your code, then...
well, you get the point! =)

HTH,

-- 
Felipe.