Marcus Goldfish wrote:
So what do you consider when making this decision
Python style tends away from validating what doesn't need to be validated. The argument seems to be that the additional validating code comes at the price of legibility, and perhaps flexibility.
It's common in Python to use public attributes - even though an irresponsible user could set an object to a harmful or meaningless state.
Instead, the Python mantra is that it is 'better to ask forgiveness than permission', i.e., to handle exceptions gracefully rather than trying to avoid them.
One form of validation that is particularly discouraged in Python is type-checking.
That said, of course there are cases where data must be validated. For data stored in one type, and used in many contexts - I would validate where stored. If the reverse, then validate at the point it is used. In the likely case of 'somewhere in between these extremes', personal taste and project conventions.
and do these factors differ between python and C#/Java?
I have rarely used C#, and never Java.
C#'s strongly typed method signatures imply a contract to act properly on objects passed to the method. Given this, it seems sensible to me to include data validation in the source object.
Python's lack of type-checking of callable arguments invites the use of 'duck types'. On the margin, I think, this argues for putting data validation at the point where the data is used, rather than when it is stored.
Michael
-- http://mail.python.org/mailman/listinfo/python-list
