Returning pointers

The program can easily test the pointer for validity: compare with NULL:

Here is an important question: is failing to find an ID likely to happen, or is it an exceptional event?

Or even more concise:

Hi,

If returning NULL is an acceptable output from the function, (e.g. you're trying to find something, but it's fine that it might fail sometimes) then you're going to have to handle NULL in the calling code.

However, if it's an error that it should fail to find something, it should still return NULL, but should probably assert or throw an exception when that happens, so you know something has gone wrong.

It's more up to you at that point whether you would prefer your program to crash and force you to fix some problem, or write extra code that attempts to handle the unexpected situation and carry on functioning in whatever way it can.

If it's really bad for it to fail, beware silently coping with it returning NULL - this may lead to you missing a really bad bug in your code [smile]

It is a perfectly acceptable situation not to find anything, so I guess checking for NULL pointers is really the only to go, I just thought there may have been a cleaner way. Thanks for the input. :)

PS: I never thought of using an if statement like that DevFred, interesting.

Quote:

Original post by DevFred Or even more concise: *** Source Snippet Removed ***

That code has a bad smell to it, if you ask me. I'd stick to rip-off's version - the loss of a single vertical line isn't particularly significant, but seeing an assignment in an if-statement is bound to set off some alarms for somebody later down the line.

Consider the Null Object design pattern.

Quote:

Original post by Sc4Freak Quote: Original post by DevFred Or even more concise: *** Source Snippet Removed *** That code has a bad smell to it, if you ask me. I'd stick to rip-off's version - the loss of a single vertical line isn't particularly significant, but seeing an assignment in an if-statement is bound to set off some alarms for somebody later down the line.

What's wrong with it? If anything, I'd say DevFreds version is cleaner as it restricts the scope of the null object. It's a common C++ paradigm and unlikely to be confused with an equality test (which won't even compile).