![\"\"](\"https:\/\/www.assoc-amazon.com\/e\/ir?t=extcas-20&l=ur2&o=1\")
<\/p>\nI hate C++ pointers. I switch from C to C++ to Java a lot, so I tend to need refreshing once in a while when I go from Java to C++.
\nI was updating some old code the other day and came across a function which was supposed to delete a pointer passed to it and then set it to null. When the function returned, the pointer was deleted, but was not null. Reason why: the function had been set up wrong:<\/p>\n
void deletePointer(XYZ* _xyz)
\n{
\n delete _xyz;
\n _xyz = 0;
\n}<\/p>\n
\nvoid callDeletePointer()
\n{
\n XYZ* xyz = new XYZ();
\n …
\n …
\n …
\n deletePointer(xyz);
\n}<\/p>\n
The way deletePointer really needed to be set up is like this:<\/p>\n
void deletePointer(XYZ*& _xyz<\/b>)
\n{
\n delete _xyz;
\n _xyz = 0;
\n}<\/p>\n
Here’s why:
\nThe first deletePointer receives a copy of the pointer to _xyz, which allows the user to modify the object addressed by the pointer.
\nThe second deletePointer receives the actual pointer to _xyz, which allows the user to modify the pointer itself.<\/p>\n
Oh and one more tip:
\nWhen deallocating memory do this: <\/p>\n
XYZ* xyz = new XYZ();
\ndelete xyz;
\nxyz = 0;<\/p>\n
instead of:<\/p>\n
XYZ* xyz = new XYZ();
\nif(xyz != 0)
\n{
\n delete xyz;
\n xyz = 0;
\n}<\/b><\/p>\n
Apparently, C++ guarantees operator delete() won’t be called by a delete if the pointer is 0, so testing for null is not necessary. In fact, if you do include a pointer null test, the test will be done twice.<\/p>\n
Source: C++ Primer, Third Edition, by Stanley B. Lippman, Josee Lajoie<\/a>