TLDR: always put const AFTER the element you want to specify as constant
We all love const variables and parameters. They are, well, constant. You're sure they won't be changed by a side-effect of your code (or someone else's).
What you usually see in C++ is something like:
const int myVar(42);
This is pretty clear, right? myVar is stores an integer, 42 in this case, and this value won't change until we get out of the current scope. So if we put something like:
const int myVar(42);
myVar++;
your compiler will yell at you with something like error: increment of read-only variable ‘myVar’.
So far so good. But I once again stumbled across an issue using const that I, even with my 10+ years of experience, still haven't stuck in my head (but I hope that writing this down will help so I'll guess I should share it).
Let's say you want to declare a constant pointer to an integer. The no-brainer line of code that gets out is:
const int *myVarPtr;
Following the same logic as before, if I try to modify it:
int myVar(42);
const int *myVarPtr;
myVarPtr++;
the compiler doesn't complain! And it's normal because, if you stick to its definition, const applies to the element to its left, unless it's the first element. Then it applies to its right.
So what we declared earlier was actually a non-const pointer to a const integer. The original goal would be achieved with this code:
int * const myVarPtr;
I hate exceptions. This is the typical case where the exception makes the concept confusing. Now I just make sure I'm never in this case.
If you have only one thing to learn from this article it would be this: always put const AFTER the element you want to specify as constant.
Top comments (3)
I totally agree, by placing the const on the right side we can read from right to left:
int const& const foo;doesn't compile in C++.It is impossible make a reference "points" to another variable.
A 'const reference' means that the reference considers it "points" a const object.
I disagree, while it does read slightly better right-to-left, which is required for determining what is const, it isn't clear. For instance in your example, you did not modify myVar, (assuming that pointer was set to the address of myVar) - but you modified the *. To me it reads much cleaner as:
const int* myVarPtrassuming you don't want to change the value, which is 99% of the time - generally we don't care if the pointer gets modified, but the contents it points to... now that is important.
In the 1% of the time that both, or just the pointer, is important to be const, then the imo ugly:
const int* const myVarPtr;int* const myPtr;
Will stop and make the reader / maintainer look twice. If the const was ALWAYS on the right,
int const value;int* const valuePtr;
it is easy, very easy to slip up in maintaining or accidental writing of a const pointer to a non const value, when it should have been a pointer (const or not) to a const value.