Discussion on: When to use const in C++? Part IV: parameters

[Pierre Gradot]

Hello,

I have 2 remarks:

1/ "POD" has a specific meaning (until C++20). "POD" is not a synonym for "primitive data types" (I have just learned that the dedicated name seems to be "Fundamental types" ^^)

See:

• en.cppreference.com/w/cpp/named_re...
• en.cppreference.com/w/cpp/language...

2/ It's worth noting that const for parameters is only relevant for functions definitions, not function declarations.

For example, the following code is valid:

void f(const int);

void f(int a) {
    a = 42;
}

clang-tidy warns about it:

Parameter 1 is const-qualified in the function declaration; const-qualification of parameters only has an effect in function definitions

---

Personally, I don't use const for base types. But for other types const& is of course the default way :)

[Sandor Dargo]

Thanks, for your remarks!

I used "POD" for Plain Old Data, but indeed I should have used Fundamental types. We both learned something :)

Your second point is wow, interesting... Even without clang-tidy, I'd warn anyone to not match the declaration with the definition. I'd find it really disturbing :(

I agree, that for small functions, using const value parameters don't add much readability, but in case of longer ones, it gives some extra safety and then I prefer to be consistent.

Thanks again for your captivating remarks!

[Pierre Gradot]

You can learned new stuffs everyday, and I like that :)

Interesting indeed... "Scary" sounds more accurate to me 😅

In fact, there are other scary stuffs regarding declarations, definitions and redeclaration (yes, you can re-declare a function without exactly matching the previous declarations). Since you live in France, I suppose you speak French so take a look at the section "Le grand bazar des redéclarations" in this article I wrote about default parameters: "Arguments par défaut d'une fonction en C++". You will get psyched 😂

You're welcome! Glad you found them interesting :)

[Sandor Dargo]

Thanks, I'm going to read it during the weekend. I just had a glance at it, and yeah, the default parameters for virtual can be problematic. I also wrote a draft about it after we had discussed that in our team a few weeks ago when we activated a new Sonar profile...

By the way, luckily the const mismatch doesn't work with references or pointers to const data.