Narrowing conversion size_t to long

Asked Jul 18 '22 at 07:24

Active Jul 18 '22 at 07:30

Viewed 47 times

How should I write this to appease clang-tidy?

#include <vector>
int main()
{
    std::vector<int> v(10);
    size_t index = 2;
    auto it = v.begin() + index;
    auto rollover = (index + 1) % v.size();
}

clang-tidy complains

Clang-Tidy: Narrowing conversion from 'size_t' (aka 'unsigned long') to signed type '__gnu_cxx::__normal_iteratorstd::basic_string<char *, std::vectorstd::basic_string<char>>::difference_type' (aka 'long') is implementation-defined

If I use long for index, the mod operation narrows v.size() if I use the more sensible size_t for index the v.begin() + index narrows index. (Why is the iterators operator+ only defined for a signed type anyways? g++11.2.0)

Is there a way to get around this without using static_cast<long> all over the place?

C++ Insights demo

edited Jul 18 '22 at 07:30

asked Jul 18 '22 at 07:24

Captain Giraffe

14,407
6
39
67

1

its not the iterator that is signed but the argument to its `operator+` – 463035818_is_not_an_ai Jul 18 '22 at 07:29
Related: [C++: What is the correct cast for offseting to std::vector iterator?](https://stackoverflow.com/questions/44480794/c-what-is-the-correct-cast-for-offseting-to-stdvector-iterator) – Jason Jul 18 '22 at 07:29
Maybe choose one or the other when it comes to indices or iterators. This example is a bit contrived, and it would be better to see a real example. As for the iterator, see [LegacyRandomAccessIterator](https://en.cppreference.com/w/cpp/named_req/RandomAccessIterator) or similar documentation. Operations on such iterators are signed. – paddy Jul 18 '22 at 07:30
@463035818_is_not_a_number Corrected, thanks. – Captain Giraffe Jul 18 '22 at 07:31
In any case, just using `+1ULL` could already work. Or `+1UL`. – infinitezero Jul 18 '22 at 07:32
1

@paddy The non-contrived code sits at 760 lines of code. I'm pretty sure no-one here is interested in that =) – Captain Giraffe Jul 18 '22 at 07:36
I meant that seeing some kind of real-world example where you need to mix these types in such a way as to need loads of casts might help understand your goals. – paddy Jul 18 '22 at 07:54
@paddy I'm iterating over a file-list. There are next(), previous() that uses the second construct. A current() that uses the first. And plenty of similar uses. – Captain Giraffe Jul 18 '22 at 08:02
C-style casts are nice and short (just sayin') – Paul Sanders Jul 18 '22 at 08:44
@PaulSanders I ended up doing just that. – Captain Giraffe Jul 18 '22 at 09:08
@CaptainGiraffe Sensible move. – Paul Sanders Jul 18 '22 at 09:09
Have you tried using `std::vector::difference_type`? – Goswin von Brederlow Jul 18 '22 at 11:20
@GoswinvonBrederlow Yes, that is actually what I'm using now. Casting with an alias for that difference type. – Captain Giraffe Jul 18 '22 at 11:54

Narrowing conversion size_t to long

0 Answers0