Question

Did you ever went through the process of replacing the BOOST_FOREACH by the range-based for loop? Are there any gotchas when doing it mechanically (i.e. not doing a deep study of the code around).

Background: I do have sources with about 250K lines, compiled with Microsoft VC++ 10 (Visual Studio 2010). The code uses also BOOST_FOREACH from the boost library. I am just trying the VC++ 11 (Visual Studio 2012 RC) that implements also the range-based for loop with the auto.

So far, I have met the situation where the element was declared outside the BOOST_FOREACH (no declaration in the parentheses):

BOOST_FOREACH (element, container) {
    ...
}

… and the element was expected to have the value set in the loop after the break. Can you remember similar problems?

When the BOOST_FOREACH contains the declaration, would you choose to replace the explicit type by auto? Is there any advantage to do so when one just mechanically rewrites the BOOST_FOREACH instead of leaving the explicit declaration there?

Thanks for your help,
Petr

Answer 1

The advantage of auto over using the actual type or reference type is that the code will continue to work if the container type changes. This may not be what you want if you’re intentionally performing a type conversion on the result of indirecting the container iterator.

Example:

std::vector<int> v;
for (int i: v) ...

If v changes to std::vector<long> or std::vector<unsigned> then the type conversion may be unwanted; still, a compiler warning should catch this (especially for a narrowing conversion). On the other hand, you could have code where a (widening) conversion is intended:

std::vector<char> v;
for (int i: v) ...