Opened 12 years ago
Closed 12 years ago
#5011 closed Bugs (wontfix)
comparison/equal_to should work recursively
Reported by: | Sebastian Redl | Owned by: | Joel de Guzman |
---|---|---|---|
Milestone: | To Be Determined | Component: | fusion |
Version: | Boost 1.45.0 | Severity: | Problem |
Keywords: | Cc: |
Description
Suppose I have a struct that is adapted to Fusion, and it contains an array:
typedef int i100[100]; need a typedef due to ADAPT_STRUCT limitation struct with_ar {
double d1; double d2; i100 ar;
}; BOOST_FUSION_ADAPT_STRUCT(::with_ar, (double d1)(double d2)(i100, ar));
Now I want to use the Fusion comparisons with this:
#include <boost/fusion/comparison/equal_to.hpp> with_ar wa1 = { 0.0, 0.0, {1, 2 } },
wa2 = { 0.0, 0.0, {1, 2 } };
assert(boost::fusion::equal_to(wa1, wa2));
What I expect is that equal_to compares d1 and d2, and then compares ar as a Fusion sequence (the array adapter is included). However, the assertion fails, because equal_to compares the arrays with ==, which compares addresses. equal_to should be recursive, using equal_to to compare elements that are Fusion sequences, and == only for other things.
I have hacked my local Fusion to do this, and it seems to work fine, but the hack is very ugly because of the recursive dependency between comparison/equal_to.hpp and comparison/detail/equal_to.hpp. Basically, I just add another version of equal_to that just does == on the arguments. SFINAE on is_sequence is used to decide between the two. The element comparison in sequence_equal_to then calls equal_to on the elements instead of using ==.
The only problem I can see would be that equal_to would be used for adapted structs that have a specialized ==, but I think that would be a rather strange case.
This is not possible without breaking compatibility with Boost.Tuples. I'll be closing this ticket as no-fix. If you have objections, let's discuss this on list. There are alternate solutions.