Refactor[MWC]: explicit operator= in mwcc containers

[678098]

Removing the custom copy constructor OrderedHashMap_SequentialIterator(const NcIter& other); doesn't work since it produces compilation errors

In file included from /blazingmq/src/groups/bmq/bmqimp/bmqimp_brokersession.cpp:17:
In file included from /blazingmq/src/groups/bmq/bmqimp/bmqimp_brokersession.h:37:
In file included from /blazingmq/src/groups/bmq/bmqimp/bmqimp_messagecorrelationidcontainer.h:39:
/blazingmq/src/groups/bmq/bmqp/bmqp_requestmanager.h:1509:30: error: no viable conversion from 'OrderedHashMap_SequentialIterator<pair<...>>' to 'OrderedHashMap_SequentialIterator<const pair<...>>'
    for (RequestMapConstIter it = requestsCopy.begin();
                             ^    ~~~~~~~~~~~~~~~~~~~~

Implemented explicit operator= for OrderedHashMap_SequentialIterator and OrderedHashMapWithHistory_Iterator

[678098]

Note that I decided to not to add checks for self-assignment, since we don't assign many fields, and just assign will be much faster than if-branch.

We don't check for self-assignment for some small classes, for example

template <class VALUE>
inline WeakMemFnResult<VALUE>&
WeakMemFnResult<VALUE>::operator=(const WeakMemFnResult& original)
{
    d_value.object() = original.d_value.object();
    return *this;
}

[pniedzielski]

This makes sense, this is such a cheap object that we shouldn't worry about self-assignment.

[ddragan-bloomberg]

From what I gathered the OrderedHashMap_SequentialIterator& operator=(const NcIter&) definition exists so that a const iterator is constructible from a non-const itertator, but the same definition also might act as the copy constructor. Took me a while to figure this out. I think this is convoluted. A better way would be to:

• not define the copy constructor, not the copy operator;
• define a special from-const-iterator constructor.

[678098]

• not define the copy constructor, not the copy operator;
• define a special from-const-iterator constructor.

If I remove the original OrderedHashMapWithHistory_Iterator(const NcIter& other); copy constructor, I will get compilation errors:

blazingmq/src/groups/mwc/mwcc/mwcc_orderedhashmapwithhistory.h:698:12: error: no matching conversion for functional-style cast from 'const iterator' (aka 'const OrderedHashMapWithHistory_Iterator<Value>') to 'const_iterator' (aka 'OrderedHashMapWithHistory_Iterator<const Value>')
    return const_iterator(d_first);

These errors will remain even if I define from-const-iterator constructor

typedef typename bsl::add_const<VALUE>::type      CType;
typedef OrderedHashMapWithHistory_Iterator<CType> CIter;
OrderedHashMapWithHistory_Iterator(const CIter& other);

[ddragan-bloomberg]

I just realized that the assignment operator is also double-purpose. It allows to assign a non-const iterator to a const one, and services as the copy-assignment operator. So it should be explicitly declared and implemented.

-- this code is fine

[pniedzielski]

Small nit, if you want to fix, go ahead and I'll reapprove; otherwise looks good.

[pniedzielski]

Nit: Because of the documentation comment, I think we want an rhs name here (like in the other commit in this PR):
OrderedHashMap_SequentialIterator& operator=(const NcIter& rhs);

[678098]

Oh, I already force-pushed a fix for this. Might be GH lagging

[pniedzielski]

Huh, you're right! Looks good, go ahead and merge.

[pniedzielski]

This makes sense, this is such a cheap object that we shouldn't worry about self-assignment.