[flat.map.modifiers] Remove redundancy in 'insert(sorted_unique, i, j)'

[jwakely]

We can specify this in terms of the overload without the sorted_unique tag. That is consistent with how the equivalent functions in flat_set and flat_multiset are specified.

LWG asked for this in the telecon today.

[jwakely]

To give some more context: the wording for insert(i, j) is not word-for-word identical to the wording for insert(sorted_unique, i, j) because the latter has the additional text "sorts the range of newly inserted elements with respect to value_comp();" which implies it does more work.

However, if the contents of the range [i, j) is already sorted then that step is a no-op. The implementation can elide that step, because it doesn't say the effects are equivalent to actually calling ranges::sort, it just says "sorts the range". If the implementation happens to know it's already sorted, it can skip that step.

So specifying insert(sorted_unique, i, j) in terms of insert(i, j) with a more restrictive complexity guarantee completely described the effects, and specifies the complexity.

And [flat.set.modifiers] p8 and [flat.multiset.modifiers] p7 already take exactly this specification approach, just deferring to insert(i, j) instead of duplicating 95% of its wording.

[jwakely]

And [structure.specifications] p4 says that the Complexity: element here overrules the implied Complexity: that we would otherwise have from using "Effects: Equivalent to ..."

[tkoeppe]

Thanks for the explanations! Yes, it certainly seems technically correct, and a good deduplication. The more restrictive complexity may be a bit subtle, but then again, the Standard is not a tuturial, and implementers will know how to read this.