US313 25.02 [algorithms.requirements] Properly define scope of subclause to cover [specialized.algorithms]

[wg21bot]

This subclause describes general purpose wording that applies to all algorithms in the standard, without defining algorithm. It generally applies such definitions to algorithms "in this clause", but the wording for specialized algorithms in 20.10.11 [specialized.algorithms] relies on this wording too, especially to provide definitions for its ranges overloads.

Proposed change:
Revise this subclause to include a definition of algorithm, so that all the wording that applies to this subclause instead applies to all algorithms. Reasonable definitions of algorithm (for the purposes of the standard library) might be all function templates in clause 25, and clause 20.10.11, or some definition constructed around function templates having arguments of iterator or range type. The former is likely a simpler fix for C++20, the latter would avoid having to update the list of locations for algorithms in the future,

[jwakely]

Reviewed by LWG small group on Monday morning.

[BillyONeal]

Reviewed in LWG this afternoon. @BillyONeal to write a paper which

• moves the algorithms from [specialized.algorithms] to [algorithms], following the algorithms that were moved from numerics in this standardization cycle as a model
• leaves addressof, construct_at, and destroy_at in their existing location; they aren't 'classical STL algorithms'
• deletes duplicated front matter in [specialized.algorithms] covered by the existing [algorithms] front matter

[jensmaurer]

@BillyONeal , the suggested changes appear to all be editorial. Please clarify with @mclow whether you want LWG review for the changes or whether the editors can just do it.

[BillyONeal]

I already wrote D1963R0.pdf and put it on the LWG wiki. It didn't end up being quite editorial.

[mclow]

LWG on Saturday in Belfast: Resolved by adoption of P1963 to be adopted in Prague.