P3086 R5 Proxy: A Pointer-Semantics-Based Polymorphism Library

[wg21bot]

P3086R0 Proxy: A Pointer-Semantics-Based Polymorphism Library (Mingxin Wang)

[inbal2l]

Related to P0957 (#306)
@nliber @billy-baker - please look at this first, preferably before Tokyo

[billy-baker]

2024-03-18 Library Evolution Incubator Telecon

P3086R1: A Pointer-Semantics-Based Polymorphism Library

2024-03-18 Library Evolution Incubator Minutes

Champion: Mingxin Wang

Chair: Billy Baker

Minute Taker: Elias Kosunen

Start: 2024-03-18 13:40 UTC + 9

Does this paper have:

• Why this facility should go into the standard library?
• Examples? Yes
• Field experience?
  • Implementation experience? Yes
  • Usage experience? Yes
  • Deployment experience?
• Performance considerations? Yes
• Discussion of prior art? Yes
• Changes Library Evolution previously requested?
• Support for "std::format", "hash", and allocators (if required)? N/A
• Wording? Yes
• Feature test macro? Yes
• Freestanding? No

POLL: The problem presented in P3086R1 Proxy: A Pointer-Semantics-Based Polymorphism Library is a problem worth solving.

Attendance: 11

# of Authors: 1

Author Position:

Outcome: Unanimous consent to continue with P3086.

POLL: Remove make_proxy from P3086.

Attendance: 11

# of Authors: 1

Author Position:

Outcome: Unanimous consent for the next revision to remove make_proxy.

End: 16:15 UTC + 9

Summary

While there has been usage of the github library, it is unknown if any of the usage was from embedded development.

The non-intrusive capability was well-liked.

The paper shows macros (not proposed for standardization) to avoid the boilerplate necessary to define a facade. Some non-macro solution in this area should be considered for future standardization.

Removal of make_proxy for lack of allocator support improves consensus. A separate future paper could explore this more. The pattern of make_shared/allocate_shared was given as a possible solution.

The paper proposes adding a header. The facilities of P3019 "indirect and polymorphic: Vocabulary Types for Composite Class Design" were added to the header. There should be more discussion on where the facilities of P3086 should reside.

Next Steps

Another revision of the paper is needed that removes make_proxy. Some discussion of P3019 "indirect and polymorphic: Vocabulary Types for Composite Class Design" would be good to have in the next revision. Alternate naming suggestions, if possible, would also be good to have in the next revision. As noted, the wording needs to use tuple-like in two places.