P1998 Simple Facility for Lossless Integer Conversion

[wg21bot]

P1998R0 Simple Facility for Lossless Integer Conversion (Ryan McDougall)

[wg21bot]

P1998R1 Simple Facility for Lossless Integer Conversion (Ryan McDougall)

[Cpp-Lisa]

Reviewed by SG6 Monday in Prague. We generally approve. The "narrow" naming caused some confusion about the intent and limitations of the function (e.g., can it be used to perform a strictly widening conversion?).

[brycelelbach]

Prague 2020-02 LEWGI Minutes

P1998R0 std::narrow[_cast]: Design Review

Chair: Billy Baker

Champion: Ryan McDougall

Minute Taker: David Olsen

Start Review: 2020-02-11 9:19

Implementation experience:

• Yes, for everything but the compile time literals (explicit mention in paper).

Usage experience:

• Yes, within Ryan's company (explicitly mention in paper).

Prior art:

• boost::numeric_cast (add to paper).
• gsl::narrow (add to paper).

convertible not convertable.

It seems novel that this proposal suggests that the noexcept narrow should specify that it terminates, instead of just saying UB on precondition failure.

narrow_or in its current form shouldn't be noexcept, under the current noexcept guidance.

Explore other prior art, such as gsl::narrow_cast.

Drop the literals?

Start Polling: 9:40

POLL: We should promise more committee time to pursuing P1998, knowing that our time is scarce and this will leave less time for other work.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
5	14	2	0	0

Attendance: 22

# of Authors: 1

Author Position: SF

POLL: For a lossless integer conversion, a solution that throws is desired (i.e. std::narrow_cast).

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
4	8	4	1	0

Attendance: 22

# of Authors: 1

Author Position: Abstained

The consensus is that a solution that throws is desired.

POLL: For a lossless integer conversion, a solution that aborts the program in some manner is desired (i.e. std::narrow).

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
1	5	7	3	2

Attendance: 22

# of Authors: 1

Author Position: SF

SA: Should not set precedent before contracts show up

N: Just the precondition would have been fine, but uncertain when contracts are not here

A: You can do this with std::narrow_or.

No consensus to provide a solution that aborts a program.

POLL: For a lossless integer conversion, a solution that allows invoking a Callable is desired (i.e. std::narrow_or).

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
4	9	5	0	0

Attendance: 22

# of Authors: 1

Author Position: SF

The consensus is that a solution that allows invoking a Callable is desired.

POLL: For a lossless integer conversion, a solution that provides literal suffixes is desired.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
0	2	9	6	1

Attendance: 22

# of Authors: 1

Author Position: Not recorded

SA: I don't think it is fundamental to the proposal (could be added later)

WA: Teaching people to use both is harder than teaching a single context (i.e. the throwing one)

No consensus to provide literal suffixes.

End: 09:59

Consensus: LEWGI sends P1998R1 (Simple Facility for Lossless Integer Conversion), with the guidance below, to LEWG for the Numerics TS.

• Remove the proposed solution that aborts a program (narrow).
• Remove literal suffixes.