Removing SG6, adding LWG

SG6 brought up a question for LEWG here about a separate bug.

SG6 brought up a question for LEWG here about a separate bug.

We need to know what decisions Library Evolution needs to make here; this is unclear to me from the paper. Whatever question came up in SG6 needs to be clearly defined.

The authors have identified the questions for Library Evolution.

Is generate_canonical's second template parameter really meant to be "bits" or should it be "digits in the radix of the floating-point type in question" (as would be consistent with the "numeric_limits<RealType>::digits" used in conjunction with it).

LEWG Mailing list review, 2020-07

We didn't have sufficient participation to be very affirmative about anything, but there is some agreement that generate_canonical's second template parameter should be "bits"However, there are some concerns about the definition of the function and its inherent bias, and lack of uniformity.We will defer to SG-6 to answer these questions.

Sending to SG6

P0952R0: A new specification for std::generate_canonical

2022-03-17 SG6 Telecon Minutes

Chair: Matthias Kretz

Champion: Thomas Köppe and Davis Herring

Minute Taker: Dietmar Kühl

Summary

The main question from LWG to LEWG, how to interpret the second template parameter, was deferred from LEWG to SG6. SG6 made a unanimous decision to interpret the template parameter as digits in the floating-point radix. SG6 discussed a few more details of the paper but did not request any changes.

Polls

POLL: The second template argument for generate_canonical should be interpreted as a number of digits in the floating-point radix.

No objections to unanimous consent.

# of Authors: 2
# of Participants: 8

POLL: Deprecate generate_canonical.

# of Authors: 2
# of Participants: 8
Outcome: no consent
Comments: First merge the functionality into something else, then consider deprecation.

POLL: Let generate_canonical produce a member of an arithmetic progression of floats with equal probability.

No objections to unanimous consent.

# of Authors: 2
# of Participants: 8

POLL: Choosing a member of the arithmetic progression is QoI.

# of Authors: 2
# of Participants: 8
Outcome: no consent

POLL: Instruct the author to apply the bits/digits wording change and send a revision with wording of the paper to LWG.

No objections to unanimous consent.

# of Authors: 2
# of Participants: 8

P0952R1 A new specification for std::generate_canonical (Thomas Koeppe)

Notes: https://wiki.edg.com/bin/view/Wg21kona2023/LibraryWorkingGroup

Poll: Set LWG2524 to tentatively ready by adopting P0952r2?

|F|A|N|
|10|0|0|

LWG 2023-11-10: Approved for plenary straw polls in Kona.

P0952R2 A new specification for std::generate_canonical (Thomas Köppe, Davis Herring)

@JeffGarland: LWG2524 still shows as Open.

@JeffGarland: LWG2524 still shows as Open.

I've closed it as Resolved