P2592 Hashing support for std::chrono value classes

[wg21bot]

P2592R0 Hashing support for std::chrono value classes (Giuseppe D'Angelo)

[ben-craig]

P2592R0: Hashing support for std::chrono value classes

P2592R0: Hashing support for std::chrono value classes

2022-06-14 Library Evolution Telecon Minutes

Chair: Ben Craig

Minute Taker: Inbal Levi

Champion: Giuseppe D'Angelo

Summary

day can hold unspecified values. This may be the first time that hash has had to deal with that, so the wording needs to address the issue. What happens if you hash an unspecified value?

What about heterogenous lookups? Durations can have different periods, but still compare equal. For example, 1s and 1000ms compare equal. If someone has a map with seconds as the key, do we want to be able to index into the map with milliseconds?

Next Steps

Bring back a new revision that addresses the topics discussed. "Addressing" the topic doesn't necessarily require solving the issue, but the pros and cons need to be discussed at a minimum.

[wg21bot]

P2592R1 Hashing support for std::chrono value classes (Giuseppe D'Angelo)

[brycelelbach]

2022-09-06 Library Evolution Telecon

P2592R1: Hashing support for std::chrono value classes

2022-09-06 Library Evolution Telecon Minutes

Chair: Fabio Fracassi

Minute Taker: Mark Hoemmen

Champion: Giuseppe D'Angelo

POLL: Update P2592R1 with the discussed issues addressed and forward it to C++26 (to be confirmed by Electronic Polling).

SF	F	N	A	SA
6	7	0	0	1

SA: This is fundamentally unsound that information in the type system that represents these values is not represented in the hash. These things in the type system are part of the distinguishment; they should be hashed.

Outcome: Consensus in favor

Author position: SF

Present: 20

Summary

TODO: @FabioFracassi

Next Steps

Update P2592R1 (Hashing support for std::chrono value classes) with the discussed issues addressed and then take an electronic poll to send it to Library for C++26 classified as B2 - Improvement.