P0554 Composition of Arithmetic Types

[wg21bot]

P0554R0 Composition of Arithmetic Types (John McFarlane)

[jensmaurer]

LEWG-I in Kona: Design feedback.

[johnmcfarlane]

LEWGI review

[wg21bot]

P0554R1 Composition of Arithmetic Types (John McFarlane)

[brycelelbach]

Kona 2019-02 LEWGI Minutes

P0828R1 Elastic Integers
P0037R6 Fixed Point Integers
P0554R0 Composition of Arithmetic Types

Champion: Davis Herring

Minute Taker: Mark Hoemmen

Start P0828R1 Overview: 10:06

Start P0037R6 Overview: 10:30

Consistency with std::chrono durations and std::ratio?

Start P0554R0 Overview: 11:16

Start Polling: 11:28

POLL: Given what we know right now, we like the general library design of the compositional approach in P0554R0/etc (example: fixed_point<elastic_integer, Exponent>)

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
6	3	3	1	0

Attendance: 16

That has consensus.

Jeffrey (A): Worried about complexity; having all of these options available might confuse people.

Break for Lunch: 11:33

Resume after Lunch: 13:07

End: 13:52

[brycelelbach]

Cologne 2019-07 LEWGI Minutes

P0554R1 Compositional Numeric Types: Design Feedback

P1751R0 Numeric Type Families: Design Feedback

Champion: John McFarlane

Minute Taker: Billy O'Neal

Start Overview: 07-16 13:42

Start Review: 14:03

Composition design - Unary adaptor types that compose vs an Nary type with multiple properties.

// Unary adaptor types that compose.
template <typename Rep> struct a_integer;
template <typename Rep> struct b_integer;
template <typename Rep> struct c_integer;

vs

// Nary type with ordered properties.
struct a_tag {};
struct b_tag {};
struct c_tag {};

template <typename... Properties>
struct integer;

POLL: We want unary adaptor types that compose instead of exploring a type with nary ordered properties.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
4	7	2	2	1

A: Would love to see other designs explored.

That has consensus.

Attendance: 18

Explore a compositional helper.

POLL: We should promise more committee time to pursuing new arithmetic types in the standard library that satisfy the use cases described in P0554, knowing that our time is scarce and this will leave less time for other work.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
6	7	2	1	0

Attendance: 16

This has strong consensus.

elastic_integer<20, int> // Status quo.
elastic_integer<70, wide_integer<31>> // Status quo, works through traits.
elastic_integer<20, int_family>

POLL: We are okay with representing families of arithmetic types with a single, unique member of that family and traits (e.g. int for signed integers, unsigned for unsigned integers).

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
0	2	5	7	2

Attendance: 18

End: 15:35

CONSENSUS: Bring a revision of P0554R1, with the guidance below, to LEWGI for further design feedback.

• Consider merging the various other integer type papers targeting the Numeric TS into P0554.
• Produce draft wording.
• Explore an alternative design to numeric type families which does not rely on using a single unique member of each family and traits to represent the family.

[johnmcfarlane]

Cologne minutes

[johnmcfarlane]

Cologne slides