P1371 Pattern Matching

[jensmaurer]

P1371R0 Pattern Matching (Sergei Murzin, Michael Park, David Sankel, Dan Sarginson)

[jfbastien]

P1260R0 Pattern Matching #363 and P1308R0 Pattern Matching #364 were seen by EWGI in SAN.
Strong support, authors got feedback to address.

[jensmaurer]

EWG in Kona

[wg21bot]

P1371R1 Pattern Matching (Sergei Murzin, Michael Park, David Sankel, Dan Sarginson)

[jensmaurer]

EWG in Cologne

[jfbastien]

EWG saw this: http://wiki.edg.com/bin/view/Wg21belfast/P1371-EWG

EWG is interested in hearing more about pattern matching in the general direction presented

SF	F	N	A	SA
20	9	1	0	0

[wg21bot]

P1371R2 Pattern Matching (Sergei Murzin, Michael Park, David Sankel, Dan Sarginson)

[jfbastien]

EWG Prague on Tuesday afternoon:

Pattern matching should require a case that always matches.

SF	F	N	A	SA
8	10	12	5	4

Assuming we don’t mandate a case that always matches, if no case matched, we get UB.

SF	F	N	A	SA
8	14	7	6	9

Drop the expression form.

SF	F	N	A	SA
6	7	8	7	12

Drop the statement form.

SF	F	N	A	SA
0	1	5	12	23

Pattern matching shouldn’t have arbitrary lookahead.

SF	F	N	A	SA
31	10	3	1	0

[jfbastien]

EWG Prague Friday afternoon, discuss design tuning: an update from earlier this week, Pattern_Matching_Updates.pdf

Non-void inspect expressions should have at least one pattern which, on its own, can match every possible value.

SF	F	N	A	SA
0	0	2	11	13

[jfbastien]

We discussed exhaustiveness checking in today's EWG telecon, and took the following poll:

POLL We like the presented approach to exhaustiveness checking in pattern matching.

SF	F	N	A	SA
6	11	5	2	1

[jfbastien]

EWG discussed "inspect is always an expression" in a telecon today.

POLL: Allow break and continue leaving from inspect expressions.

SF	F	N	A	SA
7	5	6	1	1

Lots of other feedback was given to the authors, and everyone agreed to more exploration. We did not poll these.

[wg21bot]

P1371R3 Pattern Matching (Michael Park, Bruno Cardoso Lopes, Sergei Murzin, David Sankel, Dan Sarginson, Bjarne Stroustrup)

[jfbastien]

Today in the EWG telecon, we discussed goto in pattern actions. Polls:

Support goto jumping into a pattern.

SF	F	N	A	SA
0	0	1	4	19

Support goto jumping out of a pattern, when the goto jumps forward, not across a variable declaration or into a nested scope.

SF	F	N	A	SA
1	12	6	3	1

Support goto jumping out of a pattern, without additional constraints.

SF	F	N	A	SA
2	3	11	5	1

Support goto which remains strictly within its pattern.

SF	F	N	A	SA
7	8	6	0	0

Bjarne wrote a short paper to start our discussion, and will come back with a bigger paper which covers all the cases and feedback received.

[jfbastien]

Bruno presented his implementation experience to EWG on 2021-02-25.
Godbolt demo: https://godbolt.org/z/fdd5j4
Feedback was given to the authors.

[jfbastien]

EWG discussed Pattern matching identifier patterns syntax: let vs auto vs none on April 8th.

POLL: annotate id-pattern with auto.

inspect (e) { // like this:
 
  auto&& x => binds

  x => matches

}

SF	F	N	A	SA
0	3	4	3	8

POLL: if we annotate id-patterns with let.

inspect (e) { // like this:

  let x => binds

  x => matches

}

SF	F	N	A	SA
3	6	6	2	2

POLL: keep the current p1371r3 proposal with case.

inspect (e) { // like this:

  x => // binds

  case x => matches

}

SF	F	N	A	SA
2	6	8	0	3

[erichkeane]

Adding needs-revision, this got EWG feedback in April of '21.