[CP] [parser] Support nullable record types in patterns

[stereotype441]

Commit(s) to merge

284296f

Target

Stable

Prepared changelist for beta/stable

https://dart-review.googlesource.com/c/sdk/+/305824

Issue Description

If a variable pattern or wildcard pattern has an explicit type which is a nullable record type, there is a parse error. For example, this code:

f(Object? x) {
  switch (x) {
    case ((int, int)? x, String y):
      break;
  }
}

is rejected with the error message:

.../test.dart:3:23: Error: Expected ')' before this.
    case ((int, int)? x, String y):
                      ^

What is the fix

The parser's disambiguation rules are modified so that if a pattern starts with (, and the matching ) is followed by ? identifier, an attempt is made to parse the pattern as a variable pattern. Previously, the disambiguation rules only attempted to parse a pattern starting with ( as a variable pattern if the matching ) was followed by an identifier (i.e. a variable pattern with a non-nullable record type).

Why cherry-pick

Without this fix, users cannot express a pattern match using a variable or wildcard pattern with a nullable record type. It is not clear from the error message why the code is being rejected.

If we don't cherry-pick this fix, the fix will be released as part of Dart 3.1, so there will be a danger of a package author releasing code that is intended to be compatible with Dart 3.0, but which fails to work due to its use of a variable pattern with an explicit nullable record type.

The fix is low risk, because it is confined to the disambiguation logic that fires specifically for patterns that begin with (. The only syntax that is newly recognized as a variable pattern is (...)? identifier. Any ambiguities that might arise from this interpretation are also problems for the syntax (...) identifier (e.g. (...) when or (...) as), and the resolution of these ambiguities is already taken care of and well tested.

Risk

low

Issue link(s)

#52439

Extra Info

No response

[mit-mit]

SGTM