[Sema] Introduce a couple of ExprPattern requests #64174
Conversation
This replaces `synthesizeTildeEqualsOperatorApplication`, and synthesizes the match expression and var on-demand. Additionally, it pushes the lookup logic into pre-checking.
It's not clear to me why it was ever set this way, but it prevents IUOs from working with the match operator. Unset it, and let pre-checking correctly assign it a `FunctionRefKind::SingleApply`.
This is needed to ensure we only ever synthesize a single unique ExprPattern when solving an EnumElementPattern that we failed to lookup a member for.
hamishknight
requested review from
hborla,
xedin,
slavapestov,
ahoppen,
bnbarham,
CodaFi,
DougGregor and
rintaro
as code owners
|
@swift-ci please test |
|
@swift-ci please test source compatibility |
| @@ -214,7 +214,8 @@ deriveBodyCodingKey_enum_stringValue(AbstractFunctionDecl *strValDecl, void *) { | |||
| for (auto *elt : elements) { | |||
| auto *baseTE = TypeExpr::createImplicit(enumType, C); | |||
| auto *pat = new (C) EnumElementPattern(baseTE, SourceLoc(), DeclNameLoc(), | |||
| DeclNameRef(), elt, nullptr); | |||
| DeclNameRef(), elt, nullptr, | |||
| /*DC*/ strValDecl); | |||
There was a problem hiding this comment.
Nit: I think the preferred style for these kinds of named arguments is the following, which is detected by clang-format and removes the space after the comment.
Same below
| /*DC*/ strValDecl); | |
| /*DC=*/strValDecl); |
There was a problem hiding this comment.
Seems like clang-format is happy with either style, if the = isn't present, it inserts a space, otherwise it removes it. Both styles seem to be widely in use in the codebase:
swift on pattern-requests
❯ rg "/\*\w+\*/" lib | wc -l
2331
swift on pattern-requests
❯ rg "/\*\w+=\*/" lib | wc -l
2893
Personally I much prefer the variant without the = and with the space, it feels less claustrophobic. https://llvm.org/docs/CodingStandards.html#comment-formatting does give an example of a comment that uses the =, but it's unclear if it's saying that it's the preferred style.
I can change this if you want, but I'll do it as a follow-up.
There was a problem hiding this comment.
I prefer /*<ID>=*/ because that's what we mostly use in the solver but I don't usually pay much attention to formatting there.
There was a problem hiding this comment.
Seems like clang-format is happy with either style, if the = isn't present, it inserts a space, otherwise it removes it.
FWIW I'm pretty sure the "happy with either style" is really just because it adds whitespace around /**/ normally (as opposed to the special case when it has =). I'm surprised how many non-= there are, I've basically always seen = 🤔
There was a problem hiding this comment.
I think it varies by component, e.g:
With = |
Without = |
|
|---|---|---|
| IDE | 170 | 41 |
| IRGen | 420 | 144 |
| SILOptimizer | 122 | 260 |
| Serialization | 46 | 96 |
There was a problem hiding this comment.
Hmm, OK. I didn’t know it was this mixed. It’s not a blocker for me either, just something I noticed. I’m happy to merge it as-is as well.
Introduce a request to synthesize the
~=application for an ExprPattern, and a request to synthesize an ExprPattern for the fallback type-checking of an EnumElementPattern when we are unable to lookup its member. This has been split off #63963.