[CodeCompletion] Provide basic code completion support for Swift KeyPath. rdar://31768743 #9467
Conversation
|
@swift-ci please smoke test |
|
@swift-ci please smoke test |
| // PERSON-MEMBER-OPT: Begin completions, 7 items | ||
| // PERSON-MEMBER-OPT-NEXT: Decl[InstanceVar]/CurrNominal/Erase[1]: ?.name[#String#]; name=name | ||
| // PERSON-MEMBER-OPT-NEXT: Decl[InstanceVar]/CurrNominal/Erase[1]: ?.friends[#[Person]#]; name=friends | ||
| // PERSON-MEMBER-OPT-NEXT: Decl[InstanceVar]/CurrNominal/Erase[1]: ?.bestFriend[#Person?#]; name=bestFriend |
There was a problem hiding this comment.
Aren't these incorrect for PART_4? The result will be \Person.bestFriend??.name, etc. Shouldn't it be just one ??
There was a problem hiding this comment.
The generic property keyPath isn't landed in the compiler yet. I will drop the test case.
lib/Parse/ParseExpr.cpp
Outdated
| @@ -1136,8 +1139,17 @@ Parser::parseExprPostfixSuffix(ParserResult<Expr> Result, bool isExprBasic, | |||
|
|
|||
| // Handle "x.<tab>" for code completion. | |||
| if (Tok.is(tok::code_complete)) { | |||
| if (CodeCompletion && Result.isNonNull()) | |||
| if (CodeCompletion && Result.isNonNull()) { | |||
| if (InSwiftKeyPath) | |||
There was a problem hiding this comment.
For clarity, please put {} braces around these since they're split on several lines and it's important to know that the completeDoctExpr is not part of the else.
| @@ -525,6 +525,7 @@ ParserResult<Expr> Parser::parseExprUnary(Diag<> Message, bool isExprBasic) { | |||
| ParserResult<Expr> Parser::parseExprKeyPath() { | |||
| // Consume '\'. | |||
| SourceLoc backslashLoc = consumeToken(tok::backslash); | |||
| llvm::SaveAndRestore<SourceLoc> slashLoc(SwiftKeyPathSlashLoc, backslashLoc); | |||
There was a problem hiding this comment.
Why not use backslashLoc directly? It's not being modified.
There was a problem hiding this comment.
This basically to save SwiftKeyPathSlashLoc in the parser context so parseExprPostfixSuffix can use it to create an instance of KeyPathExpr.
| @@ -538,6 +539,8 @@ ParserResult<Expr> Parser::parseExprKeyPath() { | |||
| } | |||
|
|
|||
| if (startsWithSymbol(Tok, '.')) { | |||
| llvm::SaveAndRestore<Expr*> S(SwiftKeyPathRoot, rootResult.getPtrOrNull()); | |||
There was a problem hiding this comment.
Same question for rootResult
There was a problem hiding this comment.
If we are not using SaveAndRestore, we are not able to use this part in parseExprPostfixSuffix.
There was a problem hiding this comment.
Ah, sorry I misread this. For some reason I didn't notice they were in completely different functions ,etc.
lib/IDE/CodeCompletion.cpp
Outdated
| case CompletionKind::SwiftKeyPath: { | ||
| Lookup.setHaveDot(DotLoc); | ||
| Lookup.setIsSwiftKeyPathExpr(); | ||
| Lookup.getValueExprCompletions((*ExprType)->getAs<BoundGenericType>()-> |
There was a problem hiding this comment.
What's going on here with BoundGenericType and getGenericArgs()[1]?
|
@swift-ci please smoke test |
|
Maybe put a FIXME in the test for the missing case so it's obvious that PART 4 isn't actually checking anything. Otherwise, LGTM thanks! |
|
@swift-ci please smoke test |
…ath. rdar://31768743 (#9467)
This patch starts libIDE support for code completing keyPath expression; after each
.in key path component, code completion provides type-sensitive member properties and subscripts. This patch doesn't implement code completion for contextual-inferred keyPath root type like\.property.[0].