[ASTPrinter] Intorduce 'IntroducerKeyword' name kind #38521
Conversation
|
@swift-ci Please smoke test |
lib/AST/ASTPrinter.cpp
Outdated
| if (!Options.SkipIntroducerKeywords) { | ||
| if (decl->isExplicitActor()) { | ||
| Printer.printKeyword("actor", Options, " "); | ||
| } else { | ||
| Printer << tok::kw_class << " "; | ||
| } | ||
| } | ||
| if (!Options.SkipIntroducerKeywords) | ||
| Printer.printIntroducerKeyword( | ||
| decl->isExplicitActor() ? "actor" : "class"); |
There was a problem hiding this comment.
Since actor is not declared as a keyword, not all introducer are "keyword". There's no reason to use tok::kw_* for introducer printing anymore.
include/swift/AST/ASTPrinter.h
Outdated
| @@ -217,6 +219,14 @@ class ASTPrinter { | |||
| *this << Suffix; | |||
| } | |||
|
|
|||
| void printIntroducerKeyword(StringRef name, | |||
| StringRef Suffix = " ") { | |||
There was a problem hiding this comment.
I think we should keep the default Suffix the same between printKeyword and printIntroducerKeyword, otherwise it's confusing.
| @@ -1292,9 +1293,8 @@ void PrintAST::printPattern(const Pattern *pattern) { | |||
|
|
|||
| case PatternKind::Binding: | |||
| if (!Options.SkipIntroducerKeywords) | |||
| Printer << (cast<BindingPattern>(pattern)->isLet() ? tok::kw_let | |||
There was a problem hiding this comment.
Should we move checking SkipIntroducerKeywords into printIntroducerKeyword? That way you cannot forget, and also we could give the method the same signature as printKeyword, which also takes options.
rintaro
force-pushed
the
astprinter-introducerkeyword
branch
2 times, most recently
from
9222db0 to
c535893
Compare
|
@swift-ci Please smoke test |
lib/AST/ASTPrinter.cpp
Outdated
| Printer.printKeyword("__consuming", Options, " "); | ||
| } | ||
| Printer << tok::kw_func << " "; | ||
| printMutabilityModifiersIfNeeded(decl); |
There was a problem hiding this comment.
Will these (mutability and __consuming) now get printed in places they didn't used to due to not checking SkipIntroducerKeywords? Not sure if that's desirable or not 🤔
There was a problem hiding this comment.
Renamed printMutabilityModifiersIfNeeded() to printSelfAccessKindModifiersIfNeeded() and moved __consuming printing into it. Also, added a printing option whether to print mutating/nonmutating/__consuming.
So far SkipIntroducerKeywords is used only in <RelatedName /> printing in SourceKit.
rintaro
force-pushed
the
astprinter-introducerkeyword
branch
2 times, most recently
from
d5ae9c9 to
c113ba0
Compare
|
@swift-ci Please smoke test |
For more fine grained annoations. For now, it's handled as the same as 'Keyword' name kind. Fix an issue where 'extension' wasn't marked as "keyword". Also, move 'static' priting out of 'SkipIntroducerKeywords' guard because 'static' is not an declaration introducer.
rintaro
force-pushed
the
astprinter-introducerkeyword
branch
from
c113ba0 to
6e69a44
Compare
|
@swift-ci Please smoke test |
| break; | ||
| case SelfAccessKind::NonMutating: | ||
| if (AD && AD->isExplicitNonMutating() && | ||
| !Options.excludeAttrKind(DAK_NonMutating)) |
There was a problem hiding this comment.
This was checking DAK_Mutating before, which seems like a latent bug, but I don't think it's observable since the only place I found that adds DAK_Mutating but not DAK_NonMutating is protocol conformance diagnostics for non-accessors, but this code only prints nonmutating on accessors. Oh well, at least it's correct now 🤷
For more fine grained annotations. For now, it's handled as the same as
Keywordname kind.Fix an issue where
extensionwasn't marked as "keyword".Also, move
staticpriting out ofSkipIntroducerKeywordsguard because 'static' is not an declaration introducer.(Preparation for #38453)