[WIP][Macro] Emit error if AccessorMacro is applied to node that is not a variable #2624
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change | ||||
|---|---|---|---|---|---|---|
|
|
@@ -633,6 +633,7 @@ private enum MacroApplicationError: DiagnosticMessage, Error { | |||||
| case accessorMacroOnVariableWithMultipleBindings | ||||||
| case peerMacroOnVariableWithMultipleBindings | ||||||
| case malformedAccessor | ||||||
| case macroAttachedToInvalidDecl(String, String) | ||||||
|
|
||||||
| var diagnosticID: MessageID { | ||||||
| return MessageID(domain: diagnosticDomain, id: "\(self)") | ||||||
|
|
@@ -650,6 +651,8 @@ private enum MacroApplicationError: DiagnosticMessage, Error { | |||||
| return """ | ||||||
| macro returned a malformed accessor. Accessors should start with an introducer like 'get' or 'set'. | ||||||
| """ | ||||||
| case let .macroAttachedToInvalidDecl(macroType, declType): | ||||||
| return "'\(macroType)' macro cannot be attached to \(declType)" | ||||||
| } | ||||||
| } | ||||||
| } | ||||||
|
|
@@ -666,6 +669,7 @@ private class MacroApplication<Context: MacroExpansionContext>: SyntaxRewriter { | |||||
| /// Store expanded extension while visiting member decls. This should be | ||||||
| /// added to top-level 'CodeBlockItemList'. | ||||||
| var extensions: [CodeBlockItemSyntax] = [] | ||||||
| var expandedAttributes: [AttributeSyntax] = [] | ||||||
|
There was a problem hiding this comment. I think this can be Also, I just tried to suggest a doc comment for it and couldn’t really come up with a great explanation, which is usually a sign that the design isn’t quite right (maybe you can find a good explanation). Thinking about it made me wonder if it would it be easier to
/// Attributes that look like macro attributes and should be expanded.
///
/// Attributes are added to this list from `visitAny`. Any attributes that weren’t actually expanded (eg. because they are accessor macros but not attached to a variable or subscript), will be diagnosed from `visitAny`.
var attributesToExpand: [AttributeSyntax] = []
Do you think that would work? There was a problem hiding this comment.
I found that the id is still different. The It seems that in There was a problem hiding this comment. I also delved into the implementation of There was a problem hiding this comment. I think I’ll need to play around with the code myself a little to figure out how we want to continue, but am too busy with other things at the moment. I’ll definitely come back to you but it might take a few weeks. |
||||||
|
|
||||||
| init( | ||||||
| macroSystem: MacroSystem, | ||||||
|
|
@@ -715,14 +719,55 @@ private class MacroApplication<Context: MacroExpansionContext>: SyntaxRewriter { | |||||
| let visitedNode = self.visit(declSyntax) | ||||||
| skipVisitAnyHandling.remove(Syntax(declSyntax)) | ||||||
|
|
||||||
| let attributesToRemove = self.macroAttributes(attachedTo: visitedNode) | ||||||
| for (attribute, spec) in attributesToRemove { | ||||||
| if let index = self.expandedAttributes.firstIndex(where: { expandedAttribute in | ||||||
| expandedAttribute.position == attribute.position | ||||||
| }) { | ||||||
| self.expandedAttributes.remove(at: index) | ||||||
|
There was a problem hiding this comment. Is removing necessary for correctness or do we just do it to keep the number of elements in There was a problem hiding this comment. It will be incorrect if it's possible that two attributes with same position are actually different during the process. For example, the list still contains an attribute expanded in the last visit, and in the same position there is attribute not expanded in this visit, I think it'll go wrong. I think we should remove it from the list if it has been checked to make sure it's correct There was a problem hiding this comment. Ah, I thought we were fine because macros only ever introduce code after the the macro attribute itself, so I thought that the position would still uniquely identify a macro within the source file but thinking about it again, I realize that isn’t true because the macro itself gets removed, which shifts positions forward and thus allows another macro to have the same position. For example, you could have the following @NoOpMacroAB
@MacroThatExpandsToMemberAttributesMacroAndWhichIsAlsoAnAccessorMacro
sturct Foo {
func foo() {}
}which expands to sturct Foo {
@MacroThatExpandsToMemberAttributesMacroAndWhichIsAlsoAnAccessorMacro
func foo() {}
}And the two And yes, I think with the removal of attributes from I hope that I’m making sense here. I think it would be good to have a test case for what I described above and make sure that we do actually diagnose that |
||||||
| } else { | ||||||
| self.diagnosticForUnexpandedMacro(attribute: attribute, macroType: spec.type, attachedTo: declSyntax) | ||||||
| } | ||||||
| } | ||||||
| return AttributeRemover(removingWhere: { attributesToRemove.map(\.attributeNode).contains($0) }).rewrite( | ||||||
| visitedNode | ||||||
| ) | ||||||
| } | ||||||
|
|
||||||
| return nil | ||||||
| } | ||||||
|
|
||||||
| private func diagnosticForUnexpandedMacro(attribute: AttributeSyntax, macroType: Macro.Type, attachedTo: DeclSyntax) { | ||||||
| var diagnosticError: Error? = nil | ||||||
| var fixitMessage: FixItMessage? = nil | ||||||
| switch macroType { | ||||||
| case is AccessorMacro.Type: | ||||||
| if (!attachedTo.is(VariableDeclSyntax.self) && !attachedTo.is(SubscriptDeclSyntax.self)) { | ||||||
|
There was a problem hiding this comment.
Suggested change
|
||||||
| diagnosticError = MacroApplicationError.macroAttachedToInvalidDecl( | ||||||
| MacroRole.accessor.description, | ||||||
| attachedTo.kind.nameForDiagnostics ?? "" | ||||||
| ) | ||||||
| fixitMessage = SwiftSyntaxMacros.MacroExpansionFixItMessage( | ||||||
| "Remove '\(attribute.trimmedDescription)'" | ||||||
| ) | ||||||
| } | ||||||
| default: | ||||||
| break | ||||||
|
There was a problem hiding this comment. I think your error message There was a problem hiding this comment. If that's the case, I believe there's no need to diagnose based on the type of |
||||||
| } | ||||||
| if let diagnosticError, let fixitMessage { | ||||||
| contextGenerator(Syntax(attachedTo)).addDiagnostics( | ||||||
| from: diagnosticError, | ||||||
| node: attachedTo, | ||||||
| fixIts: [ | ||||||
| FixIt( | ||||||
| message: fixitMessage, | ||||||
| changes: [.replace(oldNode: Syntax(attribute), newNode: Syntax(AttributeListSyntax()))] | ||||||
| ) | ||||||
| ] | ||||||
| ) | ||||||
| } | ||||||
| } | ||||||
|
|
||||||
| /// Visit for both the body and preamble macros. | ||||||
| func visitBodyAndPreambleMacros<Node: DeclSyntaxProtocol & WithOptionalCodeBlockSyntax>( | ||||||
| _ node: Node | ||||||
|
|
@@ -1023,6 +1068,7 @@ extension MacroApplication { | |||||
| macroAttribute.conformanceList | ||||||
| ) { | ||||||
| result += expanded | ||||||
| self.expandedAttributes.append(macroAttribute.attributeNode) | ||||||
| } | ||||||
| } catch { | ||||||
| contextGenerator(Syntax(decl)).addDiagnostics(from: error, node: macroAttribute.attributeNode) | ||||||
|
|
@@ -1170,6 +1216,7 @@ extension MacroApplication { | |||||
| from: newAccessors, | ||||||
| indentationWidth: self.indentationWidth | ||||||
| ) | ||||||
| self.expandedAttributes.append(macro.attributeNode) | ||||||
| } | ||||||
| } else if let newAccessors = try expandAccessorMacroWithoutExistingAccessors( | ||||||
| definition: macro.definition, | ||||||
|
|
@@ -1192,6 +1239,7 @@ extension MacroApplication { | |||||
| } else { | ||||||
| newAccessorsBlock = newAccessors | ||||||
| } | ||||||
| self.expandedAttributes.append(macro.attributeNode) | ||||||
| } | ||||||
| } catch { | ||||||
| contextGenerator(Syntax(storage)).addDiagnostics(from: error, node: macro.attributeNode) | ||||||
|
|
||||||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could you add labels for these associated values so it’s easier to tell what the two strings do?