Assert that tokens in the syntax tree match token_choices and text_choices
#1180
Conversation
ahoppen
changed the title
token_choices and text_choices 🚥 #1173token_choices and text_choices
ahoppen
force-pushed
the
ahoppen/assert-token-choices
branch
3 times, most recently
from
897e9b7 to
3f64e22
Compare
ahoppen
changed the title
token_choices and text_choicestoken_choices and text_choices 🚥 #1289
| @@ -8531,6 +8680,7 @@ public struct RawAccessorDeclSyntax: RawDeclSyntaxNodeProtocol { | |||
| _ unexpectedAfterBody: RawUnexpectedNodesSyntax? = nil, | |||
| arena: __shared SyntaxArena | |||
| ) { | |||
| assert(accessorKind.tokenKind == .keyword(.get) || accessorKind.tokenKind == .keyword(.set) || accessorKind.tokenKind == .keyword(.didSet) || accessorKind.tokenKind == .keyword(.willSet) || accessorKind.tokenKind == .keyword(.unsafeAddress) || accessorKind.tokenKind == .keyword(.addressWithOwner) || accessorKind.tokenKind == .keyword(.addressWithNativeOwner) || accessorKind.tokenKind == .keyword(.unsafeMutableAddress) || accessorKind.tokenKind == .keyword(.mutableAddressWithOwner) || accessorKind.tokenKind == .keyword(.mutableAddressWithNativeOwner) || accessorKind.tokenKind == .keyword(._read) || accessorKind.tokenKind == .keyword(._modify), "Received \(accessorKind.tokenKind)") | |||
There was a problem hiding this comment.
These should overlap with the subset right? Wonder if we should autogenerate subsets when there's > 1 kind. I imagine that would be a lot though, maybe only > 3 and then at can have overloads for up to 3 kinds or a subset?
There was a problem hiding this comment.
That’s the next item on my To-Do list after this PR is merge: “Generate RawTokenKindSubset”. I’m still not entirely sure how exactly we want to decide whether to generate the RawTokenKindSubset, but it’s definitely something we should do.
| @_dynamicReplacement(for: <#identifier#>) | ||
| @_dynamicReplacement(for : <#identifier#>) |
There was a problem hiding this comment.
What's the whitespace for?
There was a problem hiding this comment.
Oh, I wanted to fix that. It’s because for is a keyword now and keywords always have trailing space in BasicFormat at the moment. Improving that is a slightly larger change in BasicFormat that I would like to do in a follow-up PR.
ahoppen
force-pushed
the
ahoppen/assert-token-choices
branch
from
3f64e22 to
d581360
Compare
…oices Previously, we specified text_choices and token_chocies in the Python files but we weren’t actually verifying that the tokens we pass into the syntax tree match these expectations. Add assertions that check the token kinds. rdar://103448567
ahoppen
force-pushed
the
ahoppen/assert-token-choices
branch
from
d581360 to
1cde192
Compare
ahoppen
changed the title
token_choices and text_choices 🚥 #1289token_choices and text_choices
|
@swift-ci Please test |
We currently specify
token_choicesandtext_choicesin the Python files but don't enforce that the tokens in the syntax tree actually match these – effectively we are ignoring these properties.Add assertions that the actual tokens in the syntax tree match the specification and fix a few cases where these assertions were violated.
rdar://103448567