Improve modeling of Child in CodeGeneration
#1289
Conversation
|
@swift-ci Please test |
| case .token(choices: let choices): | ||
| if choices.count == 1 { | ||
| switch choices.first! { | ||
| case .keyword(text: _): return "KeywordToken" |
There was a problem hiding this comment.
Could text also be replaced with _
ahoppen
force-pushed
the
ahoppen/assert-token-choices-preparation
branch
from
6208ab4 to
2467d12
Compare
|
@swift-ci Please test |
| @@ -81,14 +66,11 @@ public let AVAILABILITY_NODES: [Node] = [ | |||
| kind: "Syntax", | |||
| children: [ | |||
| Child(name: "Platform", | |||
| kind: "IdentifierToken", | |||
| kind: .token(choices: [.token(tokenKind: "IdentifierToken")]), | |||
| description: "The name of the OS on which the availability should berestricted or 'swift' if the availability should berestricted based on a Swift version.", | |||
There was a problem hiding this comment.
I'm not actually sure what these are used for. But it appears we're not adding a space when we drop the newlines and indentation (separate to this PR, just noticed it here).
There was a problem hiding this comment.
What exactly do you mean?
There was a problem hiding this comment.
"The name of the OS on which the availability should berestricted or 'swift' if the availability should berestricted based on a Swift version."
Note the two "berestricted". This is coming from the python where it's defined as a multiline string. But the lines are being joined as is (well, with indentation removed), without a space separator.
There was a problem hiding this comment.
Oh, I see, this wasn’t related to the line I changed.
Interesting that I haven’t noticed this before. I fixed it now.
| # This will force validation logic to check the text passed into the | ||
| # token against the choices. | ||
| self.text_choices = text_choices or [] | ||
| self.token_choices.append((token, choice_text)) |
There was a problem hiding this comment.
Is there a case where token != KeywordToken and kind != KeywordToken? Can't we just check/use kind instead?
There was a problem hiding this comment.
What exactly are you suggesting? Also, this is going to be removed in a month or so anyway when all gyb files are migrated to CodeGeneration so maybe it doesn’t matter too much anyway.
ahoppen
force-pushed
the
ahoppen/assert-token-choices-preparation
branch
from
2467d12 to
8984572
Compare
…Generation # Conflicts: # CodeGeneration/Sources/SyntaxSupport/gyb_generated/AttributeNodes.swift # CodeGeneration/Sources/SyntaxSupport/gyb_generated/AvailabilityNodes.swift # CodeGeneration/Sources/SyntaxSupport/gyb_generated/DeclNodes.swift
|
@swift-ci Please test |
|
@swift-ci please test macOS |
|
@swift-ci Please test macOS |
Instead of having
tokenChoices,textChoices,nodeChoicesandcollectionElementNameinChild, which are all pretty much mutually exclusive with each other, model the kind of a child as an enum that be either ofThe syntax in the Python definition files is a little awkward with this because you specify keywords e.g. as
Keyword|asyncbut since I anticipate that they will soon be gone anyway once we've moved all code generation to use SwiftSyntaxBuilder, I don't think we need to put too much effort into making the Python files look nice.This is a prerequisite for #1180, which will assert that tokens in a syntax tree match the specified token kind and text choices.