Merge pull request #3843 from thomasmarshall/error-recovery-nodes

Validate expected nodes

Author: kddnewton

.github/workflows/main.yml

@@ -51,8 +51,8 @@ jobs:
       with:
         ruby-version: "3.4"
         bundler-cache: true
-    - name: Check Sorbet
-      run: bundle exec rake typecheck:tapioca typecheck:sorbet
+    # - name: Check Sorbet
+    #   run: bundle exec rake typecheck:tapioca typecheck:sorbet
     - name: Check Steep
       run: bundle exec rake typecheck:steep
     - name: Check field kinds

config.yml

@@ -814,8 +814,6 @@ nodes:
           - GlobalVariableReadNode
           - BackReferenceReadNode
           - NumberedReferenceReadNode
-          - on error: SymbolNode # alias $a b
-          - on error: MissingNode # alias $a 42
         comment: |
           Represents the old name of the global variable that can be used before aliasing.
 
@@ -856,8 +854,6 @@ nodes:
         kind:
           - SymbolNode
           - InterpolatedSymbolNode
-          - on error: GlobalVariableReadNode # alias a $b
-          - on error: MissingNode # alias a 42
         comment: |
           Represents the old name of the method that will be aliased.
 
@@ -1871,7 +1867,6 @@ nodes:
         kind:
           - ConstantReadNode
           - ConstantPathNode
-          - on error: CallNode # class 0.X end
       - name: inheritance_operator_loc
         type: location?
         comment: |
@@ -2425,6 +2420,15 @@ nodes:
           ^^^^^^
             bar
           end
+  - name: ErrorRecoveryNode
+    fields:
+      - name: unexpected
+        type: node?
+        kind: Node
+        comment: |
+          The unexpected node that was found in the tree, if there was one.
+    comment: |
+      Represents a node that is either missing or unexpected and results in a syntax error.
   - name: FalseNode
     comment: |
       Represents the use of the literal `false` keyword.
@@ -2467,9 +2471,7 @@ nodes:
                                ^^^^^^
       - name: right
         type: node
-        kind:
-          - SplatNode
-          - on error: MissingNode
+        kind: SplatNode
         comment: |
           Represents the second wildcard node in the pattern.
 
@@ -2552,9 +2554,6 @@ nodes:
           - CallTargetNode
           - IndexTargetNode
           - MultiTargetNode
-          - on error: BackReferenceReadNode # for $& in a end
-          - on error: NumberedReferenceReadNode # for $1 in a end
-          - on error: MissingNode # for in 1..10; end
         comment: |
           The index expression for `for` loops.
 
@@ -3308,10 +3307,6 @@ nodes:
           - EmbeddedStatementsNode
           - EmbeddedVariableNode
           - InterpolatedStringNode # `"a" "#{b}"`
-          - on error: XStringNode # `<<`FOO` "bar"
-          - on error: InterpolatedXStringNode
-          - on error: SymbolNode
-          - on error: InterpolatedSymbolNode
       - name: closing_loc
         type: location?
     newline: parts
@@ -3689,9 +3684,6 @@ nodes:
 
           /(?<foo>bar)/ =~ baz
           ^^^^^^^^^^^^^^^^^^^^
-  - name: MissingNode
-    comment: |
-      Represents a node that is missing from the source and results in a syntax error.
   - name: ModuleNode
     fields:
       - name: locals
@@ -3703,7 +3695,6 @@ nodes:
         kind:
           - ConstantReadNode
           - ConstantPathNode
-          - on error: MissingNode # module Parent module end
       - name: body
         type: node?
         kind:
@@ -3733,8 +3724,6 @@ nodes:
           - IndexTargetNode
           - MultiTargetNode
           - RequiredParameterNode # def m((a,b)); end
-          - on error: BackReferenceReadNode # a, (b, $&) = z
-          - on error: NumberedReferenceReadNode # a, (b, $1) = z
         comment: |
           Represents the targets expressions before a splat node.
 
@@ -3778,8 +3767,6 @@ nodes:
           - IndexTargetNode
           - MultiTargetNode
           - RequiredParameterNode # def m((*,b)); end
-          - on error: BackReferenceReadNode # a, (*, $&) = z
-          - on error: NumberedReferenceReadNode # a, (*, $1) = z
         comment: |
           Represents the targets expressions after a splat node.
 
@@ -3823,8 +3810,6 @@ nodes:
           - CallTargetNode
           - IndexTargetNode
           - MultiTargetNode
-          - on error: BackReferenceReadNode # $&, = z
-          - on error: NumberedReferenceReadNode # $1, = z
         comment: |
           Represents the targets expressions before a splat node.
 
@@ -3867,8 +3852,6 @@ nodes:
           - CallTargetNode
           - IndexTargetNode
           - MultiTargetNode
-          - on error: BackReferenceReadNode # *, $& = z
-          - on error: NumberedReferenceReadNode # *, $1 = z
         comment: |
           Represents the targets expressions after a splat node.
 
@@ -4066,14 +4049,6 @@ nodes:
         kind:
           - RequiredParameterNode
           - MultiTargetNode
-          # On parsing error of `f(**kwargs, ...)` or `f(**nil, ...)`, the keyword_rest value is moved here:
-          - on error: KeywordRestParameterNode
-          - on error: NoKeywordsParameterNode
-          # On parsing error of `f(..., ...)`, the first forwarding parameter is moved here:
-          - on error: ForwardingParameterNode
-          # On parsing error of `f(&nil, &foo)`/`f(&foo, &nil)`, the first forwarding parameter is moved here:
-          - on error: BlockParameterNode
-          - on error: NoBlockParameterNode
       - name: keywords
         type: node[]
         kind:
@@ -4160,7 +4135,6 @@ nodes:
           - BackReferenceReadNode # foo in ^$&
           - NumberedReferenceReadNode # foo in ^$1
           - ItLocalVariableReadNode # proc { 1 in ^it }
-          - on error: MissingNode # foo in ^Bar
         comment: |
           The variable used in the pinned expression
 
@@ -4243,7 +4217,7 @@ nodes:
 
               1...foo
                   ^^^
-          If neither right-hand or left-hand side was included, this will be a MissingNode.
+          If neither right-hand or left-hand side was included, this will be an ErrorRecoveryNode.
       - name: operator_loc
         type: location
         comment: |
@@ -4358,9 +4332,6 @@ nodes:
           - ConstantPathTargetNode
           - CallTargetNode
           - IndexTargetNode
-          - on error: BackReferenceReadNode # => begin; rescue => $&; end
-          - on error: NumberedReferenceReadNode # => begin; rescue => $1; end
-          - on error: MissingNode # begin; rescue =>; end
       - name: then_keyword_loc
         type: location?
       - name: statements

docs/mapping.md

@@ -112,6 +112,6 @@ The following table shows how the various CRuby nodes are mapped to prism nodes.
 | `NODE_ARYPTN` | `PM_ARRAY_PATTERN_NODE` |
 | `NODE_HSHPTN` | `PM_HASH_PATTERN_NODE` |
 | `NODE_FNDPTN` | `PM_FIND_PATTERN_NODE` |
-| `NODE_ERROR` | `PM_MISSING_NODE` |
+| `NODE_ERROR` | `PM_ERROR_RECOVERY_NODE` |
 | `NODE_LAST` | |
 ```

lib/prism/node_ext.rb

@@ -221,10 +221,10 @@ class ConstantPathNode < Node
     # local variable
     class DynamicPartsInConstantPathError < StandardError; end
 
-    # An error class raised when missing nodes are found while computing a
+    # An error class raised when error recovery nodes are found while computing a
     # constant path's full name. For example:
     # Foo:: -> raises because the constant path is missing the last part
-    class MissingNodesInConstantPathError < StandardError; end
+    class ErrorRecoveryNodesInConstantPathError < StandardError; end
 
     # Returns the list of parts for the full name of this constant path.
     # For example: [:Foo, :Bar]
@@ -237,7 +237,7 @@ def full_name_parts
       while current.is_a?(ConstantPathNode)
         name = current.name
         if name.nil?
-          raise MissingNodesInConstantPathError, "Constant path contains missing nodes. Cannot compute full name"
+          raise ErrorRecoveryNodesInConstantPathError, "Constant path contains error recovery nodes. Cannot compute full name"
         end
 
         parts.unshift(name)
@@ -277,7 +277,7 @@ def full_name_parts
         end
 
       if (name = self.name).nil?
-        raise ConstantPathNode::MissingNodesInConstantPathError, "Constant target path contains missing nodes. Cannot compute full name"
+        raise ConstantPathNode::ErrorRecoveryNodesInConstantPathError, "Constant target path contains error recovery nodes. Cannot compute full name"
       end
 
       parts.push(name)

lib/prism/translation/parser/compiler.rb

@@ -1324,7 +1324,7 @@ def visit_match_write_node(node)
         # A node that is missing from the syntax tree. This is only used in the
         # case of a syntax error. The parser gem doesn't have such a concept, so
         # we invent our own here.
-        def visit_missing_node(node)
+        def visit_error_recovery_node(node)
           ::AST::Node.new(:missing, [], location: ::Parser::Source::Map.new(srange(node.location)))
         end
 

lib/prism/translation/ripper.rb

@@ -2511,8 +2511,8 @@ def visit_match_write_node(node)
 
       # A node that is missing from the syntax tree. This is only used in the
       # case of a syntax error.
-      def visit_missing_node(node)
-        raise "Cannot visit missing nodes directly."
+      def visit_error_recovery_node(node)
+        raise "Cannot visit error recovery nodes directly."
       end
 
       # module Foo; end

lib/prism/translation/ruby_parser.rb

@@ -1071,8 +1071,8 @@ def visit_match_write_node(node)
         # A node that is missing from the syntax tree. This is only used in the
         # case of a syntax error. The parser gem doesn't have such a concept, so
         # we invent our own here.
-        def visit_missing_node(node)
-          raise "Cannot visit missing node directly"
+        def visit_error_recovery_node(node)
+          raise "Cannot visit error recovery node directly"
         end
 
         # module Foo; end

rakelib/lint.rake

@@ -29,7 +29,7 @@ task :lint do
     exit(1)
   end
 
-  if (uncommented = nodes.select { |node| !%w[MissingNode ProgramNode].include?(node.fetch("name")) && !node.fetch("comment").match?(/^\s{4}/) }).any?
+  if (uncommented = nodes.select { |node| !%w[ErrorRecoveryNode ProgramNode].include?(node.fetch("name")) && !node.fetch("comment").match?(/^\s{4}/) }).any?
     names = uncommented.map { |node| node.fetch("name") }
     warn("Expected all nodes to have an example, missing comments for #{names.join(", ")}")
     exit(1)