NodePattern: Add ` for descendant search [#7693]

lib/rubocop/node_pattern.rb

@@ -78,6 +78,8 @@ module RuboCop
   #                         # matching process starts
   #     '^^send'            # each ^ ascends one level in the AST
   #                         # so this matches against the grandparent node
+  #     '`send'             # descends any number of level in the AST
+  #                         # so this matches against any descendant node
   #     '#method'           # we call this a 'funcall'; it calls a method in the
   #                         # context where a pattern-matching method is defined
   #                         # if that returns a truthy value, the match succeeds
@@ -112,7 +114,7 @@ class Compiler
       SYMBOL       = %r{:(?:[\w+@*/?!<>=~|%^-]+|\[\]=?)}.freeze
       IDENTIFIER   = /[a-zA-Z_][a-zA-Z0-9_-]*/.freeze
       META         = Regexp.union(
-        %w"( ) { } [ ] $< < > $... $ ! ^ ... + * ?"
+        %w"( ) { } [ ] $< < > $... $ ! ^ ` ... + * ?"
       ).freeze
       NUMBER       = /-?\d+(?:\.\d+)?/.freeze
       STRING       = /".+?"/.freeze
@@ -223,6 +225,7 @@ def compile_expr(token = tokens.shift)
         when '!'       then compile_negation
         when '$'       then compile_capture
         when '^'       then compile_ascend
+        when '`'       then compile_descend
         when WILDCARD  then compile_wildcard(token[1..-1])
         when FUNCALL   then compile_funcall(token)
         when LITERAL   then compile_literal(token)
@@ -551,6 +554,19 @@ def compile_ascend
         with_context("#{CUR_NODE} && #{compile_expr}", "#{CUR_NODE}.parent")
       end
 
+      def compile_descend
+        with_temp_variables do |descendant|
+          pattern = with_context(compile_expr, descendant,
+                                 use_temp_node: false)
+          [
+            "RuboCop::NodePattern.descend(#{CUR_ELEMENT}).",
+            "any? do |#{descendant}|",
+            "  #{pattern}",
+            'end'
+          ].join("\n")
+        end
+      end
+
       def compile_wildcard(name)
         if name.empty?
           'true'
@@ -850,6 +866,22 @@ def ==(other)
     def to_s
       "#<#{self.class} #{pattern}>"
     end
+
+    # Yields its argument and any descendants, depth-first.
+    #
+    def self.descend(element, &block)
+      return to_enum(__method__, element) unless block_given?
+
+      yield element
+
+      if element.is_a?(::RuboCop::AST::Node)
+        element.children.each do |child|
+          descend(child, &block)
+        end
+      end
+
+      nil
+    end
   end
 end
 # rubocop:enable Metrics/ClassLength, Metrics/CyclomaticComplexity

manual/node_pattern.md

@@ -230,6 +230,34 @@ For example, the previous example is basically the same as:
 
 ```
 (pair ^^hash $_value)
+
+## `` ` `` for descendants
+
+The `` ` `` character can be used to search a node and all its descendants.
+For example if looking for a `return` statement anywhere within a method definition,
+we can write:
+
+```
+(def _method_name _args `return)
+```
+
+This would match both of these methods `foo` and `bar`, even though
+these `return` for `foo` and `bar` are not at the same level.
+
+```
+def foo              # (def :foo
+  return 42          #   (args)
+end                  #   (return
+                     #     (int 42)))
+
+def bar              # (def :bar
+  return 42 if foo   #   (args)
+  nil                #   (begin
+end                  #     (if
+                     #       (send nil :foo)
+                     #       (return
+                     #         (int 42)) nil)
+                     #     (nil)))
 ```
 
 ## Predicate methods

spec/rubocop/ast/node_spec.rb

@@ -1,9 +1,9 @@
 # frozen_string_literal: true
 
 RSpec.describe RuboCop::AST::Node do
-  describe '#value_used?' do
-    let(:node) { RuboCop::ProcessedSource.new(src, ruby_version).ast }
+  let(:node) { RuboCop::ProcessedSource.new(src, ruby_version).ast }
 
+  describe '#value_used?' do
     before do
       module RuboCop
         module AST
@@ -116,8 +116,6 @@ def used?
   end
 
   describe '#recursive_basic_literal?' do
-    let(:node) { RuboCop::ProcessedSource.new(src, ruby_version).ast }
-
     shared_examples 'literal' do |source|
       let(:src) { source }
 
@@ -164,8 +162,6 @@ def used?
   end
 
   describe '#pure?' do
-    let(:node) { RuboCop::ProcessedSource.new(src, ruby_version).ast }
-
     context 'for a method call' do
       let(:src) { 'obj.method(arg1, arg2)' }
 
@@ -337,8 +333,6 @@ def used?
   end
 
   describe '#sibling_index' do
-    let(:node) { RuboCop::ProcessedSource.new(src, ruby_version).ast }
-
     let(:src) do
       [
         'def foo; end',

spec/rubocop/node_pattern_spec.rb

@@ -1638,6 +1638,46 @@ def withargs(foo, bar, qux)
     end
   end
 
+  describe 'descend' do
+    let(:ruby) { '[1, [[2, 3, [[5]]], 4]]' }
+
+    context 'with an immediate match' do
+      let(:pattern) { '(array `$int _)' }
+
+      let(:captured_val) { s(:int, 1) }
+
+      it_behaves_like 'single capture'
+    end
+
+    context 'with a match multiple levels, depth first' do
+      let(:pattern) { '(array (int 1) `$int)' }
+
+      let(:captured_val) { s(:int, 2) }
+
+      it_behaves_like 'single capture'
+    end
+
+    context 'nested' do
+      let(:pattern) { '(array (int 1) `(array <`(array $int) ...>))' }
+
+      let(:captured_val) { s(:int, 5) }
+
+      it_behaves_like 'single capture'
+    end
+
+    context 'with a literal match' do
+      let(:pattern) { '(array (int 1) `4)' }
+
+      it_behaves_like 'matching'
+    end
+
+    context 'without match' do
+      let(:pattern) { '(array `$str ...)' }
+
+      it_behaves_like 'nonmatching'
+    end
+  end
+
   describe 'bad syntax' do
     context 'with empty parentheses' do
       let(:pattern) { '()' }
@@ -1711,4 +1751,20 @@ def withargs(foo, bar, qux)
       it_behaves_like 'invalid'
     end
   end
+
+  describe '.descend' do
+    let(:ruby) { '[[1, 2], 3]' }
+
+    it 'yields all children depth first' do
+      e = described_class.descend(node)
+      expect(e.instance_of?(Enumerator)).to be(true)
+      array, three = node.children
+      one, two = array.children
+      expect(e.to_a).to eq([node, array, one, 1, two, 2, three, 3])
+    end
+
+    it 'yields the given argument if it is not a Node' do
+      expect(described_class.descend(42).to_a).to eq([42])
+    end
+  end
 end