Search updates

.gitignore

@@ -9,3 +9,4 @@
 /vendor/
 
 test.rb
+query.txt

README.md

@@ -15,6 +15,7 @@ It is built with only standard library dependencies. It additionally ships with
 - [CLI](#cli)
   - [ast](#ast)
   - [check](#check)
+  - [expr](#expr)
   - [format](#format)
   - [json](#json)
   - [match](#match)
@@ -26,6 +27,7 @@ It is built with only standard library dependencies. It additionally ships with
   - [SyntaxTree.read(filepath)](#syntaxtreereadfilepath)
   - [SyntaxTree.parse(source)](#syntaxtreeparsesource)
   - [SyntaxTree.format(source)](#syntaxtreeformatsource)
+  - [SyntaxTree.search(source, query, &block)](#syntaxtreesearchsource-query-block)
 - [Nodes](#nodes)
   - [child_nodes](#child_nodes)
   - [Pattern matching](#pattern-matching)
@@ -129,6 +131,24 @@ To change the print width that you are checking against, specify the `--print-wi
 stree check --print-width=100 path/to/file.rb
 ```
 
+### expr
+
+This command will output a Ruby case-match expression that would match correctly against the first expression of the input.
+
+```sh
+stree expr path/to/file.rb
+```
+
+For a file that contains `1 + 1`, you will receive:
+
+```ruby
+SyntaxTree::Binary[
+  left: SyntaxTree::Int[value: "1"],
+  operator: :+,
+  right: SyntaxTree::Int[value: "1"]
+]
+```
+
 ### format
 
 This command will output the formatted version of each of the listed files. Importantly, it will not write that content back to the source files. It is meant to display the formatted version only.
@@ -312,6 +332,10 @@ This function takes an input string containing Ruby code and returns the syntax
 
 This function takes an input string containing Ruby code, parses it into its underlying syntax tree, and formats it back out to a string. You can optionally pass a second argument to this method as well that is the maximum width to print. It defaults to `80`.
 
+### SyntaxTree.search(source, query, &block)
+
+This function takes an input string containing Ruby code, an input string containing a valid Ruby `in` clause expression that can be used to match against nodes in the tree (can be generated using `stree expr`, `stree match`, or `Node#construct_keys`), and a block. Each node that matches the given query will be yielded to the block. The block will receive the node as its only argument.
+
 ## Nodes
 
 There are many different node types in the syntax tree. They are meant to be treated as immutable structs containing links to child nodes with minimal logic contained within their implementation. However, for the most part they all respond to a certain set of APIs, listed below.

lib/syntax_tree.rb

@@ -21,6 +21,7 @@
 require_relative "syntax_tree/visitor/with_environment"
 
 require_relative "syntax_tree/parser"
+require_relative "syntax_tree/pattern"
 require_relative "syntax_tree/search"
 
 # Syntax Tree is a suite of tools built on top of the internal CRuby parser. It
@@ -74,4 +75,10 @@ def self.read(filepath)
 
     File.read(filepath, encoding: encoding)
   end
+
+  # Searches through the given source using the given pattern and yields each
+  # node in the tree that matches the pattern to the given block.
+  def self.search(source, query, &block)
+    Search.new(Pattern.new(query).compile).scan(parse(source), &block)
+  end
 end

lib/syntax_tree/cli.rb

@@ -188,6 +188,20 @@ def run(item)
       end
     end
 
+    # An action of the CLI that outputs a pattern-matching Ruby expression that
+    # would match the first expression of the input given.
+    class Expr < Action
+      def run(item)
+        case item.handler.parse(item.source)
+        in Program[statements: Statements[body: [expression]]]
+          puts expression.construct_keys
+        else
+          warn("The input to `stree expr` must be a single expression.")
+          exit(1)
+        end
+      end
+    end
+
     # An action of the CLI that formats the input source and prints it out.
     class Format < Action
       def run(item)
@@ -219,10 +233,15 @@ class Search < Action
 
       def initialize(query)
         query = File.read(query) if File.readable?(query)
-        @search = SyntaxTree::Search.new(query)
-      rescue SyntaxTree::Search::UncompilableError => error
-        warn(error.message)
-        exit(1)
+        pattern =
+          begin
+            Pattern.new(query).compile
+          rescue Pattern::CompilationError => error
+            warn(error.message)
+            exit(1)
+          end
+
+        @search = SyntaxTree::Search.new(pattern)
       end
 
       def run(item)
@@ -281,6 +300,10 @@ def run(item)
       #{Color.bold("stree doc [--plugins=...] [-e SCRIPT] FILE")}
         Print out the doc tree that would be used to format the given files
 
+      #{Color.bold("stree expr [-e SCRIPT] FILE")}
+        Print out a pattern-matching Ruby expression that would match the first
+        expression of the given files
+
       #{Color.bold("stree format [--plugins=...] [--print-width=NUMBER] [-e SCRIPT] FILE")}
         Print out the formatted version of the given files
 
@@ -436,6 +459,8 @@ def run(argv)
             Debug.new(options)
           when "doc"
             Doc.new(options)
+          when "e", "expr"
+            Expr.new(options)
           when "f", "format"
             Format.new(options)
           when "help"

lib/syntax_tree/pattern.rb

@@ -0,0 +1,172 @@
+# frozen_string_literal: true
+
+module SyntaxTree
+  # A pattern is an object that wraps a Ruby pattern matching expression. The
+  # expression would normally be passed to an `in` clause within a `case`
+  # expression or a rightward assignment expression. For example, in the
+  # following snippet:
+  #
+  #     case node
+  #     in Const[value: "SyntaxTree"]
+  #     end
+  #
+  # the pattern is the `Const[value: "SyntaxTree"]` expression. Within Syntax
+  # Tree, every node generates these kinds of expressions using the
+  # #construct_keys method.
+  #
+  # The pattern gets compiled into an object that responds to call by running
+  # the #compile method. This method itself will run back through Syntax Tree to
+  # parse the expression into a tree, then walk the tree to generate the
+  # necessary callable objects. For example, if you wanted to compile the
+  # expression above into a callable, you would:
+  #
+  #     callable = SyntaxTree::Pattern.new("Const[value: 'SyntaxTree']").compile
+  #     callable.call(node)
+  #
+  # The callable object returned by #compile is guaranteed to respond to #call
+  # with a single argument, which is the node to match against. It also is
+  # guaranteed to respond to #===, which means it itself can be used in a `case`
+  # expression, as in:
+  #
+  #     case node
+  #     when callable
+  #     end
+  #
+  # If the query given to the initializer cannot be compiled into a valid
+  # matcher (either because of a syntax error or because it is using syntax we
+  # do not yet support) then a SyntaxTree::Pattern::CompilationError will be
+  # raised.
+  class Pattern
+    # Raised when the query given to a pattern is either invalid Ruby syntax or
+    # is using syntax that we don't yet support.
+    class CompilationError < StandardError
+      def initialize(repr)
+        super(<<~ERROR)
+          Syntax Tree was unable to compile the pattern you provided to search
+          into a usable expression. It failed on to understand the node
+          represented by:
+
+          #{repr}
+
+          Note that not all syntax supported by Ruby's pattern matching syntax
+          is also supported by Syntax Tree's code search. If you're using some
+          syntax that you believe should be supported, please open an issue on
+          GitHub at https://github.com/ruby-syntax-tree/syntax_tree/issues/new.
+        ERROR
+      end
+    end
+
+    attr_reader :query
+
+    def initialize(query)
+      @query = query
+    end
+
+    def compile
+      program =
+        begin
+          SyntaxTree.parse("case nil\nin #{query}\nend")
+        rescue Parser::ParseError
+          raise CompilationError, query
+        end
+
+      compile_node(program.statements.body.first.consequent.pattern)
+    end
+
+    private
+
+    def combine_and(left, right)
+      ->(node) { left.call(node) && right.call(node) }
+    end
+
+    def combine_or(left, right)
+      ->(node) { left.call(node) || right.call(node) }
+    end
+
+    def compile_node(root)
+      case root
+      in AryPtn[constant:, requireds:, rest: nil, posts: []]
+        compiled_constant = compile_node(constant) if constant
+
+        preprocessed = requireds.map { |required| compile_node(required) }
+
+        compiled_requireds = ->(node) do
+          deconstructed = node.deconstruct
+
+          deconstructed.length == preprocessed.length &&
+            preprocessed
+              .zip(deconstructed)
+              .all? { |(matcher, value)| matcher.call(value) }
+        end
+
+        if compiled_constant
+          combine_and(compiled_constant, compiled_requireds)
+        else
+          compiled_requireds
+        end
+      in Binary[left:, operator: :|, right:]
+        combine_or(compile_node(left), compile_node(right))
+      in Const[value:] if SyntaxTree.const_defined?(value)
+        clazz = SyntaxTree.const_get(value)
+
+        ->(node) { node.is_a?(clazz) }
+      in Const[value:] if Object.const_defined?(value)
+        clazz = Object.const_get(value)
+
+        ->(node) { node.is_a?(clazz) }
+      in ConstPathRef[
+           parent: VarRef[value: Const[value: "SyntaxTree"]], constant:
+         ]
+        compile_node(constant)
+      in DynaSymbol[parts: []]
+        symbol = :""
+
+        ->(node) { node == symbol }
+      in DynaSymbol[parts: [TStringContent[value:]]]
+        symbol = value.to_sym
+
+        ->(attribute) { attribute == value }
+      in HshPtn[constant:, keywords:, keyword_rest: nil]
+        compiled_constant = compile_node(constant)
+
+        preprocessed =
+          keywords.to_h do |keyword, value|
+            raise NoMatchingPatternError unless keyword.is_a?(Label)
+            [keyword.value.chomp(":").to_sym, compile_node(value)]
+          end
+
+        compiled_keywords = ->(node) do
+          deconstructed = node.deconstruct_keys(preprocessed.keys)
+
+          preprocessed.all? do |keyword, matcher|
+            matcher.call(deconstructed[keyword])
+          end
+        end
+
+        if compiled_constant
+          combine_and(compiled_constant, compiled_keywords)
+        else
+          compiled_keywords
+        end
+      in RegexpLiteral[parts: [TStringContent[value:]]]
+        regexp = /#{value}/
+
+        ->(attribute) { regexp.match?(attribute) }
+      in StringLiteral[parts: []]
+        ->(attribute) { attribute == "" }
+      in StringLiteral[parts: [TStringContent[value:]]]
+        ->(attribute) { attribute == value }
+      in SymbolLiteral[value:]
+        symbol = value.value.to_sym
+
+        ->(attribute) { attribute == symbol }
+      in VarRef[value: Const => value]
+        compile_node(value)
+      in VarRef[value: Kw[value: "nil"]]
+        ->(attribute) { attribute.nil? }
+      end
+    rescue NoMatchingPatternError
+      raise CompilationError, PP.pp(root, +"").chomp
+    end
+  end
+end