Merge pull request #3849 from Earlopain/ripper-shim-irb-no-crash

Make irb work with the ripper shim

Author: Earlopain

lib/prism/translation/ripper/lexer.rb

@@ -6,7 +6,7 @@
 module Prism
   module Translation
     class Ripper
-      class Lexer # :nodoc:
+      class Lexer < Ripper # :nodoc:
         # :stopdoc:
         class State
 
@@ -39,6 +39,92 @@ def allbits?(i) to_int.allbits?(i) end
           def anybits?(i) to_int.anybits?(i) end
           def nobits?(i) to_int.nobits?(i) end
         end
+
+        class Elem
+          attr_accessor :pos, :event, :tok, :state, :message
+
+          def initialize(pos, event, tok, state, message = nil)
+            @pos = pos
+            @event = event
+            @tok = tok
+            @state = State.new(state)
+            @message = message
+          end
+
+          def [](index)
+            case index
+            when 0, :pos
+              @pos
+            when 1, :event
+              @event
+            when 2, :tok
+              @tok
+            when 3, :state
+              @state
+            when 4, :message
+              @message
+            else
+              nil
+            end
+          end
+
+          def inspect
+            "#<#{self.class}: #{event}@#{pos[0]}:#{pos[1]}:#{state}: #{tok.inspect}#{": " if message}#{message}>"
+          end
+
+          alias to_s inspect
+
+          def pretty_print(q)
+            q.group(2, "#<#{self.class}:", ">") {
+              q.breakable
+              q.text("#{event}@#{pos[0]}:#{pos[1]}")
+              q.breakable
+              state.pretty_print(q)
+              q.breakable
+              q.text("token: ")
+              tok.pretty_print(q)
+              if message
+                q.breakable
+                q.text("message: ")
+                q.text(message)
+              end
+            }
+          end
+
+          def to_a
+            if @message
+              [@pos, @event, @tok, @state, @message]
+            else
+              [@pos, @event, @tok, @state]
+            end
+          end
+        end
+
+        def initialize(...)
+          super
+          @lex_compat = Prism.lex_compat(@source, filepath: filename, line: lineno)
+        end
+
+        # Returns the lex_compat result wrapped in `Elem`. Errors are omitted.
+        # Since ripper is a streaming parser, tokens are expected to be emitted in the order
+        # that the parser encounters them. This is not implemented.
+        def parse(raise_errors: false)
+          if @lex_compat.failure? && raise_errors
+            raise SyntaxError, @lex_compat.errors.first.message
+          else
+            @lex_compat.value.map do |position, event, token, state|
+              Elem.new(position, event, token, state.to_int)
+            end
+          end
+        end
+
+        # Similar to parse but ripper sorts the elements by position in the source. Also
+        # includes errors. Since prism does error recovery, in cases of syntax errors
+        # the result may differ greatly compared to ripper.
+        def scan(...)
+          parse(...)
+        end
+
         # :startdoc:
       end
     end

test/prism/ruby/ripper_test.rb

@@ -38,7 +38,7 @@ class RipperTest < TestCase
     end
 
     # Skip these tests that we haven't implemented yet.
-    omitted = [
+    omitted_sexp_raw = [
       "dos_endings.txt",
       "heredocs_with_fake_newlines.txt",
       "heredocs_with_ignored_newlines.txt",
@@ -59,8 +59,29 @@ class RipperTest < TestCase
       "whitequark/slash_newline_in_heredocs.txt"
     ]
 
-    Fixture.each_for_current_ruby(except: incorrect | omitted) do |fixture|
-      define_method(fixture.test_name) { assert_ripper(fixture.read) }
+    omitted_lexer_parse = [
+      "comments.txt",
+      "heredoc_percent_q_newline_delimiter.txt",
+      "heredoc_with_escaped_newline_at_start.txt",
+      "heredocs_with_fake_newlines.txt",
+      "indented_file_end.txt",
+      "seattlerb/TestRubyParserShared.txt",
+      "seattlerb/class_comments.txt",
+      "seattlerb/module_comments.txt",
+      "seattlerb/parse_line_block_inline_comment_leading_newlines.txt",
+      "seattlerb/parse_line_block_inline_multiline_comment.txt",
+      "spanning_heredoc_newlines.txt",
+      "strings.txt",
+      "whitequark/dedenting_heredoc.txt",
+      "whitequark/procarg0.txt",
+    ]
+
+    Fixture.each_for_current_ruby(except: incorrect | omitted_sexp_raw) do |fixture|
+      define_method("#{fixture.test_name}_sexp_raw") { assert_ripper_sexp_raw(fixture.read) }
+    end
+
+    Fixture.each_for_current_ruby(except: incorrect | omitted_lexer_parse) do |fixture|
+      define_method("#{fixture.test_name}_lexer_parse") { assert_ripper_lexer_parse(fixture.read) }
     end
 
     # Check that the hardcoded values don't change without us noticing.
@@ -76,8 +97,27 @@ def test_internals
 
     private
 
-    def assert_ripper(source)
+    def assert_ripper_sexp_raw(source)
       assert_equal Ripper.sexp_raw(source), Prism::Translation::Ripper.sexp_raw(source)
     end
+
+    def assert_ripper_lexer_parse(source)
+      prism = Translation::Ripper::Lexer.new(source).parse
+      ripper = Ripper::Lexer.new(source).parse
+      ripper.reject! { |elem| elem.event == :on_sp } # Prism doesn't emit on_sp
+      ripper.sort_by!(&:pos) # Prism emits tokens by their order in the code, not in parse order
+
+      [prism.size, ripper.size].max.times do |i|
+        expected = ripper[i].to_a
+        actual = prism[i].to_a
+        # Since tokens related to heredocs are not emitted in the same order,
+        # the state also doesn't line up.
+        if expected[1] == :on_heredoc_end && actual[1] == :on_heredoc_end
+          expected[3] = actual[3] = nil
+        end
+
+        assert_equal(expected, actual)
+      end
+    end
   end
 end