Fix dstr unparsing

* This is an entirely new approach.
* Instead to find the "correct" dstr segments we simply try all and unparse the first one
  that round trips.
* This so far guarantees we always get good concrete syntax, but it can be time intensive as
  the combinatoric space of possible dynamic string sequence is quadratic with the dstr children size.
* For this reason we try above (currently) dstr children to unparse as heredoc first.
* Passes the entire corpus and fixes bugs.

[fix #249]

Author: mbj

Changelog.md

@@ -1,3 +1,10 @@
+# v0.7.0 2024-09-16
+
+[#366](https://github.com/mbj/unparser/pull/366)
+
+* Fix all known dstring issues.
+* Interface changes.
+
 # v0.6.15 2024-06-10
 
 [#373](https://github.com/mbj/unparser/pull/373)

Gemfile

@@ -2,4 +2,6 @@
 
 source 'https://rubygems.org'
 
+gem 'mutant', path: '../mutant'
+
 gemspec

Gemfile.lock

@@ -1,3 +1,13 @@
+PATH
+  remote: ../mutant
+  specs:
+    mutant (0.12.4)
+      diff-lcs (~> 1.3)
+      parser (~> 3.3.0)
+      regexp_parser (~> 2.9.0)
+      sorbet-runtime (~> 0.5.0)
+      unparser (~> 0.6.14)
+
 PATH
   remote: .
   specs:
@@ -12,14 +22,8 @@ GEM
     diff-lcs (1.5.1)
     json (2.7.2)
     language_server-protocol (3.17.0.3)
-    mutant (0.12.3)
-      diff-lcs (~> 1.3)
-      parser (~> 3.3.0)
-      regexp_parser (~> 2.9.0)
-      sorbet-runtime (~> 0.5.0)
-      unparser (~> 0.6.14)
-    mutant-rspec (0.12.3)
-      mutant (= 0.12.3)
+    mutant-rspec (0.12.4)
+      mutant (= 0.12.4)
       rspec-core (>= 3.8.0, < 4.0.0)
     parallel (1.25.1)
     parser (3.3.2.0)
@@ -70,7 +74,7 @@ PLATFORMS
   ruby
 
 DEPENDENCIES
-  mutant (~> 0.12.2)
+  mutant!
   mutant-rspec (~> 0.12.2)
   rspec (~> 3.9)
   rspec-core (~> 3.9)

bin/corpus

@@ -1,10 +1,15 @@
 #!/usr/bin/env ruby
 # frozen_string_literal: true
 
+require 'etc'
 require 'mutant'
 require 'optparse'
+require 'pathname'
 require 'unparser'
 
+Thread.abort_on_exception = true
+Thread.report_on_exception = true
+
 module Unparser
   module Corpus
     ROOT = Pathname.new(__dir__).parent
@@ -17,16 +22,85 @@ module Unparser
       #
       # @return [Boolean]
       def verify
+        puts("Verifiying: #{name}")
         checkout
-        command = %W[unparser #{repo_path}]
-        exclude.each do |name|
-          command.push('--ignore', repo_path.join(name).to_s)
+
+        paths = Pathname.glob(Pathname.new(repo_path).join('**/*.rb'))
+
+        driver = Mutant::Parallel.async(
+          config: Mutant::Parallel::Config.new(
+            block:            method(:verify_path),
+            jobs:             Etc.nprocessors,
+            on_process_start: ->(*) {},
+            process_name:     'unparser-corpus-test',
+            sink:             Sink.new,
+            source:           Mutant::Parallel::Source::Array.new(jobs: paths),
+            thread_name:      'unparser-corpus-test',
+            timeout:          nil
+          ),
+          world:  Mutant::WORLD
+        )
+
+        loop do
+          status = driver.wait_timeout(1)
+
+          puts("Processed: #{status.payload.total}")
+
+          status.payload.errors.each do |report|
+            puts report
+            fail
+          end
+
+          break if status.done?
         end
-        Kernel.system(*command)
+
+        true
       end
 
     private
 
+      class Sink
+        include Mutant::Parallel::Sink
+
+        attr_reader :errors, :total
+
+        def initialize
+          @errors = []
+          @total  = 0
+        end
+
+        def stop?
+          !@errors.empty?
+        end
+
+        def status
+          self
+        end
+
+        def response(response)
+          if response.error
+            Mutant::WORLD.stderr.puts(response.log)
+            fail response.error
+          end
+
+          @total += 1
+
+          if response.result
+            @errors << response.result
+          end
+        end
+      end
+
+      def verify_path(path)
+        validation = Validation.from_path(path)
+
+        if original_syntax_error?(validation) || generated_encoding_error?(validation) || validation.success?
+          return
+        end
+
+        validation.report
+      end
+
       def checkout
         TMP.mkdir unless TMP.directory?
 
@@ -50,6 +124,21 @@ module Unparser
         TMP.join(name)
       end
 
+    private
+
+      # This happens if the original source contained a non UTF charset meta comment.
+      # These are not exposed to the AST in a way unparser could know about to generate a non UTF-8
+      # target and emit that meta comment itself.
+      # For the purpose of corpus testing these cases are ignored.
+      def generated_encoding_error?(validation)
+        exception = validation.generated_node.from_left { return false }
+        exception.instance_of?(Parser::SyntaxError) && exception.message.eql?('literal contains escape sequences incompatible with UTF-8')
+      end
+
+      def original_syntax_error?(validation)
+        validation.original_node.from_left { return false }.instance_of?(Parser::SyntaxError)
+      end
+
       def system(arguments)
         return if Kernel.system(*arguments)
 

bin/parser-round-trip-test

@@ -40,7 +40,8 @@ class Test
     :rubies
   )
 
-  EXPECT_FAILURE = {}.freeze
+  STATIC_LOCAL_VARIABLES = %i[foo bar baz].freeze
+  EXPECT_FAILURE         = {}.freeze
 
   def legacy_attributes
     default_builder_attributes.reject do |attribute_name, value|
@@ -77,9 +78,9 @@ class Test
 
   # rubocop:disable Metrics/AbcSize
   def validation
-    identification   = name.to_s
+    identification = name.to_s
 
-    generated_source = Unparser.unparse_either(node)
+    generated_source = Unparser.unparse_either(node, static_local_variables: STATIC_LOCAL_VARIABLES)
       .fmap { |string| string.dup.force_encoding(parser_source.encoding).freeze }
 
     generated_node = generated_source.bind do |source|
@@ -99,7 +100,7 @@ class Test
 
   def parser
     Unparser.parser.tap do |parser|
-      %w[foo bar baz].each(&parser.static_env.method(:declare))
+      STATIC_LOCAL_VARIABLES.each(&parser.static_env.method(:declare))
     end
   end