Compiles but really really broken

shard.yml

@@ -14,6 +14,8 @@ dependencies:
   sixteen:
     github: ralsina/sixteen
     branch: main
+  cre2:
+    git: "https://git.ralsina.me/ralsina/cre2.git"
 
 crystal: ">= 1.13.0"
 

src/actions.cr

@@ -1,3 +1,5 @@
+require "xml"
+
 # These are Lexer actions. When a rule matches, it will
 # perform a list of actions. These actions can emit tokens
 # or change the state machine.
@@ -24,11 +26,12 @@ module Tartrazine
     end
 
     # ameba:disable Metrics/CyclomaticComplexity
-    def emit(match : Regex::MatchData?, lexer : Lexer, match_group = 0) : Array(Token)
+    def emit(match : MatchData,
+             lexer : Lexer, match_group = 0) : Array(Token)
       case type
       when "token"
-        raise Exception.new "Can't have a token without a match" if match.nil?
-        [Token.new(type: xml["type"], value: match[match_group])]
+        raise Exception.new "Can't have a token without a match" if match.nil? || match[0].size == 0
+        [Token.new(type: xml["type"], value: match[0])]
       when "push"
         states_to_push = xml.attributes.select { |attrib|
           attrib.name == "state"
@@ -61,35 +64,37 @@ module Tartrazine
       when "bygroups"
         # FIXME: handle
         # ><bygroups>
-        # <token type="Punctuation"/>
+        # <token type="Punctuation"/>https://github.com/google/re2/wiki/Syntax
         # None
         # <token type="LiteralStringRegex"/>
         #
         # where that None means skipping a group
         #
-        raise Exception.new "Can't have a token without a match" if match.nil?
+        raise Exception.new "Can't have a bygroups without a match" if match.nil? || match[0].size == 0
 
         # Each group matches an action. If the group match is empty,
         # the action is skipped.
         result = [] of Token
         @actions.each_with_index do |e, i|
-          next if match[i + 1]?.nil?
-          result += e.emit(match, lexer, i + 1)
+          next if match[i].size == 0
+          result += e.emit(match, lexer, i)
         end
         result
       when "using"
         # Shunt to another lexer entirely
-        return [] of Token if match.nil?
+        return [] of Token if match.nil? || match[0].size == 0
         lexer_name = xml["lexer"].downcase
-        Log.trace { "to tokenize: #{match[match_group]}" }
-        Tartrazine.lexer(lexer_name).tokenize(match[match_group], usingself: true)
+        # Log.trace { "to tokenize: #{match[match_group]}" }
+        to_tokenize = match[match_group]
+        Tartrazine.lexer(lexer_name).tokenize(to_tokenize, usingself: true)
       when "usingself"
         # Shunt to another copy of this lexer
-        return [] of Token if match.nil?
+        return [] of Token if match.nil? || match[0].size == 0
 
         new_lexer = Lexer.from_xml(lexer.xml)
-        Log.trace { "to tokenize: #{match[match_group]}" }
-        new_lexer.tokenize(match[match_group], usingself: true)
+        # Log.trace { "to tokenize: #{match[match_group]}" }
+        to_tokenize = match[match_group]
+        new_lexer.tokenize(to_tokenize, usingself: true)
       when "combined"
         # Combine two states into one anonymous state
         states = xml.attributes.select { |attrib|

src/rules.cr

@@ -1,4 +1,5 @@
 require "./actions"
+# require "cre2"
 
 # These are lexer rules. They match with the text being parsed
 # and perform actions, either emitting tokens or changing the
@@ -6,8 +7,12 @@ require "./actions"
 module Tartrazine
   # This rule matches via a regex pattern
 
+  # alias Regex = CRe2::Regex
+  # alias MatchData = CRe2::MatchDataLike | Regex::MatchData | Nil
+  alias MatchData = Regex::MatchData | Nil
+
   class Rule
-    property pattern : Regex = Re2.new ""
+    property pattern : Regex = Regex.new ""
     property actions : Array(Action) = [] of Action
     property xml : String = "foo"
 
@@ -15,7 +20,8 @@ module Tartrazine
       match = pattern.match(text, pos)
       # We don't match if the match doesn't move the cursor
       # because that causes infinite loops
-      return false, pos, [] of Token if match.nil? || match.end == 0
+
+      return false, pos, [] of Token if match.nil?
       # Log.trace { "#{match}, #{pattern.inspect}, #{text}, #{pos}" }
       tokens = [] of Token
       # Emit the tokens
@@ -23,18 +29,17 @@ module Tartrazine
         # Emit the token
         tokens += action.emit(match, lexer)
       end
-      Log.trace { "#{xml}, #{match.end}, #{tokens}" }
-      return true, match.end, tokens
+      # Log.trace { "#{xml}, #{match.end}, #{tokens}" }
+      return true, match[0].size, tokens
     end
 
     def initialize(node : XML::Node, multiline, dotall, ignorecase)
       @xml = node.to_s
-      @pattern = Re2.new(
-        node["pattern"],
-        multiline,
-        dotall,
-        ignorecase,
-        anchored: true)
+      options = Regex::Options::ANCHORED
+      options |= Regex::Options::MULTILINE if multiline
+      options |= Regex::Options::DOTALL if dotall
+      options |= Regex::Options::IGNORE_CASE if ignorecase
+      @pattern = Regex.new(node["pattern"], options)
       add_actions(node)
     end
 
@@ -86,24 +91,4 @@ module Tartrazine
       add_actions(node)
     end
   end
-
-  # This is a hack to workaround that Crystal seems to disallow
-  # having regexes multiline but not dot_all
-  class Re2 < Regex
-    @source = "fa"
-    @options = Regex::Options::None
-    @jit = true
-
-    def initialize(pattern : String, multiline = false, dotall = false, ignorecase = false, anchored = false)
-      flags = LibPCRE2::UTF | LibPCRE2::DUPNAMES |
-              LibPCRE2::UCP
-      flags |= LibPCRE2::MULTILINE if multiline
-      flags |= LibPCRE2::DOTALL if dotall
-      flags |= LibPCRE2::CASELESS if ignorecase
-      flags |= LibPCRE2::ANCHORED if anchored
-      @re = Regex::PCRE2.compile(pattern, flags) do |error_message|
-        raise Exception.new(error_message)
-      end
-    end
-  end
 end

src/tartrazine.cr

@@ -63,8 +63,6 @@ module Tartrazine
       tokens = [] of Token
       pos = 0
       matched = false
-      time = 0
-      count = 0
 
       # Respect the `ensure_nl` config option
       if text.size > 0 && text[-1] != '\n' && config[:ensure_nl] && !usingself