Execute user code blocks assigned to tokens

assets/parser.d.erb

@@ -119,6 +119,7 @@ class <%= @classname %>
             uint transition_table_index;
             uint n_transitions;
             uint accepts;
+            uint code_id;
         }
 
 <% transition_table, state_table = @lexer.build_tables %>
@@ -130,7 +131,10 @@ class <%= @classname %>
 
         private static const State states[] = [
 <% state_table.each do |state_table_entry| %>
-            State(<%= state_table_entry[:transition_table_index] %>u, <%= state_table_entry[:n_transitions] %>u, <%= state_table_entry[:accepts] %>u),
+            State(<%= state_table_entry[:transition_table_index] %>u,
+                  <%= state_table_entry[:n_transitions] %>u,
+                  <%= state_table_entry[:accepts] %>u,
+                  <%= state_table_entry[:code_id] %>u),
 <% end %>
         ];
 
@@ -166,6 +170,31 @@ class <%= @classname %>
             }
         }
 
+        /**
+         * Execute user code associated with a lexer pattern.
+         *
+         * @param code_id The ID of the user code block to execute.
+         *
+         * @return Token ID to accept, or _TOKEN_NONE if the user code does
+         *   not explicitly return a token.
+         */
+        private uint user_code(uint code_id)
+        {
+            switch (code_id)
+            {
+<% @grammar.patterns.each do |pattern| %>
+<%   if pattern.code_id %>
+            case <%= pattern.code_id %>u: {
+<%= pattern.code %>
+            } break;
+<%   end %>
+<% end %>
+            default: break;
+            }
+
+            return _TOKEN_NONE;
+        }
+
         private LexedToken attempt_lex_token()
         {
             LexedToken lt = LexedToken(m_input_row, m_input_col, 0, _TOKEN_NONE);
@@ -175,6 +204,7 @@ class <%= @classname %>
                 size_t delta_row;
                 size_t delta_col;
                 uint token;
+                uint code_id;
             }
             LexedTokenState last_accepts_info;
             last_accepts_info.token = _TOKEN_NONE;
@@ -209,6 +239,7 @@ class <%= @classname %>
                         if (states[current_state].accepts != _TOKEN_NONE)
                         {
                             attempt_info.token = states[current_state].accepts;
+                            attempt_info.code_id = states[current_state].code_id;
                             last_accepts_info = attempt_info;
                         }
                     }
@@ -220,10 +251,24 @@ class <%= @classname %>
                 }
                 if (!lex_continue)
                 {
-                    if (last_accepts_info.token != _TOKEN_NONE)
+                    bool pattern_accepted = false;
+                    uint token_to_accept = last_accepts_info.token;
+                    if (last_accepts_info.code_id != 0xFFFF_FFFFu)
                     {
-                        lt.token = last_accepts_info.token;
-                        lt.length = last_accepts_info.length;
+                        uint user_code_token = user_code(last_accepts_info.code_id);
+                        /* A return of _TOKEN_NONE from user_code() means
+                         * that the user code did not explicitly return a
+                         * token. So only override the token to return if the
+                         * user code does explicitly return a token. */
+                        if (user_code_token != _TOKEN_NONE)
+                        {
+                            token_to_accept = user_code_token;
+                        }
+                        pattern_accepted = true;
+                    }
+                    if (pattern_accepted || (token_to_accept != _TOKEN_NONE))
+                    {
+                        /* Update the input position tracking. */
                         m_input_position += last_accepts_info.length;
                         m_input_row += last_accepts_info.delta_row;
                         if (last_accepts_info.delta_row != 0u)
@@ -235,7 +280,13 @@ class <%= @classname %>
                             m_input_col += last_accepts_info.delta_col;
                         }
                     }
-                    break;
+                    if (token_to_accept != _TOKEN_NONE)
+                    {
+                        /* We have a token to accept. */
+                        lt.token = last_accepts_info.token;
+                        lt.length = last_accepts_info.length;
+                        break;
+                    }
                 }
             }
             return lt;

lib/propane/grammar.rb

@@ -12,6 +12,7 @@ class Propane
       @patterns = []
       @tokens = []
       @rules = []
+      @code_id = 0
       input = input.gsub("\r\n", "\n")
       parse_grammar(input)
     end
@@ -29,7 +30,7 @@ class Propane
           @modulename = $1
         elsif sliced = input.slice!(/\Aclass\s+(\S+)\s*;/)
           @classname = $1
-        elsif sliced = input.slice!(/\Atoken\s+(\S+?)(?:\s+(.+?))?\s*(?:;|<<\n(.*?)^>>\n)/m)
+        elsif sliced = input.slice!(/\Atoken\s+(\S+?)(?:\s+([^\n]+?))?\s*(?:;|<<\n(.*?)^>>\n)/m)
           name, pattern, code = $1, $2, $3
           if pattern.nil?
             pattern = name
@@ -39,7 +40,13 @@ class Propane
           end
           token = Token.new(name: name, id: @tokens.size, line_number: line_number)
           @tokens << token
-          pattern = Pattern.new(pattern: pattern, token: token, line_number: line_number)
+          if code
+            code_id = @code_id
+            @code_id += 1
+          else
+            code_id = nil
+          end
+          pattern = Pattern.new(pattern: pattern, token: token, line_number: line_number, code: code, code_id: code_id)
           @patterns << pattern
         elsif sliced = input.slice!(/\Atokenid\s+(\S+?)\s*;/m)
           name = $1
@@ -51,7 +58,7 @@ class Propane
         elsif sliced = input.slice!(/\Adrop\s+(\S+)\s*;/)
           pattern = $1
           @patterns << Pattern.new(pattern: pattern, line_number: line_number, drop: true)
-        elsif sliced = input.slice!(/\A(\S+)\s*->\s*(.*?)(?:;|<<\n(.*?)^>>\n)/m)
+        elsif sliced = input.slice!(/\A(\S+)\s*->\s*([^\n]*?)(?:;|<<\n(.*?)^>>\n)/m)
           rule_name, components, code = $1, $2, $3
           unless rule_name =~ /^[a-zA-Z_][a-zA-Z_0-9]*$/
             raise Error.new("Invalid rule name #{name.inspect}")

lib/propane/lexer.rb

@@ -22,10 +22,17 @@ class Propane
           else
             state.accepts.token.id
           end
+        code_id =
+          if state.accepts && state.accepts.code_id
+            state.accepts.code_id
+          else
+            0xFFFF_FFFF
+          end
         state_table << {
           transition_table_index: transition_table.size,
           n_transitions: state.transitions.size,
           accepts: accepts,
+          code_id: code_id,
         }
         state.transitions.each do |transition|
           transition_table << {

lib/propane/pattern.rb

@@ -2,6 +2,14 @@ class Propane
 
   class Pattern
 
+    # @return [String, nil]
+    #   Code block to execute when the pattern is matched.
+    attr_reader :code
+
+    # @option options [Integer, nil] :code_id
+    #   Code block ID.
+    attr_reader :code_id
+
     # @return [String, nil]
     #   Pattern.
     attr_reader :pattern
@@ -22,6 +30,10 @@ class Propane
     #
     # @param options [Hash]
     #   Optional parameters.
+    # @option options [String, nil] :code
+    #   Code block to execute when the pattern is matched.
+    # @option options [Integer, nil] :code_id
+    #   Code block ID.
     # @option options [Boolean] :drop
     #   Whether this is a drop pattern.
     # @option options [String, nil] :pattern
@@ -31,6 +43,8 @@ class Propane
     # @option options [Integer, nil] :line_number
     #   Line number where the token was defined in the input grammar.
     def initialize(options)
+      @code = options[:code]
+      @code_id = options[:code_id]
       @drop = options[:drop]
       @pattern = options[:pattern]
       @token = options[:token]

spec/propane_spec.rb

@@ -95,4 +95,19 @@ EOF
     compile("spec/test_d_lexer2.d")
     run
   end
+
+  it "executes user code when matching lexer token" do
+    write_grammar <<EOF
+token abc <<
+  writeln("abc!");
+>>
+token def;
+Start -> Abcs def;
+Abcs -> ;
+Abcs -> abc Abcs;
+EOF
+    build_parser
+    compile("spec/test_user_code.d")
+    run
+  end
 end

spec/test_user_code.d

@@ -0,0 +1,20 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "abcdef";
+    auto parser = new Testparser.Parser(cast(const(ubyte) *)input.ptr, input.length);
+    assert(parser.parse() == true);
+    writeln("pass1");
+
+    input = "abcabcdef";
+    parser = new Testparser.Parser(cast(const(ubyte) *)input.ptr, input.length);
+    assert(parser.parse() == true);
+    writeln("pass2");
+}