Merge pull request #1755 from Shopify/non-parsing-comment

Don't parse nodes inside a comment tag

lib/liquid/block_body.rb

@@ -6,6 +6,7 @@ module Liquid
   class BlockBody
     LiquidTagToken      = /\A\s*(#{TagName})\s*(.*?)\z/o
     FullToken           = /\A#{TagStart}#{WhitespaceControl}?(\s*)(#{TagName})(\s*)(.*?)#{WhitespaceControl}?#{TagEnd}\z/om
+    FullTokenPossiblyInvalid = /\A(.*)#{TagStart}#{WhitespaceControl}?\s*(\w+)\s*(.*)?#{WhitespaceControl}?#{TagEnd}\z/om
     ContentOfVariable   = /\A#{VariableStart}#{WhitespaceControl}?(.*?)#{WhitespaceControl}?#{VariableEnd}\z/om
     WhitespaceOrNothing = /\A\s*\z/
     TAGSTART            = "{%"

lib/liquid/tags/comment.rb

@@ -15,6 +15,8 @@ module Liquid
   #   {% endcomment %}
   # @liquid_syntax_keyword content The content of the comment.
   class Comment < Block
+    TAG_DELIMITER = /\A(.*)#{TagStart}#{WhitespaceControl}?\s*(endcomment)\s*(.*)?#{WhitespaceControl}?#{TagEnd}\z/om
+
     def render_to_output_buffer(_context, output)
       output
     end
@@ -25,6 +27,71 @@ module Liquid
     def blank?
       true
     end
+
+    private
+
+    def parse_body(body, tokenizer)
+      if parse_context.depth >= MAX_DEPTH
+        raise StackLevelError, "Nesting too deep"
+      end
+
+      parse_context.depth += 1
+      comment_tag_depth = 1
+
+      begin
+        # Consume tokens without creating child nodes.
+        # The children tag doesn't require to be a valid Liquid except the comment and raw tag.
+        # The child comment and raw tag must be closed.
+        while (token = tokenizer.send(:shift))
+          tag_name = if tokenizer.for_liquid_tag
+            next if token.empty? || token.match?(BlockBody::WhitespaceOrNothing)
+
+            tag_name_match = BlockBody::LiquidTagToken.match(token)
+
+            next if tag_name_match.nil?
+
+            tag_name_match[1]
+          elsif TAG_DELIMITER.match?(token)
+            # aggressively match comment delimiter
+            "endcomment"
+          elsif token =~ BlockBody::FullToken && Regexp.last_match(2) == "comment"
+            # aggressively match comment tag
+            "comment"
+          else
+            tag_name_match = BlockBody::FullTokenPossiblyInvalid.match(token)
+
+            next if tag_name_match.nil?
+
+            tag_name_match[2]
+          end
+
+          case tag_name
+          when "raw"
+            parse_raw_tag_body(tokenizer)
+          when "comment"
+            comment_tag_depth += 1
+          when "endcomment"
+            comment_tag_depth -= 1
+
+            return false if comment_tag_depth.zero?
+          end
+        end
+
+        raise_tag_never_closed(block_name)
+      ensure
+        parse_context.depth -= 1
+      end
+
+      false
+    end
+
+    def parse_raw_tag_body(tokenizer)
+      while (token = tokenizer.send(:shift))
+        return if token =~ BlockBody::FullTokenPossiblyInvalid && "endraw" == Regexp.last_match(2)
+      end
+
+      raise_tag_never_closed("raw")
+    end
   end
 
   Template.register_tag('comment', Comment)

lib/liquid/tags/raw.rb

@@ -14,7 +14,6 @@ module Liquid
   # @liquid_syntax_keyword expression The expression to be output without being rendered.
   class Raw < Block
     Syntax = /\A\s*\z/
-    FullTokenPossiblyInvalid = /\A(.*)#{TagStart}#{WhitespaceControl}?\s*(\w+)\s*(.*)?#{WhitespaceControl}?#{TagEnd}\z/om
 
     def initialize(tag_name, markup, parse_context)
       super
@@ -25,7 +24,7 @@ module Liquid
     def parse(tokens)
       @body = +''
       while (token = tokens.shift)
-        if token =~ FullTokenPossiblyInvalid && block_delimiter == Regexp.last_match(2)
+        if token =~ BlockBody::FullTokenPossiblyInvalid && block_delimiter == Regexp.last_match(2)
           parse_context.trim_whitespace = (token[-3] == WhitespaceControl)
           @body << Regexp.last_match(1) if Regexp.last_match(1) != ""
           return