Refactor and improve performance of RDoc::Markup::Parser

This change introduces a wrapper of StringScanner that is aware of the
current position (column and lineno).
It has two advantages: faster and more modular.

The old code frequently runs `@input.byteslice(0, byte_offset).length`
to get the current position, but it was painfully slow.  This change
keeps track of the position at each scan, which reduces about half of
time of "Generating RI format into ..." in Ruby's `make rdoc`
(5.5 sec -> 3.0 sec).

And the old code used four instance variables (`@input`, `@line`,
`@line_pos`, and `@s`) to track the position.  This change factors them
out into MyStringScanner, so now only one variable (`@s`) is needed.

2 files changed, 65 insertions, 49 deletions

diff --git a/lib/rdoc/markup/parser.rb b/lib/rdoc/markup/parser.rb
index 14f1f6c719..600eb841ac 100644
--- a/lib/rdoc/markup/parser.rb
+++ b/lib/rdoc/markup/parser.rb
@@ -80,10 +80,6 @@ class RDoc::Markup::Parser
     @binary_input   = nil
     @current_token  = nil
     @debug          = false
-    @input          = nil
-    @input_encoding = nil
-    @line           = 0
-    @line_pos       = 0
     @s              = nil
     @tokens         = []
   end
@@ -320,13 +316,6 @@ class RDoc::Markup::Parser
   end
 
   ##
-  # The character offset for the input string at the given +byte_offset+
-
-  def char_pos byte_offset
-    @input.byteslice(0, byte_offset).length
-  end
-
-  ##
   # Pulls the next token from the stream.
 
   def get
@@ -425,14 +414,53 @@ class RDoc::Markup::Parser
   end
 
   ##
+  # A simple wrapper of StringScanner that is aware of the current column and lineno
+
+  class MyStringScanner
+    def initialize(input)
+      @line = @column = 0
+      @s = StringScanner.new input
+    end
+
+    def scan(re)
+      prev_pos = @s.pos
+      ret = @s.scan(re)
+      @column += ret.length if ret
+      ret
+    end
+
+    def unscan(s)
+      @s.pos -= s.bytesize
+      @column -= s.length
+    end
+
+    def pos
+      [@column, @line]
+    end
+
+    def newline!
+      @column = 0
+      @line += 1
+    end
+
+    def eos?
+      @s.eos?
+    end
+
+    def matched
+      @s.matched
+    end
+
+    def [](i)
+      @s[i]
+    end
+  end
+
+  ##
   # Creates the StringScanner
 
   def setup_scanner input
-    @line     = 0
-    @line_pos = 0
-    @input    = input.dup
-
-    @s = StringScanner.new input
+    @s = MyStringScanner.new input
   end
 
   ##
@@ -467,31 +495,30 @@ class RDoc::Markup::Parser
       @tokens << case
                  # [CR]LF => :NEWLINE
                  when @s.scan(/\r?\n/) then
-                   token = [:NEWLINE, @s.matched, *token_pos(pos)]
-                   @line_pos = char_pos @s.pos
-                   @line += 1
+                   token = [:NEWLINE, @s.matched, *pos]
+                   @s.newline!
                    token
                  # === text => :HEADER then :TEXT
                  when @s.scan(/(=+)(\s*)/) then
                    level = @s[1].length
-                   header = [:HEADER, level, *token_pos(pos)]
+                   header = [:HEADER, level, *pos]
 
                    if @s[2] =~ /^\r?\n/ then
-                     @s.pos -= @s[2].length
+                     @s.unscan(@s[2])
                      header
                    else
                      pos = @s.pos
                      @s.scan(/.*/)
                      @tokens << header
-                     [:TEXT, @s.matched.sub(/\r$/, ''), *token_pos(pos)]
+                     [:TEXT, @s.matched.sub(/\r$/, ''), *pos]
                    end
                  # --- (at least 3) and nothing else on the line => :RULE
                  when @s.scan(/(-{3,}) *\r?$/) then
-                   [:RULE, @s[1].length - 2, *token_pos(pos)]
+                   [:RULE, @s[1].length - 2, *pos]
                  # * or - followed by white space and text => :BULLET
                  when @s.scan(/([*-]) +(\S)/) then
-                   @s.pos -= @s[2].bytesize # unget \S
-                   [:BULLET, @s[1], *token_pos(pos)]
+                   @s.unscan(@s[2])
+                   [:BULLET, @s[1], *pos]
                  # A. text, a. text, 12. text => :UALPHA, :LALPHA, :NUMBER
                  when @s.scan(/([a-z]|\d+)\. +(\S)/i) then
                    # FIXME if tab(s), the column will be wrong
@@ -500,7 +527,7 @@ class RDoc::Markup::Parser
                    # before (and provide a check for that at least in debug
                    # mode)
                    list_label = @s[1]
-                   @s.pos -= @s[2].bytesize # unget \S
+                   @s.unscan(@s[2])
                    list_type =
                      case list_label
                      when /[a-z]/ then :LALPHA
@@ -509,24 +536,24 @@ class RDoc::Markup::Parser
                      else
                        raise ParseError, "BUG token #{list_label}"
                      end
-                   [list_type, list_label, *token_pos(pos)]
+                   [list_type, list_label, *pos]
                  # [text] followed by spaces or end of line => :LABEL
                  when @s.scan(/\[(.*?)\]( +|\r?$)/) then
-                   [:LABEL, @s[1], *token_pos(pos)]
+                   [:LABEL, @s[1], *pos]
                  # text:: followed by spaces or end of line => :NOTE
                  when @s.scan(/(.*?)::( +|\r?$)/) then
-                   [:NOTE, @s[1], *token_pos(pos)]
+                   [:NOTE, @s[1], *pos]
                  # >>> followed by end of line => :BLOCKQUOTE
                  when @s.scan(/>>> *(\w+)?$/) then
-                   [:BLOCKQUOTE, @s[1], *token_pos(pos)]
+                   [:BLOCKQUOTE, @s[1], *pos]
                  # anything else: :TEXT
                  else
                    @s.scan(/(.*?)(  )?\r?$/)
-                   token = [:TEXT, @s[1], *token_pos(pos)]
+                   token = [:TEXT, @s[1], *pos]
 
                    if @s[2] then
                      @tokens << token
-                     [:BREAK, @s[2], *token_pos(pos + @s[1].length)]
+                     [:BREAK, @s[2], pos[0] + @s[1].length, pos[1]]
                    else
                      token
                    end
@@ -537,16 +564,6 @@ class RDoc::Markup::Parser
   end
 
   ##
-  # Calculates the column (by character) and line of the current token based
-  # on +byte_offset+.
-
-  def token_pos byte_offset
-    offset = char_pos byte_offset
-
-    [offset - @line_pos, @line]
-  end
-
-  ##
   # Returns the current token to the token stream
 
   def unget
diff --git a/lib/rdoc/tom_doc.rb b/lib/rdoc/tom_doc.rb
index 625a6b5cfa..e161fcf42f 100644
--- a/lib/rdoc/tom_doc.rb
+++ b/lib/rdoc/tom_doc.rb
@@ -242,19 +242,18 @@ class RDoc::TomDoc < RDoc::Markup::Parser
 
       @tokens << case
                  when @s.scan(/\r?\n/) then
-                   token = [:NEWLINE, @s.matched, *token_pos(pos)]
-                   @line_pos = char_pos @s.pos
-                   @line += 1
+                   token = [:NEWLINE, @s.matched, *pos]
+                   @s.newline!
                    token
                  when @s.scan(/(Examples|Signature)$/) then
-                   @tokens << [:HEADER, 3, *token_pos(pos)]
+                   @tokens << [:HEADER, 3, *pos]
 
-                   [:TEXT, @s[1], *token_pos(pos)]
+                   [:TEXT, @s[1], *pos]
                  when @s.scan(/([:\w][\w\[\]]*)[ ]+- /) then
-                   [:NOTE, @s[1], *token_pos(pos)]
+                   [:NOTE, @s[1], *pos]
                  else
                    @s.scan(/.*/)
-                   [:TEXT, @s.matched.sub(/\r$/, ''), *token_pos(pos)]
+                   [:TEXT, @s.matched.sub(/\r$/, ''), *pos]
                  end
     end